feat: 建立版本化模型目录与规范字段生成链路 (#116)

AGENTS.md

@@ -56,7 +56,9 @@
 
 ## 修改与来源记录规则
 
-- 数据主文件仍是 `compute/**` JSON。共享规格不包含价格；接入面推理 effort、回放能力不能混入共享规格；官网没有支持的新字段先查客户端和 frozen schema。
+- 模型目录唯一作者数据是 `catalog/v2/catalog.json`，退役注册在 `catalog/v2/retirements.json`；`compute/v2/**` 以及根 `compute/providers`、`compute/coding-plans`、`compute/model-specs` 的模型正文由 `npm run catalog:compile` 生成，禁止手工双写。索引、service-map、pricing 等非投影文件仍分别维护。共享规格不包含价格；接入面推理 effort、回放能力不能混入共享规格；官网没有支持的新字段先查客户端和 frozen schema。
+- 每次目录更新先检查 `FIELD-DECISIONS`、`SPEC-DECISIONS` 和 `ADAPTER-DECISIONS`，精确绑定或显式例外必须保留。未知事实维持 unknown；未适配协议保持禁止调用，不能用将来的版本号替代缺失适配器。生成后运行 `npm run catalog:check`，核对根 V1 投影保真及 340 条实际解析对照。
+- 候选目录 `publication.state=candidate` 不启用 V2。切换 `published` 前，必须检查正式客户端 tag、合入代码和发布产物，填写真实 `requiredClientVersion`；只完成代码合并或构建不能作为已经正式发布的证据。
 - 来源使用细分目录：`docs/model-sources/providers/<supplier>/SOURCES.md` 保存官网证据；`models/<model-id>.md` 每个模型独立记录；`access/<category>--<config-id>.md` 每个接入面独立记录；README 只保留导航。
 - 一个模型在原厂、订阅、套餐和聚合平台上的差异分段记录，不合并价格/窗口/可用性。同名不等于同一接入合同；文件名规范见维护指南。
 - 基于已有文件增量编辑，保留既有证据、日期、待核实项和历史说明。批量迁移目录只能迁移记录，不能凭迁移提高核验状态或刷新核验日期。

__tests__/catalog-v2.test.mjs

@@ -0,0 +1,148 @@
+import test from 'node:test';
+import assert from 'node:assert/strict';
+import fs from 'node:fs';
+import path from 'node:path';
+import os from 'node:os';
+import {fileURLToPath} from 'node:url';
+import {compileCatalog,projectV1,stableJson,validateCatalog,synchronizeRootProjection} from '../scripts/catalog-v2/compile.mjs';
+import {resolveEffectiveModel} from '../scripts/catalog-v2/effective-model.mjs';
+import {classifyExtra} from '../scripts/catalog-v2/migrate.mjs';
+const root=path.resolve(path.dirname(fileURLToPath(import.meta.url)),'..');
+const catalog=JSON.parse(fs.readFileSync(path.join(root,'catalog/v2/catalog.json'),'utf8'));
+const spec={modelRef:'vendor:exact',identity:{vendor:'vendor',canonicalModelId:'exact'},facts:{contextWindow:1000,maxInputTokens:800,maxOutputTokens:500,capabilities:['chat'],constraints:{thinkingOnly:true}},routing:{reasoning:{supportedModes:['off','low','high']}},lifecycle:{status:'active'}};
+const model={apiModelId:'exact',modelRef:spec.modelRef,binding:{kind:'exact'},overrides:{contextWindow:{value:700,reason:'platform-cap'}},request:{reasoning:{supportedEfforts:['none','low','high','max'],defaultEffort:'none'}},billing:[{amount:-1,currency:'USD',unit:'million-tokens'}],lifecycle:{status:'active'},metadata:{recordedExtra:{capabilities:['vision']}}};
+const options={publicationState:'published',requiredClientVersion:'10.0.200',clientVersion:'10.0.200',connection:{apiFormat:'openai-completions'}};
+test('canonical projections reproduce each V1 field including missing fields and metadata',()=>{
+ for(const [file,data]of Object.entries(projectV1(catalog))) assert.equal(stableJson(data),stableJson(JSON.parse(fs.readFileSync(path.join(root,file),'utf8'))),file);
+});
+test('compiler is deterministic and does not depend on provider files',()=>assert.equal(stableJson(compileCatalog(catalog)),stableJson(compileCatalog(structuredClone(catalog)))));
+test('candidate and invalid versions fail closed while budget/output remain independent',()=>{
+ assert.ok(resolveEffectiveModel(spec,model,{...options,publicationState:'candidate'}).blockedReasons.includes('catalog-not-published'));
+ assert.ok(resolveEffectiveModel(spec,model,{...options,clientVersion:'garbage'}).blockedReasons.includes('invalid-client-version'));
+ assert.ok(resolveEffectiveModel(spec,model,{...options,clientVersion:'10.0.199'}).blockedReasons.includes('client-update-required'));
+ const effective=resolveEffectiveModel(spec,model,options);assert.equal(effective.facts.contextWindow,700);assert.equal(effective.facts.maxOutputTokens,500);assert.equal(effective.billing[0].amount,-1);
+});
+test('resolver uses explicit access set without capability union or metadata consumption',()=>{
+ const override=structuredClone(model);override.overrides.capabilities={value:['chat'],reason:'no vision access'};
+ const result=resolveEffectiveModel(spec,override,{...options,userOverrides:{capabilities:['vision'],contextWindow:2000,maxInputTokens:600}});
+ assert.deepEqual(result.facts.capabilities,['chat']);assert.equal(result.facts.contextWindow,700);assert.equal(result.facts.maxInputTokens,600);assert.deepEqual(result.request.reasoning.supportedEfforts,['low','high']);assert.equal(result.request.reasoning.defaultEffort,undefined);
+});
+test('misbound identities and retired records cannot authorize calls',()=>{
+ assert.ok(resolveEffectiveModel(spec,{...model,binding:{kind:'dynamic'}},options).blockedReasons.includes('unresolved-model-binding'));
+ assert.ok(resolveEffectiveModel({...spec,lifecycle:{status:'historical'}},model,options).blockedReasons.includes('retired-model'));
+});
+test('legacy runtime keys and unexplained overrides are refused',()=>{
+ const copy=structuredClone(catalog);copy.accesses[0].models[0].request.thinkingDefault=true;assert.throws(()=>validateCatalog(copy),/Old runtime key|Unknown managed request key|Invalid V2 schema/);
+ assert.throws(()=>resolveEffectiveModel(spec,{...model,overrides:{contextWindow:{value:500}}},options),/Unexplained/);
+});
+test('all inventoried fields have one owner and nonconsumed defaults remain metadata',()=>{
+ const decisions=JSON.parse(fs.readFileSync(path.join(root,'catalog/v2/field-decisions.json'),'utf8'));
+ assert.equal(decisions.length,149);assert.equal(new Set(decisions.map(d=>d.path)).size,149);
+ assert.equal(classifyExtra('thinkingDefault',['fake-read']), 'metadata');assert.equal(classifyExtra('cachePricing'), 'billing');
+ const base=compileCatalog(catalog).output;for(const s of base.specs)assert.equal(s.metadata.projection,undefined);
+});
+
+test('X2 independent budget is not a total context assumption and V1 remains frozen',()=>{
+ const access=catalog.accesses.find(a=>a.metadata.sourceFile==='compute/providers/xunfei-x2.json');
+ const model=access.models.find(m=>m.apiModelId==='spark-x');
+ assert.equal(model.overrides.maxInputTokens.value,65536);assert.equal(model.overrides.contextWindow.value,null);assert.equal(model.overrides.maxOutputTokens.value,131072);
+ assert.equal(projectV1(catalog)[access.metadata.sourceFile].models[0].contextWindow,65536);
+});
+test('all dependency gates and lifecycle deadline enforce actual calls',()=>{
+ for(const layer of ['provider','protocol','spec']){
+  const newer='10.0.201'; const s=structuredClone(spec),m=structuredClone(model),o={...options,connection:{apiFormat:'anthropic-messages'}};
+  if(layer==='provider')o.connection.requiredClientVersion=newer;
+  if(layer==='protocol')m.protocol={state:'supported',requiredClientVersion:newer};
+  if(layer==='spec')s.requiredClientVersion=newer;
+  assert.ok(resolveEffectiveModel(s,m,o).blockedReasons.includes('client-update-required'),layer);
+ }
+ const dated={...spec,lifecycle:{status:'deprecated',retiredAt:'2027-04-01'}};
+ assert.equal(resolveEffectiveModel(dated,model,{...options,now:'2027-03-31T23:59:59Z'}).eligible,true);
+ assert.ok(resolveEffectiveModel(dated,model,{...options,now:'2027-04-01T00:00:00Z'}).blockedReasons.includes('retired-model'));
+});
+test('all intrinsic hard constraints survive access false and scoped adaptive applies to dynamic exact identity',()=>{
+ const s={...spec,facts:{...spec.facts,constraints:{thinkingOnly:true,samplingParametersDeprecated:true,forcedToolChoiceUnsupported:true}},protocolProfiles:[{apiFormats:['anthropic-messages'],request:{adaptiveThinking:true}}]};
+ const m={...model,request:{adaptiveThinking:false,constraints:{thinkingOnly:false,samplingParametersDeprecated:false,forcedToolChoiceUnsupported:false}}};
+ const result=resolveEffectiveModel(s,m,{...options,connection:{apiFormat:'anthropic-messages'}});
+ assert.deepEqual(result.facts.constraints,s.facts.constraints);assert.equal(result.request.adaptiveThinking,true);
+ assert.equal(resolveEffectiveModel(s,{...m,request:{}},{...options,connection:{apiFormat:'openai-completions'}}).request.adaptiveThinking,undefined);
+});
+test('unverified budget expansion cannot enlarge effective hard budget; official platform contract can',()=>{
+ const s={...spec,facts:{...spec.facts,maxOutputTokens:20}},m={...model,overrides:{maxOutputTokens:{value:2000,reason:'old-assumption',verification:'inherited-not-reverified'}}};
+ const result=resolveEffectiveModel(s,m,options);assert.equal(result.facts.maxOutputTokens,20);assert.equal(result.pendingOverrides.length,1);assert.equal(result.automaticRoutingEligible,false);
+ const verified=structuredClone(m);verified.overrides.maxOutputTokens.verification='officially-verified-access-contract';assert.equal(resolveEffectiveModel(s,verified,options).facts.maxOutputTokens,2000);
+});
+
+test('typed schema rejects unknown runtime fields, unsafe versions and invalid billing dimensions',()=>{
+ const corruptions=[
+  c=>c.specs[0].facts.unreadBehavior=true,
+  c=>c.specs[0].facts.maxOutputTokens='128000',
+  c=>c.accesses[0].connection.hiddenProtocol='anything',
+  c=>c.accesses[0].connection.requiredClientVersion='tomorrow',
+  c=>c.accesses[0].models[0].request.reasoning.supportedEfforts=['made-up'],
+  c=>c.accesses.flatMap(a=>a.models).find(m=>m.billing.length).billing[0].amount=true,
+  c=>c.accesses.flatMap(a=>a.models).find(m=>m.billing.length).billing[0].unit='anything',
+  c=>c.specs[0].lifecycle.retiredAt='not-a-date',
+  c=>c.retirements.push({id:'invalid',scope:'endpoint'})
+ ];
+ for(const corrupt of corruptions){const c=structuredClone(catalog);corrupt(c);assert.throws(()=>validateCatalog(c),/Invalid V2 schema/);}
+});
+
+test('explicit access-local/dynamic contracts stay callable on known adapters but never auto route or inherit',()=>{
+ for(const kind of ['access-local','dynamic']){
+  const m={...model,modelRef:null,binding:{kind},overrides:{contextWindow:{value:123,reason:'existing-access-contract'}}};
+  const result=resolveEffectiveModel(spec,m,options);assert.equal(result.eligibleForCall,true);assert.equal(result.eligibleForAutoRouting,false);assert.equal(result.manualOnly,true);assert.equal(result.facts.contextWindow,123);assert.equal(result.facts.maxOutputTokens,undefined);assert.equal(result.identity,null);
+  assert.ok(resolveEffectiveModel(spec,m,{...options,connection:{apiFormat:'not-implemented'}}).blockedReasons.includes('unknown-adapter'));
+ }
+});
+test('all 36 confirmed unimplemented native models remain visible and cannot call or auto route',()=>{
+ const decisions=JSON.parse(fs.readFileSync(path.join(root,'docs/plans/catalog-contract-migration/ADAPTER-DECISIONS.json'),'utf8')).decisions;
+ assert.equal(decisions.length,36);let count=0;
+ for(const decision of decisions){
+  const access=catalog.accesses.find(a=>a.metadata.sourceFile===decision.file),m=access.models.find(m=>m.apiModelId===decision.apiModelId),s=catalog.specs.find(s=>s.modelRef===m.modelRef);
+  assert.equal(m.protocol.state,'unsupported',`${decision.file}:${m.apiModelId}`);assert.ok(m.protocol.reason);assert.equal(m.protocol.requiredClientVersion,undefined);
+  const result=resolveEffectiveModel(s,m,{...options,connection:access.connection});assert.equal(result.eligibleForCall,false);assert.equal(result.eligibleForAutoRouting,false);assert.ok(result.blockedReasons.includes('unsupported-adapter'));count++;
+ }
+ assert.equal(count,36);
+});
+test('actual outgoing aliases survive canonical transport IDs and V1 projection',()=>{
+ const expected={ 'glm-5.2':'glm-5.2[1m]','doubao-seed-2.1-pro':'doubao-seed-2-1-pro-260628','doubao-seed-2.1-turbo':'doubao-seed-2-1-turbo-260628'};
+ for(const [stable,wire]of Object.entries(expected)){
+  const access=catalog.accesses.find(a=>a.models.some(m=>m.apiModelId===stable&&m.request.upstreamModelId===wire));const m=access.models.find(m=>m.apiModelId===stable),s=catalog.specs.find(s=>s.modelRef===m.modelRef);
+  assert.equal(m.request.upstreamModelId,wire);assert.equal(m.metadata.attributes.apiModelId,undefined);assert.equal(resolveEffectiveModel(s,m,{...options,connection:access.connection}).apiModelId,wire);
+  assert.equal(projectV1(catalog)[access.metadata.sourceFile].models.find(m=>m.modelName===stable).apiModelId,wire);
+ }
+});
+test('search unit price ownership is billing at the leaf, while protocol group remains request',()=>{
+ assert.equal(classifyExtra('serverSideWebSearch.searchRequestPriceUsd'),'billing');
+ assert.equal(classifyExtra('serverSideWebSearch',['consumer.ts']),'request');
+ const decisions=JSON.parse(fs.readFileSync(path.join(root,'catalog/v2/field-decisions.json'),'utf8'));assert.equal(decisions.find(x=>x.path==='extra.serverSideWebSearch.searchRequestPriceUsd').owner,'billing');
+});
+test('six product-deprecated Seed choices retain explicit calls without reentering selectors or automatic routing',()=>{
+ const access=catalog.accesses.find(a=>a.id==='provider-volcengine-001'),models=access.models.filter(m=>m.lifecycle.deprecated===true);assert.equal(models.length,6);
+ for(const model of models){const spec=catalog.specs.find(s=>s.modelRef===model.modelRef);const result=resolveEffectiveModel(spec,model,{...options,connection:access.connection});assert.equal(result.eligibleForCall,true);assert.equal(result.selectable,false);assert.equal(result.eligibleForAutoRouting,false);assert.ok(!result.blockedReasons.includes('retired-model'));}
+});
+test('implemented Responses format handles explicit override and mandatory Responses-only without accepting arbitrary adapters',()=>{
+ const direct={...model,request:{apiFormatOverride:'openai-responses'}};assert.equal(resolveEffectiveModel(spec,direct,options).eligibleForCall,true);
+ const only={...model,request:{responsesOnly:true}};const result=resolveEffectiveModel(spec,only,options);assert.equal(result.apiFormat,'openai-responses');assert.equal(result.request.apiFormatOverride,'openai-responses');assert.equal(result.eligibleForCall,true);
+ assert.ok(resolveEffectiveModel(spec,{...model,request:{apiFormatOverride:'unimplemented-native-api'}},options).blockedReasons.includes('unknown-adapter'));
+ const listed=catalog.accesses.flatMap(a=>a.models).filter(m=>m.request.responsesOnly);assert.equal(listed.length,3);for(const m of listed)assert.equal(m.metadata.recordedExtra.responsesOnly,undefined);
+});
+
+test('root V1 is generated from canonical, rejects manual drift and never rewrites indexes or unrelated files',()=>{
+ const temp=fs.mkdtempSync(path.join(os.tmpdir(),'catalog-root-projection-'));
+ try{
+  const projection=projectV1(catalog),file=Object.keys(projection)[0];
+  const indexes=['compute/providers/_index.json','compute/model-specs/_index.json'];
+  for(const index of indexes){fs.mkdirSync(path.dirname(path.join(temp,index)),{recursive:true});fs.writeFileSync(path.join(temp,index),'{"sentinel":"user-index"}');}
+  fs.mkdirSync(path.join(temp,'logs'));fs.writeFileSync(path.join(temp,'logs','user.log'),'preserve');
+  synchronizeRootProjection(temp,projection);synchronizeRootProjection(temp,projection,{check:true});
+  const changed=JSON.parse(fs.readFileSync(path.join(temp,file),'utf8'));changed.specs[0].spec.contextWindow=42;fs.writeFileSync(path.join(temp,file),JSON.stringify(changed));
+  assert.throws(()=>synchronizeRootProjection(temp,projection,{check:true}),/Stale V1 projection/);
+  synchronizeRootProjection(temp,projection);synchronizeRootProjection(temp,projection,{check:true});
+  assert.equal(stableJson(JSON.parse(fs.readFileSync(path.join(temp,file),'utf8'))),stableJson(projection[file]));
+  for(const index of indexes)assert.equal(fs.readFileSync(path.join(temp,index),'utf8'),'{'+'"sentinel":"user-index"}');
+  assert.equal(fs.readFileSync(path.join(temp,'logs/user.log'),'utf8'),'preserve');
+  assert.throws(()=>synchronizeRootProjection(temp,{'compute/providers/_index.json':{}},{check:true}),/Invalid V1 publication path/);
+ }finally{fs.rmSync(temp,{recursive:true,force:true});}
+});

__tests__/spec-identity-governance.test.mjs

@@ -0,0 +1,27 @@
+import {test} from 'node:test'
+import assert from 'node:assert/strict'
+import {readFileSync} from 'node:fs'
+const read=path=>JSON.parse(readFileSync(new URL('../'+path,import.meta.url),'utf8'))
+test('GPT-5.4 base cannot swallow distinct Nano, Pro or future variants',()=>{
+ const rows=read('compute/model-specs/openai.json').specs
+ const base=rows.find(x=>x.id==='gpt-5.4')
+ assert.equal(base.match.patterns,undefined)
+ assert.ok(base.match.exact.includes('gpt-5.4'))
+ for(const [id,window] of [['gpt-5.4-nano',400000],['gpt-5.4-pro',1050000]]){
+  const spec=rows.find(x=>x.id===id);assert.equal(spec.spec.contextWindow,window);assert.equal(spec.spec.maxOutputTokens,128000);assert.ok(spec.match.exact.includes(id))
+ }
+})
+test('identity-only coding variants cannot inherit family context, output or routing',()=>{
+ const rows=read('compute/model-specs/qwen.json').specs
+ for(const id of ['qwen3-coder-plus','qwen3-coder-next']){
+  const spec=rows.find(x=>x.id===id);assert.ok(spec.match.exact.includes(id));assert.deepEqual(Object.keys(spec.spec),['description']);assert.equal(spec.routing,undefined)
+ }
+})
+test('all recorded baseline differences and exceptions have explicit decisions',()=>{
+ const d=read('docs/plans/catalog-contract-migration/SPEC-DECISIONS.json')
+ assert.equal(d.bindings.length,24);assert.equal(d.differences.length,45);assert.equal(d.specOnly.length,58)
+ for(const row of [...d.differences,...d.specOnly]){assert.ok(row.classification);assert.ok(row.decision);assert.ok(row.reason)}
+ for(const row of d.accessExceptions){assert.equal(row.inheritFamily,false);assert.equal(row.automaticRouting,false);assert.equal(row.effectiveBinding,null)}
+ assert.ok(d.differences.some(r=>r.id==='gpt-5.4-nano'&&!r.currentDifferences.length))
+ assert.ok(d.differences.some(r=>r.id==='MiniMax-M3'&&r.verifiedOverride===false))
+})

catalog/v2/README.md

@@ -0,0 +1,25 @@
+# V2作者数据
+
+`catalog.json` 是迁移后的唯一模型/接入作者数据。`field-decisions.json` 为149路径静态消费核查记录，`retirements.json` 为独立退役注册（若存在）。编译器只读取这里，不从旧Provider再次推导。
+
+- `node scripts/catalog-v2/compile.mjs` 生成 `compute/v2/catalog.json`、版本manifest、诊断、`compute/v2/compatibility/**`，并将同一冻结V1投影发布到根 `compute/providers`、`compute/coding-plans`、`compute/model-specs` 正文。索引、manifest、来源记录仍按各自维护规则更新，编译器不重写。
+- `node scripts/catalog-v2/compile.mjs --check` 校验可复现结果、根V1完整字段与canonical投影一致、frozen schema、旧键回流、绑定和覆盖说明。根V1任何手工数值更改都会失败，不能将那里作为第二个作者位置。根文件已有序列化不影响语义一致时保留，避免无意义的key顺序变更。
+- `migrate.mjs --write --client-root <client repo>` 是首次摄取工具，**不能作为日常编译入口**。重新摄取会丢掉人工canonical编辑，必须先保存基线、review差异；当前批次允许在并行官网规格修正结束后重新摄取一次。
+
+当前 `publication.state=candidate` 与 `requiredClientVersion=null` 是刻意失败关闭，尚未启用线上V2。正式发布版本和行为验收后才能将发布状态及最低版本一起更新；未知协议/动态身份不能借发布变成精确可调用。
+
+metadata不消费。V1布局及历史未消费extra保存在canonical metadata，完整回投影；运行时产物删除布局信息。迁移不刷新官方核验日期、不提升proof等级。价格币种和原数值保持，复杂价阶/注释保留sourceKey和源单位；没有转换为零或统一按token收费。
+
+客户端用户配置迁移不由这个仓库对用户磁盘执行，由客户端迁移器保留秘密、用户覆写与一次回滚基线；本工具只迁移受管作者数据。
+
+## 实施后的执行边界
+
+- `access-local` / `dynamic` 是明确的接入合同，不虚构原厂Spec。已知适配器上的现存显式模型可手动调用；保留全部340条记录，禁止Spec/family继承、Smart自动路由。未知/未实现适配器仍门控，不将无Spec条目静默丢弃。
+- 原厂硬约束（强制思考、拒绝采样、拒绝强制tool_choice）不能被接入false放松。未核的更大输入/窗口/输出不突破已核Spec限制；保留原候选与pending解释，只允许本次已核的独立平台合同扩大。
+- `protocolProfiles` 保存严格作用于指定协议的合同，例如8条adaptive thinking仅适用于anthropic-messages。V1 adaptive副本从profile生成，动态云目录的精确身份也必须经过同一协议resolver。
+- 只有执行性 `lifecycle.retiredAt` 会按UTC日00:00门控；旧且缺字段证明的日期保存在 `metadata.recordedFacts`，不冒充正式停服执行证据。Nano官网退役日期已纳入执行性生命周期。
+- `compute/v2/resolver-parity.json` 为真实340条配置生成消费摘要；其中999.0.0/0.0.0只用于测试，不是实际发布或最低版本。客户端用相同纯resolver与固定时间验证，不能维护第二套不同算法。
+
+## 日常修改顺序
+
+只编辑 `catalog/v2` 中相应事实、接入覆盖或计价，运行 `npm run catalog:compile` 同步根V1与V2产物；随后更新对应官方来源记录、规格索引缓存标记及发布manifest，运行全套校验。禁止改根Provider/Spec后通过重新import制造双写或提升证明等级。`catalog:check` 会拒绝根V1漂移；用户本地compute与秘密不在生成范围。

catalog/v2/retirements.json

@@ -0,0 +1,24 @@
+{
+  "description": "本产品已下架的受管预置接入。规则限定Provider身份与原始端点；不证明厂家根域停服，不删除用户凭据。",
+  "rules": [
+    {
+      "id": "config-center-lingyiwanwu-access-retirement",
+      "kind": "product-access-retired",
+      "scope": "managed-provider",
+      "providerId": "provider-lingyiwanwu-001",
+      "origin": {
+        "exactHost": "api.lingyiwanwu.com",
+        "pathPrefix": "/v1"
+      },
+      "effectiveAt": "2026-08-28T06:52:05Z",
+      "reason": "config-center #84 已移除零一万物预置接入；存量受管配置不再作为官方可调用接入。",
+      "evidence": {
+        "kind": "product-repository",
+        "url": "https://github.com/desirecore/config-center/commit/a998e18e6c4bdcc161f0baf82cbf534cc9d53627",
+        "checkedAt": "2026-10-03"
+      },
+      "preserveCredentials": true,
+      "vendorDomainShutdownConfirmed": false
+    }
+  ]
+}

compute/model-specs/_index.json

@@ -1,5 +1,5 @@
 {
-  "description": "模型规格库加载顺序索引，以及智能路由三档的同域定义。模型级路由策略仅写入各 ModelSpec 的 routing 字段，不另建 smart-routing 目录。 数据批次 126：官方字段复核与临时、历史规格边界。",
+  "description": "模型规格库加载顺序索引，以及智能路由三档的同域定义。模型级路由策略仅写入各 ModelSpec 的 routing 字段，不另建 smart-routing 目录。 数据批次 126：官方字段复核与临时、历史规格边界。 精确身份治理：隔离Nano/实时/固定变体，identity-only不继承family参数；发布批次由manifest统一标记。",
   "order": [
     "anthropic",
     "openai",

compute/model-specs/openai.json

@@ -341,10 +341,8 @@
       "family": "gpt-5",
       "match": {
         "exact": [
-          "gpt-5.4"
+          "gpt-5.4",
-        ],
+          "gpt-5.4-2026-03-05"
-        "patterns": [
-          "gpt-5.4*"
         ]
       },
       "spec": {
@@ -1242,6 +1240,82 @@
           "defaultMode": "auto"
         }
       }
+    },
+    {
+      "id": "gpt-5.4-nano",
+      "displayName": "GPT-5.4 Nano",
+      "family": "gpt-5.4-nano",
+      "match": {
+        "exact": [
+          "gpt-5.4-nano",
+          "gpt-5.4-nano-2026-03-17"
+        ]
+      },
+      "spec": {
+        "contextWindow": 400000,
+        "maxOutputTokens": 128000,
+        "description": "GPT-5.4 Nano；官网确认400000总窗口、128000最大输出。2026-10-01宣布deprecated，停止服务日2027-04-01；当前不据此退役。独立输入上限未由本页确认。"
+      }
+    },
+    {
+      "id": "gpt-realtime-2.1",
+      "displayName": "gpt-realtime-2.1",
+      "family": "gpt-realtime-2.1",
+      "match": {
+        "exact": [
+          "gpt-realtime-2.1"
+        ]
+      },
+      "spec": {
+        "contextWindow": 128000,
+        "maxOutputTokens": 32000,
+        "description": "官方精确型号与窗口/输出已核；独立实时协议适配与账号验收由接入面决定，不继承通用Chat合同。"
+      }
+    },
+    {
+      "id": "gpt-realtime-2.1-mini",
+      "displayName": "gpt-realtime-2.1-mini",
+      "family": "gpt-realtime-2.1-mini",
+      "match": {
+        "exact": [
+          "gpt-realtime-2.1-mini"
+        ]
+      },
+      "spec": {
+        "contextWindow": 128000,
+        "maxOutputTokens": 32000,
+        "description": "官方精确型号与窗口/输出已核；独立实时协议适配与账号验收由接入面决定，不继承通用Chat合同。"
+      }
+    },
+    {
+      "id": "gpt-realtime-translate",
+      "displayName": "gpt-realtime-translate",
+      "family": "gpt-realtime-translate",
+      "match": {
+        "exact": [
+          "gpt-realtime-translate"
+        ]
+      },
+      "spec": {
+        "contextWindow": 16000,
+        "maxOutputTokens": 2000,
+        "description": "官方精确型号与窗口/输出已核；独立实时协议适配与账号验收由接入面决定，不继承通用Chat合同。"
+      }
+    },
+    {
+      "id": "gpt-5.4-pro",
+      "displayName": "GPT-5.4 Pro",
+      "family": "gpt-5.4-pro",
+      "match": {
+        "exact": [
+          "gpt-5.4-pro"
+        ]
+      },
+      "spec": {
+        "contextWindow": 1050000,
+        "maxOutputTokens": 128000,
+        "description": "官方独立Pro规格；Responses协议限制由实际接入面执行，不沿用基础型号或其他代际family。"
+      }
     }
   ]
 }

compute/model-specs/qwen.json

@@ -1153,6 +1153,32 @@
         ],
         "description": "千问图像生成 2.0 标准版"
       }
+    },
+    {
+      "id": "qwen3-coder-plus",
+      "displayName": "qwen3-coder-plus",
+      "family": "qwen3-coder-plus",
+      "match": {
+        "exact": [
+          "qwen3-coder-plus"
+        ]
+      },
+      "spec": {
+        "description": "精确绑定已有接入型号，参数未核实；identity-only规格不继承旧family能力、窗口、协议或自动路由资格。"
+      }
+    },
+    {
+      "id": "qwen3-coder-next",
+      "displayName": "qwen3-coder-next",
+      "family": "qwen3-coder-next",
+      "match": {
+        "exact": [
+          "qwen3-coder-next"
+        ]
+      },
+      "spec": {
+        "description": "精确绑定已有接入型号，参数未核实；identity-only规格不继承旧family能力、窗口、协议或自动路由资格。"
+      }
     }
   ]
 }

compute/v2/compatibility/coding-plans/baidu-coding.json

@@ -0,0 +1,175 @@
+{
+  "id": "provider-baidu-coding-001",
+  "provider": "baidu",
+  "brandGroup": "baidu",
+  "label": "百度千帆 Coding Plan",
+  "baseUrl": "https://qianfan.baidubce.com/v2/coding",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "accessMode": "coding-plan",
+  "services": [
+    "chat",
+    "reasoning"
+  ],
+  "codingPlan": {
+    "quotas": {},
+    "usageTracking": {
+      "method": "none",
+      "consoleUrl": "https://console.bce.baidu.com/qianfan/codingplan"
+    }
+  },
+  "models": [
+    {
+      "modelName": "qianfan-code-latest",
+      "displayName": "Auto (自动路由)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "千帆 Coding Plan 自动路由模型，后端自动选择最优模型",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "deepseek-v3.2",
+      "displayName": "DeepSeek V3.2",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "DeepSeek V3.2（千帆 Coding Plan 渠道），倍率 ×1",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "tool_use"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "deepseek-v4-flash",
+      "displayName": "DeepSeek V4 Flash",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "DeepSeek V4 Flash（千帆 Coding Plan 渠道），倍率 ×1",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "fast"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "ernie-4.5-turbo-20260402",
+      "displayName": "文心 ERNIE 4.5 Turbo",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百度 ERNIE 4.5 Turbo（千帆 Coding Plan 渠道）；套餐当前不支持图像理解",
+      "capabilities": [
+        "chat",
+        "code",
+        "fast"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "glm-5",
+      "displayName": "GLM-5",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "智谱 GLM-5（千帆 Coding Plan 渠道）",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "agent"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "glm-5.1",
+      "displayName": "GLM-5.1",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "智谱 GLM-5.1（千帆 Coding Plan 渠道），高峰抵扣 ×4 / 低峰 ×3；高峰时段随流量动态调整",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "agent"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "kimi-k2.5",
+      "displayName": "Kimi K2.5",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Kimi K2.5（千帆 Coding Plan 渠道）",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "agent",
+        "long_context"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "kimi-k2.6",
+      "displayName": "Kimi K2.6",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Kimi K2.6（千帆 Coding Plan 渠道），倍率 ×5",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "agent",
+        "long_context"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "minimax-m2.5",
+      "displayName": "MiniMax M2.5",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "MiniMax M2.5（千帆 Coding Plan 渠道）",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "minimax-m2.7",
+      "displayName": "MiniMax M2.7",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "MiniMax M2.7（千帆 Coding Plan 渠道），倍率 ×2",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "agent"
+      ],
+      "source": "preset"
+    }
+  ]
+}

compute/v2/compatibility/coding-plans/dashscope-coding.json

@@ -0,0 +1,209 @@
+{
+  "id": "provider-dashscope-coding-001",
+  "provider": "dashscope",
+  "brandGroup": "dashscope",
+  "label": "百炼 Coding Plan",
+  "baseUrl": "https://coding.dashscope.aliyuncs.com/v1",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "accessMode": "coding-plan",
+  "services": [
+    "chat",
+    "reasoning"
+  ],
+  "codingPlan": {
+    "quotas": {},
+    "usageTracking": {
+      "method": "none",
+      "consoleUrl": "https://bailian.console.aliyun.com/"
+    }
+  },
+  "models": [
+    {
+      "modelName": "qwen3.7-plus",
+      "displayName": "Qwen3.7 Plus (Coding)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百炼 Coding Plan，通义千问 3.7 Plus，多模态智能体模型，1M 上下文",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "qwen3.6-plus",
+      "displayName": "Qwen3.6 Plus (Coding)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百炼 Coding Plan，通义千问 3.6 Plus，1M 上下文",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "qwen3.5-plus",
+      "displayName": "Qwen3.5 Plus (Coding)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百炼 Coding Plan，通义千问 3.5 Plus，1M 上下文",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "long_context"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "qwen3-coder-plus",
+      "displayName": "Qwen3 Coder Plus (Coding)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百炼 Coding Plan，Qwen3 编程增强版，1M 上下文",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "code",
+        "tool_use"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "qwen3-coder-next",
+      "displayName": "Qwen3 Coder Next (Coding)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百炼 Coding Plan，Qwen3 编程下一代实验模型，262K 上下文",
+      "contextWindow": 262144,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "code",
+        "tool_use"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "qwen3-max-2026-01-23",
+      "displayName": "Qwen3 Max (Coding)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百炼 Coding Plan，Qwen3 Max 旗舰模型，262K 上下文",
+      "contextWindow": 262144,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual",
+        "tool_use",
+        "long_context",
+        "vision"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "kimi-k2.5",
+      "displayName": "Kimi K2.5 (Coding)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百炼 Coding Plan，月之暗面 Kimi K2.5，262K 上下文",
+      "contextWindow": 262144,
+      "maxOutputTokens": 32768,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "tool_use",
+        "agent",
+        "vision",
+        "long_context"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "glm-5",
+      "displayName": "GLM-5 (Coding)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百炼 Coding Plan，智谱 GLM-5，200K 上下文",
+      "contextWindow": 202752,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "tool_use",
+        "agent"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "glm-4.7",
+      "displayName": "GLM-4.7 (Coding)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百炼 Coding Plan，智谱 GLM-4.7，200K 上下文",
+      "contextWindow": 202752,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "MiniMax-M2.5",
+      "displayName": "MiniMax M2.5 (Coding)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百炼 Coding Plan，MiniMax M2.5，196K 上下文",
+      "contextWindow": 196608,
+      "maxOutputTokens": 8192,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "tool_use",
+        "long_context"
+      ],
+      "source": "preset"
+    }
+  ]
+}

compute/v2/compatibility/coding-plans/dashscope-token-plan.json

@@ -0,0 +1,549 @@
+{
+  "id": "provider-dashscope-token-plan-001",
+  "provider": "dashscope",
+  "brandGroup": "dashscope",
+  "label": "百炼 Token Plan",
+  "baseUrl": "https://token-plan.cn-beijing.maas.aliyuncs.com/compatible-mode/v1",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "accessMode": "token-plan",
+  "services": [
+    "chat",
+    "reasoning"
+  ],
+  "tombstones": [
+    "qwen3.8-max-preview",
+    "qwen3.5-plus",
+    "qwen3-coder-plus"
+  ],
+  "codingPlan": {
+    "quotas": {},
+    "usageTracking": {
+      "method": "none",
+      "consoleUrl": "https://bailian.console.aliyun.com/"
+    }
+  },
+  "models": [
+    {
+      "modelName": "deepseek-v4.1-flash",
+      "displayName": "DeepSeek V4.1 Flash",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 384000,
+      "capabilities": [
+        "chat",
+        "code",
+        "reasoning",
+        "deep_thinking",
+        "vision",
+        "multilingual",
+        "tool_use"
+      ],
+      "serviceType": [
+        "chat",
+        "reasoning",
+        "vision"
+      ],
+      "extra": {
+        "supportsThinking": true,
+        "thinkingDefault": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "high"
+        }
+      },
+      "description": "DeepSeek V4.1 Flash，支持文本与图像输入，100 万上下文",
+      "source": "preset"
+    },
+    {
+      "modelName": "wan2.7-image-pro",
+      "displayName": "通义万相 2.7 Pro",
+      "capabilities": [
+        "image_generation",
+        "chinese_optimized",
+        "high_quality"
+      ],
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "万相最新旗舰文生图模型，支持 4K 分辨率",
+      "source": "preset",
+      "extra": {}
+    },
+    {
+      "modelName": "wan2.7-image",
+      "displayName": "通义万相 2.7",
+      "capabilities": [
+        "image_generation",
+        "chinese_optimized"
+      ],
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "万相最新标准文生图模型，高画质",
+      "source": "preset",
+      "extra": {}
+    },
+    {
+      "modelName": "qwen-image-3.0-pro",
+      "displayName": "qwen-image-3.0-pro",
+      "capabilities": [
+        "image_generation",
+        "image_editing",
+        "text_rendering"
+      ],
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "qwen-image-3.0-pro",
+      "source": "preset"
+    },
+    {
+      "modelName": "happyhorse-1.1-t2v",
+      "displayName": "HappyHorse 1.1 T2V",
+      "capabilities": [
+        "video_generation",
+        "text_to_video",
+        "high_quality"
+      ],
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "HappyHorse 1.1 文生视频，进一步提升文本词义理解、镜头调度与动态生成表现",
+      "source": "preset",
+      "extra": {}
+    },
+    {
+      "modelName": "happyhorse-1.1-r2v",
+      "displayName": "HappyHorse 1.1 R2V",
+      "capabilities": [
+        "video_generation",
+        "reference_to_video",
+        "multi_image",
+        "high_quality"
+      ],
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "HappyHorse 1.1 参考生视频，支持最多9张图片参考，更加稳定的主体与场景参考",
+      "source": "preset",
+      "extra": {}
+    },
+    {
+      "modelName": "happyhorse-1.1-i2v",
+      "displayName": "HappyHorse 1.1 I2V",
+      "capabilities": [
+        "video_generation",
+        "image_to_video",
+        "high_quality"
+      ],
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "HappyHorse 1.1 图生视频，进一步提升画面质感、动态表现与跨片段一致性",
+      "source": "preset",
+      "extra": {}
+    },
+    {
+      "modelName": "glm-5.3",
+      "displayName": "GLM-5.3",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "math",
+        "multilingual",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "serviceType": [
+        "chat"
+      ],
+      "description": "智谱 GLM-5.3 大推理模型，面向复杂软件工程与长程 Agent 任务，100 万上下文；强制思考，无法关闭",
+      "extra": {
+        "thinkingDefault": true,
+        "thinkingOnly": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "max"
+        }
+      },
+      "source": "preset"
+    },
+    {
+      "modelName": "glm-5.2",
+      "displayName": "GLM-5.2",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 32768,
+      "capabilities": [
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "multilingual",
+        "tool_use",
+        "long_context"
+      ],
+      "serviceType": [
+        "chat"
+      ],
+      "description": "智谱 GLM-5.2 大推理模型，100 万上下文",
+      "source": "preset",
+      "extra": {
+        "reasoning": {
+          "supportedEfforts": [
+            "minimal",
+            "low",
+            "medium",
+            "high",
+            "xhigh"
+          ],
+          "defaultEffort": "high"
+        }
+      }
+    },
+    {
+      "modelName": "deepseek-v4-pro-0813",
+      "displayName": "DeepSeek V4 Pro 0813",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 384000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "math",
+        "science",
+        "multilingual",
+        "tool_use"
+      ],
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "extra": {
+        "supportsThinking": true,
+        "thinkingDefault": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "high"
+        }
+      },
+      "description": "DeepSeek V4 Pro 0813 版本，1.6T MoE 架构，100 万上下文",
+      "source": "preset"
+    },
+    {
+      "modelName": "deepseek-v4-pro",
+      "displayName": "DeepSeek V4 Pro",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 384000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "math",
+        "science",
+        "multilingual",
+        "tool_use"
+      ],
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "extra": {
+        "supportsThinking": true,
+        "thinkingDefault": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "high"
+        }
+      },
+      "description": "DeepSeek V4 Pro，1.6T MoE 架构，100 万上下文",
+      "source": "preset"
+    },
+    {
+      "modelName": "deepseek-v4-flash-0731",
+      "displayName": "DeepSeek V4 Flash 0731",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 384000,
+      "capabilities": [
+        "chat",
+        "code",
+        "reasoning",
+        "deep_thinking",
+        "multilingual",
+        "tool_use"
+      ],
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "extra": {
+        "supportsThinking": true,
+        "thinkingDefault": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "high",
+            "max"
+          ],
+          "defaultEffort": "high"
+        }
+      },
+      "description": "DeepSeek V4 Flash 0731 固定版本，100 万上下文",
+      "source": "preset"
+    },
+    {
+      "modelName": "qwen3.8-max",
+      "displayName": "Qwen3.8 Max (Token Plan)",
+      "serviceType": [
+        "chat",
+        "reasoning",
+        "vision"
+      ],
+      "description": "百炼 Token Plan，通义千问 3.8 Max 正式版，支持文本、图像和视频输入",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "vision",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "defaultTemperature": 0.6,
+      "extra": {
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "xhigh"
+          ],
+          "defaultEffort": "xhigh"
+        },
+        "supportsThinking": true,
+        "thinkingDefault": true,
+        "thinkingMaxTokens": 262144,
+        "preserveThinkingDefault": true,
+        "supportsParallelToolCalls": false
+      },
+      "source": "preset"
+    },
+    {
+      "modelName": "qwen3.8-flash",
+      "displayName": "Qwen3.8 Flash (Token Plan)",
+      "serviceType": [
+        "chat",
+        "reasoning",
+        "vision"
+      ],
+      "description": "百炼 Token Plan，通义千问 3.8 Flash 正式版，高速多模态推理",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "vision",
+        "tool_use",
+        "agent",
+        "long_context",
+        "fast"
+      ],
+      "defaultTemperature": 0.6,
+      "extra": {
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "xhigh"
+          ],
+          "defaultEffort": "xhigh"
+        },
+        "supportsThinking": true,
+        "thinkingDefault": true,
+        "thinkingMaxTokens": 262144,
+        "preserveThinkingDefault": true,
+        "supportsParallelToolCalls": false
+      },
+      "source": "preset"
+    },
+    {
+      "modelName": "qwen3.7-max",
+      "displayName": "Qwen3.7 Max (Token Plan)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百炼 Token Plan，通义千问 3.7 Max，1M 上下文，面向复杂推理与长程 Agent 任务",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "qwen3.7-plus",
+      "displayName": "Qwen3.7 Plus (Token Plan)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百炼 Token Plan，通义千问 3.7 Plus，多模态智能体模型，1M 上下文",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "qwen3.6-plus",
+      "displayName": "Qwen3.6 Plus (Token Plan)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百炼 Token Plan，通义千问 3.6 Plus，1M 上下文，支持图像理解",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "qwen3.6-flash",
+      "displayName": "Qwen3.6 Flash (Token Plan)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百炼 Token Plan，通义千问 3.6 Flash，高性价比多模态模型，1M 上下文",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "long_context",
+        "fast"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "kimi-k2.5",
+      "displayName": "Kimi K2.5 (Token Plan)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百炼 Token Plan，Kimi K2.5，支持图像理解",
+      "contextWindow": 262144,
+      "maxOutputTokens": 32768,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "glm-5",
+      "displayName": "GLM-5 (Token Plan)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百炼 Token Plan，智谱 GLM-5，200K 上下文",
+      "contextWindow": 202752,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "tool_use",
+        "agent"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "MiniMax-M2.5",
+      "displayName": "MiniMax M2.5 (Token Plan)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百炼 Token Plan，MiniMax M2.5，196K 上下文",
+      "contextWindow": 196608,
+      "maxOutputTokens": 8192,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "tool_use",
+        "long_context"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "glm-4.7",
+      "displayName": "GLM-4.7 (Token Plan)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百炼 Token Plan，智谱 GLM-4.7，200K 上下文",
+      "contextWindow": 200000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "tool_use"
+      ],
+      "source": "preset"
+    }
+  ]
+}

compute/v2/compatibility/coding-plans/infini-coding.json

@@ -0,0 +1,43 @@
+{
+  "id": "provider-infini-coding-001",
+  "provider": "infini",
+  "brandGroup": "infini",
+  "label": "无问芯穹 Token (Coding) Plan",
+  "baseUrl": "https://cloud.infini-ai.com/maas/coding/v1",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "accessMode": "coding-plan",
+  "services": [
+    "chat",
+    "reasoning"
+  ],
+  "codingPlan": {
+    "quotas": {},
+    "usageTracking": {
+      "method": "none",
+      "consoleUrl": "https://cloud.infini-ai.com/console"
+    }
+  },
+  "models": [
+    {
+      "modelName": "deepseek-v3",
+      "displayName": "DeepSeek V3 (芯穹 Coding)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "无问芯穹 Coding Plan，DeepSeek V3 编程优化",
+      "contextWindow": 128000,
+      "maxOutputTokens": 8192,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual"
+      ],
+      "source": "preset"
+    }
+  ]
+}

compute/v2/compatibility/coding-plans/kwai-coding.json

@@ -0,0 +1,38 @@
+{
+  "id": "provider-kwai-coding-001",
+  "provider": "kwai",
+  "brandGroup": "kwai",
+  "label": "快手 KwaiKAT Token (Coding) Plan",
+  "baseUrl": "https://wanqing.streamlakeapi.com/api/gateway/coding/v1",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "accessMode": "coding-plan",
+  "services": [
+    "chat"
+  ],
+  "codingPlan": {
+    "quotas": {},
+    "usageTracking": {
+      "method": "none",
+      "consoleUrl": "https://console.streamlake.ai/wanqing/"
+    }
+  },
+  "models": [
+    {
+      "modelName": "kwai-coder",
+      "displayName": "KwaiKAT Coder (Coding)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "快手 KwaiKAT Coding Plan，编程辅助模型",
+      "capabilities": [
+        "chat",
+        "code"
+      ],
+      "source": "preset"
+    }
+  ]
+}

compute/v2/compatibility/coding-plans/minimax-coding.json

@@ -0,0 +1,226 @@
+{
+  "id": "provider-minimax-coding-001",
+  "provider": "minimax",
+  "brandGroup": "minimax",
+  "label": "MiniMax Token (Coding) Plan",
+  "baseUrl": "https://api.minimaxi.com/anthropic",
+  "apiFormat": "anthropic-messages",
+  "mediaBaseUrl": "https://api.minimaxi.com/v1",
+  "apiKeyRef": "",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "accessMode": "coding-plan",
+  "services": [
+    "chat",
+    "fast",
+    "reasoning",
+    "image_gen",
+    "tts",
+    "video_gen",
+    "music_gen"
+  ],
+  "codingPlan": {
+    "quotas": {},
+    "usageTracking": {
+      "method": "rest-api",
+      "endpoint": "https://api.minimaxi.com/v1/token_plan/remains",
+      "consoleUrl": "https://platform.minimaxi.com/"
+    }
+  },
+  "models": [
+    {
+      "modelName": "MiniMax-M3.1-Flash-Preview",
+      "displayName": "MiniMax M3.1 Flash Preview",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "video_understanding",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "serviceType": [
+        "chat",
+        "reasoning",
+        "vision"
+      ],
+      "description": "MiniMax M3.1 Flash Preview，目前仅通过 M Plan / MiniMax Code 提供",
+      "contextWindow": 1000000,
+      "extra": {
+        "adaptiveThinking": true,
+        "thinkingOnly": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultEffort": "max"
+        }
+      },
+      "source": "preset"
+    },
+    {
+      "modelName": "MiniMax-M3",
+      "displayName": "MiniMax M3 (Token Plan)",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "MiniMax M3，Token Plan 核心文本模型，1M 上下文，面向 Coding 与 Agent",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "tool_use",
+        "vision",
+        "agent",
+        "long_context"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "MiniMax-M2.7",
+      "displayName": "MiniMax M2.7 (Token Plan)",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "MiniMax 新一代旗舰模型，Token Plan 核心文本模型，5 小时滚动窗口计费。Starter 600次/5h，Plus 1500次/5h，Max 4500次/5h",
+      "contextWindow": 204800,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "tool_use"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "MiniMax-M2.7-highspeed",
+      "displayName": "MiniMax M2.7 高速版 (Token Plan)",
+      "serviceType": [
+        "fast"
+      ],
+      "description": "MiniMax M2.7 极速版，极速套餐专用。Plus-极速 1500次/5h，Max-极速 4500次/5h，Ultra-极速 30000次/5h",
+      "contextWindow": 204800,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "tool_use",
+        "fast"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "speech-2.8-hd",
+      "displayName": "Speech 2.8 HD (Token Plan)",
+      "serviceType": [
+        "tts"
+      ],
+      "description": "MiniMax 最新高清语音合成，Token Plan Plus 及以上可用。Plus 4000字符/天，Max 11000字符/天",
+      "capabilities": [
+        "tts",
+        "multilingual",
+        "emotion_control",
+        "voice_clone"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "speech-2.8-turbo",
+      "displayName": "Speech 2.8 Turbo (Token Plan)",
+      "serviceType": [
+        "tts"
+      ],
+      "description": "MiniMax 快速语音合成，Token Plan Plus 及以上可用",
+      "capabilities": [
+        "tts",
+        "multilingual",
+        "emotion_control",
+        "fast"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "image-01",
+      "displayName": "Image 01 (Token Plan)",
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "MiniMax 文生图/图生图，Token Plan Plus 及以上可用。Plus 50张/天，Max 120张/天",
+      "capabilities": [
+        "image_generation",
+        "subject_reference"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "image-01-live",
+      "displayName": "Image 01 Live (Token Plan)",
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "MiniMax 画风控制文生图，Token Plan Plus 及以上可用",
+      "capabilities": [
+        "image_generation",
+        "style_control",
+        "subject_reference"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "MiniMax-Hailuo-2.3",
+      "displayName": "海螺视频 2.3 (Token Plan)",
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "海螺视频 2.3，文/图生视频，Token Plan Max 及以上可用。Max 2个/天",
+      "capabilities": [
+        "video_generation",
+        "text_to_video",
+        "image_to_video",
+        "camera_control"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "MiniMax-Hailuo-2.3-Fast",
+      "displayName": "海螺视频 2.3 快速版 (Token Plan)",
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "海螺视频快速版，Token Plan Max 及以上可用。Max 2个/天",
+      "capabilities": [
+        "video_generation",
+        "image_to_video",
+        "fast"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "music-2.6",
+      "displayName": "Music 2.6 (Token Plan)",
+      "serviceType": [
+        "music_gen"
+      ],
+      "description": "MiniMax AI 音乐生成，所有 Token Plan 套餐免费使用（100 首/天，≤5分钟/首）",
+      "capabilities": [
+        "music_generation",
+        "lyrics_input",
+        "instrumental"
+      ],
+      "source": "preset"
+    }
+  ]
+}

compute/v2/compatibility/coding-plans/moonshot-coding.json

@@ -0,0 +1,47 @@
+{
+  "id": "provider-moonshot-coding-001",
+  "provider": "moonshot",
+  "brandGroup": "moonshot",
+  "label": "Kimi Code",
+  "baseUrl": "https://api.kimi.com/coding/v1",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "accessMode": "coding-plan",
+  "services": [
+    "chat",
+    "reasoning"
+  ],
+  "codingPlan": {
+    "quotas": {},
+    "usageTracking": {
+      "method": "none",
+      "consoleUrl": "https://platform.moonshot.cn/"
+    }
+  },
+  "models": [
+    {
+      "modelName": "kimi-for-coding",
+      "displayName": "Kimi For Coding (Auto)",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "Kimi Code 稳定路由 ID，自动映射到 Kimi Code 当前模型，262K 上下文",
+      "contextWindow": 262144,
+      "maxOutputTokens": 32768,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "tool_use",
+        "agent",
+        "long_context",
+        "vision"
+      ],
+      "source": "preset"
+    }
+  ]
+}

compute/v2/compatibility/coding-plans/moorethread-coding.json

@@ -0,0 +1,38 @@
+{
+  "id": "provider-moorethread-coding-001",
+  "provider": "moorethread",
+  "brandGroup": "moorethread",
+  "label": "摩尔线程 Token (Coding) Plan",
+  "baseUrl": "https://api.mthreads.com/v1",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "accessMode": "coding-plan",
+  "services": [
+    "chat"
+  ],
+  "codingPlan": {
+    "quotas": {},
+    "usageTracking": {
+      "method": "none",
+      "consoleUrl": "https://www.mthreads.com/console"
+    }
+  },
+  "models": [
+    {
+      "modelName": "mt-coder",
+      "displayName": "MT Coder (Coding)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "摩尔线程 Coding Plan，国产 GPU 推理",
+      "capabilities": [
+        "chat",
+        "code"
+      ],
+      "source": "preset"
+    }
+  ]
+}

compute/v2/compatibility/coding-plans/tencent-token.json

@@ -0,0 +1,178 @@
+{
+  "id": "provider-tencent-token-001",
+  "provider": "tencent",
+  "brandGroup": "tencent",
+  "label": "腾讯云 Token Plan",
+  "baseUrl": "https://api.lkeap.cloud.tencent.com/plan/v3",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "accessMode": "token-plan",
+  "services": [
+    "chat",
+    "reasoning"
+  ],
+  "codingPlan": {
+    "quotas": {},
+    "usageTracking": {
+      "method": "none",
+      "consoleUrl": "https://console.cloud.tencent.com/hunyuan"
+    }
+  },
+  "models": [
+    {
+      "modelName": "tc-code-latest",
+      "displayName": "Auto (自动路由)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "腾讯 Token Plan 自动路由模型，后端自动选择最优模型",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "hunyuan-2.0-instruct",
+      "displayName": "混元 2.0 Instruct",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "腾讯混元 2.0 指令模型（Token Plan）",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "long_context",
+        "tool_use"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "hunyuan-2.0-thinking",
+      "displayName": "混元 2.0 Think",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "腾讯混元 2.0 推理模型（Token Plan）",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "deep_thinking",
+        "long_context"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "hunyuan-t1",
+      "displayName": "混元 T1",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "腾讯混元 T1 深度推理模型（Token Plan）",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "deep_thinking"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "hunyuan-turbos",
+      "displayName": "混元 Turbo S",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "腾讯混元 Turbo S 高速模型（Token Plan）",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "fast"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "glm-5.1",
+      "displayName": "GLM-5.1",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "智谱 GLM-5.1（腾讯 Token Plan 渠道）",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "agent"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "glm-5",
+      "displayName": "GLM-5",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "智谱 GLM-5（腾讯 Token Plan 渠道）",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "agent"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "minimax-m2.7",
+      "displayName": "MiniMax M2.7",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "MiniMax M2.7（腾讯 Token Plan 渠道）",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "agent"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "minimax-m2.5",
+      "displayName": "MiniMax M2.5",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "MiniMax M2.5（腾讯 Token Plan 渠道）",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "agent"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "kimi-k2.5",
+      "displayName": "Kimi K2.5",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Kimi K2.5（腾讯 Token Plan 渠道）",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "agent",
+        "long_context"
+      ],
+      "source": "preset"
+    }
+  ]
+}

compute/v2/compatibility/coding-plans/volcengine-coding.json

@@ -0,0 +1,258 @@
+{
+  "id": "provider-volcengine-coding-001",
+  "provider": "volcengine",
+  "brandGroup": "volcengine",
+  "label": "火山引擎 Token (Coding) Plan",
+  "baseUrl": "https://ark.cn-beijing.volces.com/api/coding/v3",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "accessMode": "coding-plan",
+  "services": [
+    "chat",
+    "reasoning"
+  ],
+  "codingPlan": {
+    "quotas": {},
+    "usageTracking": {
+      "method": "response-header",
+      "headerKeys": {
+        "remaining": "x-ratelimit-remaining",
+        "limit": "x-ratelimit-limit",
+        "reset": "x-ratelimit-reset"
+      },
+      "consoleUrl": "https://console.volcengine.com/ark"
+    },
+    "modelIdOverride": "ark-code-latest"
+  },
+  "models": [
+    {
+      "modelName": "doubao-seed-2.1-pro",
+      "displayName": "豆包 Seed 2.1 Pro",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山方舟 Coding Plan，豆包 Seed 2.1 Pro，生产级 Coding 与 Agent 模型",
+      "contextWindow": 256000,
+      "maxOutputTokens": 262144,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "doubao-seed-2.1-turbo",
+      "displayName": "豆包 Seed 2.1 Turbo",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山方舟 Coding Plan，豆包 Seed 2.1 Turbo，高吞吐 Coding 与 Agent 模型",
+      "contextWindow": 256000,
+      "maxOutputTokens": 262144,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "agent",
+        "fast",
+        "long_context"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "ark-code-latest",
+      "displayName": "Ark Code Latest (Auto)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "自动路由最优编程模型（Doubao Seed Code / GLM-4.7 / DeepSeek-V3.2 / Kimi-K2.5 等）",
+      "capabilities": [
+        "code",
+        "reasoning",
+        "chat"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "doubao-seed-2.0-code",
+      "displayName": "豆包 Seed 2.0 Code",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山方舟 Coding Plan 中的豆包 Seed 2.0 编程模型",
+      "capabilities": [
+        "code",
+        "reasoning",
+        "chat",
+        "long_context"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "doubao-seed-2.0-pro",
+      "displayName": "豆包 Seed 2.0 Pro",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山方舟 Coding Plan 中的豆包 Seed 2.0 旗舰模型",
+      "capabilities": [
+        "code",
+        "reasoning",
+        "chat",
+        "long_context",
+        "vision"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "doubao-seed-2.0-lite",
+      "displayName": "豆包 Seed 2.0 Lite",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山方舟 Coding Plan 中的豆包 Seed 2.0 均衡模型",
+      "capabilities": [
+        "code",
+        "reasoning",
+        "chat",
+        "long_context",
+        "vision"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "doubao-seed-code",
+      "displayName": "豆包 Seed Code",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山方舟 Coding Plan 中的豆包编程模型",
+      "capabilities": [
+        "code",
+        "reasoning",
+        "chat",
+        "long_context"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "glm-4.7",
+      "displayName": "GLM-4.7",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山方舟 Coding Plan 中的智谱 GLM-4.7 模型",
+      "capabilities": [
+        "code",
+        "reasoning",
+        "chat",
+        "long_context"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "deepseek-v3.2",
+      "displayName": "DeepSeek V3.2",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山方舟 Coding Plan 中的 DeepSeek V3.2 模型",
+      "capabilities": [
+        "code",
+        "reasoning",
+        "chat"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "kimi-k2.5",
+      "displayName": "Kimi K2.5",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山方舟 Coding Plan 中的 Kimi K2.5 模型",
+      "capabilities": [
+        "code",
+        "reasoning",
+        "chat",
+        "agent",
+        "long_context"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "glm-5.1",
+      "displayName": "GLM-5.1",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山方舟 Coding Plan 中的智谱 GLM-5.1 旗舰模型（2026.4 新增）",
+      "capabilities": [
+        "code",
+        "reasoning",
+        "chat",
+        "long_context",
+        "agent"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "kimi-k2.6",
+      "displayName": "Kimi K2.6",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山方舟 Coding Plan 中的 Kimi K2.6 模型（2026.4 新增）",
+      "capabilities": [
+        "code",
+        "reasoning",
+        "chat",
+        "agent",
+        "long_context",
+        "vision"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "minimax-m2.5",
+      "displayName": "MiniMax M2.5",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山方舟 Coding Plan 中的 MiniMax M2.5 模型",
+      "capabilities": [
+        "code",
+        "reasoning",
+        "chat",
+        "agent"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "minimax-m2.7",
+      "displayName": "MiniMax M2.7",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山方舟 Coding Plan 中的 MiniMax M2.7 模型（2026.4 新增），230B 总参/10B 激活",
+      "capabilities": [
+        "code",
+        "reasoning",
+        "chat",
+        "agent",
+        "long_context"
+      ],
+      "source": "preset"
+    }
+  ]
+}

compute/v2/compatibility/coding-plans/zhipu-coding.json

@@ -0,0 +1,195 @@
+{
+  "id": "provider-zhipu-coding-001",
+  "provider": "zhipu",
+  "brandGroup": "zhipu",
+  "label": "智谱 Token (Coding) Plan",
+  "baseUrl": "https://api.z.ai/api/anthropic",
+  "apiFormat": "anthropic-messages",
+  "apiKeyRef": "",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "accessMode": "coding-plan",
+  "services": [
+    "chat",
+    "reasoning"
+  ],
+  "codingPlan": {
+    "quotas": {},
+    "usageTracking": {
+      "method": "none",
+      "consoleUrl": "https://open.bigmodel.cn/console/overview"
+    }
+  },
+  "tombstones": [
+    "glm-5.1",
+    "glm-5",
+    "glm-4.5-air"
+  ],
+  "models": [
+    {
+      "modelName": "glm-5.3",
+      "displayName": "GLM-5.3",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "math",
+        "multilingual",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "serviceType": [
+        "chat"
+      ],
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "description": "智谱 GLM-5.3 大推理模型，面向复杂软件工程与长程 Agent 任务，100 万上下文；强制思考，无法关闭",
+      "extra": {
+        "thinkingDefault": true,
+        "thinkingOnly": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "max"
+        },
+        "pricingNotes": "国内官方标准价，人民币/百万 tokens；Flash 限时优惠以控制台实际结算为准。"
+      },
+      "source": "preset"
+    },
+    {
+      "modelName": "glm-5.3-flash",
+      "displayName": "GLM-5.3-Flash (Coding)",
+      "serviceType": [
+        "chat",
+        "reasoning",
+        "vision"
+      ],
+      "description": "智谱 GLM-5.3-Flash Coding Plan 版本，原生多模态高效模型，1M 上下文，128K 最大输出，强制开启思考",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "vision",
+        "image_understanding",
+        "video_understanding",
+        "long_context",
+        "tool_use",
+        "agent",
+        "multilingual",
+        "fast"
+      ],
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "supportsThinking": true,
+        "thinkingDefault": true,
+        "thinkingOnly": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "max"
+        }
+      },
+      "source": "preset"
+    },
+    {
+      "modelName": "glm-5.2",
+      "apiModelId": "glm-5.2[1m]",
+      "displayName": "GLM-5.2 (Coding)",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "智谱 GLM-5.2 Coding Plan 版本，官方最新可用模型，1M 上下文，128K 最大输出",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "deep_thinking",
+        "long_context",
+        "tool_use",
+        "agent",
+        "math",
+        "multilingual"
+      ],
+      "extra": {
+        "codingPlanLongContextModelId": "glm-5.2[1m]",
+        "reasoning": {
+          "supportedEfforts": [
+            "max",
+            "xhigh",
+            "high",
+            "medium",
+            "low",
+            "minimal",
+            "none"
+          ],
+          "defaultEffort": "max"
+        }
+      },
+      "source": "preset"
+    },
+    {
+      "modelName": "glm-5-turbo",
+      "displayName": "GLM-5-Turbo (Coding)",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "智谱 GLM-5-Turbo Coding Plan 版本，官方可用高阶模型，200K 上下文，128K 最大输出",
+      "contextWindow": 200000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "deep_thinking",
+        "long_context",
+        "tool_use",
+        "agent",
+        "math",
+        "multilingual",
+        "fast"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "glm-4.7",
+      "displayName": "GLM-4.7 (Coding)",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "智谱 GLM-4.7 Coding Plan 版本，官方可用模型，200K 上下文，128K 最大输出",
+      "contextWindow": 200000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "deep_thinking",
+        "long_context",
+        "tool_use",
+        "agent",
+        "multilingual"
+      ],
+      "source": "preset"
+    }
+  ]
+}

compute/v2/compatibility/model-specs/anthropic.json

@@ -0,0 +1,419 @@
+{
+  "description": "Anthropic Claude 系列模型规格。参数来源：config-center compute/providers/anthropic.json。",
+  "specs": [
+    {
+      "displayName": "Claude Opus 5.5",
+      "family": "claude-opus-5-5",
+      "match": {
+        "exact": [
+          "claude-opus-5-5",
+          "anthropic/claude-opus-5.5"
+        ]
+      },
+      "id": "claude-opus-5-5",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 128000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "tool_use",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "supportsReasoning": true,
+        "description": "Claude Opus 5.5，始终开启自适应思考，默认 medium effort",
+        "extra": {
+          "adaptiveThinking": true,
+          "thinkingOnly": true,
+          "samplingParametersDeprecated": true,
+          "forcedToolChoiceUnsupported": true
+        },
+        "releasedAt": "2026-09-22"
+      },
+      "routing": {
+        "tier": "flagship",
+        "routingPriority": 22,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultMode": "medium"
+        }
+      }
+    },
+    {
+      "displayName": "Claude Fable 5.1",
+      "family": "claude-fable",
+      "match": {
+        "exact": [
+          "claude-fable-5-1"
+        ],
+        "patterns": [
+          "claude-fable-5-1*"
+        ]
+      },
+      "id": "claude-fable-5-1",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 128000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "deep_thinking",
+          "code",
+          "vision",
+          "tool_use",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "supportsReasoning": true,
+        "description": "Claude Fable 5.1，面向高难推理和长程智能体任务，始终开启自适应思考",
+        "extra": {
+          "adaptiveThinking": true,
+          "thinkingOnly": true,
+          "samplingParametersDeprecated": true,
+          "forcedToolChoiceUnsupported": true
+        }
+      },
+      "routing": {
+        "tier": "flagship",
+        "routingPriority": 20,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultMode": "high"
+        }
+      }
+    },
+    {
+      "displayName": "Claude Opus 5",
+      "family": "claude-opus-5",
+      "match": {
+        "exact": [
+          "claude-opus-5"
+        ]
+      },
+      "id": "claude-opus-5",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 128000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "tool_use",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "supportsReasoning": true,
+        "description": "Claude Opus 5，Anthropic Opus 系列当前旗舰；思考默认开启，effort 支持到 max",
+        "extra": {
+          "adaptiveThinking": true,
+          "samplingParametersDeprecated": true
+        }
+      },
+      "routing": {
+        "tier": "flagship",
+        "routingPriority": 25,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultMode": "xhigh"
+        }
+      }
+    },
+    {
+      "displayName": "Claude Opus 4.8",
+      "family": "claude-opus-4-8",
+      "match": {
+        "patterns": [
+          "claude-opus-4-8*"
+        ]
+      },
+      "id": "claude-opus-4-8",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 128000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "tool_use",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "supportsReasoning": true,
+        "description": "Claude Opus 4.8，Anthropic Opus 系列最新旗舰",
+        "extra": {
+          "adaptiveThinking": true,
+          "samplingParametersDeprecated": true
+        }
+      }
+    },
+    {
+      "displayName": "Claude Opus 4.7",
+      "family": "claude-opus-4-7",
+      "match": {
+        "patterns": [
+          "claude-opus-4-7*"
+        ]
+      },
+      "id": "claude-opus-4-7",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 128000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": null,
+        "supportsReasoning": true,
+        "description": "Anthropic 当前最强通用模型，适合复杂推理和智能体编码任务",
+        "extra": {
+          "adaptiveThinking": true,
+          "samplingParametersDeprecated": true
+        }
+      }
+    },
+    {
+      "family": "claude-opus",
+      "id": "claude-opus",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 128000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "tool_use",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "supportsReasoning": true
+      }
+    },
+    {
+      "displayName": "Claude Sonnet 5.5",
+      "family": "claude-sonnet-5-5",
+      "match": {
+        "exact": [
+          "claude-sonnet-5-5",
+          "anthropic/claude-sonnet-5.5"
+        ]
+      },
+      "id": "claude-sonnet-5-5",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 128000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "tool_use",
+          "computer_use",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "supportsReasoning": true,
+        "description": "Claude Sonnet 5.5，速度与智能平衡，默认 high effort，自适应思考",
+        "extra": {
+          "adaptiveThinking": true,
+          "samplingParametersDeprecated": true,
+          "forcedToolChoiceUnsupported": true
+        },
+        "releasedAt": "2026-09-28"
+      },
+      "routing": {
+        "tier": "balanced",
+        "routingPriority": 18,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultMode": "high"
+        }
+      }
+    },
+    {
+      "displayName": "Claude Sonnet 5",
+      "family": "claude-sonnet-5",
+      "match": {
+        "patterns": [
+          "claude-sonnet-5*"
+        ]
+      },
+      "id": "claude-sonnet-5",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 128000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "tool_use",
+          "computer_use",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "supportsReasoning": true,
+        "description": "Claude Sonnet 5，速度与智能平衡；思考默认开启，effort 支持到 max",
+        "extra": {
+          "adaptiveThinking": true,
+          "samplingParametersDeprecated": true
+        }
+      },
+      "routing": {
+        "tier": "balanced",
+        "routingPriority": 20,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultMode": "xhigh"
+        }
+      }
+    },
+    {
+      "displayName": "Claude Sonnet 4.6",
+      "family": "claude-sonnet-4-6",
+      "match": {
+        "patterns": [
+          "claude-sonnet-4-6*"
+        ]
+      },
+      "id": "claude-sonnet-4-6",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 64000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "tool_use",
+          "computer_use"
+        ],
+        "serviceType": [
+          "chat",
+          "computer_use"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "description": "Anthropic 高智能高速度模型，适合编码、工具使用和智能体任务"
+      }
+    },
+    {
+      "displayName": "Claude Sonnet 4.5",
+      "family": "claude-sonnet-4-5",
+      "match": {
+        "patterns": [
+          "claude-sonnet-4-5*"
+        ]
+      },
+      "id": "claude-sonnet-4-5",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 64000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "description": "Anthropic 主力模型，长文本与代码任务表现优异"
+      }
+    },
+    {
+      "family": "claude-sonnet",
+      "id": "claude-sonnet",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 128000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "tool_use",
+          "computer_use",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "supportsReasoning": true
+      }
+    }
+  ]
+}

compute/v2/compatibility/model-specs/baichuan.json

@@ -0,0 +1,111 @@
+{
+  "description": "百川智能 Baichuan 系列模型规格。参数来源：config-center compute/providers/baichuan.json。M3/M2 系列为医疗增强大模型，235B 参数，支持多模态。",
+  "specs": [
+    {
+      "displayName": "百川 M3-Plus",
+      "family": "baichuan-m3",
+      "match": {
+        "exact": [
+          "Baichuan-M3-Plus"
+        ]
+      },
+      "id": "Baichuan-M3-Plus",
+      "spec": {
+        "contextWindow": 32000,
+        "maxOutputTokens": 32000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "multilingual",
+          "vision",
+          "long_context",
+          "medical"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 0.3,
+        "description": "百川智能M3-Plus医疗增强模型，235B参数，幻觉率仅2.6%"
+      }
+    },
+    {
+      "displayName": "百川 M3",
+      "family": "baichuan-m3",
+      "match": {
+        "exact": [
+          "Baichuan-M3"
+        ]
+      },
+      "id": "Baichuan-M3",
+      "spec": {
+        "contextWindow": 32000,
+        "maxOutputTokens": 32000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "multilingual",
+          "vision",
+          "medical"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 0.3,
+        "description": "百川智能M3新一代开源医疗增强模型，235B参数"
+      }
+    },
+    {
+      "displayName": "百川 M2-Plus",
+      "family": "baichuan-m2",
+      "match": {
+        "exact": [
+          "Baichuan-M2-Plus"
+        ]
+      },
+      "id": "Baichuan-M2-Plus",
+      "spec": {
+        "contextWindow": 32000,
+        "maxOutputTokens": 32000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "multilingual",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 0.3,
+        "description": "百川智能M2-Plus旗舰模型，最强推理和创作能力"
+      }
+    },
+    {
+      "displayName": "百川 M2",
+      "family": "baichuan-m2",
+      "match": {
+        "exact": [
+          "Baichuan-M2"
+        ]
+      },
+      "id": "Baichuan-M2",
+      "spec": {
+        "contextWindow": 32000,
+        "maxOutputTokens": 32000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "multilingual"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 0.3,
+        "description": "百川智能M2高性能通用模型"
+      }
+    }
+  ]
+}

compute/v2/compatibility/model-specs/baidu.json

@@ -0,0 +1,145 @@
+{
+  "description": "百度文心 ERNIE 系列模型规格。参数来源：config-center compute/providers/baidu.json。ernie-x1.1 为纯推理模型，defaultTemperature 为 null；ernie-5.0-thinking-latest 同时支持 chat 和 reasoning。",
+  "specs": [
+    {
+      "displayName": "文心 ERNIE 5.0",
+      "family": "ernie-5",
+      "match": {
+        "exact": [
+          "ernie-5.0-thinking-latest"
+        ],
+        "patterns": [
+          "ernie-5.0-thinking*"
+        ]
+      },
+      "id": "ernie-5.0-thinking-latest",
+      "spec": {
+        "contextWindow": 128000,
+        "maxOutputTokens": 65536,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "vision",
+          "multimodal",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning"
+        ],
+        "supportsReasoning": true,
+        "description": "百度最新旗舰模型，支持深度推理、多模态和工具调用"
+      }
+    },
+    {
+      "displayName": "文心 ERNIE 5.0 非思考",
+      "family": "ernie-5",
+      "match": {
+        "exact": [
+          "ernie-5.0"
+        ]
+      },
+      "id": "ernie-5.0",
+      "spec": {
+        "contextWindow": 131072,
+        "maxOutputTokens": 65536,
+        "capabilities": [
+          "chat",
+          "vision",
+          "multimodal",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 0.95,
+        "description": "百度 ERNIE 5.0 旗舰模型，128K 上下文，最大输出 65536 token"
+      }
+    },
+    {
+      "displayName": "文心 ERNIE 4.5 Turbo",
+      "family": "ernie-4.5",
+      "match": {
+        "exact": [
+          "ernie-4.5-turbo-128k"
+        ],
+        "patterns": [
+          "ernie-4.5-turbo-128k*"
+        ]
+      },
+      "id": "ernie-4.5-turbo-128k",
+      "spec": {
+        "contextWindow": 131072,
+        "maxOutputTokens": 12288,
+        "capabilities": [
+          "chat",
+          "code",
+          "vision",
+          "long_context",
+          "fast"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 0.8,
+        "description": "百度高性价比长上下文模型，128K 窗口，最大输出 12288 token"
+      }
+    },
+    {
+      "displayName": "文心 ERNIE 4.5 Turbo 20260402",
+      "family": "ernie-4.5",
+      "match": {
+        "exact": [
+          "ernie-4.5-turbo-20260402"
+        ]
+      },
+      "id": "ernie-4.5-turbo-20260402",
+      "spec": {
+        "contextWindow": 131072,
+        "maxOutputTokens": 12288,
+        "capabilities": [
+          "chat",
+          "code",
+          "vision",
+          "long_context",
+          "fast"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 0.8,
+        "description": "百度 ERNIE 4.5 Turbo 20260402 快照版，128K 上下文"
+      }
+    },
+    {
+      "displayName": "文心 ERNIE X1.1",
+      "family": "ernie-x1",
+      "match": {
+        "exact": [
+          "ernie-x1.1"
+        ],
+        "patterns": [
+          "ernie-x1*"
+        ]
+      },
+      "id": "ernie-x1.1",
+      "spec": {
+        "contextWindow": 65536,
+        "maxOutputTokens": 65536,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "deep_thinking",
+          "math",
+          "code"
+        ],
+        "serviceType": [
+          "reasoning"
+        ],
+        "defaultTemperature": null,
+        "supportsReasoning": true,
+        "description": "百度 ERNIE X1.1 深度思考推理模型，64K 上下文"
+      }
+    }
+  ]
+}

compute/v2/compatibility/model-specs/cohere.json

@@ -0,0 +1,297 @@
+{
+  "description": "Cohere 系列模型规格。参数来源：config-center compute/providers/cohere.json。包含对话、Embedding 和 Rerank 三类服务。",
+  "specs": [
+    {
+      "displayName": "Rerank 4 Fast",
+      "family": "rerank-v4.0-fast",
+      "match": {
+        "exact": [
+          "rerank-v4.0-fast"
+        ]
+      },
+      "id": "rerank-v4.0-fast",
+      "spec": {
+        "capabilities": [
+          "rerank",
+          "multilingual"
+        ],
+        "serviceType": [
+          "rerank"
+        ],
+        "description": "Rerank 4 Fast"
+      }
+    },
+    {
+      "displayName": "Embed 5 Fast",
+      "family": "embed-v5.0-fast",
+      "match": {
+        "exact": [
+          "embed-v5.0-fast"
+        ]
+      },
+      "id": "embed-v5.0-fast",
+      "spec": {
+        "capabilities": [
+          "text_embedding",
+          "vision",
+          "multilingual"
+        ],
+        "serviceType": [
+          "embedding"
+        ],
+        "description": "Embed 5 Fast",
+        "contextWindow": 128000,
+        "extra": {
+          "dimensions": [
+            256,
+            512,
+            768,
+            1024,
+            1536,
+            2048
+          ],
+          "defaultDimension": 2048
+        },
+        "releasedAt": "2026-09-30"
+      }
+    },
+    {
+      "displayName": "Rerank 4 Pro",
+      "family": "rerank-v4.0-pro",
+      "match": {
+        "exact": [
+          "rerank-v4.0-pro"
+        ]
+      },
+      "id": "rerank-v4.0-pro",
+      "spec": {
+        "capabilities": [
+          "rerank",
+          "multilingual"
+        ],
+        "serviceType": [
+          "rerank"
+        ],
+        "description": "Rerank 4 Pro"
+      }
+    },
+    {
+      "displayName": "Embed 5 Pro",
+      "family": "embed-v5.0-pro",
+      "match": {
+        "exact": [
+          "embed-v5.0-pro"
+        ]
+      },
+      "id": "embed-v5.0-pro",
+      "spec": {
+        "capabilities": [
+          "text_embedding",
+          "vision",
+          "multilingual"
+        ],
+        "serviceType": [
+          "embedding"
+        ],
+        "description": "Embed 5 Pro",
+        "contextWindow": 128000,
+        "extra": {
+          "dimensions": [
+            256,
+            512,
+            768,
+            1024,
+            1536,
+            2048
+          ],
+          "defaultDimension": 2048
+        },
+        "releasedAt": "2026-09-30"
+      }
+    },
+    {
+      "displayName": "Command A+",
+      "family": "command-a-plus-05-2026",
+      "match": {
+        "exact": [
+          "command-a-plus-05-2026"
+        ]
+      },
+      "id": "command-a-plus-05-2026",
+      "spec": {
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "vision",
+          "tool_use",
+          "multilingual",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "description": "Command A+",
+        "contextWindow": 128000,
+        "maxOutputTokens": 64000,
+        "supportsReasoning": true
+      }
+    },
+    {
+      "displayName": "North Mini Code",
+      "family": "north-mini-code-1-0",
+      "match": {
+        "exact": [
+          "north-mini-code-1-0",
+          "north-mini-code"
+        ]
+      },
+      "id": "north-mini-code-1-0",
+      "spec": {
+        "contextWindow": 256000,
+        "maxOutputTokens": 64000,
+        "capabilities": [
+          "chat",
+          "code",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "description": "Cohere North Mini Code 1.0，原厂API精确ID为north-mini-code-1-0；north-mini-code仅保留本仓历史配置匹配，不声明原厂接受该简称"
+      }
+    },
+    {
+      "displayName": "North Small Translate 1.0",
+      "family": "north-small-translate-1-0",
+      "match": {
+        "exact": [
+          "north-small-translate-1-0"
+        ]
+      },
+      "id": "north-small-translate-1-0",
+      "spec": {
+        "serviceType": [
+          "chat"
+        ],
+        "capabilities": [
+          "chat",
+          "multilingual"
+        ],
+        "description": "Cohere原厂机器翻译模型，支持50多种语言；官网窗口与输出为16k简写，本轮不猜测精确整数。只登记共享规格，原生接入与账号权限需独立验证"
+      }
+    },
+    {
+      "displayName": "Command A",
+      "family": "command-a-03-2025",
+      "match": {
+        "exact": [
+          "command-a-03-2025",
+          "cohere/command-a"
+        ]
+      },
+      "id": "command-a-03-2025",
+      "spec": {
+        "contextWindow": 256000,
+        "maxOutputTokens": 8000,
+        "capabilities": [
+          "chat",
+          "code",
+          "tool_use",
+          "rag",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 0.3,
+        "description": "Cohere Command A 文本模型，256K 上下文，支持 RAG 和工具调用"
+      }
+    },
+    {
+      "displayName": "Command R7B",
+      "family": "command-r7b",
+      "match": {
+        "exact": [
+          "command-r7b-12-2024"
+        ],
+        "patterns": [
+          "command-r7b*"
+        ]
+      },
+      "id": "command-r7b-12-2024",
+      "spec": {
+        "contextWindow": 128000,
+        "maxOutputTokens": 4000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "tool_use",
+          "rag",
+          "fast"
+        ],
+        "serviceType": [
+          "fast"
+        ],
+        "defaultTemperature": 0.3,
+        "description": "Cohere 小型高速对话模型，适合高吞吐、低延迟场景"
+      }
+    },
+    {
+      "displayName": "Embed V4",
+      "family": "embed-v4",
+      "match": {
+        "exact": [
+          "embed-v4.0"
+        ],
+        "patterns": [
+          "embed-v4*"
+        ]
+      },
+      "id": "embed-v4.0",
+      "spec": {
+        "contextWindow": 128000,
+        "capabilities": [
+          "text_embedding",
+          "multilingual"
+        ],
+        "serviceType": [
+          "embedding"
+        ],
+        "description": "Cohere 最新 Embedding 模型，128K 上下文，多语言支持",
+        "extra": {
+          "dimensions": [
+            256,
+            512,
+            1024,
+            1536
+          ],
+          "defaultDimension": 1536
+        }
+      }
+    },
+    {
+      "displayName": "Rerank V3.5",
+      "family": "rerank-v3",
+      "match": {
+        "exact": [
+          "rerank-v3.5"
+        ],
+        "patterns": [
+          "rerank-v3*"
+        ]
+      },
+      "id": "rerank-v3.5",
+      "spec": {
+        "contextWindow": 4096,
+        "capabilities": [
+          "rerank",
+          "semantic_reranking"
+        ],
+        "serviceType": [
+          "rerank"
+        ],
+        "description": "Cohere 语义重排序模型，用于检索结果精排"
+      }
+    }
+  ]
+}

compute/v2/compatibility/model-specs/deepseek.json

@@ -0,0 +1,354 @@
+{
+  "description": "DeepSeek 系列模型规格。参数来源：config-center compute/providers/deepseek.json。",
+  "specs": [
+    {
+      "displayName": "DeepSeek V4 Pro",
+      "family": "deepseek-v4-pro",
+      "match": {
+        "exact": [
+          "deepseek-v4-pro"
+        ]
+      },
+      "id": "deepseek-v4-pro",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 384000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "deep_thinking",
+          "code",
+          "math",
+          "science",
+          "multilingual",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "extra": {
+          "supportsThinking": true,
+          "thinkingDefault": true
+        },
+        "description": "DeepSeek V4 Pro，1.6T MoE 架构，100 万上下文"
+      },
+      "routing": {
+        "tier": "flagship",
+        "routingPriority": 30,
+        "eligibleForAgent": true,
+        "defaultReference": true,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultMode": "high"
+        }
+      }
+    },
+    {
+      "displayName": "DeepSeek V4 Pro 0813",
+      "family": "deepseek-v4-pro-0813",
+      "match": {
+        "exact": [
+          "deepseek-v4-pro-0813"
+        ]
+      },
+      "id": "deepseek-v4-pro-0813",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 384000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "deep_thinking",
+          "code",
+          "math",
+          "science",
+          "multilingual",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "extra": {
+          "supportsThinking": true,
+          "thinkingDefault": true
+        },
+        "description": "DeepSeek V4 Pro 0813 版本，1.6T MoE 架构，100 万上下文"
+      },
+      "routing": {
+        "tier": "flagship",
+        "routingPriority": 30,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultMode": "high"
+        }
+      }
+    },
+    {
+      "displayName": "DeepSeek Chat",
+      "family": "deepseek-chat",
+      "match": {
+        "patterns": [
+          "deepseek-chat*"
+        ]
+      },
+      "id": "deepseek-chat",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 384000,
+        "capabilities": [
+          "chat",
+          "code",
+          "reasoning",
+          "multilingual",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1
+      }
+    },
+    {
+      "displayName": "DeepSeek Reasoner",
+      "family": "deepseek-reasoner",
+      "match": {
+        "patterns": [
+          "deepseek-reasoner*"
+        ]
+      },
+      "id": "deepseek-reasoner",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 384000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "deep_thinking",
+          "code",
+          "math",
+          "tool_use"
+        ],
+        "serviceType": [
+          "reasoning"
+        ],
+        "defaultTemperature": null,
+        "supportsReasoning": true
+      }
+    },
+    {
+      "displayName": "DeepSeek V4.1 Flash",
+      "family": "deepseek-flash",
+      "match": {
+        "exact": [
+          "deepseek-flash",
+          "deepseek-v4.1-flash",
+          "deepseek/deepseek-v4.1-flash"
+        ]
+      },
+      "id": "deepseek-flash",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 384000,
+        "capabilities": [
+          "chat",
+          "code",
+          "reasoning",
+          "deep_thinking",
+          "vision",
+          "multilingual",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning",
+          "vision"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "extra": {
+          "supportsThinking": true,
+          "thinkingDefault": true
+        },
+        "description": "DeepSeek V4.1 Flash，支持文本与图像输入，100 万上下文"
+      },
+      "routing": {
+        "tier": "lightweight",
+        "routingPriority": 35,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultMode": "high"
+        }
+      }
+    },
+    {
+      "displayName": "DeepSeek V4 Flash",
+      "family": "deepseek-v4-flash",
+      "match": {
+        "exact": [
+          "deepseek-v4-flash"
+        ]
+      },
+      "id": "deepseek-v4-flash",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 384000,
+        "capabilities": [
+          "chat",
+          "code",
+          "reasoning",
+          "deep_thinking",
+          "multilingual",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "extra": {
+          "supportsThinking": true,
+          "thinkingDefault": true
+        }
+      },
+      "routing": {
+        "tier": "lightweight",
+        "routingPriority": 35,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultMode": "high"
+        }
+      }
+    },
+    {
+      "displayName": "DeepSeek V4 Flash 0731",
+      "family": "deepseek-v4-flash-0731",
+      "match": {
+        "exact": [
+          "deepseek-v4-flash-0731"
+        ]
+      },
+      "id": "deepseek-v4-flash-0731",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 384000,
+        "capabilities": [
+          "chat",
+          "code",
+          "reasoning",
+          "deep_thinking",
+          "multilingual",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "extra": {
+          "supportsThinking": true,
+          "thinkingDefault": true
+        },
+        "description": "DeepSeek V4 Flash 0731 固定版本，100 万上下文"
+      },
+      "routing": {
+        "tier": "lightweight",
+        "routingPriority": 35,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "high",
+            "max"
+          ],
+          "defaultMode": "high"
+        }
+      }
+    },
+    {
+      "displayName": "DeepSeek V4 Flash Vision Exp",
+      "family": "deepseek-v4-flash-vision-exp",
+      "match": {
+        "exact": [
+          "deepseek-v4-flash-vision-exp"
+        ]
+      },
+      "id": "deepseek-v4-flash-vision-exp",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 384000,
+        "capabilities": [
+          "chat",
+          "code",
+          "reasoning",
+          "deep_thinking",
+          "vision",
+          "multilingual",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning",
+          "vision"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "releasedAt": "2026-08-21",
+        "extra": {
+          "supportsThinking": true,
+          "thinkingDefault": true
+        },
+        "description": "DeepSeek V4 Flash 实验性视觉版本，支持文本与图像输入，文本能力与 V4 Flash 一致"
+      },
+      "routing": {
+        "tier": "lightweight",
+        "routingPriority": 35,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultMode": "high"
+        }
+      }
+    }
+  ]
+}

compute/v2/compatibility/model-specs/google.json

@@ -0,0 +1,535 @@
+{
+  "description": "Google Gemini 系列模型规格。参数来源：config-center compute/providers/google.json。",
+  "specs": [
+    {
+      "displayName": "Gemini Embedding 2",
+      "family": "gemini-embedding-2",
+      "match": {
+        "exact": [
+          "gemini-embedding-2"
+        ]
+      },
+      "id": "gemini-embedding-2",
+      "spec": {
+        "contextWindow": 8192,
+        "capabilities": [
+          "text_embedding",
+          "vision",
+          "audio_understanding",
+          "video_understanding",
+          "multilingual"
+        ],
+        "serviceType": [
+          "embedding"
+        ],
+        "description": "Gemini Embedding 2，原生多模态嵌入，8192 token 输入，共享文本/图像/音视频向量空间",
+        "extra": {
+          "dimensions": [
+            768,
+            1536,
+            3072
+          ],
+          "defaultDimension": 3072
+        }
+      }
+    },
+    {
+      "displayName": "gemini-3.5-transcribe-live",
+      "family": "gemini-3.5-transcribe-live",
+      "match": {
+        "exact": [
+          "gemini-3.5-transcribe-live"
+        ]
+      },
+      "id": "gemini-3.5-transcribe-live",
+      "spec": {
+        "capabilities": [
+          "asr",
+          "realtime"
+        ],
+        "serviceType": [
+          "asr"
+        ],
+        "description": "gemini-3.5-transcribe-live，官方模型目录；原生音频协议需客户端适配"
+      }
+    },
+    {
+      "displayName": "gemini-3.5-transcribe",
+      "family": "gemini-3.5-transcribe",
+      "match": {
+        "exact": [
+          "gemini-3.5-transcribe"
+        ]
+      },
+      "id": "gemini-3.5-transcribe",
+      "spec": {
+        "capabilities": [
+          "asr"
+        ],
+        "serviceType": [
+          "asr"
+        ],
+        "description": "gemini-3.5-transcribe，官方模型目录；原生音频协议需客户端适配"
+      }
+    },
+    {
+      "displayName": "gemini-3.8-live-extended-thinking",
+      "family": "gemini-3.8-live-extended-thinking",
+      "match": {
+        "exact": [
+          "gemini-3.8-live-extended-thinking"
+        ]
+      },
+      "id": "gemini-3.8-live-extended-thinking",
+      "spec": {
+        "capabilities": [
+          "realtime",
+          "speech_to_speech",
+          "reasoning"
+        ],
+        "serviceType": [
+          "realtime_voice"
+        ],
+        "description": "gemini-3.8-live-extended-thinking，官方模型目录；原生音频协议需客户端适配"
+      }
+    },
+    {
+      "displayName": "gemini-3.8-live",
+      "family": "gemini-3.8-live",
+      "match": {
+        "exact": [
+          "gemini-3.8-live"
+        ]
+      },
+      "id": "gemini-3.8-live",
+      "spec": {
+        "capabilities": [
+          "realtime",
+          "speech_to_speech"
+        ],
+        "serviceType": [
+          "realtime_voice"
+        ],
+        "description": "gemini-3.8-live，官方模型目录；原生音频协议需客户端适配"
+      }
+    },
+    {
+      "displayName": "gemini-3.8-flash-lite-tts",
+      "family": "gemini-3.8-flash-lite-tts",
+      "match": {
+        "exact": [
+          "gemini-3.8-flash-lite-tts"
+        ]
+      },
+      "id": "gemini-3.8-flash-lite-tts",
+      "spec": {
+        "capabilities": [
+          "tts",
+          "fast"
+        ],
+        "serviceType": [
+          "tts"
+        ],
+        "description": "gemini-3.8-flash-lite-tts，官方模型目录；原生音频协议需客户端适配"
+      }
+    },
+    {
+      "displayName": "gemini-3.8-flash-tts",
+      "family": "gemini-3.8-flash-tts",
+      "match": {
+        "exact": [
+          "gemini-3.8-flash-tts"
+        ]
+      },
+      "id": "gemini-3.8-flash-tts",
+      "spec": {
+        "capabilities": [
+          "tts"
+        ],
+        "serviceType": [
+          "tts"
+        ],
+        "description": "gemini-3.8-flash-tts，官方模型目录；原生音频协议需客户端适配",
+        "contextWindow": 8192,
+        "maxOutputTokens": 16384
+      }
+    },
+    {
+      "displayName": "Gemini 3.8 Flash",
+      "family": "gemini-3.8-flash",
+      "match": {
+        "exact": [
+          "gemini-3.8-flash"
+        ]
+      },
+      "id": "gemini-3.8-flash",
+      "spec": {
+        "contextWindow": 1048576,
+        "maxOutputTokens": 65536,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "video_understanding",
+          "audio_understanding",
+          "ultra_long_context",
+          "tool_use",
+          "fast"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "description": "Gemini 3.8 Flash，稳定版，支持多模态输入和 1M 上下文"
+      },
+      "routing": {
+        "tier": "balanced",
+        "routingPriority": 10,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "low",
+            "medium",
+            "high"
+          ],
+          "defaultMode": "medium"
+        }
+      }
+    },
+    {
+      "displayName": "Gemini 3.7 Flash",
+      "family": "gemini-3.7-flash",
+      "match": {
+        "exact": [
+          "gemini-3.7-flash"
+        ]
+      },
+      "id": "gemini-3.7-flash",
+      "spec": {
+        "contextWindow": 1048576,
+        "maxOutputTokens": 65536,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "video_understanding",
+          "audio_understanding",
+          "ultra_long_context",
+          "tool_use",
+          "fast"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "description": "Gemini 3.7 Flash，稳定版，支持多模态输入和 1M 上下文"
+      },
+      "routing": {
+        "tier": "balanced",
+        "routingPriority": 20,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "low",
+            "medium",
+            "high"
+          ],
+          "defaultMode": "medium"
+        }
+      }
+    },
+    {
+      "displayName": "Gemini 3.5 Flash-Lite",
+      "family": "gemini-3.5-flash-lite",
+      "match": {
+        "exact": [
+          "gemini-3.5-flash-lite"
+        ]
+      },
+      "id": "gemini-3.5-flash-lite",
+      "spec": {
+        "contextWindow": 1048576,
+        "maxOutputTokens": 65536,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "video_understanding",
+          "audio_understanding",
+          "ultra_long_context",
+          "tool_use",
+          "fast"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "description": "Gemini 3.5 Flash-Lite，稳定版，支持多模态输入和 1M 上下文"
+      },
+      "routing": {
+        "tier": "lightweight",
+        "routingPriority": 20,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "low",
+            "medium",
+            "high"
+          ],
+          "defaultMode": "medium"
+        }
+      }
+    },
+    {
+      "displayName": "Gemini 3.1 Flash Lite Image (Nano Banana 2 Lite)",
+      "family": "gemini-3.1-flash-lite-image",
+      "match": {
+        "exact": [
+          "gemini-3.1-flash-lite-image"
+        ]
+      },
+      "id": "gemini-3.1-flash-lite-image",
+      "spec": {
+        "contextWindow": 65536,
+        "maxOutputTokens": 4096,
+        "capabilities": [
+          "chat",
+          "vision",
+          "image_generation",
+          "image_editing",
+          "tool_use"
+        ],
+        "serviceType": [
+          "image_gen"
+        ],
+        "supportsReasoning": true,
+        "description": "Nano Banana 2 Lite，Google 面向高吞吐图像生成与编辑的轻量模型",
+        "releasedAt": "2026-06-30"
+      }
+    },
+    {
+      "displayName": "Gemini 3.5 Flash",
+      "family": "gemini-3.5",
+      "match": {
+        "exact": [
+          "gemini-3.5-flash"
+        ],
+        "patterns": [
+          "gemini-3.5-flash*"
+        ]
+      },
+      "id": "gemini-3.5-flash",
+      "spec": {
+        "contextWindow": 1048576,
+        "maxOutputTokens": 65536,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "video_understanding",
+          "audio_understanding",
+          "ultra_long_context",
+          "tool_use",
+          "fast"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "description": "Google Gemini 3.5 Flash，全模态高性价比模型，100 万上下文"
+      }
+    },
+    {
+      "displayName": "Gemini 3.1 Flash Image (Nano Banana 2)",
+      "family": "gemini-3.1",
+      "match": {
+        "exact": [
+          "gemini-3.1-flash-image"
+        ],
+        "patterns": [
+          "gemini-3.1-flash-image*"
+        ]
+      },
+      "id": "gemini-3.1-flash-image",
+      "spec": {
+        "contextWindow": 131072,
+        "maxOutputTokens": 32768,
+        "capabilities": [
+          "image_generation",
+          "image_editing",
+          "text_rendering"
+        ],
+        "serviceType": [
+          "image_gen"
+        ],
+        "description": "Google Nano Banana 2，最新文生图与编辑模型"
+      }
+    },
+    {
+      "displayName": "Gemini 3.1 Pro",
+      "family": "gemini-3.1",
+      "id": "gemini-3.1-pro-preview",
+      "spec": {
+        "contextWindow": 1048576,
+        "maxOutputTokens": 65536,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "ultra_long_context",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true
+      }
+    },
+    {
+      "displayName": "Gemini 3 Flash",
+      "family": "gemini-3",
+      "id": "gemini-3-flash-preview",
+      "spec": {
+        "contextWindow": 1048576,
+        "maxOutputTokens": 65536,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "ultra_long_context",
+          "tool_use",
+          "fast"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true
+      }
+    },
+    {
+      "displayName": "Gemini 2.5 Pro",
+      "family": "gemini-2.5",
+      "match": {
+        "patterns": [
+          "gemini-2.5-pro*"
+        ]
+      },
+      "id": "gemini-2.5-pro",
+      "spec": {
+        "contextWindow": 1048576,
+        "maxOutputTokens": 65536,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "ultra_long_context",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true
+      }
+    },
+    {
+      "displayName": "Gemini 2.5 Flash",
+      "family": "gemini-2.5",
+      "match": {
+        "patterns": [
+          "gemini-2.5-flash*"
+        ]
+      },
+      "id": "gemini-2.5-flash",
+      "spec": {
+        "contextWindow": 1048576,
+        "maxOutputTokens": 65536,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "ultra_long_context",
+          "tool_use",
+          "fast"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true
+      }
+    },
+    {
+      "displayName": "Gemini 3.1 Flash-Lite Preview",
+      "family": "gemini-3.1",
+      "match": {
+        "exact": [
+          "gemini-3.1-flash-lite-preview"
+        ],
+        "patterns": [
+          "gemini-3.1-flash-lite*"
+        ]
+      },
+      "id": "gemini-3.1-flash-lite-preview",
+      "spec": {
+        "contextWindow": 1048576,
+        "maxOutputTokens": 65536,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "ultra_long_context",
+          "tool_use",
+          "fast"
+        ],
+        "serviceType": [
+          "fast"
+        ],
+        "defaultTemperature": 1,
+        "description": "Google Gemini 3.1 Flash-Lite Preview，低成本高吞吐多模态模型"
+      }
+    },
+    {
+      "displayName": "Text Embedding 005",
+      "family": "text-embedding",
+      "match": {
+        "exact": [
+          "text-embedding-005"
+        ]
+      },
+      "id": "text-embedding-005",
+      "spec": {
+        "contextWindow": 2048,
+        "capabilities": [
+          "text_embedding",
+          "semantic_search"
+        ],
+        "serviceType": [
+          "embedding"
+        ],
+        "description": "Google 文本嵌入模型，768 维度"
+      }
+    }
+  ]
+}

compute/v2/compatibility/model-specs/happyhorse.json

@@ -0,0 +1,184 @@
+{
+  "description": "HappyHorse（快乐马）视频生成系列模型规格。通过阿里云百炼平台提供推理服务。官网：https://happyhorse.com。所有模型均为视频生成（video_gen），无上下文窗口概念。",
+  "specs": [
+    {
+      "displayName": "HappyHorse 1.1 T2V",
+      "family": "happyhorse-t2v",
+      "match": {
+        "exact": [
+          "happyhorse-1.1-t2v"
+        ],
+        "patterns": [
+          "happyhorse-1.1-t2v*"
+        ]
+      },
+      "id": "happyhorse-1.1-t2v",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "text_to_video",
+          "high_quality"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "HappyHorse 1.1 文生视频，进一步提升文本词义理解、镜头调度与动态生成表现"
+      },
+      "routing": {
+        "tier": "flagship",
+        "routingPriority": 116,
+        "eligibleForAgent": false,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto"
+          ],
+          "defaultMode": "auto"
+        }
+      }
+    },
+    {
+      "displayName": "HappyHorse 1.0 T2V",
+      "family": "happyhorse-t2v",
+      "match": {
+        "exact": [
+          "happyhorse-1.0-t2v"
+        ],
+        "patterns": [
+          "happyhorse-1.0-t2v*"
+        ]
+      },
+      "id": "happyhorse-1.0-t2v",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "text_to_video"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "HappyHorse 1.0 文生视频，具备高度还原的动态画面生成能力"
+      }
+    },
+    {
+      "displayName": "HappyHorse 1.1 I2V",
+      "family": "happyhorse-i2v",
+      "match": {
+        "exact": [
+          "happyhorse-1.1-i2v"
+        ],
+        "patterns": [
+          "happyhorse-1.1-i2v*"
+        ]
+      },
+      "id": "happyhorse-1.1-i2v",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "image_to_video",
+          "high_quality"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "HappyHorse 1.1 图生视频，进一步提升画面质感、动态表现与跨片段一致性"
+      },
+      "routing": {
+        "tier": "flagship",
+        "routingPriority": 115,
+        "eligibleForAgent": false,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto"
+          ],
+          "defaultMode": "auto"
+        }
+      }
+    },
+    {
+      "displayName": "HappyHorse 1.0 I2V",
+      "family": "happyhorse-i2v",
+      "match": {
+        "exact": [
+          "happyhorse-1.0-i2v"
+        ],
+        "patterns": [
+          "happyhorse-1.0-i2v*"
+        ]
+      },
+      "id": "happyhorse-1.0-i2v",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "image_to_video"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "HappyHorse 1.0 图生视频，具备高度还原的动态画面生成能力"
+      }
+    },
+    {
+      "displayName": "HappyHorse 1.1 R2V",
+      "family": "happyhorse-r2v",
+      "match": {
+        "exact": [
+          "happyhorse-1.1-r2v"
+        ],
+        "patterns": [
+          "happyhorse-1.1-r2v*"
+        ]
+      },
+      "id": "happyhorse-1.1-r2v",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "reference_to_video",
+          "multi_image",
+          "high_quality"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "HappyHorse 1.1 参考生视频，支持最多9张图片参考，更加稳定的主体与场景参考"
+      },
+      "routing": {
+        "tier": "flagship",
+        "routingPriority": 117,
+        "eligibleForAgent": false,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto"
+          ],
+          "defaultMode": "auto"
+        }
+      }
+    },
+    {
+      "displayName": "HappyHorse 1.0 R2V",
+      "family": "happyhorse-r2v",
+      "match": {
+        "exact": [
+          "happyhorse-1.0-r2v"
+        ],
+        "patterns": [
+          "happyhorse-1.0-r2v*"
+        ]
+      },
+      "id": "happyhorse-1.0-r2v",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "reference_to_video",
+          "multi_image"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "HappyHorse 1.0 参考生视频，支持最多9张图片参考"
+      }
+    }
+  ]
+}

compute/v2/compatibility/model-specs/kling.json

@@ -0,0 +1,133 @@
+{
+  "description": "快手可灵视频生成系列模型规格。参数来源：config-center compute/providers/kling.json。所有模型均为视频生成（video_gen），无上下文窗口概念。",
+  "specs": [
+    {
+      "displayName": "可灵 3.0 Turbo",
+      "family": "kling-3.0-turbo",
+      "match": {
+        "exact": [
+          "kling-3.0-turbo"
+        ]
+      },
+      "id": "kling-3.0-turbo",
+      "spec": {
+        "serviceType": [
+          "video_gen"
+        ],
+        "capabilities": [
+          "video_generation",
+          "image_to_video"
+        ],
+        "description": "可灵 3.0 Turbo，新版 /image-to-video/kling-3.0-turbo API，需独立协议适配"
+      }
+    },
+    {
+      "displayName": "可灵 3.0",
+      "family": "kling-v3",
+      "match": {
+        "exact": [
+          "kling-v3"
+        ]
+      },
+      "id": "kling-v3",
+      "spec": {
+        "serviceType": [
+          "video_gen"
+        ],
+        "capabilities": [
+          "video_generation",
+          "image_to_video"
+        ],
+        "description": "可灵 3.0，通过现有 /v1/videos/text2video 任务 API 的 model_name 调用"
+      }
+    },
+    {
+      "displayName": "可灵 V2.5 Turbo",
+      "family": "kling-v2-5",
+      "match": {
+        "exact": [
+          "kling-v2-5-turbo"
+        ],
+        "patterns": [
+          "kling-v2-5-turbo*"
+        ]
+      },
+      "id": "kling-v2-5-turbo",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "chinese_optimized",
+          "high_quality",
+          "fast"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "可灵最新视频生成模型，高性价比"
+      }
+    },
+    {
+      "displayName": "可灵 V2.5 Turbo Pro",
+      "family": "kling-v2-5",
+      "match": {
+        "exact": [
+          "kling-v2-5-turbo-pro"
+        ]
+      },
+      "id": "kling-v2-5-turbo-pro",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "chinese_optimized",
+          "ultra_quality"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "可灵高品质视频生成，支持 4K 分辨率"
+      }
+    },
+    {
+      "displayName": "可灵 V2",
+      "family": "kling-v2",
+      "match": {
+        "exact": [
+          "kling-v2"
+        ]
+      },
+      "id": "kling-v2",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "chinese_optimized",
+          "high_quality"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "可灵文生视频旗舰模型"
+      }
+    },
+    {
+      "displayName": "可灵 V2 Master",
+      "family": "kling-v2",
+      "match": {
+        "exact": [
+          "kling-v2-master"
+        ]
+      },
+      "id": "kling-v2-master",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "chinese_optimized",
+          "ultra_quality"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "可灵最高品质视频生成，支持 4K 分辨率"
+      }
+    }
+  ]
+}

compute/v2/compatibility/model-specs/minimax.json

@@ -0,0 +1,762 @@
+{
+  "description": "MiniMax 系列模型规格。参数来源：config-center compute/providers/minimax.json。id 保留上游 model_name 原始写法（含大小写与点号），匹配时由匹配器归一化处理。",
+  "specs": [
+    {
+      "displayName": "MiniMax H3 Max",
+      "family": "MiniMax-H3-Max",
+      "match": {
+        "exact": [
+          "MiniMax-H3-Max"
+        ]
+      },
+      "id": "MiniMax-H3-Max",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "image_to_video"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "MiniMax H3 Max，使用原生视频 V2 API（/v1/video_generation_v2），不适用 Anthropic Messages",
+        "extra": {
+          "maxVideoDuration": 15
+        }
+      }
+    },
+    {
+      "displayName": "MiniMax H3",
+      "family": "MiniMax-H3",
+      "match": {
+        "exact": [
+          "MiniMax-H3"
+        ]
+      },
+      "id": "MiniMax-H3",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "image_to_video"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "MiniMax H3，使用原生视频 V2 API（/v1/video_generation_v2），不适用 Anthropic Messages",
+        "extra": {
+          "maxVideoDuration": 15
+        }
+      }
+    },
+    {
+      "displayName": "MiniMax M3.1 Flash Preview",
+      "family": "MiniMax-M3.1-Flash-Preview",
+      "match": {
+        "exact": [
+          "MiniMax-M3.1-Flash-Preview"
+        ]
+      },
+      "id": "MiniMax-M3.1-Flash-Preview",
+      "spec": {
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "video_understanding",
+          "tool_use",
+          "agent",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning",
+          "vision"
+        ],
+        "description": "MiniMax M3.1 Flash Preview",
+        "contextWindow": 1000000,
+        "extra": {
+          "adaptiveThinking": true,
+          "thinkingOnly": true
+        },
+        "supportsReasoning": true
+      }
+    },
+    {
+      "displayName": "MiniMax M3",
+      "family": "minimax-m3",
+      "match": {
+        "exact": [
+          "MiniMax-M3"
+        ],
+        "patterns": [
+          "MiniMax-M3*"
+        ]
+      },
+      "id": "MiniMax-M3",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 512000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "video_understanding",
+          "tool_use",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "description": "MiniMax M3 全模态基础模型，100 万上下文"
+      },
+      "routing": {
+        "tier": "balanced",
+        "routingPriority": 30,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "off",
+            "minimal",
+            "low",
+            "medium",
+            "high",
+            "xhigh"
+          ],
+          "defaultMode": "medium"
+        }
+      }
+    },
+    {
+      "displayName": "MiniMax M2.7",
+      "family": "minimax-m2.7",
+      "id": "MiniMax-M2.7",
+      "spec": {
+        "contextWindow": 204800,
+        "maxOutputTokens": 131072,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true
+      }
+    },
+    {
+      "displayName": "MiniMax M2.5",
+      "family": "minimax-m2.5",
+      "id": "MiniMax-M2.5",
+      "spec": {
+        "contextWindow": 204800,
+        "maxOutputTokens": 131072,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "tool_use",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true
+      }
+    },
+    {
+      "displayName": "MiniMax M2",
+      "family": "minimax-m2",
+      "id": "MiniMax-M2",
+      "spec": {
+        "contextWindow": 204800,
+        "maxOutputTokens": 131072,
+        "capabilities": [
+          "chat",
+          "code",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1
+      }
+    },
+    {
+      "displayName": "MiniMax M2.7 高速版",
+      "family": "minimax-m2.7",
+      "match": {
+        "exact": [
+          "MiniMax-M2.7-highspeed"
+        ]
+      },
+      "id": "MiniMax-M2.7-highspeed",
+      "spec": {
+        "contextWindow": 204800,
+        "maxOutputTokens": 131072,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "tool_use",
+          "fast"
+        ],
+        "serviceType": [
+          "fast"
+        ],
+        "defaultTemperature": 1,
+        "description": "MiniMax M2.7 极速版，效果一致，输出更快，适合低延迟场景"
+      }
+    },
+    {
+      "displayName": "MiniMax M2.5 高速版",
+      "family": "minimax-m2.5",
+      "match": {
+        "exact": [
+          "MiniMax-M2.5-highspeed"
+        ]
+      },
+      "id": "MiniMax-M2.5-highspeed",
+      "spec": {
+        "contextWindow": 204800,
+        "maxOutputTokens": 131072,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "tool_use",
+          "fast"
+        ],
+        "serviceType": [
+          "fast"
+        ],
+        "defaultTemperature": 1,
+        "description": "MiniMax M2.5 高速推理版本（100TPS），适合低延迟场景"
+      }
+    },
+    {
+      "displayName": "MiniMax M2.1",
+      "family": "minimax-m2.1",
+      "match": {
+        "exact": [
+          "MiniMax-M2.1"
+        ]
+      },
+      "id": "MiniMax-M2.1",
+      "spec": {
+        "contextWindow": 204800,
+        "maxOutputTokens": 131072,
+        "capabilities": [
+          "chat",
+          "code",
+          "tool_use",
+          "reasoning"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "description": "MiniMax 上一代主力模型，200K 上下文，性价比优秀"
+      }
+    },
+    {
+      "displayName": "MiniMax M2.1 高速版",
+      "family": "minimax-m2.1",
+      "match": {
+        "exact": [
+          "MiniMax-M2.1-highspeed"
+        ]
+      },
+      "id": "MiniMax-M2.1-highspeed",
+      "spec": {
+        "contextWindow": 204800,
+        "maxOutputTokens": 131072,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "tool_use",
+          "fast"
+        ],
+        "serviceType": [
+          "fast"
+        ],
+        "defaultTemperature": 1,
+        "description": "MiniMax M2.1 高速推理版本，适合低延迟场景"
+      }
+    },
+    {
+      "displayName": "MiniMax M2 Her",
+      "family": "minimax-m2",
+      "match": {
+        "exact": [
+          "M2-her"
+        ]
+      },
+      "id": "M2-her",
+      "spec": {
+        "contextWindow": 204800,
+        "maxOutputTokens": 131072,
+        "capabilities": [
+          "chat",
+          "roleplay"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "description": "MiniMax 角色扮演对话模型，专为多轮角色对话设计"
+      }
+    },
+    {
+      "displayName": "MiniMax Text 01",
+      "family": "minimax-text",
+      "match": {
+        "exact": [
+          "MiniMax-Text-01"
+        ]
+      },
+      "id": "MiniMax-Text-01",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 131072,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 0.1,
+        "description": "MiniMax 百万级长上下文文本模型，擅长长文档处理"
+      }
+    },
+    {
+      "displayName": "MiniMax Image 01",
+      "family": "minimax-image",
+      "match": {
+        "exact": [
+          "image-01"
+        ]
+      },
+      "id": "image-01",
+      "spec": {
+        "capabilities": [
+          "image_generation",
+          "subject_reference",
+          "chinese_optimized"
+        ],
+        "serviceType": [
+          "image_gen"
+        ],
+        "description": "MiniMax 文生图/图生图模型，支持主题参考（角色一致性）"
+      },
+      "routing": {
+        "tier": "balanced",
+        "routingPriority": 113,
+        "eligibleForAgent": false,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto"
+          ],
+          "defaultMode": "auto"
+        }
+      }
+    },
+    {
+      "displayName": "MiniMax Image 01 Live",
+      "family": "minimax-image",
+      "match": {
+        "exact": [
+          "image-01-live"
+        ]
+      },
+      "id": "image-01-live",
+      "spec": {
+        "capabilities": [
+          "image_generation",
+          "style_control",
+          "subject_reference",
+          "chinese_optimized"
+        ],
+        "serviceType": [
+          "image_gen"
+        ],
+        "description": "MiniMax 文生图模型（Live 版），支持画风控制"
+      }
+    },
+    {
+      "displayName": "MiniMax Speech 2.8 HD",
+      "family": "minimax-speech",
+      "match": {
+        "exact": [
+          "speech-2.8-hd"
+        ]
+      },
+      "id": "speech-2.8-hd",
+      "spec": {
+        "capabilities": [
+          "tts",
+          "multilingual",
+          "emotion_control",
+          "voice_clone",
+          "chinese_optimized"
+        ],
+        "serviceType": [
+          "tts"
+        ],
+        "description": "MiniMax 最新高清语音合成，40+ 语言、100+ 预置音色"
+      }
+    },
+    {
+      "displayName": "MiniMax Speech 2.8 Turbo",
+      "family": "minimax-speech",
+      "match": {
+        "exact": [
+          "speech-2.8-turbo"
+        ]
+      },
+      "id": "speech-2.8-turbo",
+      "spec": {
+        "capabilities": [
+          "tts",
+          "multilingual",
+          "emotion_control",
+          "fast",
+          "chinese_optimized"
+        ],
+        "serviceType": [
+          "tts"
+        ],
+        "description": "MiniMax 快速语音合成模型，速度优先，适合实时场景"
+      }
+    },
+    {
+      "displayName": "MiniMax Speech 2.6 HD",
+      "family": "minimax-speech",
+      "match": {
+        "exact": [
+          "speech-2.6-hd"
+        ]
+      },
+      "id": "speech-2.6-hd",
+      "spec": {
+        "capabilities": [
+          "tts",
+          "multilingual",
+          "emotion_control",
+          "voice_clone",
+          "chinese_optimized"
+        ],
+        "serviceType": [
+          "tts"
+        ],
+        "description": "MiniMax 高清语音合成（2.6 版），音质优秀，生成速度快"
+      }
+    },
+    {
+      "displayName": "MiniMax Speech 2.6 Turbo",
+      "family": "minimax-speech",
+      "match": {
+        "exact": [
+          "speech-2.6-turbo"
+        ]
+      },
+      "id": "speech-2.6-turbo",
+      "spec": {
+        "capabilities": [
+          "tts",
+          "multilingual",
+          "emotion_control",
+          "fast",
+          "chinese_optimized"
+        ],
+        "serviceType": [
+          "tts"
+        ],
+        "description": "MiniMax 快速语音合成（2.6 版），低延迟响应"
+      }
+    },
+    {
+      "displayName": "MiniMax Speech 02 HD",
+      "family": "minimax-speech",
+      "match": {
+        "exact": [
+          "speech-02-hd"
+        ]
+      },
+      "id": "speech-02-hd",
+      "spec": {
+        "capabilities": [
+          "tts",
+          "multilingual",
+          "voice_clone",
+          "chinese_optimized"
+        ],
+        "serviceType": [
+          "tts"
+        ],
+        "description": "MiniMax 高清语音合成（历史版本），韵律稳定性优秀"
+      }
+    },
+    {
+      "displayName": "MiniMax Speech 02 Turbo",
+      "family": "minimax-speech",
+      "match": {
+        "exact": [
+          "speech-02-turbo"
+        ]
+      },
+      "id": "speech-02-turbo",
+      "spec": {
+        "capabilities": [
+          "tts",
+          "multilingual",
+          "fast",
+          "chinese_optimized"
+        ],
+        "serviceType": [
+          "tts"
+        ],
+        "description": "MiniMax 快速语音合成（历史版本），增强小语种支持"
+      }
+    },
+    {
+      "displayName": "海螺视频 2.3",
+      "family": "minimax-hailuo",
+      "match": {
+        "exact": [
+          "MiniMax-Hailuo-2.3"
+        ]
+      },
+      "id": "MiniMax-Hailuo-2.3",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "text_to_video",
+          "image_to_video",
+          "camera_control",
+          "chinese_optimized",
+          "high_quality"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "MiniMax 最新视频生成模型，支持 15 种运镜指令控制"
+      },
+      "routing": {
+        "tier": "balanced",
+        "routingPriority": 102,
+        "eligibleForAgent": false,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto"
+          ],
+          "defaultMode": "auto"
+        }
+      }
+    },
+    {
+      "displayName": "海螺视频 2.3 快速版",
+      "family": "minimax-hailuo-2.3-fast",
+      "match": {
+        "exact": [
+          "MiniMax-Hailuo-2.3-Fast",
+          "MiniMax-Hailuo-2.3-fast"
+        ]
+      },
+      "id": "MiniMax-Hailuo-2.3-Fast",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "image_to_video",
+          "camera_control",
+          "fast",
+          "chinese_optimized"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "海螺视频快速版本（仅图生视频），生成速度更快"
+      },
+      "routing": {
+        "tier": "lightweight",
+        "routingPriority": 101,
+        "eligibleForAgent": false,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto"
+          ],
+          "defaultMode": "auto"
+        }
+      }
+    },
+    {
+      "displayName": "海螺视频 Director",
+      "family": "minimax-hailuo",
+      "match": {
+        "exact": [
+          "T2V-01-Director"
+        ]
+      },
+      "id": "T2V-01-Director",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "text_to_video",
+          "camera_control",
+          "chinese_optimized"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "导演级文生视频模型，精准运镜控制（历史版本）"
+      }
+    },
+    {
+      "displayName": "海螺视频 02",
+      "family": "minimax-hailuo",
+      "match": {
+        "exact": [
+          "MiniMax-Hailuo-02"
+        ]
+      },
+      "id": "MiniMax-Hailuo-02",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "text_to_video",
+          "image_to_video",
+          "chinese_optimized"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "海螺视频 02，原生 1080P 生成（历史版本）"
+      }
+    },
+    {
+      "displayName": "海螺视频 主体参考",
+      "family": "minimax-hailuo",
+      "match": {
+        "exact": [
+          "S2V-01"
+        ]
+      },
+      "id": "S2V-01",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "subject_reference",
+          "chinese_optimized"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "主体参考视频生成，基于参考图保持角色一致性"
+      }
+    },
+    {
+      "displayName": "MiniMax Music 2.6",
+      "family": "minimax-music",
+      "match": {
+        "exact": [
+          "music-2.6"
+        ]
+      },
+      "id": "music-2.6",
+      "spec": {
+        "capabilities": [
+          "music_generation",
+          "lyrics_input",
+          "instrumental",
+          "chinese_optimized"
+        ],
+        "serviceType": [
+          "music_gen"
+        ],
+        "description": "MiniMax AI 音乐生成模型，支持音乐灵感描述和歌词输入；音乐付费 API 自2026-08-20起仅历史付费用户可继续使用，不向新用户开放；免费音乐接口已停止服务"
+      }
+    },
+    {
+      "displayName": "MiniMax Music 2.5+",
+      "family": "minimax-music",
+      "match": {
+        "exact": [
+          "music-2.5+"
+        ]
+      },
+      "id": "music-2.5+",
+      "spec": {
+        "capabilities": [
+          "music_generation",
+          "lyrics_input",
+          "instrumental",
+          "chinese_optimized"
+        ],
+        "serviceType": [
+          "music_gen"
+        ],
+        "description": "MiniMax AI 音乐生成（2.5+ 版），音质提升；音乐付费 API 自2026-08-20起仅历史付费用户可继续使用，不向新用户开放；免费音乐接口已停止服务"
+      }
+    },
+    {
+      "displayName": "MiniMax Music 2.5",
+      "family": "minimax-music",
+      "match": {
+        "exact": [
+          "music-2.5"
+        ]
+      },
+      "id": "music-2.5",
+      "spec": {
+        "capabilities": [
+          "music_generation",
+          "lyrics_input",
+          "instrumental",
+          "chinese_optimized"
+        ],
+        "serviceType": [
+          "music_gen"
+        ],
+        "description": "MiniMax AI 音乐生成（2.5 版）；音乐付费 API 自2026-08-20起仅历史付费用户可继续使用，不向新用户开放；免费音乐接口已停止服务"
+      }
+    },
+    {
+      "displayName": "MiniMax Music Cover",
+      "family": "minimax-music",
+      "match": {
+        "exact": [
+          "music-cover"
+        ]
+      },
+      "id": "music-cover",
+      "spec": {
+        "capabilities": [
+          "music_generation",
+          "cover",
+          "chinese_optimized"
+        ],
+        "serviceType": [
+          "music_gen"
+        ],
+        "description": "MiniMax AI 翻唱模型，基于参考音频生成翻唱版本；音乐付费 API 自2026-08-20起仅历史付费用户可继续使用，不向新用户开放；免费音乐接口已停止服务"
+      }
+    }
+  ]
+}

compute/v2/compatibility/model-specs/mistral.json

@@ -0,0 +1,107 @@
+{
+  "description": "Mistral 系列模型规格。参数来源：config-center compute/providers/mistral.json。",
+  "specs": [
+    {
+      "displayName": "Mistral Medium 3.5",
+      "family": "mistral-medium-3-5",
+      "match": {
+        "exact": [
+          "mistral-medium-3-5",
+          "mistral-medium-3.5",
+          "mistral-medium-latest"
+        ]
+      },
+      "id": "mistral-medium-3.5",
+      "spec": {
+        "contextWindow": 262144,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "description": "Mistral Medium 3.5，官方 API ID mistral-medium-3-5，256K 上下文"
+      }
+    },
+    {
+      "displayName": "Mistral Large",
+      "family": "mistral-large-2512",
+      "match": {
+        "exact": [
+          "mistral-large-latest",
+          "mistral-large-2512"
+        ]
+      },
+      "id": "mistral-large-latest",
+      "spec": {
+        "contextWindow": 256000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "tool_use",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 0.7
+      }
+    },
+    {
+      "displayName": "Mistral Small 4",
+      "family": "mistral-small-2603",
+      "match": {
+        "exact": [
+          "mistral-small-latest",
+          "mistral-small-2603"
+        ]
+      },
+      "id": "mistral-small-latest",
+      "spec": {
+        "contextWindow": 262144,
+        "capabilities": [
+          "chat",
+          "code",
+          "vision",
+          "reasoning",
+          "fast",
+          "tool_use"
+        ],
+        "serviceType": [
+          "fast"
+        ],
+        "defaultTemperature": 0.7
+      }
+    },
+    {
+      "displayName": "Codestral",
+      "family": "codestral",
+      "match": {
+        "patterns": [
+          "codestral*"
+        ]
+      },
+      "id": "codestral-latest",
+      "spec": {
+        "contextWindow": 128000,
+        "maxOutputTokens": 32768,
+        "capabilities": [
+          "code",
+          "fill_in_the_middle"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 0.7
+      }
+    }
+  ]
+}

compute/v2/compatibility/model-specs/moonshot.json

@@ -0,0 +1,255 @@
+{
+  "description": "Moonshot Kimi 系列模型规格。参数来源：config-center compute/providers/moonshot.json。注意 kimi-k2 存在多个子版本（k2.5/k2.6/k2-thinking），故 kimi-k2 仅用 exact 主键匹配，不用宽 pattern。",
+  "specs": [
+    {
+      "displayName": "Kimi K3",
+      "family": "kimi-k3",
+      "match": {
+        "exact": [
+          "kimi-k3",
+          "moonshotai/kimi-k3"
+        ]
+      },
+      "id": "kimi-k3",
+      "spec": {
+        "contextWindow": 1048576,
+        "maxOutputTokens": 1048576,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "tool_use",
+          "agent",
+          "long_context",
+          "vision",
+          "video_understanding"
+        ],
+        "serviceType": [
+          "reasoning"
+        ],
+        "supportsReasoning": true,
+        "description": "Kimi K3，1M 上下文，始终开启思考，low/high/max 三档（默认 max）；输出上限与输入共享窗口，默认 max_completion_tokens=131072",
+        "extra": {
+          "thinkingOnly": true,
+          "samplingParametersDeprecated": true
+        }
+      },
+      "routing": {
+        "tier": "flagship",
+        "routingPriority": 35,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultMode": "max"
+        }
+      }
+    },
+    {
+      "displayName": "Kimi K2.7 Code",
+      "family": "kimi-k2.7",
+      "match": {
+        "exact": [
+          "kimi-k2.7-code"
+        ],
+        "patterns": [
+          "kimi-k2.7-code*"
+        ]
+      },
+      "id": "kimi-k2.7-code",
+      "spec": {
+        "contextWindow": 262144,
+        "maxOutputTokens": 16384,
+        "capabilities": [
+          "chat",
+          "code",
+          "vision",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "description": "Moonshot Kimi K2.7 Code 编程专用模型"
+      },
+      "routing": {
+        "tier": "balanced",
+        "routingPriority": 50,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto"
+          ],
+          "defaultMode": "auto"
+        }
+      }
+    },
+    {
+      "displayName": "Kimi K2.6",
+      "family": "kimi-k2.6",
+      "id": "kimi-k2.6",
+      "spec": {
+        "contextWindow": 262144,
+        "maxOutputTokens": 16384,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "tool_use",
+          "agent",
+          "long_context",
+          "vision"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true
+      },
+      "routing": {
+        "tier": "balanced",
+        "routingPriority": 45,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "off",
+            "minimal",
+            "low",
+            "medium",
+            "high",
+            "xhigh"
+          ],
+          "defaultMode": "medium"
+        }
+      }
+    },
+    {
+      "displayName": "Kimi K2.5",
+      "family": "kimi-k2.5",
+      "id": "kimi-k2.5",
+      "spec": {
+        "contextWindow": 256000,
+        "maxOutputTokens": 32768,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "tool_use",
+          "agent",
+          "long_context",
+          "vision"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true
+      }
+    },
+    {
+      "displayName": "Kimi K2 Thinking",
+      "family": "kimi-k2-thinking",
+      "id": "kimi-k2-thinking",
+      "spec": {
+        "contextWindow": 256000,
+        "maxOutputTokens": 16384,
+        "capabilities": [
+          "reasoning",
+          "math",
+          "code",
+          "deep_thinking",
+          "long_context"
+        ],
+        "serviceType": [
+          "reasoning"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true
+      }
+    },
+    {
+      "displayName": "Kimi K2",
+      "family": "kimi-k2",
+      "id": "kimi-k2",
+      "spec": {
+        "contextWindow": 256000,
+        "maxOutputTokens": 8192,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "tool_use",
+          "agent",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 0.6,
+        "supportsReasoning": true
+      }
+    },
+    {
+      "displayName": "Moonshot v1 8K",
+      "family": "moonshot-v1",
+      "id": "moonshot-v1-8k",
+      "spec": {
+        "contextWindow": 8192,
+        "maxOutputTokens": 4096,
+        "capabilities": [
+          "chat",
+          "code"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 0
+      }
+    },
+    {
+      "displayName": "Moonshot v1 32K",
+      "family": "moonshot-v1",
+      "id": "moonshot-v1-32k",
+      "spec": {
+        "contextWindow": 32768,
+        "maxOutputTokens": 4096,
+        "capabilities": [
+          "chat",
+          "code",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 0
+      }
+    },
+    {
+      "displayName": "Moonshot v1 128K",
+      "family": "moonshot-v1",
+      "id": "moonshot-v1-128k",
+      "spec": {
+        "contextWindow": 131072,
+        "maxOutputTokens": 4096,
+        "capabilities": [
+          "chat",
+          "code",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 0
+      }
+    }
+  ]
+}

compute/v2/compatibility/model-specs/perplexity.json

@@ -0,0 +1,85 @@
+{
+  "description": "Perplexity Sonar 系列模型规格。参数来源：config-center compute/providers/perplexity.json。所有模型均内置实时联网搜索能力。",
+  "specs": [
+    {
+      "displayName": "Sonar Pro",
+      "family": "sonar-pro",
+      "match": {
+        "exact": [
+          "sonar-pro"
+        ]
+      },
+      "id": "sonar-pro",
+      "spec": {
+        "contextWindow": 200000,
+        "maxOutputTokens": 8192,
+        "capabilities": [
+          "chat",
+          "web_search",
+          "reasoning",
+          "citation"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "description": "Perplexity 旗舰搜索增强模型，内置实时联网搜索和引用，200K 上下文"
+      }
+    },
+    {
+      "displayName": "Sonar Reasoning Pro",
+      "family": "sonar-reasoning",
+      "match": {
+        "exact": [
+          "sonar-reasoning-pro"
+        ],
+        "patterns": [
+          "sonar-reasoning*"
+        ]
+      },
+      "id": "sonar-reasoning-pro",
+      "spec": {
+        "contextWindow": 128000,
+        "maxOutputTokens": 8192,
+        "capabilities": [
+          "chat",
+          "web_search",
+          "reasoning",
+          "deep_thinking"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": null,
+        "supportsReasoning": true,
+        "description": "Perplexity 深度推理模型，内置联网搜索和深度思考"
+      }
+    },
+    {
+      "displayName": "Sonar",
+      "family": "sonar",
+      "match": {
+        "exact": [
+          "sonar"
+        ],
+        "patterns": [
+          "sonar-*"
+        ]
+      },
+      "id": "sonar",
+      "spec": {
+        "contextWindow": 128000,
+        "maxOutputTokens": 4096,
+        "capabilities": [
+          "chat",
+          "web_search",
+          "citation",
+          "fast"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "description": "Perplexity 轻量搜索模型，低成本联网搜索"
+      }
+    }
+  ]
+}

compute/v2/compatibility/model-specs/stability.json

@@ -0,0 +1,50 @@
+{
+  "description": "Stability AI 图像生成系列模型规格。参数来源：config-center compute/providers/stability.json。无上下文窗口（图像生成模型）。",
+  "specs": [
+    {
+      "displayName": "Stable Audio 3.0",
+      "family": "stable-audio-3.0",
+      "match": {
+        "exact": [
+          "stable-audio-3.0"
+        ]
+      },
+      "id": "stable-audio-3.0",
+      "spec": {
+        "capabilities": [
+          "music_generation",
+          "audio_generation"
+        ],
+        "serviceType": [
+          "music_gen"
+        ],
+        "releasedAt": "2026-05-20",
+        "description": "Stable Audio 3.0，通过原生 /v2beta/audio/stable-audio API 提供，不使用 Stability 图像协议"
+      }
+    },
+    {
+      "displayName": "Stable Diffusion 3.5 Large",
+      "family": "stable-diffusion-3",
+      "match": {
+        "exact": [
+          "stable-diffusion-3.5-large"
+        ],
+        "patterns": [
+          "stable-diffusion-3.5*"
+        ]
+      },
+      "id": "stable-diffusion-3.5-large",
+      "spec": {
+        "capabilities": [
+          "image_generation",
+          "high_quality",
+          "controllable"
+        ],
+        "serviceType": [
+          "image_gen"
+        ],
+        "description": "Stability AI 旗舰文生图模型，支持多种分辨率和风格"
+      }
+    }
+  ]
+}

compute/v2/compatibility/model-specs/stealth.json

@@ -0,0 +1,46 @@
+{
+  "description": "Stealth 系列模型规格。参数来源：OpenRouter Models API。这些是厂商匿名发布的临时别名，揭晓后用 spec.extra.modelOrigin 标注真身，条目本身保留——云端算力可能仍以旧别名下发该模型。",
+  "specs": [
+    {
+      "displayName": "Ox Alpha (Historical)",
+      "family": "ox-alpha",
+      "match": {
+        "exact": [
+          "ox-alpha",
+          "stealth/ox-alpha"
+        ]
+      },
+      "id": "ox-alpha",
+      "spec": {
+        "contextWindow": 1048576,
+        "maxOutputTokens": 131072,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "deep_thinking",
+          "code",
+          "tool_use",
+          "agent",
+          "long_context",
+          "vision",
+          "image_understanding",
+          "video_understanding"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning",
+          "vision"
+        ],
+        "defaultTemperature": 1,
+        "defaultTopP": 0.95,
+        "supportsReasoning": true,
+        "description": "Ox Alpha 历史匿名模型兼容记录；当前官方目录未列出。旧记录曾关联 GLM-5.3-Flash，但本轮未取得官方身份揭示证据，不下发未证实的原厂身份映射；历史参数不代表当前可用性",
+        "extra": {
+          "thinkingDefault": true,
+          "thinkingOnly": true
+        },
+        "releasedAt": "2026-08-21"
+      }
+    }
+  ]
+}

compute/v2/compatibility/model-specs/tencent.json

@@ -0,0 +1,265 @@
+{
+  "description": "腾讯混元系列模型规格。参数来源：config-center compute/providers/tencent.json。hunyuan-2.0-thinking 和 hunyuan-t1 为推理模型。",
+  "specs": [
+    {
+      "displayName": "Tencent Hy4 Preview",
+      "family": "hy4-preview",
+      "match": {
+        "exact": [
+          "hy4-preview",
+          "tencent/hy4-preview"
+        ]
+      },
+      "id": "hy4-preview",
+      "spec": {
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "tool_use",
+          "agent",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "description": "Tencent Hy4 Preview，通过 TokenHub 提供；既有 Hunyuan Provider 端点可用性尚未确认",
+        "contextWindow": 1048576,
+        "releasedAt": "2026-08-28"
+      }
+    },
+    {
+      "displayName": "混元 Hy3",
+      "family": "hy3",
+      "match": {
+        "exact": [
+          "hy3"
+        ]
+      },
+      "id": "hy3",
+      "spec": {
+        "contextWindow": 262144,
+        "maxOutputTokens": 128000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "tool_use",
+          "agent"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning"
+        ],
+        "defaultTemperature": 0.9,
+        "supportsReasoning": true,
+        "description": "腾讯混元 Hy3 正式版，295B MoE（21B active），支持可配置推理强度",
+        "releasedAt": "2026-07-06"
+      }
+    },
+    {
+      "displayName": "混元 Hy3 Preview",
+      "family": "hy3",
+      "match": {
+        "exact": [
+          "hy3-preview"
+        ],
+        "patterns": [
+          "hy3-preview*"
+        ]
+      },
+      "id": "hy3-preview",
+      "spec": {
+        "contextWindow": 262144,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "multilingual",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "description": "腾讯混元 Hy3 预览版，高效 MoE 模型"
+      }
+    },
+    {
+      "displayName": "混元 2.0 Think",
+      "family": "hunyuan-2",
+      "match": {
+        "exact": [
+          "hunyuan-2.0-thinking-20251109"
+        ],
+        "patterns": [
+          "hunyuan-2.0-thinking*"
+        ]
+      },
+      "id": "hunyuan-2.0-thinking-20251109",
+      "spec": {
+        "contextWindow": 196608,
+        "maxOutputTokens": 65536,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "deep_thinking",
+          "long_context"
+        ],
+        "serviceType": [
+          "reasoning"
+        ],
+        "defaultTemperature": null,
+        "supportsReasoning": true,
+        "description": "腾讯混元最新推理模型，MoE架构406B总参数，深度思考"
+      }
+    },
+    {
+      "displayName": "混元 2.0 Instruct",
+      "family": "hunyuan-2",
+      "match": {
+        "exact": [
+          "hunyuan-2.0-instruct-20251111"
+        ],
+        "patterns": [
+          "hunyuan-2.0-instruct*"
+        ]
+      },
+      "id": "hunyuan-2.0-instruct-20251111",
+      "spec": {
+        "contextWindow": 147456,
+        "maxOutputTokens": 16384,
+        "capabilities": [
+          "chat",
+          "code",
+          "long_context",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "description": "腾讯混元 2.0 指令模型，支持联网搜索和 Function Calling"
+      }
+    },
+    {
+      "displayName": "混元 Turbo S",
+      "family": "hunyuan-turbo",
+      "match": {
+        "exact": [
+          "hunyuan-turbos-latest"
+        ],
+        "patterns": [
+          "hunyuan-turbos*"
+        ]
+      },
+      "id": "hunyuan-turbos-latest",
+      "spec": {
+        "contextWindow": 32768,
+        "maxOutputTokens": 16384,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "fast",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "description": "腾讯混元高速模型，支持推理和工具调用"
+      }
+    },
+    {
+      "displayName": "混元 T1",
+      "family": "hunyuan-t1",
+      "match": {
+        "exact": [
+          "hunyuan-t1-latest"
+        ],
+        "patterns": [
+          "hunyuan-t1*"
+        ]
+      },
+      "id": "hunyuan-t1-latest",
+      "spec": {
+        "contextWindow": 262144,
+        "maxOutputTokens": 32768,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "deep_thinking",
+          "code",
+          "math",
+          "long_context"
+        ],
+        "serviceType": [
+          "reasoning"
+        ],
+        "defaultTemperature": null,
+        "supportsReasoning": true,
+        "description": "腾讯混元 T1 深度推理模型，256K 上下文，专攻数学/代码/复杂推理"
+      }
+    },
+    {
+      "displayName": "混元 T1 Vision",
+      "family": "hunyuan-t1",
+      "match": {
+        "exact": [
+          "hunyuan-t1-vision"
+        ]
+      },
+      "id": "hunyuan-t1-vision",
+      "spec": {
+        "contextWindow": 131072,
+        "maxOutputTokens": 32768,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "deep_thinking",
+          "code",
+          "math",
+          "vision",
+          "image_understanding"
+        ],
+        "serviceType": [
+          "reasoning"
+        ],
+        "defaultTemperature": null,
+        "supportsReasoning": true,
+        "description": "腾讯混元 T1 视觉推理模型，支持图片理解 + 深度推理"
+      }
+    },
+    {
+      "displayName": "混元 Turbo S Vision",
+      "family": "hunyuan-turbo",
+      "match": {
+        "exact": [
+          "hunyuan-turbos-vision"
+        ]
+      },
+      "id": "hunyuan-turbos-vision",
+      "spec": {
+        "contextWindow": 32768,
+        "maxOutputTokens": 16384,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "image_understanding",
+          "fast",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "description": "腾讯混元 Turbo S 视觉模型，支持图片理解，高速推理"
+      }
+    }
+  ]
+}

compute/v2/compatibility/model-specs/volcengine.json

@@ -0,0 +1,808 @@
+{
+  "description": "火山引擎豆包系列模型规格。参数来源：config-center compute/providers/volcengine.json。包含豆包 Seed 系列对话、DeepSeek 托管版、Embedding、语音克隆、实时语音、同传和翻译服务。doubao-seed-1.6-thinking 为推理模型。",
+  "specs": [
+    {
+      "displayName": "Doubao Seed Evolving",
+      "family": "doubao-seed-evolving",
+      "match": {
+        "exact": [
+          "doubao-seed-evolving"
+        ]
+      },
+      "id": "doubao-seed-evolving",
+      "spec": {
+        "contextWindow": 1000000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "tool_use",
+          "agent",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning"
+        ],
+        "supportsReasoning": true,
+        "description": "Seed 系列持续迭代模型，通过固定 ID 获取最新版本，2026-09-09 更新"
+      }
+    },
+    {
+      "displayName": "豆包 Seed 2.1 Pro",
+      "family": "doubao-seed-2.1-pro",
+      "match": {
+        "exact": [
+          "doubao-seed-2.1-pro"
+        ]
+      },
+      "id": "doubao-seed-2.1-pro",
+      "spec": {
+        "serviceType": [
+          "chat",
+          "reasoning",
+          "vision"
+        ],
+        "description": "火山方舟豆包 Seed 2.1 Pro，生产级 Coding、Agent 与多模态旗舰模型",
+        "contextWindow": 256000,
+        "maxOutputTokens": 262144,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "multilingual",
+          "long_context",
+          "tool_use",
+          "agent",
+          "deep_thinking"
+        ],
+        "supportsReasoning": true
+      }
+    },
+    {
+      "displayName": "豆包 2.0 Pro",
+      "family": "doubao-seed-2",
+      "match": {
+        "exact": [
+          "doubao-seed-2.0-pro"
+        ],
+        "patterns": [
+          "doubao-seed-2.0-pro*"
+        ]
+      },
+      "id": "doubao-seed-2.0-pro",
+      "spec": {
+        "contextWindow": 256000,
+        "maxOutputTokens": 128000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "multilingual",
+          "long_context",
+          "tool_use",
+          "vision"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "description": "火山引擎豆包 Seed 2.0 旗舰模型，256K 上下文，支持复杂推理、Agent 和多模态"
+      }
+    },
+    {
+      "displayName": "豆包 2.0 Lite",
+      "family": "doubao-seed-2",
+      "match": {
+        "exact": [
+          "doubao-seed-2.0-lite"
+        ],
+        "patterns": [
+          "doubao-seed-2.0-lite*"
+        ]
+      },
+      "id": "doubao-seed-2.0-lite",
+      "spec": {
+        "contextWindow": 256000,
+        "maxOutputTokens": 32000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "multilingual",
+          "long_context",
+          "tool_use",
+          "vision",
+          "audio_understanding",
+          "video_understanding"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "description": "豆包 Seed 2.0 全模态均衡模型，支持视频/图像/音频/文本统一理解"
+      }
+    },
+    {
+      "displayName": "豆包 2.0 Mini",
+      "family": "doubao-seed-2",
+      "match": {
+        "exact": [
+          "doubao-seed-2.0-mini"
+        ],
+        "patterns": [
+          "doubao-seed-2.0-mini*"
+        ]
+      },
+      "id": "doubao-seed-2.0-mini",
+      "spec": {
+        "contextWindow": 256000,
+        "maxOutputTokens": 32000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "multilingual",
+          "long_context",
+          "tool_use",
+          "vision",
+          "audio_understanding",
+          "video_understanding",
+          "fast"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "description": "豆包 Seed 2.0 全模态轻量模型，低延迟高并发，支持全模态理解"
+      }
+    },
+    {
+      "displayName": "豆包 2.0 Code",
+      "family": "doubao-seed-2",
+      "match": {
+        "exact": [
+          "doubao-seed-2.0-code"
+        ]
+      },
+      "id": "doubao-seed-2.0-code",
+      "spec": {
+        "contextWindow": 256000,
+        "maxOutputTokens": 128000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "long_context",
+          "tool_use",
+          "vision",
+          "deep_thinking"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "description": "豆包 Seed 2.0 编程模型，面向 Agentic Coding 场景优化"
+      }
+    },
+    {
+      "displayName": "豆包 Seed-1.8",
+      "family": "doubao-seed-1",
+      "match": {
+        "exact": [
+          "doubao-seed-1.8"
+        ],
+        "patterns": [
+          "doubao-seed-1.8*"
+        ]
+      },
+      "id": "doubao-seed-1.8",
+      "spec": {
+        "contextWindow": 256000,
+        "maxOutputTokens": 16000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "multilingual",
+          "long_context",
+          "tool_use",
+          "vision"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "description": "火山引擎豆包 1.8 旗舰模型，256K 上下文，支持深度思考和多模态"
+      }
+    },
+    {
+      "displayName": "豆包 Seed-1.6",
+      "family": "doubao-seed-1",
+      "match": {
+        "exact": [
+          "doubao-seed-1.6"
+        ],
+        "patterns": [
+          "doubao-seed-1.6"
+        ]
+      },
+      "id": "doubao-seed-1.6",
+      "spec": {
+        "contextWindow": 256000,
+        "maxOutputTokens": 32000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "multilingual",
+          "long_context",
+          "tool_use",
+          "vision"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "description": "火山引擎豆包 1.6 旗舰模型，256K 上下文"
+      }
+    },
+    {
+      "displayName": "豆包 Seed-1.6 思考版",
+      "family": "doubao-seed-1",
+      "match": {
+        "exact": [
+          "doubao-seed-1.6-thinking"
+        ]
+      },
+      "id": "doubao-seed-1.6-thinking",
+      "spec": {
+        "contextWindow": 256000,
+        "maxOutputTokens": 16000,
+        "capabilities": [
+          "reasoning",
+          "math",
+          "code",
+          "deep_thinking",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": null,
+        "supportsReasoning": true,
+        "description": "火山引擎豆包 1.6 深度思考模型，256K 上下文"
+      }
+    },
+    {
+      "displayName": "豆包 Seed-1.6 Flash",
+      "family": "doubao-seed-1",
+      "match": {
+        "exact": [
+          "doubao-seed-1.6-flash"
+        ]
+      },
+      "id": "doubao-seed-1.6-flash",
+      "spec": {
+        "contextWindow": 256000,
+        "maxOutputTokens": 16000,
+        "capabilities": [
+          "chat",
+          "code",
+          "fast",
+          "long_context",
+          "vision"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "description": "火山引擎豆包 1.6 快速版，成本较前代下降 63%"
+      }
+    },
+    {
+      "displayName": "豆包 Seed-1.6 Lite",
+      "family": "doubao-seed-1",
+      "match": {
+        "exact": [
+          "doubao-seed-1.6-lite"
+        ]
+      },
+      "id": "doubao-seed-1.6-lite",
+      "spec": {
+        "contextWindow": 256000,
+        "maxOutputTokens": 32000,
+        "capabilities": [
+          "chat",
+          "code",
+          "fast"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "description": "火山引擎豆包 1.6 轻量版，极致性价比"
+      }
+    },
+    {
+      "displayName": "豆包 Seed-1.6 视觉版",
+      "family": "doubao-seed-1",
+      "match": {
+        "exact": [
+          "doubao-seed-1.6-vision"
+        ]
+      },
+      "id": "doubao-seed-1.6-vision",
+      "spec": {
+        "contextWindow": 256000,
+        "maxOutputTokens": 32000,
+        "capabilities": [
+          "chat",
+          "vision",
+          "video_understanding",
+          "gui_agent",
+          "long_context"
+        ],
+        "serviceType": [
+          "vision"
+        ],
+        "defaultTemperature": 1,
+        "description": "火山引擎豆包 1.6 视觉模型，256K 上下文，支持视频理解"
+      }
+    },
+    {
+      "displayName": "豆包 Seed-Code",
+      "family": "doubao-seed-code",
+      "match": {
+        "exact": [
+          "doubao-seed-code"
+        ]
+      },
+      "id": "doubao-seed-code",
+      "spec": {
+        "contextWindow": 256000,
+        "maxOutputTokens": 32768,
+        "capabilities": [
+          "code",
+          "reasoning",
+          "long_context",
+          "deep_thinking"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "description": "火山引擎豆包编程专用模型，256K 上下文"
+      }
+    },
+    {
+      "displayName": "DeepSeek V3.2 (火山引擎)",
+      "family": "deepseek-v3",
+      "match": {
+        "exact": [
+          "deepseek-v3.2"
+        ]
+      },
+      "id": "deepseek-v3.2",
+      "spec": {
+        "contextWindow": 128000,
+        "maxOutputTokens": 32000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "multilingual"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "description": "火山引擎托管的 DeepSeek V3.2，128K 上下文"
+      }
+    },
+    {
+      "displayName": "DeepSeek R1 (火山引擎)",
+      "family": "deepseek-r1",
+      "match": {
+        "exact": [
+          "deepseek-r1"
+        ]
+      },
+      "id": "deepseek-r1",
+      "spec": {
+        "contextWindow": 128000,
+        "maxOutputTokens": 65536,
+        "capabilities": [
+          "reasoning",
+          "math",
+          "code",
+          "science",
+          "deep_thinking"
+        ],
+        "serviceType": [
+          "reasoning"
+        ],
+        "defaultTemperature": null,
+        "supportsReasoning": true,
+        "description": "火山引擎托管的 DeepSeek R1 推理模型，128K 上下文"
+      }
+    },
+    {
+      "displayName": "豆包 Embedding",
+      "family": "doubao-embedding",
+      "match": {
+        "exact": [
+          "doubao-embedding"
+        ]
+      },
+      "id": "doubao-embedding",
+      "spec": {
+        "contextWindow": 4096,
+        "capabilities": [
+          "text_embedding",
+          "semantic_search",
+          "rag",
+          "chinese_optimized"
+        ],
+        "serviceType": [
+          "embedding"
+        ],
+        "description": "火山引擎豆包标准向量模型，默认 2048 维度，可选 512/1024/2048"
+      }
+    },
+    {
+      "displayName": "豆包 Embedding Large",
+      "family": "doubao-embedding",
+      "match": {
+        "exact": [
+          "doubao-embedding-large"
+        ]
+      },
+      "id": "doubao-embedding-large",
+      "spec": {
+        "contextWindow": 4096,
+        "capabilities": [
+          "text_embedding",
+          "semantic_search",
+          "rag",
+          "chinese_optimized"
+        ],
+        "serviceType": [
+          "embedding"
+        ],
+        "description": "火山引擎豆包大号向量模型，默认 2048 维度，可选 512/1024/2048/4096"
+      }
+    },
+    {
+      "displayName": "火山声音复刻",
+      "family": "volc-tts",
+      "match": {
+        "exact": [
+          "volc-mega-tts-clone"
+        ]
+      },
+      "id": "volc-mega-tts-clone",
+      "spec": {
+        "capabilities": [
+          "voice_clone",
+          "chinese_optimized",
+          "few_shot"
+        ],
+        "serviceType": [
+          "voice_clone"
+        ],
+        "description": "火山引擎声音复刻，支持极少样本克隆"
+      }
+    },
+    {
+      "displayName": "火山实时语音",
+      "family": "volc-voice",
+      "match": {
+        "exact": [
+          "volc-realtime-voice"
+        ]
+      },
+      "id": "volc-realtime-voice",
+      "spec": {
+        "capabilities": [
+          "realtime",
+          "speech_to_speech",
+          "low_latency",
+          "chinese_optimized"
+        ],
+        "serviceType": [
+          "realtime_voice"
+        ],
+        "description": "火山引擎实时语音交互"
+      }
+    },
+    {
+      "displayName": "火山同声传译",
+      "family": "volc-interpret",
+      "match": {
+        "exact": [
+          "volc-simultaneous"
+        ]
+      },
+      "id": "volc-simultaneous",
+      "spec": {
+        "capabilities": [
+          "simultaneous_interpretation",
+          "chinese_optimized",
+          "low_latency"
+        ],
+        "serviceType": [
+          "simultaneous_interpret"
+        ],
+        "description": "火山引擎同声传译，支持中英双向"
+      }
+    },
+    {
+      "displayName": "火山翻译",
+      "family": "volc-translation",
+      "match": {
+        "exact": [
+          "volc-translation"
+        ]
+      },
+      "id": "volc-translation",
+      "spec": {
+        "capabilities": [
+          "translation",
+          "multilingual",
+          "chinese_optimized"
+        ],
+        "serviceType": [
+          "translation"
+        ],
+        "description": "火山引擎机器翻译，支持多语种互译"
+      }
+    },
+    {
+      "displayName": "Seedance 2.0",
+      "family": "doubao-seedance-2",
+      "match": {
+        "exact": [
+          "doubao-seedance-2-0-260128",
+          "doubao-seedance-2.0"
+        ],
+        "patterns": [
+          "doubao-seedance-2-0*"
+        ]
+      },
+      "id": "doubao-seedance-2.0",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "image_to_video",
+          "text_to_video",
+          "video_editing",
+          "video_extension",
+          "audio_generation",
+          "first_last_frame"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "Seedance 2.0 豆包最强视频生成模型，4-15秒 2K 视频，原生同步音频，支持最多9张图片+3个视频+3个音频参考输入，~$0.14/秒"
+      },
+      "routing": {
+        "tier": "flagship",
+        "routingPriority": 120,
+        "eligibleForAgent": false,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto"
+          ],
+          "defaultMode": "auto"
+        }
+      }
+    },
+    {
+      "displayName": "Seedance 2.0 Fast",
+      "family": "doubao-seedance-2",
+      "match": {
+        "exact": [
+          "doubao-seedance-2-0-fast-260128",
+          "doubao-seedance-2.0-fast"
+        ],
+        "patterns": [
+          "doubao-seedance-2-0-fast*"
+        ]
+      },
+      "id": "doubao-seedance-2.0-fast",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "image_to_video",
+          "text_to_video",
+          "video_editing",
+          "video_extension",
+          "audio_generation",
+          "first_last_frame",
+          "fast"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "Seedance 2.0 快速版，音画同生，多模态生视频+编辑+延长"
+      },
+      "routing": {
+        "tier": "balanced",
+        "routingPriority": 119,
+        "eligibleForAgent": false,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto"
+          ],
+          "defaultMode": "auto"
+        }
+      }
+    },
+    {
+      "displayName": "Seedance 2.0 Mini",
+      "family": "doubao-seedance-2",
+      "match": {
+        "exact": [
+          "doubao-seedance-2-0-mini-260615",
+          "doubao-seedance-2.0-mini"
+        ],
+        "patterns": [
+          "doubao-seedance-2-0-mini*"
+        ]
+      },
+      "id": "doubao-seedance-2.0-mini",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "image_to_video",
+          "text_to_video",
+          "video_editing",
+          "video_extension",
+          "audio_generation",
+          "first_last_frame",
+          "fast",
+          "cost_effective"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "Seedance 2.0 轻量版，音画同生，高性价比"
+      },
+      "routing": {
+        "tier": "lightweight",
+        "routingPriority": 118,
+        "eligibleForAgent": false,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto"
+          ],
+          "defaultMode": "auto"
+        }
+      }
+    },
+    {
+      "displayName": "Seedance 1.5 Pro",
+      "family": "doubao-seedance-1",
+      "match": {
+        "exact": [
+          "doubao-seedance-1-5-pro-251215",
+          "doubao-seedance-1.5-pro"
+        ],
+        "patterns": [
+          "doubao-seedance-1-5*"
+        ]
+      },
+      "id": "doubao-seedance-1.5-pro",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "image_to_video",
+          "text_to_video",
+          "first_last_frame",
+          "audio_generation"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "Seedance 1.5 Pro 音画同生，12秒视频，多参考图输入（即将下线）"
+      }
+    },
+    {
+      "displayName": "Seedance 1.0 Pro",
+      "family": "doubao-seedance-1",
+      "match": {
+        "exact": [
+          "doubao-seedance-1-0-pro-250528",
+          "doubao-seedance-1.0-pro"
+        ],
+        "patterns": [
+          "doubao-seedance-1-0-pro*"
+        ]
+      },
+      "id": "doubao-seedance-1.0-pro",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "image_to_video",
+          "text_to_video",
+          "first_last_frame"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "Seedance 1.0 Pro 文生视频/图生视频，支持首尾帧控制"
+      }
+    },
+    {
+      "displayName": "Seedance 1.0 Pro Fast",
+      "family": "doubao-seedance-1",
+      "match": {
+        "exact": [
+          "doubao-seedance-1-0-pro-fast-251015",
+          "doubao-seedance-1.0-pro-fast"
+        ],
+        "patterns": [
+          "doubao-seedance-1-0-pro-fast*"
+        ]
+      },
+      "id": "doubao-seedance-1.0-pro-fast",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "image_to_video",
+          "text_to_video",
+          "fast"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "Seedance 1.0 Pro 快速版，图生视频-首帧，5秒 720P 仅需 10 秒"
+      }
+    },
+    {
+      "displayName": "Doubao Seed 2.1 Turbo",
+      "id": "doubao-seed-2.1-turbo",
+      "spec": {
+        "contextWindow": 256000,
+        "maxOutputTokens": 128000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "multilingual",
+          "long_context",
+          "tool_use",
+          "vision"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "supportsReasoning": true
+      },
+      "routing": {
+        "tier": "balanced",
+        "routingPriority": 40,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "off",
+            "minimal",
+            "low",
+            "medium",
+            "high",
+            "xhigh"
+          ],
+          "defaultMode": "medium"
+        }
+      }
+    }
+  ]
+}

compute/v2/compatibility/model-specs/wan.json

@@ -0,0 +1,106 @@
+{
+  "description": "通义万相（Wan）多模态生成模型规格。Wan 3.0 视频模型通过阿里云百炼提供，统一支持文生视频、首帧/首尾帧图生视频和多类型参考生视频。",
+  "specs": [
+    {
+      "displayName": "通义万相 3.0 视频生成高速版",
+      "family": "wan3-video",
+      "match": {
+        "exact": [
+          "wan3.0-video-prime"
+        ]
+      },
+      "id": "wan3.0-video-prime",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "text_to_video",
+          "image_to_video",
+          "reference_to_video",
+          "multi_image",
+          "reference_video",
+          "reference_audio",
+          "first_last_frame",
+          "audio_generation",
+          "high_quality",
+          "fast"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "万相 3.0 全能参考视频生成高速版，能力对齐标准版并显著提升端到端生成速度",
+        "releasedAt": "2026-08-06",
+        "extra": {
+          "maxVideoDuration": 30,
+          "fps": 30,
+          "supportedResolutions": [
+            "480P",
+            "720P",
+            "1080P"
+          ]
+        }
+      },
+      "routing": {
+        "tier": "flagship",
+        "routingPriority": 109,
+        "eligibleForAgent": false,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto"
+          ],
+          "defaultMode": "auto"
+        }
+      }
+    },
+    {
+      "displayName": "通义万相 3.0 视频生成",
+      "family": "wan3-video",
+      "match": {
+        "exact": [
+          "wan3.0-video"
+        ]
+      },
+      "id": "wan3.0-video",
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "text_to_video",
+          "image_to_video",
+          "reference_to_video",
+          "multi_image",
+          "reference_video",
+          "reference_audio",
+          "first_last_frame",
+          "audio_generation",
+          "high_quality"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "万相 3.0 全能参考视频生成标准版，统一支持文生、首帧/首尾帧图生及多类型参考生视频",
+        "releasedAt": "2026-08-06",
+        "extra": {
+          "maxVideoDuration": 30,
+          "fps": 30,
+          "supportedResolutions": [
+            "480P",
+            "720P",
+            "1080P"
+          ]
+        }
+      },
+      "routing": {
+        "tier": "flagship",
+        "routingPriority": 110,
+        "eligibleForAgent": false,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto"
+          ],
+          "defaultMode": "auto"
+        }
+      }
+    }
+  ]
+}

compute/v2/compatibility/model-specs/xai.json

@@ -0,0 +1,155 @@
+{
+  "description": "xAI Grok 系列模型规格。参数来源：config-center compute/providers/xai.json。Grok 网关命名变体较多（含日期戳/-fast/-reasoning 后缀），主要依赖 family=grok-4 兜底匹配。",
+  "specs": [
+    {
+      "displayName": "Grok 4.7",
+      "family": "grok-4.7",
+      "match": {
+        "exact": [
+          "grok-4.7"
+        ]
+      },
+      "id": "grok-4.7",
+      "spec": {
+        "contextWindow": 500000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "tool_use",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning"
+        ],
+        "supportsReasoning": true,
+        "description": "xAI Grok 4.7，500K 上下文，文本与图像输入，推理努力等级 low 到 xhigh"
+      },
+      "routing": {
+        "tier": "flagship",
+        "routingPriority": 30,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "low",
+            "medium",
+            "high",
+            "xhigh"
+          ],
+          "defaultMode": "high"
+        }
+      }
+    },
+    {
+      "displayName": "Grok 4.5",
+      "family": "grok-4.5",
+      "match": {
+        "exact": [
+          "grok-4.5"
+        ]
+      },
+      "id": "grok-4.5",
+      "spec": {
+        "contextWindow": 500000,
+        "maxOutputTokens": 450000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "long_context",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning"
+        ],
+        "supportsReasoning": true,
+        "description": "xAI Grok 4.5，面向编程、知识工作和 STEM 的旗舰推理模型",
+        "extra": {
+          "thinkingOnly": true,
+          "thinkingDefault": true
+        },
+        "releasedAt": "2026-07-08"
+      }
+    },
+    {
+      "displayName": "Grok 4.3",
+      "family": "grok-4",
+      "match": {
+        "exact": [
+          "grok-4.3"
+        ],
+        "patterns": [
+          "grok-4.3*"
+        ]
+      },
+      "id": "grok-4-3",
+      "spec": {
+        "contextWindow": 1000000,
+        "capabilities": [
+          "reasoning",
+          "code",
+          "vision",
+          "tool_use",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "supportsReasoning": true,
+        "description": "xAI Grok 4.3 推理模型，100 万上下文"
+      }
+    },
+    {
+      "displayName": "Grok 4.1 Fast",
+      "family": "grok-4",
+      "id": "grok-4-1-fast-reasoning",
+      "spec": {
+        "contextWindow": 2000000,
+        "maxOutputTokens": 16384,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "tool_use",
+          "fast",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true
+      }
+    },
+    {
+      "displayName": "Grok 4.20",
+      "family": "grok-4",
+      "id": "grok-4.20-0309-reasoning",
+      "spec": {
+        "contextWindow": 2000000,
+        "maxOutputTokens": 16384,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true
+      }
+    }
+  ]
+}

compute/v2/compatibility/model-specs/xiaomi.json

@@ -0,0 +1,583 @@
+{
+  "description": "小米 MiMo 系列模型规格。参数来源：https://mimo.mi.com/docs/zh-CN/quick-start/summary/model。注意 v2-pro/v2-omni 自 2026-06-01 起转发到 v2.5 系列，2026-06-30 正式停用，规格库仍保留供存量 NewAPI 渠道正确展示参数。",
+  "specs": [
+    {
+      "displayName": "MiMo-V2.6-Pro",
+      "family": "mimo-v2.6-pro",
+      "match": {
+        "exact": [
+          "mimo-v2.6-pro"
+        ]
+      },
+      "id": "mimo-v2.6-pro",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 131072,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "audio_understanding",
+          "video_understanding",
+          "tool_use",
+          "agent",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning",
+          "vision"
+        ],
+        "supportsReasoning": true,
+        "description": "小米全模态旗舰推理模型，适合复杂项目与长程智能体任务",
+        "extra": {
+          "thinkingDefault": true
+        },
+        "releasedAt": "2026-09-22"
+      },
+      "routing": {
+        "tier": "flagship",
+        "routingPriority": 28,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "off"
+          ],
+          "defaultMode": "auto"
+        }
+      }
+    },
+    {
+      "displayName": "MiMo-V2.6-Flash",
+      "family": "mimo-v2.6-flash",
+      "match": {
+        "exact": [
+          "mimo-v2.6-flash"
+        ]
+      },
+      "id": "mimo-v2.6-flash",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 131072,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "audio_understanding",
+          "video_understanding",
+          "tool_use",
+          "agent",
+          "long_context",
+          "fast"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning",
+          "vision"
+        ],
+        "supportsReasoning": true,
+        "description": "小米全模态高效推理模型，适合高频调用与规模化任务",
+        "extra": {
+          "thinkingDefault": true
+        },
+        "releasedAt": "2026-09-22"
+      },
+      "routing": {
+        "tier": "balanced",
+        "routingPriority": 28,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "off"
+          ],
+          "defaultMode": "auto"
+        }
+      }
+    },
+    {
+      "displayName": "MiMo-V2.6-Pro-UltraSpeed",
+      "family": "mimo-v2.6-pro-ultraspeed",
+      "match": {
+        "exact": [
+          "mimo-v2.6-pro-ultraspeed"
+        ]
+      },
+      "id": "mimo-v2.6-pro-ultraspeed",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 131072,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "audio_understanding",
+          "video_understanding",
+          "tool_use",
+          "agent",
+          "long_context",
+          "fast"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning",
+          "vision"
+        ],
+        "supportsReasoning": true,
+        "description": "MiMo-V2.6-Pro 的超高速服务，需向小米申请定制配额",
+        "extra": {
+          "thinkingDefault": true
+        },
+        "releasedAt": "2026-09-22"
+      }
+    },
+    {
+      "displayName": "MiMo-X-Flash Preview",
+      "family": "mimo-x-flash-preview",
+      "match": {
+        "exact": [
+          "mimo-x-flash-preview"
+        ]
+      },
+      "id": "mimo-x-flash-preview",
+      "spec": {
+        "contextWindow": 1000000,
+        "capabilities": [
+          "chat",
+          "vision"
+        ],
+        "serviceType": [
+          "chat",
+          "vision"
+        ],
+        "description": "小米 MiMo-X Flash 内测模型，1M 上下文，支持图片识别"
+      }
+    },
+    {
+      "displayName": "MiMo-X-Pro Preview",
+      "family": "mimo-x-pro-preview",
+      "match": {
+        "exact": [
+          "mimo-x-pro-preview"
+        ]
+      },
+      "id": "mimo-x-pro-preview",
+      "spec": {
+        "contextWindow": 1000000,
+        "capabilities": [
+          "chat",
+          "vision"
+        ],
+        "serviceType": [
+          "chat",
+          "vision"
+        ],
+        "description": "小米 MiMo-X Pro 内测模型，1M 上下文，支持图片识别"
+      }
+    },
+    {
+      "displayName": "MiMo V2.5 Pro",
+      "family": "mimo-v2.5-pro",
+      "match": {
+        "patterns": [
+          "mimo-v2.5-pro*",
+          "mimo-v2-5-pro*"
+        ]
+      },
+      "id": "mimo-v2.5-pro",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 131072,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "tool_use",
+          "code"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "description": "小米 MiMo 旗舰推理模型，支持文本/工具调用"
+      },
+      "routing": {
+        "tier": "flagship",
+        "routingPriority": 50,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "off",
+            "minimal",
+            "low",
+            "medium",
+            "high",
+            "xhigh"
+          ],
+          "defaultMode": "high"
+        }
+      }
+    },
+    {
+      "displayName": "MiMo V2.5",
+      "family": "mimo-v2.5",
+      "match": {
+        "exact": [
+          "mimo-v2.5"
+        ],
+        "patterns": [
+          "mimo-v2.5-chat*",
+          "mimo-v2-5*"
+        ]
+      },
+      "id": "mimo-v2.5",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 131072,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "vision",
+          "tool_use",
+          "code"
+        ],
+        "serviceType": [
+          "chat",
+          "vision"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "description": "小米 MiMo 全模态模型，支持图像/音频/视频理解"
+      },
+      "routing": {
+        "tier": "lightweight",
+        "routingPriority": 40,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "off",
+            "minimal",
+            "low",
+            "medium",
+            "high",
+            "xhigh"
+          ],
+          "defaultMode": "low"
+        }
+      }
+    },
+    {
+      "displayName": "MiMo V2 Pro（已停用→转发至 V2.5 Pro）",
+      "family": "mimo-v2-pro",
+      "match": {
+        "patterns": [
+          "mimo-v2-pro*"
+        ]
+      },
+      "id": "mimo-v2-pro",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 131072,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "tool_use",
+          "code"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "description": "小米 MiMo V2 Pro（自 2026-06-01 转发至 V2.5 Pro）",
+        "retiredAt": "2026-06-30"
+      }
+    },
+    {
+      "displayName": "MiMo V2 Omni（已停用→转发至 V2.5）",
+      "family": "mimo-v2-omni",
+      "match": {
+        "patterns": [
+          "mimo-v2-omni*"
+        ]
+      },
+      "id": "mimo-v2-omni",
+      "spec": {
+        "contextWindow": 256000,
+        "maxOutputTokens": 131072,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "vision",
+          "tool_use",
+          "code"
+        ],
+        "serviceType": [
+          "chat",
+          "vision"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "description": "小米 MiMo V2 全模态（自 2026-06-01 转发至 V2.5）",
+        "retiredAt": "2026-06-30"
+      }
+    },
+    {
+      "displayName": "MiMo V2 Flash",
+      "family": "mimo-v2-flash",
+      "match": {
+        "patterns": [
+          "mimo-v2-flash*"
+        ]
+      },
+      "id": "mimo-v2-flash",
+      "spec": {
+        "contextWindow": 256000,
+        "maxOutputTokens": 65536,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "tool_use",
+          "code",
+          "fast"
+        ],
+        "serviceType": [
+          "fast"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "description": "小米 MiMo 快速版，适合高频低延迟任务"
+      }
+    },
+    {
+      "displayName": "MiMo V2.5 TTS",
+      "family": "mimo-v2.5-tts",
+      "match": {
+        "exact": [
+          "mimo-v2.5-tts"
+        ]
+      },
+      "id": "mimo-v2.5-tts",
+      "spec": {
+        "contextWindow": 8192,
+        "maxOutputTokens": 8192,
+        "capabilities": [
+          "tts",
+          "multilingual",
+          "style_control"
+        ],
+        "serviceType": [
+          "tts"
+        ],
+        "description": "小米 MiMo 语音合成，支持多语言与风格控制",
+        "extra": {
+          "modelOrigin": {
+            "vendor": "xiaomi",
+            "canonicalModelId": "mimo-v2.5-tts",
+            "family": "mimo-tts"
+          },
+          "speech": {
+            "protocol": "xiaomi-mimo-chat-tts",
+            "protocolVersion": "1",
+            "modes": [
+              "batch",
+              "streaming"
+            ],
+            "outputFormats": [
+              "wav",
+              "mp3",
+              "pcm16le"
+            ],
+            "sampleRates": [
+              24000
+            ],
+            "defaultVoice": "mimo_default",
+            "voices": [
+              {
+                "id": "mimo_default",
+                "label": "MiMo 默认"
+              },
+              {
+                "id": "冰糖",
+                "label": "冰糖"
+              },
+              {
+                "id": "茉莉",
+                "label": "茉莉"
+              },
+              {
+                "id": "苏打",
+                "label": "苏打"
+              },
+              {
+                "id": "白桦",
+                "label": "白桦"
+              },
+              {
+                "id": "Mia",
+                "label": "Mia"
+              },
+              {
+                "id": "Chloe",
+                "label": "Chloe"
+              },
+              {
+                "id": "Milo",
+                "label": "Milo"
+              },
+              {
+                "id": "Dean",
+                "label": "Dean"
+              }
+            ]
+          }
+        },
+        "releasedAt": "2026-04-23"
+      },
+      "routing": {
+        "tier": "lightweight",
+        "routingPriority": 114,
+        "eligibleForAgent": false,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto"
+          ],
+          "defaultMode": "auto"
+        }
+      }
+    },
+    {
+      "displayName": "MiMo V2.5 ASR",
+      "family": "mimo-v2.5-asr",
+      "match": {
+        "exact": [
+          "mimo-v2.5-asr"
+        ]
+      },
+      "id": "mimo-v2.5-asr",
+      "spec": {
+        "contextWindow": 8192,
+        "maxOutputTokens": 2048,
+        "capabilities": [
+          "asr",
+          "multilingual"
+        ],
+        "serviceType": [
+          "asr"
+        ],
+        "description": "小米 MiMo 语音识别模型，支持中英文语言提示与整段音频识别",
+        "extra": {
+          "modelOrigin": {
+            "vendor": "xiaomi",
+            "canonicalModelId": "mimo-v2.5-asr",
+            "family": "mimo-asr"
+          },
+          "speech": {
+            "protocol": "xiaomi-mimo-chat-asr",
+            "protocolVersion": "1",
+            "modes": [
+              "batch"
+            ],
+            "inputFormats": [
+              "wav",
+              "mp3"
+            ],
+            "languageHints": true
+          }
+        },
+        "releasedAt": "2026-06-02"
+      },
+      "routing": {
+        "tier": "lightweight",
+        "routingPriority": 123,
+        "eligibleForAgent": false,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto"
+          ],
+          "defaultMode": "auto"
+        }
+      }
+    },
+    {
+      "displayName": "MiMo V2.5 TTS Voice Clone",
+      "family": "mimo-v2.5-tts",
+      "match": {
+        "exact": [
+          "mimo-v2.5-tts-voiceclone"
+        ]
+      },
+      "id": "mimo-v2.5-tts-voiceclone",
+      "spec": {
+        "contextWindow": 8192,
+        "capabilities": [
+          "tts",
+          "voice_clone",
+          "multilingual"
+        ],
+        "serviceType": [
+          "tts",
+          "voice_clone"
+        ],
+        "description": "小米 MiMo 声音复刻合成"
+      }
+    },
+    {
+      "displayName": "MiMo V2.5 TTS Voice Design",
+      "family": "mimo-v2.5-tts",
+      "match": {
+        "exact": [
+          "mimo-v2.5-tts-voicedesign"
+        ]
+      },
+      "id": "mimo-v2.5-tts-voicedesign",
+      "spec": {
+        "contextWindow": 8192,
+        "capabilities": [
+          "tts",
+          "voice_design",
+          "multilingual"
+        ],
+        "serviceType": [
+          "tts"
+        ],
+        "description": "小米 MiMo 自定义音色设计合成"
+      }
+    },
+    {
+      "displayName": "MiMo V2 TTS",
+      "family": "mimo-v2-tts",
+      "match": {
+        "exact": [
+          "mimo-v2-tts"
+        ]
+      },
+      "id": "mimo-v2-tts",
+      "spec": {
+        "contextWindow": 8192,
+        "capabilities": [
+          "tts",
+          "multilingual"
+        ],
+        "serviceType": [
+          "tts"
+        ],
+        "description": "小米 MiMo V2 语音合成（旧版）"
+      }
+    }
+  ]
+}

compute/v2/compatibility/model-specs/xunfei.json

@@ -0,0 +1,61 @@
+{
+  "description": "讯飞星火系列模型规格。参数来源：config-center compute/providers/xunfei.json。spark-x 为推理模型（serviceType: reasoning）。",
+  "specs": [
+    {
+      "displayName": "讯飞星火 X2",
+      "family": "spark-x",
+      "match": {
+        "exact": [
+          "spark-x"
+        ]
+      },
+      "id": "spark-x",
+      "spec": {
+        "contextWindow": 65536,
+        "maxOutputTokens": 131072,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "deep_thinking",
+          "math",
+          "code",
+          "tool_use"
+        ],
+        "serviceType": [
+          "reasoning"
+        ],
+        "defaultTemperature": 1.2,
+        "supportsReasoning": true,
+        "description": "讯飞星火 X2，/x2 接入合同：输入 65536 / 输出 131072；spark-x 同名在 /v2 上指向 X1.5，不能复用本输出限制"
+      }
+    },
+    {
+      "displayName": "讯飞星火 4.0 Ultra",
+      "family": "spark-4",
+      "match": {
+        "exact": [
+          "4.0Ultra"
+        ],
+        "patterns": [
+          "4.0Ultra*"
+        ]
+      },
+      "id": "4.0Ultra",
+      "spec": {
+        "contextWindow": 32768,
+        "maxOutputTokens": 32768,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "description": "讯飞星火最强通用模型，支持推理、代码和工具调用"
+      }
+    }
+  ]
+}

compute/v2/compatibility/model-specs/zhipu.json

@@ -0,0 +1,458 @@
+{
+  "description": "智谱 GLM 系列模型规格。参数来源：config-center compute/providers/zhipu.json。glm-5 与 glm-5.1/glm-5-turbo/glm-5v-turbo 前缀相近，故各自仅用 exact 主键匹配，不用宽 pattern 以防误吞。",
+  "specs": [
+    {
+      "displayName": "GLM-5.3-FlashX",
+      "family": "glm-5.3-flashx",
+      "match": {
+        "exact": [
+          "glm-5.3-flashx",
+          "z-ai/glm-5.3-flashx"
+        ]
+      },
+      "id": "glm-5.3-flashx",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 128000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "multilingual",
+          "vision",
+          "video_understanding",
+          "tool_use",
+          "long_context",
+          "fast"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning",
+          "vision"
+        ],
+        "supportsReasoning": true,
+        "description": "Z.ai GLM-5.3-FlashX，高速多模态推理模型，100 万上下文，始终开启推理",
+        "extra": {
+          "thinkingOnly": true
+        }
+      },
+      "routing": {
+        "tier": "balanced",
+        "routingPriority": 49,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultMode": "max"
+        }
+      }
+    },
+    {
+      "displayName": "GLM-5.3",
+      "family": "glm-5.3",
+      "match": {
+        "exact": [
+          "glm-5.3",
+          "z-ai/glm-5.3"
+        ]
+      },
+      "id": "glm-5.3",
+      "spec": {
+        "contextWindow": 1048576,
+        "maxOutputTokens": 131072,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "deep_thinking",
+          "code",
+          "math",
+          "multilingual",
+          "tool_use",
+          "agent",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "defaultTopP": 0.95,
+        "supportsReasoning": true,
+        "description": "智谱 GLM-5.3 大推理模型，面向复杂软件工程与长程 Agent 任务，100 万上下文；强制思考，无法关闭",
+        "extra": {
+          "thinkingDefault": true,
+          "thinkingOnly": true
+        },
+        "releasedAt": "2026-08-16"
+      },
+      "routing": {
+        "tier": "flagship",
+        "routingPriority": 50,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultMode": "max"
+        }
+      }
+    },
+    {
+      "displayName": "GLM-5.3-Flash",
+      "family": "glm-5.3-flash",
+      "match": {
+        "exact": [
+          "glm-5.3-flash",
+          "z-ai/glm-5.3-flash"
+        ]
+      },
+      "id": "glm-5.3-flash",
+      "spec": {
+        "contextWindow": 1048576,
+        "maxOutputTokens": 131072,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "deep_thinking",
+          "code",
+          "tool_use",
+          "agent",
+          "long_context",
+          "vision",
+          "image_understanding",
+          "video_understanding"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning",
+          "vision"
+        ],
+        "defaultTemperature": 1,
+        "defaultTopP": 0.95,
+        "supportsReasoning": true,
+        "description": "智谱 GLM-5.3-Flash 原生多模态模型，混合稀疏与线性注意力，面向高效编码与长程 Agent 任务，100 万上下文；强制思考，无法关闭",
+        "extra": {
+          "thinkingDefault": true,
+          "thinkingOnly": true
+        },
+        "releasedAt": "2026-08-26"
+      },
+      "routing": {
+        "tier": "balanced",
+        "routingPriority": 48,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultMode": "max"
+        }
+      }
+    },
+    {
+      "displayName": "GLM-5.2",
+      "family": "glm-5.2",
+      "match": {
+        "exact": [
+          "glm-5.2"
+        ],
+        "patterns": [
+          "glm-5.2*"
+        ]
+      },
+      "id": "glm-5.2",
+      "spec": {
+        "contextWindow": 1048576,
+        "maxOutputTokens": 32768,
+        "capabilities": [
+          "reasoning",
+          "deep_thinking",
+          "code",
+          "multilingual",
+          "tool_use",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "supportsReasoning": true,
+        "description": "智谱 GLM-5.2 大推理模型，100 万上下文"
+      },
+      "routing": {
+        "tier": "flagship",
+        "routingPriority": 45,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "off",
+            "minimal",
+            "low",
+            "medium",
+            "high",
+            "xhigh"
+          ],
+          "defaultMode": "high"
+        }
+      }
+    },
+    {
+      "displayName": "GLM-5.1",
+      "family": "glm-5.1",
+      "id": "glm-5.1",
+      "spec": {
+        "contextWindow": 200000,
+        "maxOutputTokens": 128000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "multilingual",
+          "deep_thinking",
+          "long_context",
+          "math",
+          "tool_use",
+          "agent"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true
+      },
+      "routing": {
+        "tier": "balanced",
+        "routingPriority": 65,
+        "eligibleForAgent": true,
+        "defaultReference": false,
+        "reasoning": {
+          "supportedModes": [
+            "auto",
+            "off",
+            "minimal",
+            "low",
+            "medium",
+            "high",
+            "xhigh"
+          ],
+          "defaultMode": "medium"
+        }
+      }
+    },
+    {
+      "displayName": "GLM-5",
+      "family": "glm-5",
+      "id": "glm-5",
+      "spec": {
+        "contextWindow": 200000,
+        "maxOutputTokens": 128000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "multilingual",
+          "deep_thinking",
+          "long_context",
+          "math",
+          "tool_use",
+          "agent"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true
+      }
+    },
+    {
+      "displayName": "GLM-5 Turbo",
+      "family": "glm-5-turbo",
+      "id": "glm-5-turbo",
+      "spec": {
+        "contextWindow": 200000,
+        "maxOutputTokens": 128000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "deep_thinking",
+          "long_context",
+          "tool_use",
+          "agent"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true
+      }
+    },
+    {
+      "displayName": "GLM-4.7",
+      "family": "glm-4.7",
+      "id": "glm-4.7",
+      "spec": {
+        "contextWindow": 200000,
+        "maxOutputTokens": 128000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "multilingual",
+          "deep_thinking",
+          "long_context",
+          "tool_use"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true
+      }
+    },
+    {
+      "displayName": "GLM-4.6",
+      "family": "glm-4.6",
+      "id": "glm-4.6",
+      "spec": {
+        "contextWindow": 200000,
+        "maxOutputTokens": 128000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "multilingual",
+          "deep_thinking"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true
+      }
+    },
+    {
+      "displayName": "GLM-4.7 Thinking",
+      "family": "glm-4.7",
+      "match": {
+        "exact": [
+          "glm-4.7-thinking"
+        ]
+      },
+      "id": "glm-4.7-thinking",
+      "spec": {
+        "contextWindow": 200000,
+        "maxOutputTokens": 128000,
+        "capabilities": [
+          "reasoning",
+          "math",
+          "code",
+          "deep_thinking",
+          "long_context"
+        ],
+        "serviceType": [
+          "reasoning"
+        ],
+        "defaultTemperature": null,
+        "supportsReasoning": true,
+        "description": "智谱GLM-4.7深度思考模式，交错式/保留式/轮级思考"
+      }
+    },
+    {
+      "displayName": "GLM-5V-Turbo",
+      "family": "glm-5v",
+      "match": {
+        "exact": [
+          "glm-5v-turbo"
+        ]
+      },
+      "id": "glm-5v-turbo",
+      "spec": {
+        "contextWindow": 200000,
+        "maxOutputTokens": 128000,
+        "capabilities": [
+          "chat",
+          "vision",
+          "video_understanding",
+          "image_understanding",
+          "file_understanding",
+          "reasoning",
+          "code",
+          "deep_thinking",
+          "long_context",
+          "tool_use",
+          "agent"
+        ],
+        "serviceType": [
+          "vision"
+        ],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "description": "智谱首个多模态 Coding 基座模型，支持视频、图像、文本和文件输入"
+      }
+    },
+    {
+      "displayName": "GLM-4.6V",
+      "family": "glm-4.6",
+      "match": {
+        "exact": [
+          "glm-4.6v"
+        ]
+      },
+      "id": "glm-4.6v",
+      "spec": {
+        "contextWindow": 128000,
+        "maxOutputTokens": 32768,
+        "capabilities": [
+          "chat",
+          "vision",
+          "video_understanding",
+          "image_understanding",
+          "long_context",
+          "tool_use"
+        ],
+        "serviceType": [
+          "vision"
+        ],
+        "defaultTemperature": 1,
+        "description": "智谱GLM-4.6V多模态版，106B/12B MoE，支持图像视频理解"
+      }
+    },
+    {
+      "displayName": "智谱 embedding-3",
+      "family": "zhipu-embedding",
+      "match": {
+        "exact": [
+          "embedding-3"
+        ]
+      },
+      "id": "embedding-3",
+      "spec": {
+        "contextWindow": 8192,
+        "capabilities": [
+          "text_embedding",
+          "semantic_search",
+          "rag",
+          "custom_dimensions"
+        ],
+        "serviceType": [
+          "embedding"
+        ],
+        "description": "智谱嵌入模型v3，支持自定义维度；单条输入最多 3072 tokens"
+      }
+    }
+  ]
+}

compute/v2/compatibility/providers/anthropic-claude.json

@@ -0,0 +1,239 @@
+{
+  "id": "provider-anthropic-claude-plan-001",
+  "provider": "anthropic-claude",
+  "brandGroup": "anthropic",
+  "label": "Claude 订阅 (Claude Code)",
+  "baseUrl": "https://api.anthropic.com",
+  "apiFormat": "anthropic-messages",
+  "apiKeyRef": "anthropic-claude-oauth-token",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "USD",
+  "accessMode": "coding-plan",
+  "credentialSource": "claude-oauth",
+  "requiredClientVersion": "10.0.83",
+  "codingPlan": {
+    "quotas": {},
+    "usageTracking": {
+      "method": "none",
+      "consoleUrl": "https://claude.ai/settings/usage"
+    }
+  },
+  "services": [
+    "chat",
+    "reasoning"
+  ],
+  "tombstones": [
+    "claude-opus-4-8",
+    "claude-fable-5",
+    "claude-haiku-4-5",
+    "claude-haiku-4-5-20251001"
+  ],
+  "models": [
+    {
+      "modelName": "claude-opus-5-5",
+      "displayName": "Claude Opus 5.5",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1000000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "description": "Claude 订阅额度计费，长程智能体编程与知识工作模型，1M 上下文",
+      "extra": {
+        "adaptiveThinking": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultEffort": "medium"
+        },
+        "thinkingOnly": true,
+        "thinkingRoundTrip": {
+          "enabled": true,
+          "protocol": "anthropic-messages",
+          "allowEmptySignature": false,
+          "scope": "active-tool-turn"
+        },
+        "thinkingToolTurnValidation": "accepts-thinkless"
+      },
+      "source": "preset"
+    },
+    {
+      "modelName": "claude-fable-5-1",
+      "displayName": "Claude Fable 5.1",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Claude 订阅额度计费，面向高难推理和长程智能体任务，1M 上下文，始终开启自适应思考",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "extra": {
+        "adaptiveThinking": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultEffort": "high"
+        },
+        "thinkingOnly": true,
+        "thinkingRoundTrip": {
+          "enabled": true,
+          "protocol": "anthropic-messages",
+          "allowEmptySignature": false,
+          "scope": "active-tool-turn"
+        },
+        "thinkingToolTurnValidation": "accepts-thinkless"
+      },
+      "source": "preset"
+    },
+    {
+      "modelName": "claude-opus-5",
+      "displayName": "Claude Opus 5",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Claude 订阅额度计费，Opus 5 系列模型，复杂智能体编码与长程任务首选，1M 上下文",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "extra": {
+        "adaptiveThinking": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultEffort": "high"
+        },
+        "thinkingRoundTrip": {
+          "enabled": true,
+          "protocol": "anthropic-messages",
+          "allowEmptySignature": false,
+          "scope": "active-tool-turn"
+        },
+        "thinkingToolTurnValidation": "accepts-thinkless"
+      },
+      "source": "preset"
+    },
+    {
+      "modelName": "claude-sonnet-5-5",
+      "displayName": "Claude Sonnet 5.5",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Claude Sonnet 5.5，速度与智能平衡，1M 上下文，默认 high effort，自适应思考",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "extra": {
+        "adaptiveThinking": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultEffort": "high"
+        },
+        "thinkingRoundTrip": {
+          "enabled": true,
+          "protocol": "anthropic-messages",
+          "allowEmptySignature": false,
+          "scope": "active-tool-turn"
+        },
+        "thinkingToolTurnValidation": "accepts-thinkless",
+        "samplingParametersDeprecated": true,
+        "forcedToolChoiceUnsupported": true
+      },
+      "source": "preset"
+    },
+    {
+      "modelName": "claude-sonnet-5",
+      "displayName": "Claude Sonnet 5",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Claude 订阅额度计费，速度与智能平衡模型，支持长上下文智能体任务",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "extra": {
+        "adaptiveThinking": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultEffort": "high"
+        },
+        "thinkingRoundTrip": {
+          "enabled": true,
+          "protocol": "anthropic-messages",
+          "allowEmptySignature": false,
+          "scope": "active-tool-turn"
+        },
+        "thinkingToolTurnValidation": "accepts-thinkless"
+      },
+      "source": "preset"
+    }
+  ]
+}

compute/v2/compatibility/providers/anthropic.json

@@ -0,0 +1,311 @@
+{
+  "id": "provider-anthropic-001",
+  "provider": "anthropic",
+  "label": "Anthropic Claude",
+  "baseUrl": "https://api.anthropic.com",
+  "apiFormat": "anthropic-messages",
+  "apiKeyRef": "anthropic",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "USD",
+  "services": [
+    "chat",
+    "computer_use"
+  ],
+  "tombstones": [
+    "claude-opus-4-8",
+    "claude-opus-4-7",
+    "claude-sonnet-4-6",
+    "claude-fable-5",
+    "claude-haiku-4-5",
+    "claude-haiku-4-5-20251001"
+  ],
+  "models": [
+    {
+      "modelName": "claude-opus-5-5",
+      "displayName": "Claude Opus 5.5",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Anthropic 面向长程智能体编程与知识工作的 Opus 模型，1M 上下文，始终开启自适应思考",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "inputPrice": 4,
+      "outputPrice": 20,
+      "extra": {
+        "thinkingRoundTrip": {
+          "enabled": true,
+          "protocol": "anthropic-messages",
+          "allowEmptySignature": false,
+          "scope": "active-tool-turn"
+        },
+        "thinkingToolTurnValidation": "accepts-thinkless",
+        "serverSideWebSearch": {
+          "enabled": true,
+          "dialect": "anthropic-messages",
+          "fallbackPriority": 2,
+          "toolType": "web_search_20260318",
+          "searchRequestPriceUsd": 0.01
+        },
+        "cachePricing": {
+          "write5m": 5,
+          "write1h": 8,
+          "read": 0.2
+        },
+        "promptCacheMinTokens": 512,
+        "adaptiveThinking": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultEffort": "medium"
+        },
+        "samplingParametersDeprecated": true,
+        "forcedToolChoiceUnsupported": true,
+        "thinkingOnly": true,
+        "pricingNotes": "Prices are per 1M tokens. Full 1M context is billed at standard pricing."
+      }
+    },
+    {
+      "modelName": "claude-fable-5-1",
+      "displayName": "Claude Fable 5.1",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "面向高难推理和长程智能体任务，1M 上下文，始终开启自适应思考",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "inputPrice": 10,
+      "outputPrice": 50,
+      "extra": {
+        "thinkingRoundTrip": {
+          "enabled": true,
+          "protocol": "anthropic-messages",
+          "allowEmptySignature": false,
+          "scope": "active-tool-turn"
+        },
+        "thinkingToolTurnValidation": "accepts-thinkless",
+        "serverSideWebSearch": {
+          "enabled": true,
+          "dialect": "anthropic-messages",
+          "fallbackPriority": 1,
+          "toolType": "web_search_20260318",
+          "searchRequestPriceUsd": 0.01
+        },
+        "cachePricing": {
+          "write5m": 12.5,
+          "write1h": 20,
+          "read": 0.25
+        },
+        "adaptiveThinking": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultEffort": "high"
+        },
+        "samplingParametersDeprecated": true,
+        "forcedToolChoiceUnsupported": true,
+        "pricingNotes": "Prices are per 1M tokens. Full 1M context is billed at standard pricing.",
+        "thinkingOnly": true
+      }
+    },
+    {
+      "modelName": "claude-opus-5",
+      "displayName": "Claude Opus 5",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Anthropic Opus 5 系列模型，复杂智能体编码与长程任务能力显著强于 Opus 4.8，价格不变，1M 上下文",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "inputPrice": 5,
+      "outputPrice": 25,
+      "extra": {
+        "thinkingRoundTrip": {
+          "enabled": true,
+          "protocol": "anthropic-messages",
+          "allowEmptySignature": false,
+          "scope": "active-tool-turn"
+        },
+        "thinkingToolTurnValidation": "accepts-thinkless",
+        "serverSideWebSearch": {
+          "enabled": true,
+          "dialect": "anthropic-messages",
+          "fallbackPriority": 2,
+          "toolType": "web_search_20260318",
+          "searchRequestPriceUsd": 0.01
+        },
+        "cachePricing": {
+          "write5m": 6.25,
+          "write1h": 10,
+          "read": 0.5
+        },
+        "promptCacheMinTokens": 512,
+        "adaptiveThinking": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultEffort": "high"
+        },
+        "samplingParametersDeprecated": true,
+        "pricingNotes": "Prices are per 1M tokens. Full 1M context is billed at standard pricing."
+      }
+    },
+    {
+      "modelName": "claude-sonnet-5-5",
+      "displayName": "Claude Sonnet 5.5",
+      "serviceType": [
+        "chat",
+        "computer_use"
+      ],
+      "description": "Claude Sonnet 5.5，速度与智能平衡，1M 上下文，默认 high effort，自适应思考",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "computer_use",
+        "agent",
+        "long_context"
+      ],
+      "inputPrice": 2,
+      "outputPrice": 10,
+      "extra": {
+        "thinkingRoundTrip": {
+          "enabled": true,
+          "protocol": "anthropic-messages",
+          "allowEmptySignature": false,
+          "scope": "active-tool-turn"
+        },
+        "thinkingToolTurnValidation": "accepts-thinkless",
+        "serverSideWebSearch": {
+          "enabled": true,
+          "dialect": "anthropic-messages",
+          "fallbackPriority": 3,
+          "toolType": "web_search_20260318",
+          "searchRequestPriceUsd": 0.01
+        },
+        "cachePricing": {
+          "write5m": 2.5,
+          "write1h": 4,
+          "read": 0.2
+        },
+        "adaptiveThinking": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultEffort": "high"
+        },
+        "samplingParametersDeprecated": true,
+        "forcedToolChoiceUnsupported": true,
+        "promptCacheMinTokens": 512
+      }
+    },
+    {
+      "modelName": "claude-sonnet-5",
+      "displayName": "Claude Sonnet 5",
+      "serviceType": [
+        "chat",
+        "computer_use"
+      ],
+      "description": "Anthropic Sonnet 5，速度与智能平衡模型，支持电脑使用和长上下文智能体任务",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "computer_use",
+        "agent",
+        "long_context"
+      ],
+      "inputPrice": 2,
+      "outputPrice": 10,
+      "extra": {
+        "thinkingRoundTrip": {
+          "enabled": true,
+          "protocol": "anthropic-messages",
+          "allowEmptySignature": false,
+          "scope": "active-tool-turn"
+        },
+        "thinkingToolTurnValidation": "accepts-thinkless",
+        "serverSideWebSearch": {
+          "enabled": true,
+          "dialect": "anthropic-messages",
+          "fallbackPriority": 3,
+          "toolType": "web_search_20260318",
+          "searchRequestPriceUsd": 0.01
+        },
+        "cachePricing": {
+          "write5m": 2.5,
+          "write1h": 4,
+          "read": 0.2
+        },
+        "adaptiveThinking": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultEffort": "high"
+        },
+        "samplingParametersDeprecated": true
+      }
+    }
+  ]
+}

compute/v2/compatibility/providers/baichuan.json

@@ -0,0 +1,107 @@
+{
+  "id": "provider-baichuan-001",
+  "provider": "baichuan",
+  "label": "百川 Baichuan",
+  "baseUrl": "https://api.baichuan-ai.com/v1",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "baichuan",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "CNY",
+  "services": [
+    "chat"
+  ],
+  "models": [
+    {
+      "modelName": "Baichuan-M3-Plus",
+      "displayName": "百川 M3-Plus",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百川智能M3-Plus医疗增强模型，235B参数，幻觉率仅2.6%，API价格较前代下降70%",
+      "contextWindow": 32000,
+      "maxOutputTokens": 32000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual",
+        "vision",
+        "long_context",
+        "medical"
+      ],
+      "inputPrice": 5,
+      "outputPrice": 9,
+      "defaultTemperature": 0.3,
+      "defaultTopP": 0.85,
+      "extra": {}
+    },
+    {
+      "modelName": "Baichuan-M3",
+      "displayName": "百川 M3",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百川智能M3新一代开源医疗增强模型，235B参数",
+      "contextWindow": 32000,
+      "maxOutputTokens": 32000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual",
+        "vision",
+        "medical"
+      ],
+      "inputPrice": 10,
+      "outputPrice": 30,
+      "defaultTemperature": 0.3,
+      "defaultTopP": 0.85,
+      "extra": {}
+    },
+    {
+      "modelName": "Baichuan-M2-Plus",
+      "displayName": "百川 M2-Plus",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百川智能M2-Plus旗舰模型，最强推理和创作能力",
+      "contextWindow": 32000,
+      "maxOutputTokens": 32000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual",
+        "long_context"
+      ],
+      "inputPrice": 10,
+      "outputPrice": 30,
+      "defaultTemperature": 0.3,
+      "defaultTopP": 0.85,
+      "extra": {}
+    },
+    {
+      "modelName": "Baichuan-M2",
+      "displayName": "百川 M2",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百川智能M2模型，高性能通用模型",
+      "contextWindow": 32000,
+      "maxOutputTokens": 32000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual"
+      ],
+      "inputPrice": 2,
+      "outputPrice": 20,
+      "defaultTemperature": 0.3,
+      "defaultTopP": 0.85,
+      "extra": {}
+    }
+  ]
+}

compute/v2/compatibility/providers/baidu.json

@@ -0,0 +1,157 @@
+{
+  "id": "provider-baidu-001",
+  "provider": "baidu",
+  "label": "百度 文心",
+  "baseUrl": "https://qianfan.baidubce.com/v2",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "baidu",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "CNY",
+  "services": [
+    "chat",
+    "reasoning"
+  ],
+  "models": [
+    {
+      "modelName": "ernie-5.0-thinking-latest",
+      "displayName": "文心 ERNIE 5.0",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "百度最新旗舰模型，支持深度推理、多模态和工具调用，119K输入/64K输出",
+      "contextWindow": 128000,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "vision",
+        "multimodal",
+        "tool_use"
+      ],
+      "inputPrice": 6,
+      "outputPrice": 24,
+      "extra": {
+        "pricingTiers": [
+          {
+            "maxInputTokens": 32768,
+            "inputPrice": 6,
+            "outputPrice": 24
+          },
+          {
+            "maxInputTokens": 131072,
+            "inputPrice": 10,
+            "outputPrice": 40
+          }
+        ],
+        "thinkingMaxTokens": 60000
+      }
+    },
+    {
+      "modelName": "ernie-5.0",
+      "displayName": "文心 ERNIE 5.0 非思考",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百度 ERNIE 5.0 旗舰模型，128K 上下文，119K 最大输入，最大输出 65,536 token",
+      "contextWindow": 131072,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "vision",
+        "multimodal",
+        "tool_use"
+      ],
+      "inputPrice": 6,
+      "outputPrice": 24,
+      "defaultTemperature": 0.95,
+      "defaultTopP": 0.7,
+      "extra": {
+        "pricingTiers": [
+          {
+            "maxInputTokens": 32768,
+            "inputPrice": 6,
+            "outputPrice": 24
+          },
+          {
+            "maxInputTokens": 131072,
+            "inputPrice": 10,
+            "outputPrice": 40
+          }
+        ]
+      }
+    },
+    {
+      "modelName": "ernie-4.5-turbo-128k",
+      "displayName": "文心 ERNIE 4.5 Turbo",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百度高性价比长上下文模型，128K 窗口，最大输出 12,288 token",
+      "contextWindow": 131072,
+      "maxOutputTokens": 12288,
+      "capabilities": [
+        "chat",
+        "code",
+        "vision",
+        "long_context",
+        "fast"
+      ],
+      "inputPrice": 0.8,
+      "outputPrice": 3.2,
+      "defaultTemperature": 0.8,
+      "defaultTopP": 1,
+      "extra": {
+        "cacheHitPrice": 0.2
+      }
+    },
+    {
+      "modelName": "ernie-4.5-turbo-20260402",
+      "displayName": "文心 ERNIE 4.5 Turbo 20260402",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "百度 ERNIE 4.5 Turbo 20260402 快照，128K 上下文，最大输出 12,288 token",
+      "contextWindow": 131072,
+      "maxOutputTokens": 12288,
+      "capabilities": [
+        "chat",
+        "code",
+        "vision",
+        "long_context",
+        "fast"
+      ],
+      "inputPrice": 0.8,
+      "outputPrice": 3.2,
+      "defaultTemperature": 0.8,
+      "defaultTopP": 1,
+      "extra": {
+        "cacheHitPrice": 0.2
+      }
+    },
+    {
+      "modelName": "ernie-x1.1",
+      "displayName": "文心 ERNIE X1.1",
+      "serviceType": [
+        "reasoning"
+      ],
+      "description": "百度 ERNIE X1.1 深度思考模型，64K 上下文，最大输出 65,536 token",
+      "contextWindow": 65536,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "math",
+        "code"
+      ],
+      "inputPrice": 1,
+      "outputPrice": 4,
+      "extra": {
+        "thinkingMaxTokens": 65536
+      }
+    }
+  ]
+}

compute/v2/compatibility/providers/cohere.json

@@ -0,0 +1,151 @@
+{
+  "id": "provider-cohere-001",
+  "provider": "cohere",
+  "label": "Cohere",
+  "baseUrl": "https://api.cohere.ai/compatibility/v1",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "cohere",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "USD",
+  "services": [
+    "chat",
+    "embedding",
+    "rerank"
+  ],
+  "tombstones": [
+    "rerank-v3.5"
+  ],
+  "models": [
+    {
+      "modelName": "embed-v5.0-fast",
+      "displayName": "Embed 5 Fast",
+      "capabilities": [
+        "text_embedding",
+        "multilingual"
+      ],
+      "serviceType": [
+        "embedding"
+      ],
+      "description": "Embed 5 Fast；当前 OpenAI 兼容接入仅声明文本向量能力",
+      "contextWindow": 128000,
+      "extra": {
+        "defaultDimension": 2048,
+        "pricingNotes": "Token 单价未由当前官方价格页公开确认；Model Vault 按实例计价。 Compatibility API 不接受 dimensions 参数；返回默认维度，多维度选择需原生 Embed API。"
+      }
+    },
+    {
+      "modelName": "embed-v5.0-pro",
+      "displayName": "Embed 5 Pro",
+      "capabilities": [
+        "text_embedding",
+        "multilingual"
+      ],
+      "serviceType": [
+        "embedding"
+      ],
+      "description": "Embed 5 Pro；当前 OpenAI 兼容接入仅声明文本向量能力",
+      "contextWindow": 128000,
+      "extra": {
+        "defaultDimension": 2048,
+        "pricingNotes": "Token 单价未由当前官方价格页公开确认；Model Vault 按实例计价。 Compatibility API 不接受 dimensions 参数；返回默认维度，多维度选择需原生 Embed API。"
+      }
+    },
+    {
+      "modelName": "command-a-plus-05-2026",
+      "displayName": "Command A+",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "vision",
+        "tool_use",
+        "multilingual",
+        "long_context"
+      ],
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Command A+",
+      "contextWindow": 128000,
+      "maxOutputTokens": 64000,
+      "extra": {
+        "pricingNotes": "官方文档声明限流额度内免费；生产 Model Vault 按实例单独计价，不预填 token 单价。",
+        "reasoning": {
+          "supportedEfforts": [
+            "none",
+            "high"
+          ]
+        }
+      }
+    },
+    {
+      "modelName": "command-a-03-2025",
+      "displayName": "Command A",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Cohere Command A 文本模型，256K 上下文，支持 RAG 和工具调用",
+      "contextWindow": 256000,
+      "maxOutputTokens": 8000,
+      "capabilities": [
+        "chat",
+        "code",
+        "tool_use",
+        "rag",
+        "long_context"
+      ],
+      "inputPrice": 2.5,
+      "outputPrice": 10,
+      "defaultTemperature": 0.3,
+      "defaultTopP": 0.75,
+      "extra": {
+        "pricingNotes": "Prices are per 1M tokens."
+      }
+    },
+    {
+      "modelName": "command-r7b-12-2024",
+      "displayName": "Command R7B",
+      "serviceType": [
+        "fast"
+      ],
+      "description": "Cohere 小型高速对话模型，适合高吞吐、低延迟场景",
+      "contextWindow": 128000,
+      "maxOutputTokens": 4000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "tool_use",
+        "rag",
+        "fast"
+      ],
+      "inputPrice": 0.0375,
+      "outputPrice": 0.15,
+      "defaultTemperature": 0.3,
+      "defaultTopP": 0.75,
+      "extra": {
+        "pricingNotes": "Prices are per 1M tokens."
+      }
+    },
+    {
+      "modelName": "embed-v4.0",
+      "displayName": "Embed V4",
+      "serviceType": [
+        "embedding"
+      ],
+      "description": "Cohere 最新 Embedding 模型，128K 上下文，多语言支持",
+      "contextWindow": 128000,
+      "maxOutputTokens": 0,
+      "capabilities": [
+        "text_embedding",
+        "multilingual"
+      ],
+      "inputPrice": 0.12,
+      "outputPrice": 0,
+      "extra": {
+        "defaultDimension": 1536,
+        "pricingNotes": "Embedding models are priced by embedded tokens; price is per 1M tokens. Compatibility API 不接受 dimensions 参数；返回默认维度，多维度选择需原生 Embed API。"
+      }
+    }
+  ]
+}

compute/v2/compatibility/providers/dashscope.json

@@ -0,0 +1,941 @@
+{
+  "id": "provider-dashscope-001",
+  "provider": "dashscope",
+  "label": "阿里 通义千问",
+  "baseUrl": "https://dashscope.aliyuncs.com/compatible-mode/v1",
+  "mediaBaseUrl": "https://dashscope.aliyuncs.com/api/v1",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "dashscope",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "CNY",
+  "services": [
+    "chat",
+    "embedding",
+    "rerank",
+    "tts",
+    "asr",
+    "vision",
+    "image_gen",
+    "video_gen",
+    "voice_clone",
+    "omni",
+    "translation"
+  ],
+  "tombstones": [
+    "qwen3.8-max-preview",
+    "qwen3.6-plus",
+    "qwen3.5-plus",
+    "qwen-max",
+    "qwen-plus",
+    "qwen-turbo",
+    "qwen3-max",
+    "qwen3-max-trans"
+  ],
+  "models": [
+    {
+      "modelName": "qwen-image-3.0",
+      "displayName": "qwen-image-3.0",
+      "capabilities": [
+        "image_generation",
+        "image_editing",
+        "text_rendering"
+      ],
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "qwen-image-3.0",
+      "extra": {
+        "pricingNotes": "按图片计费，北京：输入 0.02 元/张；Pro 输出 1K 0.25 元、2K 0.5 元；普通版输出 0.18 元/张。"
+      }
+    },
+    {
+      "modelName": "qwen-image-3.0-pro",
+      "displayName": "qwen-image-3.0-pro",
+      "capabilities": [
+        "image_generation",
+        "image_editing",
+        "text_rendering"
+      ],
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "qwen-image-3.0-pro",
+      "extra": {
+        "pricingNotes": "按图片计费，北京：输入 0.02 元/张；Pro 输出 1K 0.25 元、2K 0.5 元；普通版输出 0.18 元/张。"
+      }
+    },
+    {
+      "modelName": "qwen3.7-text-rerank",
+      "displayName": "Qwen3.7 Text Rerank",
+      "capabilities": [
+        "rerank",
+        "multilingual"
+      ],
+      "serviceType": [
+        "rerank"
+      ],
+      "description": "Qwen3.7 Text Rerank"
+    },
+    {
+      "modelName": "qwen3.7-text-embedding-flash",
+      "displayName": "qwen3.7-text-embedding-flash",
+      "capabilities": [
+        "text_embedding",
+        "multilingual"
+      ],
+      "serviceType": [
+        "embedding"
+      ],
+      "description": "qwen3.7-text-embedding-flash",
+      "inputPrice": 0.125,
+      "outputPrice": 0
+    },
+    {
+      "modelName": "qwen3.7-text-embedding",
+      "displayName": "qwen3.7-text-embedding",
+      "capabilities": [
+        "text_embedding",
+        "multilingual"
+      ],
+      "serviceType": [
+        "embedding"
+      ],
+      "description": "qwen3.7-text-embedding",
+      "inputPrice": 0.5,
+      "outputPrice": 0
+    },
+    {
+      "modelName": "qwen3.8-omni-flash",
+      "displayName": "Qwen3.8 Omni Flash",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "vision",
+        "audio_understanding",
+        "video_understanding",
+        "tool_use",
+        "long_context"
+      ],
+      "serviceType": [
+        "chat",
+        "vision"
+      ],
+      "description": "Qwen3.8 Omni Flash",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 131072,
+      "inputPrice": 0.8,
+      "outputPrice": 2.7,
+      "extra": {
+        "cachedInputPrice": 0.1
+      }
+    },
+    {
+      "modelName": "qwen3.8-max",
+      "displayName": "阿里云 Qwen3.8-Max",
+      "serviceType": [
+        "chat",
+        "vision"
+      ],
+      "description": "通义千问3.8 Max 稳定版，支持视觉、视频理解和工具调用，100万上下文",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual",
+        "long_context",
+        "tool_use",
+        "agent",
+        "vision",
+        "video_understanding"
+      ],
+      "inputPrice": 12,
+      "outputPrice": 36,
+      "defaultTemperature": 0.6,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cachedInputPrice": 1.5,
+        "pricingNotes": "CNY per 1M tokens, Beijing region standard price.",
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "xhigh"
+          ],
+          "defaultEffort": "xhigh"
+        },
+        "preserveThinkingDefault": true,
+        "thinkingMaxTokens": 262144
+      }
+    },
+    {
+      "modelName": "qwen3.8-flash",
+      "displayName": "阿里云 Qwen3.8-Flash",
+      "serviceType": [
+        "chat",
+        "vision"
+      ],
+      "description": "通义千问3.8 Flash 稳定版，高速多模态推理，100万上下文",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual",
+        "long_context",
+        "tool_use",
+        "agent",
+        "vision",
+        "video_understanding"
+      ],
+      "inputPrice": 0.8,
+      "outputPrice": 2.7,
+      "defaultTemperature": 0.6,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cachedInputPrice": 0.1,
+        "pricingNotes": "CNY per 1M tokens, Beijing region standard price.",
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "xhigh"
+          ],
+          "defaultEffort": "xhigh"
+        },
+        "preserveThinkingDefault": true,
+        "thinkingMaxTokens": 262144
+      }
+    },
+    {
+      "modelName": "qwen3.7-max",
+      "displayName": "阿里云 Qwen3.7-Max",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "通义千问3.7 Max，面向复杂推理与长程 Agent 任务的旗舰模型，100万上下文，支持内置工具和 Function Calling",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual",
+        "long_context",
+        "tool_use",
+        "agent"
+      ],
+      "inputPrice": 12,
+      "outputPrice": 36,
+      "defaultTemperature": 0.6,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cacheHitPrice": 2.4
+      }
+    },
+    {
+      "modelName": "qwen3.7-plus",
+      "displayName": "阿里云 Qwen3.7-Plus",
+      "serviceType": [
+        "chat",
+        "vision"
+      ],
+      "description": "通义千问3.7 Plus，多模态智能体模型，100万上下文，支持内置工具、Function Calling 和结构化输出",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual",
+        "long_context",
+        "tool_use",
+        "agent",
+        "vision"
+      ],
+      "inputPrice": 2,
+      "outputPrice": 8,
+      "defaultTemperature": 0.6,
+      "defaultTopP": 0.95,
+      "extra": {
+        "pricingTiers": [
+          {
+            "maxInputTokens": 256000,
+            "inputPrice": 2,
+            "outputPrice": 8,
+            "thinkingOutputPrice": 8
+          },
+          {
+            "maxInputTokens": 1000000,
+            "inputPrice": 6,
+            "outputPrice": 24,
+            "thinkingOutputPrice": 24
+          }
+        ],
+        "thinkingMaxTokens": 262144
+      }
+    },
+    {
+      "modelName": "qwen3.6-flash",
+      "displayName": "阿里云 Qwen3.6-Flash",
+      "serviceType": [
+        "chat",
+        "vision"
+      ],
+      "description": "通义千问3.6 Flash，高性价比多模态模型，100万上下文，支持内置工具和 Function Calling",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual",
+        "long_context",
+        "tool_use",
+        "vision",
+        "fast"
+      ],
+      "inputPrice": 1.2,
+      "outputPrice": 7.2,
+      "defaultTemperature": 0.6,
+      "defaultTopP": 0.95,
+      "extra": {
+        "pricingTiers": [
+          {
+            "maxInputTokens": 256000,
+            "inputPrice": 1.2,
+            "outputPrice": 7.2
+          },
+          {
+            "maxInputTokens": 1000000,
+            "inputPrice": 4.8,
+            "outputPrice": 28.8
+          }
+        ],
+        "thinkingMaxTokens": 131072
+      }
+    },
+    {
+      "modelName": "qwen-long",
+      "displayName": "阿里云 qwen-long",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "通义千问Long，支持1000万上下文，适合超长文档处理，降价97%",
+      "contextWindow": 10000000,
+      "maxOutputTokens": 32768,
+      "capabilities": [
+        "chat",
+        "ultra_long_context",
+        "document_processing"
+      ],
+      "inputPrice": 0.5,
+      "outputPrice": 2,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.8,
+      "extra": {}
+    },
+    {
+      "modelName": "qwen3-vl-plus",
+      "displayName": "阿里云 Qwen3-VL-Plus",
+      "serviceType": [
+        "vision"
+      ],
+      "description": "通义千问3代视觉模型Plus版，支持图像理解",
+      "contextWindow": 262144,
+      "maxOutputTokens": 32768,
+      "capabilities": [
+        "chat",
+        "vision",
+        "image_understanding",
+        "ocr",
+        "chart_analysis"
+      ],
+      "inputPrice": 1.5,
+      "outputPrice": 6,
+      "defaultTemperature": 0.7,
+      "defaultTopP": 0.8,
+      "extra": {}
+    },
+    {
+      "modelName": "qwen3-vl-flash",
+      "displayName": "阿里云 Qwen3-VL-Flash",
+      "serviceType": [
+        "vision"
+      ],
+      "description": "通义千问3代视觉模型Flash版，高性价比",
+      "contextWindow": 262144,
+      "maxOutputTokens": 32768,
+      "capabilities": [
+        "chat",
+        "vision",
+        "image_understanding",
+        "fast"
+      ],
+      "inputPrice": 0.8,
+      "outputPrice": 3,
+      "defaultTemperature": 0.7,
+      "defaultTopP": 0.8,
+      "extra": {}
+    },
+    {
+      "modelName": "text-embedding-v3",
+      "displayName": "text-embedding-v3",
+      "serviceType": [
+        "embedding"
+      ],
+      "description": "阿里云文本向量化",
+      "contextWindow": 8192,
+      "capabilities": [
+        "text_embedding",
+        "semantic_search",
+        "chinese_optimized"
+      ],
+      "inputPrice": 0.7,
+      "extra": {}
+    },
+    {
+      "modelName": "text-embedding-v4",
+      "displayName": "阿里云 text-embedding-v4",
+      "serviceType": [
+        "embedding"
+      ],
+      "description": "阿里云百炼文本向量模型v4，支持多维度与多语种",
+      "contextWindow": 8192,
+      "capabilities": [
+        "text_embedding",
+        "semantic_search",
+        "rag"
+      ],
+      "inputPrice": 0.5,
+      "extra": {}
+    },
+    {
+      "modelName": "qwen3-rerank",
+      "displayName": "Qwen3-Rerank",
+      "serviceType": [
+        "rerank"
+      ],
+      "description": "检索重排序，支持 100+ 语种",
+      "contextWindow": 120000,
+      "capabilities": [
+        "rerank",
+        "semantic_reranking",
+        "multilingual"
+      ],
+      "inputPrice": 1,
+      "extra": {}
+    },
+    {
+      "modelName": "cosyvoice-v2",
+      "displayName": "CosyVoice V2",
+      "serviceType": [
+        "tts"
+      ],
+      "description": "通义语音合成，支持中英文",
+      "capabilities": [
+        "tts",
+        "multilingual",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "voices": [
+          "longxiaochun",
+          "longxiaoxia",
+          "longlaotie",
+          "longshu"
+        ],
+        "supportedAudioFormats": [
+          "mp3",
+          "wav",
+          "pcm"
+        ]
+      }
+    },
+    {
+      "modelName": "paraformer-v2",
+      "displayName": "Paraformer V2",
+      "serviceType": [
+        "asr"
+      ],
+      "description": "通义语音识别",
+      "capabilities": [
+        "asr",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "maxAudioLength": 3600,
+        "supportedInputFormats": [
+          "wav",
+          "mp3",
+          "pcm",
+          "ogg"
+        ],
+        "streamingSupported": true
+      }
+    },
+    {
+      "modelName": "wan2.7-image-pro",
+      "displayName": "通义万相 2.7 Pro",
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "万相最新旗舰文生图模型，支持 4K 分辨率、thinking_mode 智能思考、色彩调色板控制",
+      "capabilities": [
+        "image_generation",
+        "chinese_optimized",
+        "high_quality"
+      ],
+      "extra": {
+        "endpoint": "/services/aigc/image-generation/generation",
+        "asyncHeader": "X-DashScope-Async",
+        "taskQueryEndpoint": "/tasks",
+        "supportedImageSizes": [
+          "768*768",
+          "1024*1024",
+          "1280*1280",
+          "1024*1792",
+          "1792*1024",
+          "2048*2048",
+          "4096*4096"
+        ],
+        "sizeShortcuts": {
+          "1K": "1024*1024",
+          "2K": "2048*2048",
+          "4K": "4096*4096"
+        },
+        "maxN": 4,
+        "supportsThinkingMode": true,
+        "supportsColorPalette": true
+      }
+    },
+    {
+      "modelName": "wan2.7-image",
+      "displayName": "通义万相 2.7",
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "万相最新标准文生图模型，高画质，支持 thinking_mode",
+      "capabilities": [
+        "image_generation",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "endpoint": "/services/aigc/image-generation/generation",
+        "asyncHeader": "X-DashScope-Async",
+        "taskQueryEndpoint": "/tasks",
+        "supportedImageSizes": [
+          "768*768",
+          "1024*1024",
+          "1280*1280",
+          "1024*1792",
+          "1792*1024"
+        ],
+        "maxN": 4,
+        "supportsThinkingMode": true
+      }
+    },
+    {
+      "modelName": "wan2.6-t2i",
+      "displayName": "通义万相 2.6 文生图",
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "万相 2.6 文生图专用模型，稳定可靠，支持 prompt_extend 智能改写",
+      "capabilities": [
+        "image_generation",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "endpoint": "/services/aigc/image-generation/generation",
+        "asyncHeader": "X-DashScope-Async",
+        "taskQueryEndpoint": "/tasks",
+        "supportedImageSizes": [
+          "1024*1024",
+          "1280*1280",
+          "720*1280",
+          "1280*720",
+          "768*1152",
+          "1152*768"
+        ],
+        "maxN": 4,
+        "supportsNegativePrompt": true,
+        "supportsPromptExtend": true
+      }
+    },
+    {
+      "modelName": "wan2.2-t2i-plus",
+      "displayName": "通义万相 2.2 Plus",
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "万相 2.2 增强版，画质与速度平衡",
+      "capabilities": [
+        "image_generation",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "endpoint": "/services/aigc/image-generation/generation",
+        "asyncHeader": "X-DashScope-Async",
+        "taskQueryEndpoint": "/tasks",
+        "supportedImageSizes": [
+          "1024*1024",
+          "720*1280",
+          "1280*720"
+        ],
+        "maxN": 4,
+        "supportsNegativePrompt": true
+      }
+    },
+    {
+      "modelName": "wan2.2-t2i-flash",
+      "displayName": "通义万相 2.2 Flash",
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "万相 2.2 快速版，生成速度极快，适合批量和实时场景",
+      "capabilities": [
+        "image_generation",
+        "chinese_optimized",
+        "fast"
+      ],
+      "extra": {
+        "endpoint": "/services/aigc/image-generation/generation",
+        "asyncHeader": "X-DashScope-Async",
+        "taskQueryEndpoint": "/tasks",
+        "supportedImageSizes": [
+          "1024*1024",
+          "720*1280",
+          "1280*720"
+        ],
+        "maxN": 4,
+        "supportsNegativePrompt": true
+      }
+    },
+    {
+      "modelName": "qwen-image-2.0-pro",
+      "displayName": "千问图像 2.0 Pro",
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "千问图像生成 2.0 Pro 旗舰版，高画质多风格",
+      "capabilities": [
+        "image_generation",
+        "chinese_optimized",
+        "high_quality"
+      ],
+      "extra": {
+        "endpoint": "/services/aigc/image-generation/generation",
+        "asyncHeader": "X-DashScope-Async",
+        "taskQueryEndpoint": "/tasks",
+        "supportedImageSizes": [
+          "512*512",
+          "1024*1024",
+          "1280*1280",
+          "1664*928",
+          "928*1664"
+        ],
+        "maxN": 4
+      }
+    },
+    {
+      "modelName": "qwen-image-2.0",
+      "displayName": "千问图像 2.0",
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "千问图像生成 2.0 标准版",
+      "capabilities": [
+        "image_generation",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "endpoint": "/services/aigc/image-generation/generation",
+        "asyncHeader": "X-DashScope-Async",
+        "taskQueryEndpoint": "/tasks",
+        "supportedImageSizes": [
+          "512*512",
+          "1024*1024",
+          "1280*1280",
+          "2048*2048"
+        ],
+        "maxN": 4
+      }
+    },
+    {
+      "modelName": "wan3.0-video-prime",
+      "displayName": "通义万相 3.0 视频生成高速版",
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "万相 3.0 全能参考视频生成高速版，能力对齐标准版并显著提升端到端生成速度",
+      "capabilities": [
+        "video_generation",
+        "text_to_video",
+        "image_to_video",
+        "reference_to_video",
+        "multi_image",
+        "reference_video",
+        "reference_audio",
+        "first_last_frame",
+        "audio_generation",
+        "chinese_optimized",
+        "high_quality",
+        "fast"
+      ],
+      "extra": {
+        "endpoint": "/services/aigc/video-generation/video-synthesis",
+        "asyncHeader": "X-DashScope-Async",
+        "taskQueryEndpoint": "/tasks",
+        "maxVideoDuration": 30,
+        "supportsIntelligentDuration": true,
+        "fps": 30,
+        "supportedResolutions": [
+          "480P",
+          "720P",
+          "1080P"
+        ],
+        "supportedRatios": [
+          "adaptive",
+          "16:9",
+          "4:3",
+          "1:1",
+          "3:4",
+          "9:16"
+        ],
+        "mediaInputTypes": [
+          "first_frame",
+          "last_frame",
+          "reference_image",
+          "reference_video",
+          "reference_audio",
+          "file",
+          "link"
+        ],
+        "maxReferenceImages": 10,
+        "maxReferenceVideos": 5,
+        "maxReferenceAudios": 5,
+        "supportsOutputAudio": true,
+        "supportsSeed": true,
+        "supportsPromptExtend": true,
+        "supportsWatermark": true
+      }
+    },
+    {
+      "modelName": "wan3.0-video",
+      "displayName": "通义万相 3.0 视频生成",
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "万相 3.0 全能参考视频生成标准版，统一支持文生、首帧/首尾帧图生及多类型参考生视频",
+      "capabilities": [
+        "video_generation",
+        "text_to_video",
+        "image_to_video",
+        "reference_to_video",
+        "multi_image",
+        "reference_video",
+        "reference_audio",
+        "first_last_frame",
+        "audio_generation",
+        "chinese_optimized",
+        "high_quality"
+      ],
+      "extra": {
+        "endpoint": "/services/aigc/video-generation/video-synthesis",
+        "asyncHeader": "X-DashScope-Async",
+        "taskQueryEndpoint": "/tasks",
+        "maxVideoDuration": 30,
+        "supportsIntelligentDuration": true,
+        "fps": 30,
+        "supportedResolutions": [
+          "480P",
+          "720P",
+          "1080P"
+        ],
+        "supportedRatios": [
+          "adaptive",
+          "16:9",
+          "4:3",
+          "1:1",
+          "3:4",
+          "9:16"
+        ],
+        "mediaInputTypes": [
+          "first_frame",
+          "last_frame",
+          "reference_image",
+          "reference_video",
+          "reference_audio",
+          "file",
+          "link"
+        ],
+        "maxReferenceImages": 10,
+        "maxReferenceVideos": 5,
+        "maxReferenceAudios": 5,
+        "supportsOutputAudio": true,
+        "supportsSeed": true,
+        "supportsPromptExtend": true,
+        "supportsWatermark": true
+      }
+    },
+    {
+      "modelName": "wan2.6-t2v",
+      "displayName": "通义万相 2.6 文生视频",
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "万相 2.6 文生视频模型，支持文生视频和图生视频",
+      "capabilities": [
+        "video_generation",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "maxVideoDuration": 10,
+        "supportedResolutions": [
+          "720p",
+          "1080p"
+        ],
+        "endpoint": "/services/aigc/video-generation/generation",
+        "asyncHeader": "X-DashScope-Async",
+        "taskQueryEndpoint": "/tasks"
+      }
+    },
+    {
+      "modelName": "cosyvoice-clone",
+      "displayName": "CosyVoice 声音复刻",
+      "serviceType": [
+        "voice_clone"
+      ],
+      "description": "通义声音复刻，少量样本即可克隆",
+      "capabilities": [
+        "voice_clone",
+        "chinese_optimized",
+        "few_shot"
+      ],
+      "extra": {
+        "maxCloneSamples": 5,
+        "minCloneDuration": 3
+      }
+    },
+    {
+      "modelName": "qwen-omni-turbo",
+      "displayName": "Qwen-Omni-Turbo",
+      "serviceType": [
+        "omni"
+      ],
+      "description": "通义全能模型，支持音视频文本多模态",
+      "contextWindow": 32768,
+      "maxOutputTokens": 2048,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tts",
+        "asr",
+        "multimodal",
+        "tool_use"
+      ],
+      "defaultTemperature": 0.7,
+      "defaultTopP": 0.01,
+      "extra": {}
+    },
+    {
+      "modelName": "qwen-mt-plus",
+      "displayName": "Qwen-MT-Plus",
+      "serviceType": [
+        "translation"
+      ],
+      "description": "通义千问 Qwen-MT 旗舰翻译模型，支持 92 种语言互译，适合专业领域、正式文书和技术报告",
+      "contextWindow": 16384,
+      "maxOutputTokens": 8192,
+      "capabilities": [
+        "translation",
+        "multilingual",
+        "chinese_optimized",
+        "terminology"
+      ],
+      "inputPrice": 1.8,
+      "outputPrice": 5.4,
+      "extra": {
+        "supportedLanguageCount": 92,
+        "supportsIncrementalStreaming": false
+      }
+    },
+    {
+      "modelName": "happyhorse-1.0-t2v",
+      "displayName": "快乐马 1.0 文生视频",
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "阿里 HappyHorse 文生视频，原生音视频联合生成，最长 15s，支持多语种口型",
+      "capabilities": [
+        "video_generation",
+        "text_to_video",
+        "chinese_optimized",
+        "high_quality"
+      ],
+      "extra": {
+        "endpoint": "/services/aigc/video-generation/video-synthesis",
+        "asyncHeader": "X-DashScope-Async",
+        "taskQueryEndpoint": "/tasks",
+        "maxVideoDuration": 15,
+        "supportedResolutions": [
+          "720P",
+          "1080P"
+        ]
+      }
+    },
+    {
+      "modelName": "happyhorse-1.0-i2v",
+      "displayName": "快乐马 1.0 图生视频",
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "阿里 HappyHorse 图生视频（首帧），宽高比自动跟随输入首帧",
+      "capabilities": [
+        "video_generation",
+        "image_to_video",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "endpoint": "/services/aigc/video-generation/video-synthesis",
+        "asyncHeader": "X-DashScope-Async",
+        "taskQueryEndpoint": "/tasks",
+        "maxVideoDuration": 15,
+        "supportedResolutions": [
+          "720P",
+          "1080P"
+        ]
+      }
+    },
+    {
+      "modelName": "happyhorse-1.0-r2v",
+      "displayName": "快乐马 1.0 参考生视频",
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "阿里 HappyHorse 参考生视频，支持最多 9 张参考图（多主体/元素一致性）",
+      "capabilities": [
+        "video_generation",
+        "reference_to_video",
+        "multi_image",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "endpoint": "/services/aigc/video-generation/video-synthesis",
+        "asyncHeader": "X-DashScope-Async",
+        "taskQueryEndpoint": "/tasks",
+        "maxReferenceImages": 9,
+        "maxVideoDuration": 15,
+        "supportedResolutions": [
+          "720P",
+          "1080P"
+        ]
+      }
+    }
+  ]
+}

compute/v2/compatibility/providers/deepseek.json

@@ -0,0 +1,219 @@
+{
+  "id": "provider-deepseek-001",
+  "provider": "deepseek",
+  "label": "DeepSeek",
+  "baseUrl": "https://api.deepseek.com",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "deepseek",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "CNY",
+  "services": [
+    "chat",
+    "reasoning",
+    "vision"
+  ],
+  "tombstones": [
+    "deepseek-chat",
+    "deepseek-reasoner"
+  ],
+  "models": [
+    {
+      "modelName": "deepseek-flash",
+      "displayName": "DeepSeek V4.1 Flash",
+      "serviceType": [
+        "chat",
+        "reasoning",
+        "vision"
+      ],
+      "description": "DeepSeek V4.1 Flash，支持文本与图像输入，1M 上下文，最大输出 384K，并发上限 2500",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 384000,
+      "capabilities": [
+        "chat",
+        "code",
+        "reasoning",
+        "deep_thinking",
+        "vision",
+        "long_context",
+        "multilingual",
+        "tool_use"
+      ],
+      "inputPrice": 1,
+      "outputPrice": 4,
+      "defaultTemperature": 1,
+      "defaultTopP": 1,
+      "extra": {
+        "cacheHitPrice": 0.02,
+        "concurrencyLimit": 2500,
+        "supportsThinking": true,
+        "thinkingDefault": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "high"
+        },
+        "pricingTiers": [
+          {
+            "condition": "idle",
+            "inputPrice": 1,
+            "outputPrice": 4,
+            "cacheHitPrice": 0.02
+          },
+          {
+            "condition": "peak",
+            "inputPrice": 2,
+            "outputPrice": 8,
+            "cacheHitPrice": 0.04
+          }
+        ],
+        "pricingNotes": "空闲时段为北京时间工作日 09:00-12:00、14:00-18:00 之外的时段；高峰时段价格为输入 2 元、输出 8 元、缓存命中输入 0.04 元/百万 tokens。",
+        "features": {
+          "jsonOutput": true,
+          "toolCalls": true,
+          "chatPrefixCompletion": "beta",
+          "fimCompletion": "non-thinking-only"
+        }
+      }
+    },
+    {
+      "modelName": "deepseek-v4-flash-vision-exp",
+      "displayName": "DeepSeek V4 Flash Vision Exp",
+      "serviceType": [
+        "chat",
+        "reasoning",
+        "vision"
+      ],
+      "description": "DeepSeek V4 Flash 实验性视觉版本，支持文本与图像输入，文本能力与 V4 Flash 一致，1M 上下文，并发上限 2500",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 384000,
+      "capabilities": [
+        "chat",
+        "code",
+        "reasoning",
+        "deep_thinking",
+        "vision",
+        "long_context",
+        "multilingual",
+        "tool_use"
+      ],
+      "inputPrice": 1.008,
+      "outputPrice": 2.016,
+      "defaultTemperature": 1,
+      "defaultTopP": 1,
+      "extra": {
+        "cacheHitPrice": 0.02016,
+        "concurrencyLimit": 2500,
+        "supportsThinking": true,
+        "thinkingDefault": true,
+        "features": {
+          "jsonOutput": true,
+          "toolCalls": true,
+          "chatPrefixCompletion": "beta",
+          "fimCompletion": false
+        },
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "high"
+        }
+      }
+    },
+    {
+      "modelName": "deepseek-v4-flash",
+      "displayName": "DeepSeek V4 Flash",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "DeepSeek V4 Flash，284B 总参数 / 13B 激活 MoE，1M 上下文，默认思考模式，支持非思考模式，并发上限 2500",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 384000,
+      "capabilities": [
+        "chat",
+        "code",
+        "reasoning",
+        "deep_thinking",
+        "long_context",
+        "multilingual",
+        "tool_use"
+      ],
+      "inputPrice": 1.008,
+      "outputPrice": 2.016,
+      "defaultTemperature": 1,
+      "defaultTopP": 1,
+      "extra": {
+        "cacheHitPrice": 0.02016,
+        "concurrencyLimit": 2500,
+        "supportsThinking": true,
+        "thinkingDefault": true,
+        "features": {
+          "jsonOutput": true,
+          "toolCalls": true,
+          "chatPrefixCompletion": "beta",
+          "fimCompletion": "non-thinking-only"
+        },
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "high"
+        }
+      }
+    },
+    {
+      "modelName": "deepseek-v4-pro",
+      "displayName": "DeepSeek V4 Pro",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "DeepSeek V4 Pro，1.6T 总参数 / 49B 激活 MoE，1M 上下文，默认思考模式，支持非思考模式，并发上限 500",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 384000,
+      "capabilities": [
+        "chat",
+        "code",
+        "reasoning",
+        "deep_thinking",
+        "long_context",
+        "multilingual",
+        "tool_use",
+        "math"
+      ],
+      "inputPrice": 3.132,
+      "outputPrice": 6.264,
+      "defaultTemperature": 1,
+      "defaultTopP": 1,
+      "extra": {
+        "cacheHitPrice": 0.0261,
+        "concurrencyLimit": 500,
+        "supportsThinking": true,
+        "thinkingDefault": true,
+        "features": {
+          "jsonOutput": true,
+          "toolCalls": true,
+          "chatPrefixCompletion": "beta",
+          "fimCompletion": "non-thinking-only"
+        },
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "high"
+        }
+      }
+    }
+  ]
+}

compute/v2/compatibility/providers/github-copilot.json

@@ -0,0 +1,28 @@
+{
+  "id": "provider-github-copilot-plan-001",
+  "provider": "github-copilot",
+  "brandGroup": "github",
+  "label": "GitHub Copilot 订阅",
+  "baseUrl": "https://api.githubcopilot.com",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "github-copilot-session-token",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "USD",
+  "accessMode": "coding-plan",
+  "credentialSource": "github-copilot",
+  "requiredClientVersion": "10.0.132",
+  "codingPlan": {
+    "quotas": {},
+    "usageTracking": {
+      "method": "none",
+      "consoleUrl": "https://github.com/settings/copilot/features"
+    }
+  },
+  "services": [
+    "chat",
+    "reasoning"
+  ],
+  "models": []
+}

compute/v2/compatibility/providers/google.json

@@ -0,0 +1,331 @@
+{
+  "id": "provider-google-001",
+  "provider": "google",
+  "label": "Google Gemini",
+  "baseUrl": "https://generativelanguage.googleapis.com/v1beta",
+  "apiFormat": "google-generative-ai",
+  "apiKeyRef": "google",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "USD",
+  "services": [
+    "chat",
+    "vision",
+    "embedding"
+  ],
+  "tombstones": [
+    "gemini-3.1-flash-lite-preview"
+  ],
+  "models": [
+    {
+      "modelName": "gemini-embedding-2",
+      "displayName": "Gemini Embedding 2",
+      "contextWindow": 8192,
+      "capabilities": [
+        "text_embedding",
+        "vision",
+        "audio_understanding",
+        "video_understanding",
+        "multilingual"
+      ],
+      "serviceType": [
+        "embedding"
+      ],
+      "description": "Gemini Embedding 2，原生多模态嵌入，8192 token 输入，共享文本/图像/音视频向量空间",
+      "extra": {
+        "dimensions": [
+          768,
+          1536,
+          3072
+        ],
+        "defaultDimension": 3072
+      }
+    },
+    {
+      "modelName": "gemini-3.8-flash",
+      "displayName": "Gemini 3.8 Flash",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Google 最新稳定版 Flash，面向长程编程和智能体任务，1M 上下文",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "ultra_long_context",
+        "tool_use",
+        "fast"
+      ],
+      "inputPrice": 0.75,
+      "outputPrice": 3.75,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cachePricing": {
+          "inputCacheRead": 0.075
+        },
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high"
+          ],
+          "defaultEffort": "medium"
+        },
+        "pricingNotes": "Standard paid tier introductory price through 2026-12-31; standard rates rise on 2027-01-01."
+      }
+    },
+    {
+      "modelName": "gemini-3.7-flash",
+      "displayName": "Gemini 3.7 Flash",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Google 稳定版 Flash，面向日常编程和多步工具调用，1M 上下文",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "ultra_long_context",
+        "tool_use",
+        "fast"
+      ],
+      "inputPrice": 0.75,
+      "outputPrice": 3.75,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cachePricing": {
+          "inputCacheRead": 0.075
+        },
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high"
+          ],
+          "defaultEffort": "medium"
+        },
+        "pricingNotes": "Standard paid tier introductory price through 2026-12-31; standard rates rise on 2027-01-01."
+      }
+    },
+    {
+      "modelName": "gemini-3.5-flash-lite",
+      "displayName": "Gemini 3.5 Flash-Lite",
+      "serviceType": [
+        "chat",
+        "fast"
+      ],
+      "description": "Google 低延迟轻量模型，支持多模态输入，1M 上下文",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "ultra_long_context",
+        "tool_use",
+        "fast"
+      ],
+      "inputPrice": 0.3,
+      "outputPrice": 2.5,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cachePricing": {
+          "inputCacheRead": 0.03
+        }
+      }
+    },
+    {
+      "modelName": "gemini-3.5-flash",
+      "displayName": "Gemini 3.5 Flash",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Google Gemini 3.5 Flash 稳定版，高速多模态模型，1M 上下文",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "ultra_long_context",
+        "tool_use",
+        "fast"
+      ],
+      "inputPrice": 1.5,
+      "outputPrice": 9,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cachePricing": {
+          "inputCacheRead": 0.75
+        },
+        "pricingNotes": "Standard paid tier text/image/video price; audio input uses a separate Google pricing row."
+      }
+    },
+    {
+      "modelName": "gemini-3.1-flash-lite",
+      "displayName": "Gemini 3.1 Flash-Lite",
+      "serviceType": [
+        "fast"
+      ],
+      "description": "Google Gemini 3.1 Flash-Lite 稳定版，低成本高吞吐多模态模型，替代已关闭的 Preview 版本",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "ultra_long_context",
+        "tool_use",
+        "fast"
+      ],
+      "inputPrice": 0.25,
+      "outputPrice": 1.5,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "audioInputPrice": 0.5,
+        "cachePricing": {
+          "inputCacheRead": 0.125,
+          "audioInputCacheRead": 0.25
+        }
+      }
+    },
+    {
+      "modelName": "gemini-3.1-pro-preview",
+      "displayName": "Gemini 3.1 Pro Preview",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Google Gemini 3.1 Pro Preview，面向复杂推理、编码和智能体工作流",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "ultra_long_context",
+        "tool_use"
+      ],
+      "inputPrice": 2,
+      "outputPrice": 12,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "pricingNotes": "Standard paid tier text/image/video price for prompts <= 200K tokens; prompts > 200K are $4 input and $18 output per 1M tokens.",
+        "cachePricing": {
+          "inputCacheRead": 0.2,
+          "inputCacheReadOver200k": 0.4,
+          "storagePerMillionTokensPerHour": 4.5
+        }
+      }
+    },
+    {
+      "modelName": "gemini-3-flash-preview",
+      "displayName": "Gemini 3 Flash Preview",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Google Gemini 3 Flash Preview，高速多模态推理模型",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "ultra_long_context",
+        "tool_use",
+        "fast"
+      ],
+      "inputPrice": 0.5,
+      "outputPrice": 3,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "pricingNotes": "Standard paid tier text/image/video price; audio input is $1.00 per 1M tokens.",
+        "cachePricing": {
+          "inputCacheRead": 0.05,
+          "audioInputCacheRead": 0.1,
+          "storagePerMillionTokensPerHour": 1
+        }
+      }
+    },
+    {
+      "modelName": "gemini-2.5-pro",
+      "displayName": "Gemini 2.5 Pro",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Google Gemini 2.5 Pro，1,048,576 token 上下文，多模态",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "ultra_long_context",
+        "tool_use"
+      ],
+      "inputPrice": 1.25,
+      "outputPrice": 10,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {}
+    },
+    {
+      "modelName": "gemini-2.5-flash",
+      "displayName": "Gemini 2.5 Flash",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Google 高性价比模型，1,048,576 token 上下文，速度快",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "ultra_long_context",
+        "tool_use",
+        "fast"
+      ],
+      "inputPrice": 0.3,
+      "outputPrice": 2.5,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {}
+    },
+    {
+      "modelName": "text-embedding-005",
+      "displayName": "Text Embedding 005",
+      "serviceType": [
+        "embedding"
+      ],
+      "description": "Google 文本嵌入模型，768维度",
+      "contextWindow": 2048,
+      "capabilities": [
+        "text_embedding",
+        "semantic_search"
+      ],
+      "inputPrice": 0.1,
+      "extra": {}
+    }
+  ]
+}

compute/v2/compatibility/providers/kling.json

@@ -0,0 +1,112 @@
+{
+  "id": "provider-kling-001",
+  "provider": "kling",
+  "label": "快手 可灵",
+  "baseUrl": "https://api.klingai.com/v1",
+  "apiFormat": "kling-task-api",
+  "apiKeyRef": "kling",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "CNY",
+  "services": [
+    "video_gen"
+  ],
+  "models": [
+    {
+      "modelName": "kling-v3",
+      "displayName": "可灵 3.0",
+      "serviceType": [
+        "video_gen"
+      ],
+      "capabilities": [
+        "video_generation",
+        "image_to_video"
+      ],
+      "description": "可灵 3.0，官方 legacy /v1/videos/text2video 协议已支持该 model_name"
+    },
+    {
+      "modelName": "kling-v2-5-turbo",
+      "displayName": "可灵 V2.5 Turbo",
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "可灵最新视频生成模型，高性价比",
+      "capabilities": [
+        "video_generation",
+        "chinese_optimized",
+        "high_quality",
+        "fast"
+      ],
+      "extra": {
+        "maxVideoDuration": 10,
+        "supportedResolutions": [
+          "720p",
+          "1080p"
+        ],
+        "pricePerGeneration": 1.5
+      }
+    },
+    {
+      "modelName": "kling-v2-5-turbo-pro",
+      "displayName": "可灵 V2.5 Turbo Pro",
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "可灵高品质视频生成",
+      "capabilities": [
+        "video_generation",
+        "chinese_optimized",
+        "ultra_quality"
+      ],
+      "extra": {
+        "maxVideoDuration": 10,
+        "supportedResolutions": [
+          "1080p",
+          "4k"
+        ],
+        "pricePerGeneration": 2.5
+      }
+    },
+    {
+      "modelName": "kling-v2",
+      "displayName": "可灵 V2",
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "可灵文生视频旗舰模型",
+      "capabilities": [
+        "video_generation",
+        "chinese_optimized",
+        "high_quality"
+      ],
+      "extra": {
+        "maxVideoDuration": 10,
+        "supportedResolutions": [
+          "720p",
+          "1080p"
+        ]
+      }
+    },
+    {
+      "modelName": "kling-v2-master",
+      "displayName": "可灵 V2 Master",
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "可灵最高品质视频生成",
+      "capabilities": [
+        "video_generation",
+        "chinese_optimized",
+        "ultra_quality"
+      ],
+      "extra": {
+        "maxVideoDuration": 10,
+        "supportedResolutions": [
+          "1080p",
+          "4k"
+        ]
+      }
+    }
+  ]
+}

compute/v2/compatibility/providers/local-whisper.json

@@ -0,0 +1,40 @@
+{
+  "id": "provider-local-whisper",
+  "provider": "local-whisper",
+  "label": "本地 Whisper",
+  "baseUrl": "http://localhost:8080",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "USD",
+  "services": [
+    "asr"
+  ],
+  "models": [
+    {
+      "modelName": "whisper-large-v3",
+      "displayName": "Whisper Large V3",
+      "serviceType": [
+        "asr"
+      ],
+      "description": "本地部署的 Whisper 大模型",
+      "capabilities": [
+        "asr",
+        "multilingual"
+      ],
+      "extra": {
+        "maxAudioLength": 7200,
+        "supportedInputFormats": [
+          "wav",
+          "mp3",
+          "flac",
+          "ogg",
+          "webm"
+        ],
+        "streamingSupported": true
+      }
+    }
+  ]
+}

compute/v2/compatibility/providers/minimax.json

@@ -0,0 +1,728 @@
+{
+  "id": "provider-minimax-001",
+  "provider": "minimax",
+  "label": "MiniMax",
+  "baseUrl": "https://api.minimaxi.com/anthropic",
+  "apiFormat": "anthropic-messages",
+  "mediaBaseUrl": "https://api.minimaxi.com/v1",
+  "apiKeyRef": "minimax",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "CNY",
+  "services": [
+    "chat",
+    "fast",
+    "reasoning",
+    "vision",
+    "image_gen",
+    "tts",
+    "video_gen",
+    "music_gen"
+  ],
+  "tombstones": [
+    "MiniMax-Text-01"
+  ],
+  "models": [
+    {
+      "modelName": "MiniMax-M3",
+      "displayName": "MiniMax M3",
+      "serviceType": [
+        "chat",
+        "reasoning",
+        "vision"
+      ],
+      "description": "MiniMax M3，前沿多模态 Coding 与 Agent 模型，1M 上下文",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "tool_use",
+        "agent",
+        "vision",
+        "long_context"
+      ],
+      "inputPrice": 2.1,
+      "outputPrice": 8.4,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "pricingTiers": [
+          {
+            "maxInputTokens": 512000,
+            "inputPrice": 2.1,
+            "outputPrice": 8.4,
+            "cacheReadPrice": 0.42
+          },
+          {
+            "maxInputTokens": 1000000,
+            "inputPrice": 4.2,
+            "outputPrice": 16.8,
+            "cacheReadPrice": 0.84
+          }
+        ],
+        "priorityPricingTiers": [
+          {
+            "maxInputTokens": 512000,
+            "inputPrice": 3.15,
+            "outputPrice": 12.6,
+            "cacheReadPrice": 0.63
+          },
+          {
+            "maxInputTokens": 1000000,
+            "inputPrice": 6.3,
+            "outputPrice": 25.2,
+            "cacheReadPrice": 1.26
+          }
+        ],
+        "pricingNotes": "MiniMax 国内开放平台按量计费永久五折价格。"
+      }
+    },
+    {
+      "modelName": "MiniMax-M2.7",
+      "displayName": "MiniMax M2.7",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "MiniMax 新一代旗舰模型，擅长复杂 Agent、软件工程与专业办公任务",
+      "contextWindow": 204800,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "tool_use"
+      ],
+      "inputPrice": 2.1,
+      "outputPrice": 8.4,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cacheReadPrice": 0.42,
+        "cacheWritePrice": 2.625,
+        "outputSpeedTps": 60
+      }
+    },
+    {
+      "modelName": "MiniMax-M2.7-highspeed",
+      "displayName": "MiniMax M2.7 高速版",
+      "serviceType": [
+        "fast"
+      ],
+      "description": "MiniMax M2.7 极速版，效果一致，输出更快，适合低延迟场景",
+      "contextWindow": 204800,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "tool_use",
+        "fast"
+      ],
+      "inputPrice": 4.2,
+      "outputPrice": 16.8,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cacheReadPrice": 0.42,
+        "cacheWritePrice": 2.625,
+        "outputSpeedTps": 100
+      }
+    },
+    {
+      "modelName": "MiniMax-M2.5",
+      "displayName": "MiniMax M2.5",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "MiniMax M2.5，204.8K 上下文，支持深度推理和工具调用",
+      "contextWindow": 204800,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "tool_use",
+        "long_context"
+      ],
+      "inputPrice": 2.1,
+      "outputPrice": 8.4,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cacheReadPrice": 0.21,
+        "cacheWritePrice": 2.625,
+        "outputSpeedTps": 60
+      }
+    },
+    {
+      "modelName": "MiniMax-M2.5-highspeed",
+      "displayName": "MiniMax M2.5 高速版",
+      "serviceType": [
+        "fast"
+      ],
+      "description": "MiniMax M2.5 高速推理版本(100TPS)，适合低延迟场景",
+      "contextWindow": 204800,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "tool_use",
+        "fast"
+      ],
+      "inputPrice": 4.2,
+      "outputPrice": 16.8,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cacheReadPrice": 0.21,
+        "cacheWritePrice": 2.625,
+        "outputSpeedTps": 100
+      }
+    },
+    {
+      "modelName": "MiniMax-M2.1",
+      "displayName": "MiniMax M2.1",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "MiniMax 上一代主力模型，200K 上下文，性价比优秀",
+      "contextWindow": 204800,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "code",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 2.1,
+      "outputPrice": 8.4,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cacheReadPrice": 0.21,
+        "cacheWritePrice": 2.625,
+        "outputSpeedTps": 60
+      }
+    },
+    {
+      "modelName": "MiniMax-M2.1-highspeed",
+      "displayName": "MiniMax M2.1 高速版",
+      "serviceType": [
+        "fast"
+      ],
+      "description": "MiniMax M2.1 高速推理版本，适合低延迟场景",
+      "contextWindow": 204800,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "tool_use",
+        "fast"
+      ],
+      "inputPrice": 4.2,
+      "outputPrice": 16.8,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cacheReadPrice": 0.21,
+        "cacheWritePrice": 2.625,
+        "outputSpeedTps": 100
+      }
+    },
+    {
+      "modelName": "MiniMax-M2",
+      "displayName": "MiniMax M2",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "MiniMax M2，优化编码和 Agent 工作流",
+      "contextWindow": 204800,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "code",
+        "tool_use"
+      ],
+      "inputPrice": 2.1,
+      "outputPrice": 8.4,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "legacy": true,
+        "cacheReadPrice": 0.21,
+        "cacheWritePrice": 2.625
+      }
+    },
+    {
+      "modelName": "M2-her",
+      "displayName": "MiniMax M2 Her",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "MiniMax 角色扮演对话模型，专为多轮角色对话设计",
+      "contextWindow": 204800,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "roleplay"
+      ],
+      "inputPrice": 2.1,
+      "outputPrice": 8.4,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cacheReadPrice": 0.21,
+        "cacheWritePrice": 2.625
+      }
+    },
+    {
+      "modelName": "image-01",
+      "displayName": "MiniMax Image 01",
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "MiniMax 文生图/图生图模型，支持主题参考（角色一致性），最长 1500 字符 prompt",
+      "capabilities": [
+        "image_generation",
+        "subject_reference",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "supportedAspectRatios": [
+          "1:1",
+          "16:9",
+          "9:16",
+          "4:3",
+          "3:4",
+          "3:2",
+          "2:3",
+          "21:9"
+        ],
+        "maxPromptLength": 1500,
+        "maxBatchSize": 9,
+        "responseFormats": [
+          "url",
+          "base64"
+        ],
+        "urlExpiry": "24h",
+        "pricePerImage": 0.025
+      }
+    },
+    {
+      "modelName": "image-01-live",
+      "displayName": "MiniMax Image 01 Live",
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "MiniMax 文生图模型（Live 版），支持画风控制（漫画/元气/中世纪/水彩）",
+      "capabilities": [
+        "image_generation",
+        "style_control",
+        "subject_reference",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "supportedAspectRatios": [
+          "1:1",
+          "16:9",
+          "9:16",
+          "4:3",
+          "3:4",
+          "3:2",
+          "2:3"
+        ],
+        "supportedStyles": [
+          "漫画",
+          "元气",
+          "中世纪",
+          "水彩"
+        ],
+        "maxPromptLength": 1500,
+        "maxBatchSize": 9,
+        "responseFormats": [
+          "url",
+          "base64"
+        ],
+        "urlExpiry": "24h"
+      }
+    },
+    {
+      "modelName": "speech-2.8-hd",
+      "displayName": "MiniMax Speech 2.8 HD",
+      "serviceType": [
+        "tts"
+      ],
+      "description": "MiniMax 最新高清语音合成模型，支持 40+ 语言、100+ 预置音色，毫秒级延迟",
+      "capabilities": [
+        "tts",
+        "multilingual",
+        "emotion_control",
+        "voice_clone",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "maxTextLength": 10000,
+        "supportedFormats": [
+          "mp3",
+          "wav",
+          "pcm",
+          "flac",
+          "aac",
+          "ogg"
+        ],
+        "pricePerMillionCharacters": 350
+      }
+    },
+    {
+      "modelName": "speech-2.8-turbo",
+      "displayName": "MiniMax Speech 2.8 Turbo",
+      "serviceType": [
+        "tts"
+      ],
+      "description": "MiniMax 快速语音合成模型，速度优先，适合实时场景",
+      "capabilities": [
+        "tts",
+        "multilingual",
+        "emotion_control",
+        "fast",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "maxTextLength": 10000,
+        "supportedFormats": [
+          "mp3",
+          "wav",
+          "pcm",
+          "flac",
+          "aac",
+          "ogg"
+        ],
+        "pricePerMillionCharacters": 200
+      }
+    },
+    {
+      "modelName": "speech-2.6-hd",
+      "displayName": "MiniMax Speech 2.6 HD",
+      "serviceType": [
+        "tts"
+      ],
+      "description": "MiniMax 高清语音合成模型（2.6 版），音质优秀，生成速度快",
+      "capabilities": [
+        "tts",
+        "multilingual",
+        "emotion_control",
+        "voice_clone",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "maxTextLength": 10000,
+        "supportedFormats": [
+          "mp3",
+          "wav",
+          "pcm",
+          "flac",
+          "aac",
+          "ogg"
+        ],
+        "pricePerMillionCharacters": 350
+      }
+    },
+    {
+      "modelName": "speech-2.6-turbo",
+      "displayName": "MiniMax Speech 2.6 Turbo",
+      "serviceType": [
+        "tts"
+      ],
+      "description": "MiniMax 快速语音合成模型（2.6 版），低延迟响应",
+      "capabilities": [
+        "tts",
+        "multilingual",
+        "emotion_control",
+        "fast",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "maxTextLength": 10000,
+        "supportedFormats": [
+          "mp3",
+          "wav",
+          "pcm",
+          "flac",
+          "aac",
+          "ogg"
+        ],
+        "pricePerMillionCharacters": 200
+      }
+    },
+    {
+      "modelName": "speech-02-hd",
+      "displayName": "MiniMax Speech 02 HD",
+      "serviceType": [
+        "tts"
+      ],
+      "description": "MiniMax 高清语音合成模型（历史版本），韵律稳定性优秀",
+      "capabilities": [
+        "tts",
+        "multilingual",
+        "voice_clone",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "maxTextLength": 10000,
+        "supportedFormats": [
+          "mp3",
+          "wav",
+          "pcm",
+          "flac",
+          "aac",
+          "ogg"
+        ],
+        "legacy": true,
+        "pricePerMillionCharacters": 350
+      }
+    },
+    {
+      "modelName": "speech-02-turbo",
+      "displayName": "MiniMax Speech 02 Turbo",
+      "serviceType": [
+        "tts"
+      ],
+      "description": "MiniMax 快速语音合成模型（历史版本），增强小语种支持",
+      "capabilities": [
+        "tts",
+        "multilingual",
+        "fast",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "maxTextLength": 10000,
+        "supportedFormats": [
+          "mp3",
+          "wav",
+          "pcm",
+          "flac",
+          "aac",
+          "ogg"
+        ],
+        "legacy": true,
+        "pricePerMillionCharacters": 200
+      }
+    },
+    {
+      "modelName": "MiniMax-Hailuo-2.3",
+      "displayName": "海螺视频 2.3",
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "MiniMax 最新视频生成模型，支持文生视频和图生视频，支持 15 种运镜指令控制",
+      "capabilities": [
+        "video_generation",
+        "text_to_video",
+        "image_to_video",
+        "camera_control",
+        "chinese_optimized",
+        "high_quality"
+      ],
+      "extra": {
+        "maxVideoDuration": 10,
+        "supportedResolutions": [
+          "768p",
+          "1080p"
+        ],
+        "maxPromptLength": 2000,
+        "asyncMode": true
+      }
+    },
+    {
+      "modelName": "MiniMax-Hailuo-2.3-Fast",
+      "displayName": "海螺视频 2.3 快速版",
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "海螺视频快速版本（仅图生视频），生成速度更快",
+      "capabilities": [
+        "video_generation",
+        "image_to_video",
+        "camera_control",
+        "fast",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "maxVideoDuration": 10,
+        "supportedResolutions": [
+          "768p",
+          "1080p"
+        ],
+        "asyncMode": true
+      }
+    },
+    {
+      "modelName": "T2V-01-Director",
+      "displayName": "海螺视频 Director",
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "导演级文生视频模型，精准运镜控制（历史版本）",
+      "capabilities": [
+        "video_generation",
+        "text_to_video",
+        "camera_control",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "maxVideoDuration": 6,
+        "supportedResolutions": [
+          "720p"
+        ],
+        "asyncMode": true,
+        "legacy": true
+      }
+    },
+    {
+      "modelName": "MiniMax-Hailuo-02",
+      "displayName": "海螺视频 02",
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "海螺视频 02，原生 1080P 生成，指令跟随能力强（历史版本）",
+      "capabilities": [
+        "video_generation",
+        "text_to_video",
+        "image_to_video",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "maxVideoDuration": 10,
+        "supportedResolutions": [
+          "512p",
+          "768p",
+          "1080p"
+        ],
+        "asyncMode": true,
+        "legacy": true
+      }
+    },
+    {
+      "modelName": "S2V-01",
+      "displayName": "海螺视频 主体参考",
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "主体参考视频生成模型，基于参考图保持角色一致性",
+      "capabilities": [
+        "video_generation",
+        "subject_reference",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "asyncMode": true
+      }
+    },
+    {
+      "modelName": "music-2.6",
+      "displayName": "MiniMax Music 2.6",
+      "serviceType": [
+        "music_gen"
+      ],
+      "description": "MiniMax AI 音乐生成模型，支持音乐灵感描述和歌词输入，可生成带人声或纯音乐；音乐付费 API 自2026-08-20起仅历史付费用户可继续使用，不向新用户开放；免费音乐接口已停止服务",
+      "capabilities": [
+        "music_generation",
+        "lyrics_input",
+        "instrumental",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "maxPromptLength": 2000,
+        "maxLyricsLength": 3500,
+        "outputFormats": [
+          "hex",
+          "url"
+        ],
+        "supportedLyricsStructures": [
+          "[Intro]",
+          "[Verse]",
+          "[Pre Chorus]",
+          "[Chorus]",
+          "[Interlude]",
+          "[Bridge]",
+          "[Outro]",
+          "[Hook]",
+          "[Solo]",
+          "[Inst]"
+        ],
+        "pricePerSongUpToFiveMinutes": 1
+      }
+    },
+    {
+      "modelName": "music-2.5+",
+      "displayName": "MiniMax Music 2.5+",
+      "serviceType": [
+        "music_gen"
+      ],
+      "description": "MiniMax AI 音乐生成模型（2.5+ 版），音质提升；音乐付费 API 自2026-08-20起仅历史付费用户可继续使用，不向新用户开放；免费音乐接口已停止服务",
+      "capabilities": [
+        "music_generation",
+        "lyrics_input",
+        "instrumental",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "outputFormats": [
+          "hex",
+          "url"
+        ],
+        "pricePerSongUpToFiveMinutes": 1
+      }
+    },
+    {
+      "modelName": "music-2.5",
+      "displayName": "MiniMax Music 2.5",
+      "serviceType": [
+        "music_gen"
+      ],
+      "description": "MiniMax AI 音乐生成模型（2.5 版）；音乐付费 API 自2026-08-20起仅历史付费用户可继续使用，不向新用户开放；免费音乐接口已停止服务",
+      "capabilities": [
+        "music_generation",
+        "lyrics_input",
+        "instrumental",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "outputFormats": [
+          "hex",
+          "url"
+        ],
+        "pricePerSongUpToFiveMinutes": 1
+      }
+    },
+    {
+      "modelName": "music-cover",
+      "displayName": "MiniMax Music Cover",
+      "serviceType": [
+        "music_gen"
+      ],
+      "description": "MiniMax AI 翻唱模型，基于参考音频生成翻唱版本，支持修改歌词；音乐付费 API 自2026-08-20起仅历史付费用户可继续使用，不向新用户开放；免费音乐接口已停止服务",
+      "capabilities": [
+        "music_generation",
+        "cover",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "maxAudioDuration": 360,
+        "maxAudioSize": "50MB",
+        "supportedFormats": [
+          "mp3",
+          "wav",
+          "flac"
+        ]
+      }
+    }
+  ]
+}

compute/v2/compatibility/providers/mistral.json

@@ -0,0 +1,125 @@
+{
+  "id": "provider-mistral-001",
+  "provider": "mistral",
+  "label": "Mistral AI",
+  "baseUrl": "https://api.mistral.ai/v1",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "mistral",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "USD",
+  "services": [
+    "chat",
+    "fast"
+  ],
+  "models": [
+    {
+      "modelName": "mistral-medium-3-5",
+      "displayName": "Mistral Medium 3.5",
+      "contextWindow": 262144,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use"
+      ],
+      "serviceType": [
+        "chat"
+      ],
+      "defaultTemperature": 1,
+      "description": "Mistral Medium 3.5，官方 API ID mistral-medium-3-5，256K 上下文",
+      "inputPrice": 1.5,
+      "outputPrice": 7.5
+    },
+    {
+      "modelName": "mistral-medium-latest",
+      "displayName": "Mistral Medium 3.5",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Mistral 前沿多模态模型，面向 Agentic 和 Coding 场景优化，256K 上下文",
+      "contextWindow": 262144,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "inputPrice": 1.5,
+      "outputPrice": 7.5,
+      "defaultTemperature": 0.7,
+      "defaultTopP": 1,
+      "extra": {}
+    },
+    {
+      "modelName": "mistral-large-latest",
+      "displayName": "Mistral Large 3",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Mistral 旗舰模型，256K 上下文，支持视觉和工具调用",
+      "contextWindow": 256000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "long_context"
+      ],
+      "inputPrice": 0.5,
+      "outputPrice": 1.5,
+      "defaultTemperature": 0.7,
+      "defaultTopP": 1,
+      "extra": {}
+    },
+    {
+      "modelName": "mistral-small-latest",
+      "displayName": "Mistral Small 4",
+      "serviceType": [
+        "fast"
+      ],
+      "description": "Mistral Small 4，指令、推理与编码混合模型，256K 上下文",
+      "contextWindow": 262144,
+      "capabilities": [
+        "chat",
+        "code",
+        "vision",
+        "reasoning",
+        "fast",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "inputPrice": 0.15,
+      "outputPrice": 0.6,
+      "defaultTemperature": 0.7,
+      "defaultTopP": 1,
+      "extra": {}
+    },
+    {
+      "modelName": "codestral-latest",
+      "displayName": "Codestral",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Mistral 专业代码模型，支持 Fill-in-the-Middle 补全",
+      "contextWindow": 128000,
+      "maxOutputTokens": 32768,
+      "capabilities": [
+        "code",
+        "fill_in_the_middle"
+      ],
+      "inputPrice": 0.3,
+      "outputPrice": 0.9,
+      "defaultTemperature": 0.7,
+      "defaultTopP": 1,
+      "extra": {}
+    }
+  ]
+}

compute/v2/compatibility/providers/moonshot.json

@@ -0,0 +1,419 @@
+{
+  "id": "provider-moonshot-001",
+  "provider": "moonshot",
+  "label": "月之暗面 Kimi",
+  "baseUrl": "https://api.moonshot.cn/v1",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "moonshot",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "CNY",
+  "services": [
+    "chat",
+    "reasoning",
+    "vision"
+  ],
+  "tombstones": [
+    "kimi-k2",
+    "kimi-k2-0905-preview",
+    "kimi-k2-0711-preview",
+    "kimi-k2-turbo-preview",
+    "kimi-k2-thinking",
+    "kimi-k2-thinking-turbo",
+    "kimi-latest",
+    "kimi-thinking-preview"
+  ],
+  "models": [
+    {
+      "modelName": "kimi-k3",
+      "displayName": "Kimi K3",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 1048576,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "tool_use",
+        "agent",
+        "long_context",
+        "vision",
+        "video_understanding"
+      ],
+      "serviceType": [
+        "reasoning"
+      ],
+      "description": "Kimi K3，1M 上下文，始终开启思考，low/high/max 三档（默认 max）；输出上限与输入共享窗口，默认 max_completion_tokens=131072",
+      "extra": {
+        "thinkingOnly": true,
+        "thinking": {
+          "default": "enabled",
+          "disableSupported": false,
+          "preserveReasoningContent": true
+        },
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "max"
+        },
+        "pricingNotes": "CNY per 1M tokens: input 20, output 100, cache hit 2. Cache writes: TTL 5min 20, TTL 1h 40; recorded separately from cache-hit billing.",
+        "samplingParametersDeprecated": true,
+        "cacheHitPrice": 2
+      },
+      "inputPrice": 20,
+      "outputPrice": 100
+    },
+    {
+      "modelName": "kimi-k2.7-code",
+      "displayName": "Kimi K2.7 Code",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "月之暗面 Kimi K2.7 Code，当前最强 Coding 模型，256K 上下文，支持文本/图片/视频输入，始终启用思考模式",
+      "contextWindow": 262144,
+      "maxOutputTokens": 32768,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "tool_use",
+        "agent",
+        "long_context",
+        "vision",
+        "video_understanding",
+        "image_understanding"
+      ],
+      "inputPrice": 6.5,
+      "outputPrice": 27,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "thinking": {
+          "default": "enabled",
+          "disableSupported": false,
+          "preserveReasoningContent": true
+        },
+        "cacheHitPrice": 1.3,
+        "supportedImageFormats": [
+          "png",
+          "jpeg",
+          "webp",
+          "gif"
+        ],
+        "supportedVideoFormats": [
+          "mp4",
+          "mpeg",
+          "mov",
+          "avi",
+          "x-flv",
+          "mpg",
+          "webm",
+          "wmv",
+          "3gpp"
+        ]
+      }
+    },
+    {
+      "modelName": "kimi-k2.7-code-highspeed",
+      "displayName": "Kimi K2.7 Code HighSpeed",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "月之暗面 Kimi K2.7 Code 高速版，与 K2.7 Code 参数一致，输出速度约 180 Tokens/s，短上下文可达 260 Tokens/s",
+      "contextWindow": 262144,
+      "maxOutputTokens": 32768,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "tool_use",
+        "agent",
+        "long_context",
+        "vision",
+        "video_understanding",
+        "image_understanding",
+        "fast"
+      ],
+      "inputPrice": 13,
+      "outputPrice": 54,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "thinking": {
+          "default": "enabled",
+          "disableSupported": false,
+          "preserveReasoningContent": true
+        },
+        "cacheHitPrice": 2.6,
+        "supportedImageFormats": [
+          "png",
+          "jpeg",
+          "webp",
+          "gif"
+        ],
+        "supportedVideoFormats": [
+          "mp4",
+          "mpeg",
+          "mov",
+          "avi",
+          "x-flv",
+          "mpg",
+          "webm",
+          "wmv",
+          "3gpp"
+        ]
+      }
+    },
+    {
+      "modelName": "kimi-k2.6",
+      "displayName": "Kimi K2.6",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "月之暗面 Kimi K2.6，多模态模型，长程代码编写与 Agent 自主执行能力增强，256K 上下文，支持思考/非思考模式",
+      "contextWindow": 262144,
+      "maxOutputTokens": 32768,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "tool_use",
+        "agent",
+        "long_context",
+        "vision",
+        "video_understanding",
+        "image_understanding"
+      ],
+      "inputPrice": 6.5,
+      "outputPrice": 27,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "thinking": {
+          "default": "enabled",
+          "disabledTemperature": 0.6
+        },
+        "cacheHitPrice": 1.1,
+        "supportedImageFormats": [
+          "png",
+          "jpeg",
+          "webp",
+          "gif"
+        ],
+        "supportedVideoFormats": [
+          "mp4",
+          "mpeg",
+          "mov",
+          "avi",
+          "x-flv",
+          "mpg",
+          "webm",
+          "wmv",
+          "3gpp"
+        ]
+      }
+    },
+    {
+      "modelName": "kimi-k2.5",
+      "displayName": "Kimi K2.5",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "月之暗面 Kimi K2.5，原生多模态模型，256K 上下文，支持思考/非思考模式",
+      "contextWindow": 262144,
+      "maxOutputTokens": 32768,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "tool_use",
+        "agent",
+        "long_context",
+        "vision",
+        "video_understanding",
+        "image_understanding"
+      ],
+      "inputPrice": 4,
+      "outputPrice": 21,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "thinking": {
+          "default": "enabled",
+          "disabledTemperature": 0.6
+        },
+        "cacheHitPrice": 0.7,
+        "supportedImageFormats": [
+          "png",
+          "jpeg",
+          "webp",
+          "gif"
+        ],
+        "supportedVideoFormats": [
+          "mp4",
+          "mpeg",
+          "mov",
+          "avi",
+          "x-flv",
+          "mpg",
+          "webm",
+          "wmv",
+          "3gpp"
+        ]
+      }
+    },
+    {
+      "modelName": "moonshot-v1-8k",
+      "displayName": "Moonshot V1 8K",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "月之暗面 Moonshot V1 生成模型，8K 上下文",
+      "contextWindow": 8192,
+      "maxOutputTokens": 4096,
+      "capabilities": [
+        "chat",
+        "code"
+      ],
+      "inputPrice": 2,
+      "outputPrice": 10,
+      "defaultTemperature": 0,
+      "defaultTopP": 1,
+      "extra": {}
+    },
+    {
+      "modelName": "moonshot-v1-32k",
+      "displayName": "Moonshot V1 32K",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "月之暗面 Moonshot V1 生成模型，32K 上下文",
+      "contextWindow": 32768,
+      "maxOutputTokens": 4096,
+      "capabilities": [
+        "chat",
+        "code",
+        "long_context"
+      ],
+      "inputPrice": 5,
+      "outputPrice": 20,
+      "defaultTemperature": 0,
+      "defaultTopP": 1,
+      "extra": {}
+    },
+    {
+      "modelName": "moonshot-v1-128k",
+      "displayName": "Moonshot V1 128K",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "月之暗面 Moonshot V1 生成模型，128K 上下文",
+      "contextWindow": 131072,
+      "maxOutputTokens": 4096,
+      "capabilities": [
+        "chat",
+        "code",
+        "long_context"
+      ],
+      "inputPrice": 10,
+      "outputPrice": 30,
+      "defaultTemperature": 0,
+      "defaultTopP": 1,
+      "extra": {}
+    },
+    {
+      "modelName": "moonshot-v1-8k-vision-preview",
+      "displayName": "Moonshot V1 8K Vision Preview",
+      "serviceType": [
+        "vision"
+      ],
+      "description": "月之暗面 Moonshot V1 视觉模型，理解图片内容并输出文本，8K 上下文",
+      "contextWindow": 8192,
+      "maxOutputTokens": 4096,
+      "capabilities": [
+        "chat",
+        "vision",
+        "image_understanding"
+      ],
+      "inputPrice": 2,
+      "outputPrice": 10,
+      "defaultTemperature": 0,
+      "defaultTopP": 1,
+      "extra": {
+        "supportedImageFormats": [
+          "png",
+          "jpeg",
+          "webp",
+          "gif"
+        ]
+      }
+    },
+    {
+      "modelName": "moonshot-v1-32k-vision-preview",
+      "displayName": "Moonshot V1 32K Vision Preview",
+      "serviceType": [
+        "vision"
+      ],
+      "description": "月之暗面 Moonshot V1 视觉模型，理解图片内容并输出文本，32K 上下文",
+      "contextWindow": 32768,
+      "maxOutputTokens": 4096,
+      "capabilities": [
+        "chat",
+        "vision",
+        "image_understanding",
+        "long_context"
+      ],
+      "inputPrice": 5,
+      "outputPrice": 20,
+      "defaultTemperature": 0,
+      "defaultTopP": 1,
+      "extra": {
+        "supportedImageFormats": [
+          "png",
+          "jpeg",
+          "webp",
+          "gif"
+        ]
+      }
+    },
+    {
+      "modelName": "moonshot-v1-128k-vision-preview",
+      "displayName": "Moonshot V1 128K Vision Preview",
+      "serviceType": [
+        "vision"
+      ],
+      "description": "月之暗面 Moonshot V1 视觉模型，理解图片内容并输出文本，128K 上下文",
+      "contextWindow": 131072,
+      "maxOutputTokens": 4096,
+      "capabilities": [
+        "chat",
+        "vision",
+        "image_understanding",
+        "long_context"
+      ],
+      "inputPrice": 10,
+      "outputPrice": 30,
+      "defaultTemperature": 0,
+      "defaultTopP": 1,
+      "extra": {
+        "supportedImageFormats": [
+          "png",
+          "jpeg",
+          "webp",
+          "gif"
+        ]
+      }
+    }
+  ]
+}

compute/v2/compatibility/providers/ollama.json

@@ -0,0 +1,35 @@
+{
+  "id": "provider-local-ollama",
+  "provider": "ollama",
+  "label": "本地 Ollama",
+  "baseUrl": "http://localhost:11434/v1",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "USD",
+  "services": [
+    "chat"
+  ],
+  "models": [
+    {
+      "modelName": "llama3.1:70b",
+      "displayName": "Llama 3.1 70B",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "本地运行的 Llama 3.1 70B",
+      "contextWindow": 131072,
+      "maxOutputTokens": 8192,
+      "capabilities": [
+        "chat",
+        "code",
+        "reasoning"
+      ],
+      "defaultTemperature": 0.8,
+      "defaultTopP": 0.9,
+      "extra": {}
+    }
+  ]
+}

compute/v2/compatibility/providers/openai-codex.json

@@ -0,0 +1,365 @@
+{
+  "id": "provider-openai-codex-plan-001",
+  "provider": "openai-codex",
+  "brandGroup": "openai",
+  "label": "ChatGPT 订阅 (Codex)",
+  "baseUrl": "https://chatgpt.com/backend-api",
+  "apiFormat": "openai-codex-responses",
+  "apiKeyRef": "openai-codex-access-token",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "USD",
+  "accessMode": "coding-plan",
+  "credentialSource": "codex-cli",
+  "codingPlan": {
+    "quotas": {},
+    "usageTracking": {
+      "method": "none",
+      "consoleUrl": "https://chatgpt.com/codex/settings/usage"
+    }
+  },
+  "services": [
+    "chat",
+    "reasoning"
+  ],
+  "tombstones": [
+    "gpt-5.3-codex",
+    "gpt-5.2-codex",
+    "gpt-5.1-codex-mini"
+  ],
+  "models": [
+    {
+      "modelName": "gpt-6.1-sol",
+      "displayName": "GPT-6.1 Sol",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "GPT-6.1 Sol，面向复杂编程、电脑使用和专业工作，1.05M 上下文；工具调用使用 Responses API",
+      "contextWindow": 1050000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "ultra_long_context",
+        "tool_use",
+        "agent"
+      ],
+      "extra": {
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultEffort": "medium"
+        }
+      },
+      "source": "preset"
+    },
+    {
+      "modelName": "gpt-6-sol",
+      "displayName": "GPT-6 Sol",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "ChatGPT 订阅额度计费，兼顾复杂编程、智能体工作流与成本的 GPT-6 模型；1.05M 上下文",
+      "contextWindow": 1050000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "ultra_long_context",
+        "tool_use",
+        "agent"
+      ],
+      "extra": {
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultEffort": "medium"
+        }
+      },
+      "source": "preset"
+    },
+    {
+      "modelName": "gpt-6-luna",
+      "displayName": "GPT-6 Luna",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "ChatGPT 订阅额度计费，面向明确任务和高吞吐场景的 GPT-6 轻量模型；1.05M 上下文",
+      "contextWindow": 1050000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "ultra_long_context",
+        "tool_use",
+        "fast"
+      ],
+      "extra": {
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultEffort": "medium"
+        }
+      },
+      "source": "preset"
+    },
+    {
+      "modelName": "gpt-6-astra",
+      "displayName": "GPT-6 Astra",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "ChatGPT 订阅额度计费，面向复杂推理、编程和长程智能体任务，1.05M 上下文",
+      "contextWindow": 1050000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "long_context",
+        "tool_use",
+        "agent"
+      ],
+      "extra": {
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultEffort": "medium"
+        }
+      },
+      "source": "preset"
+    },
+    {
+      "modelName": "gpt-reserve",
+      "displayName": "GPT-Reserve",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "ChatGPT 订阅额度计费，参数暂按 GPT-5.6 Luna 配置，待官方规格确认，272K 上下文",
+      "contextWindow": 272000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "long_context",
+        "tool_use",
+        "fast"
+      ],
+      "extra": {
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultEffort": "medium"
+        }
+      },
+      "source": "preset"
+    },
+    {
+      "modelName": "gpt-5.6-sol",
+      "displayName": "GPT-5.6 Sol",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "ChatGPT 订阅额度计费，新一代旗舰模型，复杂编程/计算机操作/网络安全能力最强，272K 上下文",
+      "contextWindow": 272000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "long_context",
+        "tool_use",
+        "agent"
+      ],
+      "extra": {
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultEffort": "low"
+        }
+      },
+      "source": "preset"
+    },
+    {
+      "modelName": "gpt-5.6-terra",
+      "displayName": "GPT-5.6 Terra",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "ChatGPT 订阅额度计费，均衡型模型，日常工作性能对齐 GPT-5.5 且成本更低，272K 上下文",
+      "contextWindow": 272000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "long_context",
+        "tool_use",
+        "agent"
+      ],
+      "extra": {
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultEffort": "medium"
+        }
+      },
+      "source": "preset"
+    },
+    {
+      "modelName": "gpt-5.6-luna",
+      "displayName": "GPT-5.6 Luna",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "ChatGPT 订阅额度计费，GPT-5.6 家族中最快最省成本的档位，272K 上下文",
+      "contextWindow": 272000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "long_context",
+        "tool_use",
+        "fast"
+      ],
+      "extra": {
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultEffort": "medium"
+        }
+      },
+      "source": "preset"
+    },
+    {
+      "modelName": "gpt-5.5",
+      "displayName": "GPT-5.5",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "ChatGPT 订阅额度计费，上一代旗舰模型，272K 上下文",
+      "contextWindow": 272000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "long_context",
+        "tool_use",
+        "agent"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "gpt-5.4",
+      "displayName": "GPT-5.4",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "ChatGPT 订阅额度计费，高阶通用模型，272K 上下文",
+      "contextWindow": 272000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "long_context",
+        "tool_use",
+        "agent"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "gpt-5.4-mini",
+      "displayName": "GPT-5.4 Mini",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "ChatGPT 订阅额度计费，轻量高性价比模型，是 gpt-5.1-codex-mini 的官方迁移目标，272K 上下文",
+      "contextWindow": 272000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "long_context",
+        "tool_use",
+        "fast"
+      ],
+      "source": "preset"
+    },
+    {
+      "modelName": "gpt-5.3-codex-spark",
+      "displayName": "GPT-5.3 Codex Spark（Pro 专属）",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "ChatGPT Pro 订阅专属研究预览模型，面向实时编程迭代优化，>1000 tok/s，纯文本，128K 上下文",
+      "contextWindow": 128000,
+      "capabilities": [
+        "chat",
+        "code",
+        "tool_use",
+        "fast"
+      ],
+      "source": "preset"
+    }
+  ]
+}

compute/v2/compatibility/providers/openrouter.json

@@ -0,0 +1,317 @@
+{
+  "id": "provider-openrouter-001",
+  "provider": "openrouter",
+  "label": "OpenRouter",
+  "baseUrl": "https://openrouter.ai/api/v1",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "openrouter",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "USD",
+  "services": [
+    "chat",
+    "reasoning",
+    "vision"
+  ],
+  "tombstones": [
+    "stealth/ox-alpha",
+    "openai/gpt-oss-120b:free",
+    "qwen/qwen3-coder:free"
+  ],
+  "models": [
+    {
+      "modelName": "cohere/command-a-plus",
+      "displayName": "Cohere: Command A+",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 192000,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 0.3,
+      "outputPrice": 1.5,
+      "description": "OpenRouter 官方公共目录模型，接入参数与计价以该平台为准",
+      "maxOutputTokens": 64000
+    },
+    {
+      "modelName": "mistralai/mistral-medium-3-5",
+      "displayName": "Mistral: Mistral Medium 3.5",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 262144,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 1.5,
+      "outputPrice": 7.5,
+      "description": "OpenRouter 官方公共目录模型，接入参数与计价以该平台为准",
+      "maxOutputTokens": 209715
+    },
+    {
+      "modelName": "google/gemini-3.8-flash",
+      "displayName": "Google: Gemini 3.8 Flash",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1048576,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 0.75,
+      "outputPrice": 3.75,
+      "description": "OpenRouter 官方公共目录模型，接入参数与计价以该平台为准",
+      "maxOutputTokens": 65536
+    },
+    {
+      "modelName": "minimax/minimax-m3",
+      "displayName": "MiniMax: MiniMax M3",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1048576,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 0.3,
+      "outputPrice": 1.2,
+      "description": "OpenRouter 官方公共目录模型，接入参数与计价以该平台为准",
+      "maxOutputTokens": 512000
+    },
+    {
+      "modelName": "qwen/qwen3.8-omni-flash",
+      "displayName": "Qwen: Qwen3.8 Omni Flash",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1000000,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 0.15,
+      "outputPrice": 0.47,
+      "description": "OpenRouter 官方公共目录模型，接入参数与计价以该平台为准",
+      "maxOutputTokens": 131072
+    },
+    {
+      "modelName": "qwen/qwen3.8-max-prime",
+      "displayName": "Qwen: Qwen3.8 Max Prime",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1000000,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 4,
+      "outputPrice": 12,
+      "description": "OpenRouter 官方公共目录模型，接入参数与计价以该平台为准",
+      "maxOutputTokens": 131072
+    },
+    {
+      "modelName": "deepseek/deepseek-v4.1-flash",
+      "displayName": "DeepSeek: DeepSeek V4.1 Flash",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1048576,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 0.3,
+      "outputPrice": 1.2,
+      "description": "OpenRouter 官方公共目录模型；接入参数与计价以该平台为准",
+      "extra": {
+        "pricingNotes": "USD per 1M tokens; OpenRouter access-specific pricing."
+      },
+      "maxOutputTokens": 943718
+    },
+    {
+      "modelName": "tencent/hy4-preview",
+      "displayName": "Tencent: Hy4 preview",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1048576,
+      "capabilities": [
+        "chat",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 0.834,
+      "outputPrice": 2.501,
+      "description": "OpenRouter 官方公共目录模型；接入参数与计价以该平台为准",
+      "extra": {
+        "pricingNotes": "USD per 1M tokens; OpenRouter access-specific pricing."
+      },
+      "maxOutputTokens": 64000
+    },
+    {
+      "modelName": "z-ai/glm-5.3-flash",
+      "displayName": "Z.ai: GLM 5.3 Flash",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1048576,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 0.15,
+      "outputPrice": 0.5,
+      "description": "OpenRouter 官方公共目录模型；接入参数与计价以该平台为准",
+      "extra": {
+        "pricingNotes": "USD per 1M tokens; OpenRouter access-specific pricing."
+      },
+      "maxOutputTokens": 943717
+    },
+    {
+      "modelName": "z-ai/glm-5.3",
+      "displayName": "Z.ai: GLM 5.3",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1048576,
+      "capabilities": [
+        "chat",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 1.4,
+      "outputPrice": 4.4,
+      "description": "OpenRouter 官方公共目录模型；接入参数与计价以该平台为准",
+      "extra": {
+        "pricingNotes": "USD per 1M tokens; OpenRouter access-specific pricing."
+      },
+      "maxOutputTokens": 131072
+    },
+    {
+      "modelName": "moonshotai/kimi-k3",
+      "displayName": "MoonshotAI: Kimi K3",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1048576,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 2.7,
+      "outputPrice": 13.5,
+      "description": "OpenRouter 官方公共目录模型；接入参数与计价以该平台为准",
+      "extra": {
+        "pricingNotes": "USD per 1M tokens; OpenRouter access-specific pricing."
+      },
+      "maxOutputTokens": 943718
+    },
+    {
+      "modelName": "anthropic/claude-sonnet-5.5",
+      "displayName": "Anthropic: Claude Sonnet 5.5",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1000000,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 2,
+      "outputPrice": 10,
+      "description": "OpenRouter 官方公共目录模型；接入参数与计价以该平台为准",
+      "extra": {
+        "pricingNotes": "USD per 1M tokens; OpenRouter access-specific pricing."
+      },
+      "maxOutputTokens": 128000
+    },
+    {
+      "modelName": "anthropic/claude-opus-5.5",
+      "displayName": "Anthropic: Claude Opus 5.5",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1000000,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 4,
+      "outputPrice": 20,
+      "description": "OpenRouter 官方公共目录模型；接入参数与计价以该平台为准",
+      "extra": {
+        "pricingNotes": "USD per 1M tokens; OpenRouter access-specific pricing."
+      },
+      "maxOutputTokens": 128000
+    },
+    {
+      "modelName": "openai/gpt-6.1-sol",
+      "displayName": "OpenAI: GPT-6.1 Sol",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1050000,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 2,
+      "outputPrice": 10,
+      "description": "OpenRouter 官方公共目录模型；接入参数与计价以该平台为准",
+      "extra": {
+        "pricingNotes": "USD per 1M tokens; OpenRouter access-specific pricing."
+      },
+      "maxOutputTokens": 128000
+    },
+    {
+      "modelName": "openrouter/auto",
+      "displayName": "OpenRouter Auto",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "OpenRouter 自动路由，智能选择最优模型",
+      "contextWindow": 2000000,
+      "maxOutputTokens": 16384,
+      "capabilities": [
+        "chat",
+        "auto_routing"
+      ],
+      "defaultTemperature": 1,
+      "defaultTopP": 1,
+      "extra": {
+        "pricingNotes": "OpenRouter Models API 的 prompt/completion=-1 为动态路由占位值，实际价格取决于选中的模型，不代表免费。"
+      }
+    }
+  ]
+}

compute/v2/compatibility/providers/perplexity.json

@@ -0,0 +1,95 @@
+{
+  "id": "provider-perplexity-001",
+  "provider": "perplexity",
+  "label": "Perplexity Sonar",
+  "baseUrl": "https://api.perplexity.ai",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "perplexity",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "USD",
+  "services": [
+    "chat"
+  ],
+  "models": [
+    {
+      "modelName": "sonar-pro",
+      "displayName": "Sonar Pro",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Perplexity 旗舰搜索增强模型，内置实时联网搜索和引用",
+      "contextWindow": 200000,
+      "maxOutputTokens": 8192,
+      "capabilities": [
+        "chat",
+        "web_search",
+        "reasoning",
+        "citation"
+      ],
+      "inputPrice": 3,
+      "outputPrice": 15,
+      "extra": {
+        "requestPricingPer1k": {
+          "low": 6,
+          "medium": 10,
+          "high": 14
+        },
+        "pricingNotes": "Total Sonar API cost includes token costs plus a request fee based on search context size."
+      }
+    },
+    {
+      "modelName": "sonar-reasoning-pro",
+      "displayName": "Sonar Reasoning Pro",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Perplexity 深度推理模型，内置联网搜索和深度思考",
+      "contextWindow": 128000,
+      "maxOutputTokens": 8192,
+      "capabilities": [
+        "chat",
+        "web_search",
+        "reasoning",
+        "deep_thinking"
+      ],
+      "inputPrice": 2,
+      "outputPrice": 8,
+      "extra": {
+        "requestPricingPer1k": {
+          "low": 6,
+          "medium": 10,
+          "high": 14
+        },
+        "pricingNotes": "Total Sonar API cost includes token costs plus a request fee based on search context size."
+      }
+    },
+    {
+      "modelName": "sonar",
+      "displayName": "Sonar",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Perplexity 轻量搜索模型，低成本联网搜索",
+      "contextWindow": 128000,
+      "maxOutputTokens": 4096,
+      "capabilities": [
+        "chat",
+        "web_search",
+        "citation",
+        "fast"
+      ],
+      "inputPrice": 1,
+      "outputPrice": 1,
+      "extra": {
+        "requestPricingPer1k": {
+          "low": 5,
+          "medium": 8,
+          "high": 12
+        },
+        "pricingNotes": "Total Sonar API cost includes token costs plus a request fee based on search context size."
+      }
+    }
+  ]
+}

compute/v2/compatibility/providers/siliconflow.json

@@ -0,0 +1,80 @@
+{
+  "id": "provider-siliconflow-001",
+  "provider": "siliconflow",
+  "label": "硅基流动",
+  "baseUrl": "https://api.siliconflow.cn/v1",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "siliconflow",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "CNY",
+  "services": [
+    "chat",
+    "embedding"
+  ],
+  "models": [
+    {
+      "modelName": "Qwen/Qwen3-Coder-480B-A35B-Instruct",
+      "displayName": "Qwen3 Coder 480B (硅基)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Qwen3-Coder-480B-A35B-Instruct，480B MoE（35B 激活），原生 256K 上下文，可通过外推扩展至 1M，面向 Agentic Coding",
+      "contextWindow": 262144,
+      "maxOutputTokens": 262144,
+      "capabilities": [
+        "chat",
+        "code",
+        "long_context",
+        "agent",
+        "tool_use"
+      ],
+      "inputPrice": 8,
+      "outputPrice": 16,
+      "defaultTemperature": 0.7,
+      "defaultTopP": 0.7,
+      "extra": {}
+    },
+    {
+      "modelName": "Qwen/Qwen3-235B-A22B-Instruct-2507",
+      "displayName": "Qwen3 235B (硅基)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Qwen3-235B-A22B-Instruct-2507，235B MoE（22B 激活）通用指令模型，262K 上下文",
+      "contextWindow": 262144,
+      "maxOutputTokens": 262144,
+      "capabilities": [
+        "chat",
+        "code",
+        "multilingual",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "inputPrice": 2.5,
+      "outputPrice": 10,
+      "defaultTemperature": 0.7,
+      "defaultTopP": 0.7,
+      "extra": {}
+    },
+    {
+      "modelName": "BAAI/bge-m3",
+      "displayName": "BGE-M3 (硅基)",
+      "serviceType": [
+        "embedding"
+      ],
+      "description": "多语言 Embedding 模型，支持 8K 上下文，免费版",
+      "contextWindow": 8192,
+      "maxOutputTokens": 0,
+      "capabilities": [
+        "text_embedding",
+        "multilingual"
+      ],
+      "inputPrice": 0,
+      "outputPrice": 0,
+      "extra": {}
+    }
+  ]
+}

compute/v2/compatibility/providers/stability.json

@@ -0,0 +1,45 @@
+{
+  "id": "provider-stability-001",
+  "provider": "stability",
+  "label": "Stability SDXL",
+  "baseUrl": "https://api.stability.ai/v2beta",
+  "apiFormat": "stability-rest-api",
+  "apiKeyRef": "stability",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "USD",
+  "services": [
+    "image_gen"
+  ],
+  "models": [
+    {
+      "modelName": "stable-diffusion-3.5-large",
+      "displayName": "Stable Diffusion 3.5 Large",
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "Stability AI 旗舰文生图模型",
+      "capabilities": [
+        "image_generation",
+        "high_quality",
+        "controllable"
+      ],
+      "extra": {
+        "supportedImageSizes": [
+          "1024x1024",
+          "1152x896",
+          "896x1152",
+          "1216x832",
+          "832x1216"
+        ],
+        "supportedStyles": [
+          "photographic",
+          "digital-art",
+          "anime",
+          "comic-book"
+        ]
+      }
+    }
+  ]
+}

compute/v2/compatibility/providers/tencent.json

@@ -0,0 +1,180 @@
+{
+  "id": "provider-tencent-001",
+  "provider": "tencent",
+  "label": "腾讯混元",
+  "baseUrl": "https://api.hunyuan.cloud.tencent.com/v1",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "tencent",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "CNY",
+  "services": [
+    "chat",
+    "reasoning"
+  ],
+  "models": [
+    {
+      "modelName": "hunyuan-2.0-thinking-20251109",
+      "displayName": "混元 2.0 Think",
+      "serviceType": [
+        "reasoning"
+      ],
+      "description": "腾讯混元最新推理模型，128K输入/64K输出，MoE架构406B总参数",
+      "contextWindow": 196608,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "long_context"
+      ],
+      "inputPrice": 3.975,
+      "outputPrice": 15.9,
+      "defaultTemperature": 1,
+      "defaultTopP": 1,
+      "extra": {
+        "pricingTiers": [
+          {
+            "maxInputTokens": 32768,
+            "inputPrice": 3.975,
+            "outputPrice": 15.9
+          },
+          {
+            "maxInputTokens": 131072,
+            "inputPrice": 5.3,
+            "outputPrice": 21.2
+          }
+        ]
+      }
+    },
+    {
+      "modelName": "hunyuan-2.0-instruct-20251111",
+      "displayName": "混元 2.0 Instruct",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "腾讯混元 2.0 指令模型，最大输入128K/最大输出16K，支持联网搜索和 Function Calling",
+      "contextWindow": 147456,
+      "maxOutputTokens": 16384,
+      "capabilities": [
+        "chat",
+        "code",
+        "long_context",
+        "tool_use"
+      ],
+      "inputPrice": 3.18,
+      "outputPrice": 7.95,
+      "defaultTemperature": 1,
+      "defaultTopP": 1,
+      "extra": {
+        "pricingTiers": [
+          {
+            "maxInputTokens": 32768,
+            "inputPrice": 3.18,
+            "outputPrice": 7.95
+          },
+          {
+            "maxInputTokens": 131072,
+            "inputPrice": 4.505,
+            "outputPrice": 11.13
+          }
+        ]
+      }
+    },
+    {
+      "modelName": "hunyuan-turbos-latest",
+      "displayName": "混元 Turbo S",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "腾讯混元高速模型，32K输入/16K输出，支持推理和工具调用",
+      "contextWindow": 32768,
+      "maxOutputTokens": 16384,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "fast",
+        "tool_use"
+      ],
+      "inputPrice": 0.8,
+      "outputPrice": 2,
+      "defaultTemperature": 1,
+      "defaultTopP": 1,
+      "extra": {}
+    },
+    {
+      "modelName": "hunyuan-t1-latest",
+      "displayName": "混元 T1",
+      "serviceType": [
+        "reasoning"
+      ],
+      "description": "腾讯混元 T1 深度推理模型，256K 上下文，专攻数学/代码/复杂推理",
+      "contextWindow": 262144,
+      "maxOutputTokens": 32768,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "math",
+        "long_context"
+      ],
+      "inputPrice": 5,
+      "outputPrice": 20,
+      "defaultTemperature": 1,
+      "defaultTopP": 1,
+      "extra": {}
+    },
+    {
+      "modelName": "hunyuan-t1-vision",
+      "displayName": "混元 T1 Vision",
+      "serviceType": [
+        "reasoning"
+      ],
+      "description": "腾讯混元 T1 视觉推理模型，支持图片理解 + 深度推理",
+      "contextWindow": 131072,
+      "maxOutputTokens": 32768,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "math",
+        "vision",
+        "image_understanding"
+      ],
+      "inputPrice": 5,
+      "outputPrice": 20,
+      "defaultTemperature": 1,
+      "defaultTopP": 1,
+      "extra": {}
+    },
+    {
+      "modelName": "hunyuan-turbos-vision",
+      "displayName": "混元 Turbo S Vision",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "腾讯混元 Turbo S 视觉模型，支持图片理解，高速推理",
+      "contextWindow": 32768,
+      "maxOutputTokens": 16384,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "image_understanding",
+        "fast",
+        "tool_use"
+      ],
+      "inputPrice": 0.8,
+      "outputPrice": 2,
+      "defaultTemperature": 1,
+      "defaultTopP": 1,
+      "extra": {}
+    }
+  ]
+}

compute/v2/compatibility/providers/volcengine.json

@@ -0,0 +1,712 @@
+{
+  "id": "provider-volcengine-001",
+  "provider": "volcengine",
+  "label": "火山 豆包",
+  "baseUrl": "https://ark.cn-beijing.volces.com/api/v3",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "volcengine",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "CNY",
+  "services": [
+    "chat",
+    "reasoning",
+    "vision",
+    "embedding",
+    "voice_clone",
+    "realtime_voice",
+    "simultaneous_interpret",
+    "translation",
+    "image_gen",
+    "video_gen"
+  ],
+  "models": [
+    {
+      "modelName": "doubao-seed-evolving",
+      "displayName": "Doubao Seed Evolving",
+      "contextWindow": 1000000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "Seed 系列持续迭代模型，通过固定 ID 获取最新版本，2026-09-09 更新"
+    },
+    {
+      "modelName": "doubao-seed-2.1-pro",
+      "displayName": "豆包 Seed 2.1 Pro",
+      "serviceType": [
+        "chat",
+        "reasoning",
+        "vision"
+      ],
+      "description": "火山方舟豆包 Seed 2.1 Pro，生产级 Coding、Agent 与多模态旗舰模型",
+      "contextWindow": 256000,
+      "maxOutputTokens": 262144,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "multilingual",
+        "long_context",
+        "tool_use",
+        "agent",
+        "deep_thinking"
+      ],
+      "inputPrice": 6,
+      "outputPrice": 30,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.7,
+      "apiModelId": "doubao-seed-2-1-pro-260628",
+      "extra": {
+        "cacheHitPrice": 1.2,
+        "cacheStoragePricePerMillionTokenHour": 0.017,
+        "supportsResponsesApi": true,
+        "supportsChatApi": true
+      }
+    },
+    {
+      "modelName": "doubao-seed-2.1-turbo",
+      "displayName": "豆包 Seed 2.1 Turbo",
+      "serviceType": [
+        "chat",
+        "reasoning",
+        "vision"
+      ],
+      "description": "火山方舟豆包 Seed 2.1 Turbo，高吞吐低成本生产模型，适合规模化 Agent 和 Coding 调用",
+      "contextWindow": 256000,
+      "maxOutputTokens": 262144,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "multilingual",
+        "long_context",
+        "tool_use",
+        "agent",
+        "fast"
+      ],
+      "inputPrice": 3,
+      "outputPrice": 15,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.7,
+      "apiModelId": "doubao-seed-2-1-turbo-260628",
+      "extra": {
+        "cacheHitPrice": 0.6,
+        "cacheStoragePricePerMillionTokenHour": 0.017,
+        "supportsResponsesApi": true,
+        "supportsChatApi": true
+      }
+    },
+    {
+      "modelName": "doubao-seed-2.0-pro",
+      "displayName": "豆包 2.0 Pro",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山引擎豆包 Seed 2.0 旗舰模型，256K上下文，支持复杂推理、Agent 和多模态理解",
+      "contextWindow": 256000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual",
+        "long_context",
+        "tool_use",
+        "vision"
+      ],
+      "inputPrice": 3.2,
+      "outputPrice": 16,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.7,
+      "extra": {
+        "pricingTiers": [
+          {
+            "maxInputTokens": 32000,
+            "inputPrice": 3.2,
+            "outputPrice": 16
+          },
+          {
+            "maxInputTokens": 128000,
+            "inputPrice": 4.8,
+            "outputPrice": 24
+          },
+          {
+            "maxInputTokens": 256000,
+            "inputPrice": 9.6,
+            "outputPrice": 48
+          }
+        ]
+      }
+    },
+    {
+      "modelName": "doubao-seed-2.0-lite",
+      "displayName": "豆包 2.0 Lite",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山引擎豆包 Seed 2.0 全模态均衡模型，256K上下文，支持视频/图像/音频/文本统一理解",
+      "contextWindow": 256000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual",
+        "long_context",
+        "tool_use",
+        "vision",
+        "audio_understanding",
+        "video_understanding"
+      ],
+      "inputPrice": 0.6,
+      "outputPrice": 3.6,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.7,
+      "extra": {
+        "pricingTiers": [
+          {
+            "maxInputTokens": 32000,
+            "inputPrice": 0.6,
+            "outputPrice": 3.6
+          },
+          {
+            "maxInputTokens": 128000,
+            "inputPrice": 0.9,
+            "outputPrice": 5.4
+          },
+          {
+            "maxInputTokens": 256000,
+            "inputPrice": 1.8,
+            "outputPrice": 10.8
+          }
+        ]
+      }
+    },
+    {
+      "modelName": "doubao-seed-2.0-mini",
+      "displayName": "豆包 2.0 Mini",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山引擎豆包 Seed 2.0 全模态轻量模型，256K上下文，低延迟高并发，支持全模态理解",
+      "contextWindow": 256000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual",
+        "long_context",
+        "tool_use",
+        "vision",
+        "audio_understanding",
+        "video_understanding",
+        "fast"
+      ],
+      "inputPrice": 0.2,
+      "outputPrice": 2,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.7,
+      "extra": {
+        "pricingTiers": [
+          {
+            "maxInputTokens": 32000,
+            "inputPrice": 0.2,
+            "outputPrice": 2
+          },
+          {
+            "maxInputTokens": 128000,
+            "inputPrice": 0.4,
+            "outputPrice": 4
+          },
+          {
+            "maxInputTokens": 256000,
+            "inputPrice": 0.8,
+            "outputPrice": 8
+          }
+        ]
+      }
+    },
+    {
+      "modelName": "doubao-seed-2.0-code",
+      "displayName": "豆包 2.0 Code",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山引擎豆包 Seed 2.0 编程模型，面向 Agentic Coding 场景优化",
+      "contextWindow": 256000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "long_context",
+        "tool_use",
+        "vision",
+        "deep_thinking"
+      ],
+      "inputPrice": 3.2,
+      "outputPrice": 16,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.7,
+      "extra": {
+        "pricingTiers": [
+          {
+            "maxInputTokens": 32000,
+            "inputPrice": 3.2,
+            "outputPrice": 16
+          },
+          {
+            "maxInputTokens": 128000,
+            "inputPrice": 4.8,
+            "outputPrice": 24
+          },
+          {
+            "maxInputTokens": 256000,
+            "inputPrice": 9.6,
+            "outputPrice": 48
+          }
+        ]
+      }
+    },
+    {
+      "modelName": "doubao-seed-1.8",
+      "displayName": "豆包 Seed-1.8",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山引擎豆包1.8旗舰模型，256K上下文，支持深度思考和多模态",
+      "contextWindow": 256000,
+      "maxOutputTokens": 16000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual",
+        "long_context",
+        "tool_use",
+        "vision"
+      ],
+      "inputPrice": 0.8,
+      "outputPrice": 8,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.7,
+      "extra": {
+        "pricingTiers": [
+          {
+            "maxInputTokens": 32000,
+            "inputPrice": 0.8,
+            "outputPrice": 8,
+            "pricingNotes": "Short outputs up to 0.2K tokens are priced at 2 CNY/M output tokens."
+          },
+          {
+            "maxInputTokens": 128000,
+            "inputPrice": 1.2,
+            "outputPrice": 16
+          },
+          {
+            "maxInputTokens": 256000,
+            "inputPrice": 2.4,
+            "outputPrice": 24
+          }
+        ],
+        "deprecated": true,
+        "deprecationNotice": "火山方舟 2026-07-06 模型列表中标记为即将下线，建议迁移到 Seed 2.1 或 Seed 2.0 系列。",
+        "replacedBy": "doubao-seed-2.1-turbo"
+      }
+    },
+    {
+      "modelName": "doubao-seed-1.6",
+      "displayName": "豆包 Seed-1.6",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山引擎豆包1.6旗舰模型，256K上下文",
+      "contextWindow": 256000,
+      "maxOutputTokens": 32000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual",
+        "long_context",
+        "tool_use",
+        "vision"
+      ],
+      "inputPrice": 0.8,
+      "outputPrice": 8,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.7,
+      "extra": {
+        "pricingTiers": [
+          {
+            "maxInputTokens": 32000,
+            "inputPrice": 0.8,
+            "outputPrice": 8,
+            "pricingNotes": "Short outputs up to 0.2K tokens are priced at 2 CNY/M output tokens."
+          },
+          {
+            "maxInputTokens": 128000,
+            "inputPrice": 1.2,
+            "outputPrice": 16
+          },
+          {
+            "maxInputTokens": 256000,
+            "inputPrice": 2.4,
+            "outputPrice": 24
+          }
+        ],
+        "deprecated": true,
+        "deprecationNotice": "火山方舟 2026-07-06 模型列表中标记为即将下线，建议迁移到 Seed 2.1 或 Seed 2.0 系列。",
+        "replacedBy": "doubao-seed-2.1-turbo"
+      }
+    },
+    {
+      "modelName": "doubao-seed-1.6-thinking",
+      "displayName": "豆包 Seed-1.6 思考版",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山引擎豆包1.6深度思考模型，256K上下文",
+      "contextWindow": 256000,
+      "maxOutputTokens": 16000,
+      "capabilities": [
+        "reasoning",
+        "math",
+        "code",
+        "deep_thinking",
+        "long_context"
+      ],
+      "inputPrice": 0.8,
+      "outputPrice": 8,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.7,
+      "extra": {
+        "deprecated": true,
+        "deprecationNotice": "火山方舟 2026-07-06 模型列表中标记为即将下线，建议迁移到 Seed 2.1 或 Seed 2.0 系列。",
+        "replacedBy": "doubao-seed-2.1-turbo"
+      }
+    },
+    {
+      "modelName": "doubao-seed-1.6-flash",
+      "displayName": "豆包 Seed-1.6 Flash",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山引擎豆包1.6快速版，成本较前代下降63%",
+      "contextWindow": 256000,
+      "maxOutputTokens": 16000,
+      "capabilities": [
+        "chat",
+        "code",
+        "fast",
+        "long_context",
+        "vision"
+      ],
+      "inputPrice": 0.15,
+      "outputPrice": 1.5,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.7,
+      "extra": {
+        "deprecated": true,
+        "deprecationNotice": "火山方舟 2026-07-06 模型列表中标记为即将下线，建议迁移到 Seed 2.1 或 Seed 2.0 系列。",
+        "replacedBy": "doubao-seed-2.1-turbo"
+      }
+    },
+    {
+      "modelName": "doubao-seed-1.6-lite",
+      "displayName": "豆包 Seed-1.6 Lite",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山引擎豆包1.6轻量版，极致性价比",
+      "contextWindow": 256000,
+      "maxOutputTokens": 32000,
+      "capabilities": [
+        "chat",
+        "code",
+        "fast"
+      ],
+      "inputPrice": 0.3,
+      "outputPrice": 2.4,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.7,
+      "extra": {
+        "pricingTiers": [
+          {
+            "maxInputTokens": 32000,
+            "inputPrice": 0.3,
+            "outputPrice": 2.4
+          },
+          {
+            "maxInputTokens": 128000,
+            "inputPrice": 0.6,
+            "outputPrice": 4
+          },
+          {
+            "maxInputTokens": 256000,
+            "inputPrice": 1.2,
+            "outputPrice": 12
+          }
+        ],
+        "deprecated": true,
+        "deprecationNotice": "火山方舟 2026-07-06 模型列表中标记为即将下线，建议迁移到 Seed 2.1 或 Seed 2.0 系列。",
+        "replacedBy": "doubao-seed-2.1-turbo"
+      }
+    },
+    {
+      "modelName": "doubao-seed-1.6-vision",
+      "displayName": "豆包 Seed-1.6 视觉版",
+      "serviceType": [
+        "vision"
+      ],
+      "description": "火山引擎豆包1.6视觉模型，256K上下文，支持视频理解",
+      "contextWindow": 256000,
+      "maxOutputTokens": 32000,
+      "capabilities": [
+        "chat",
+        "vision",
+        "video_understanding",
+        "gui_agent",
+        "long_context"
+      ],
+      "inputPrice": 0.8,
+      "outputPrice": 8,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.7,
+      "extra": {
+        "deprecated": true,
+        "deprecationNotice": "火山方舟 2026-07-06 模型列表中标记为即将下线，建议迁移到 Seed 2.1 或 Seed 2.0 系列。",
+        "replacedBy": "doubao-seed-2.1-turbo"
+      }
+    },
+    {
+      "modelName": "doubao-seed-code",
+      "displayName": "豆包 Seed-Code",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山引擎豆包编程专用模型，256K上下文",
+      "contextWindow": 256000,
+      "maxOutputTokens": 32768,
+      "capabilities": [
+        "code",
+        "reasoning",
+        "long_context",
+        "deep_thinking"
+      ],
+      "inputPrice": 1.2,
+      "outputPrice": 8,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.7,
+      "extra": {}
+    },
+    {
+      "modelName": "deepseek-v3.2",
+      "displayName": "DeepSeek V3.2 (火山引擎)",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "火山引擎托管的DeepSeek V3.2，128K上下文",
+      "contextWindow": 128000,
+      "maxOutputTokens": 32000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual"
+      ],
+      "inputPrice": 2,
+      "outputPrice": 3,
+      "defaultTemperature": 1,
+      "defaultTopP": 1,
+      "extra": {}
+    },
+    {
+      "modelName": "deepseek-r1",
+      "displayName": "DeepSeek R1 (火山引擎)",
+      "serviceType": [
+        "reasoning"
+      ],
+      "description": "火山引擎托管的DeepSeek R1推理模型，128K上下文",
+      "contextWindow": 128000,
+      "maxOutputTokens": 65536,
+      "capabilities": [
+        "reasoning",
+        "math",
+        "code",
+        "science",
+        "deep_thinking"
+      ],
+      "inputPrice": 4,
+      "outputPrice": 16,
+      "extra": {}
+    },
+    {
+      "modelName": "doubao-embedding",
+      "displayName": "豆包 Embedding",
+      "serviceType": [
+        "embedding"
+      ],
+      "description": "火山引擎豆包标准向量模型，默认2048维度，可选512/1024/2048维度",
+      "contextWindow": 4096,
+      "capabilities": [
+        "text_embedding",
+        "semantic_search",
+        "rag",
+        "chinese_optimized"
+      ],
+      "inputPrice": 0.5,
+      "extra": {}
+    },
+    {
+      "modelName": "doubao-embedding-large",
+      "displayName": "豆包 Embedding Large",
+      "serviceType": [
+        "embedding"
+      ],
+      "description": "火山引擎豆包大号向量模型，默认2048维度，可选512/1024/2048/4096维度",
+      "contextWindow": 4096,
+      "capabilities": [
+        "text_embedding",
+        "semantic_search",
+        "rag",
+        "chinese_optimized"
+      ],
+      "inputPrice": 0.5,
+      "extra": {}
+    },
+    {
+      "modelName": "volc-mega-tts-clone",
+      "displayName": "火山声音复刻",
+      "serviceType": [
+        "voice_clone"
+      ],
+      "description": "火山引擎声音复刻，支持极少样本",
+      "capabilities": [
+        "voice_clone",
+        "chinese_optimized",
+        "few_shot"
+      ],
+      "extra": {
+        "maxCloneSamples": 3,
+        "minCloneDuration": 5
+      }
+    },
+    {
+      "modelName": "volc-realtime-voice",
+      "displayName": "火山实时语音",
+      "serviceType": [
+        "realtime_voice"
+      ],
+      "description": "火山引擎实时语音交互",
+      "capabilities": [
+        "realtime",
+        "speech_to_speech",
+        "low_latency",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "supportedModes": [
+          "speech-to-speech"
+        ],
+        "latencyMs": 200
+      }
+    },
+    {
+      "modelName": "volc-simultaneous",
+      "displayName": "火山同声传译",
+      "serviceType": [
+        "simultaneous_interpret"
+      ],
+      "description": "火山引擎同声传译，支持中英双向",
+      "capabilities": [
+        "simultaneous_interpretation",
+        "chinese_optimized",
+        "low_latency"
+      ],
+      "extra": {
+        "streamingSupported": true
+      }
+    },
+    {
+      "modelName": "volc-translation",
+      "displayName": "火山翻译",
+      "serviceType": [
+        "translation"
+      ],
+      "description": "火山引擎机器翻译，支持多语种互译",
+      "capabilities": [
+        "translation",
+        "multilingual",
+        "chinese_optimized"
+      ],
+      "extra": {}
+    },
+    {
+      "modelName": "doubao-seedream-4-0-250828",
+      "displayName": "豆包 Seedream 4.0（文生图/图生图）",
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "字节 Seedream 图像模型，支持文生图、图生图（img2img）、局部重绘与多图融合（最多 10 张参考图）；OpenAI 风格 /images/generations 同步返回",
+      "capabilities": [
+        "image_generation",
+        "image_to_image",
+        "multi_image_fusion",
+        "chinese_optimized",
+        "high_quality"
+      ],
+      "extra": {
+        "endpoint": "/images/generations",
+        "supportedSizes": [
+          "1K",
+          "2K",
+          "4K"
+        ],
+        "maxFusionImages": 10
+      }
+    },
+    {
+      "modelName": "doubao-seedance-2-0-260128",
+      "displayName": "豆包 Seedance 2.0（文生/图生视频）",
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "字节 Seedance 视频模型，支持文生视频、图生视频、多图参考、首尾帧、参考视频/音频（多模态），异步任务",
+      "capabilities": [
+        "video_generation",
+        "text_to_video",
+        "image_to_video",
+        "first_last_frame",
+        "reference_video",
+        "reference_audio",
+        "chinese_optimized",
+        "high_quality"
+      ],
+      "extra": {
+        "endpoint": "/contents/generations/tasks",
+        "maxVideoDuration": 15,
+        "supportedResolutions": [
+          "720p",
+          "1080p"
+        ],
+        "maxReferenceImages": 9,
+        "asyncMode": true
+      }
+    }
+  ]
+}

compute/v2/compatibility/providers/xai.json

@@ -0,0 +1,143 @@
+{
+  "id": "provider-xai-001",
+  "provider": "xai",
+  "label": "xAI Grok",
+  "baseUrl": "https://api.x.ai/v1",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "xai",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "USD",
+  "services": [
+    "chat",
+    "reasoning"
+  ],
+  "tombstones": [
+    "grok-4-1-fast-reasoning"
+  ],
+  "models": [
+    {
+      "modelName": "grok-4.7",
+      "displayName": "Grok 4.7",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "xAI 面向编程与智能体任务的旗舰模型，500K 上下文，支持图像输入和可配置推理",
+      "contextWindow": 500000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "structured_output",
+        "long_context"
+      ],
+      "inputPrice": 2,
+      "outputPrice": 6,
+      "extra": {
+        "cachedInputPrice": 0.5,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high",
+            "xhigh"
+          ],
+          "defaultEffort": "high"
+        },
+        "pricingNotes": "Prices are per 1M tokens below 200K prompt tokens; above 200K, input and cache rates are 2x and output rates are 2x. No text output limit is documented."
+      }
+    },
+    {
+      "modelName": "grok-4.3",
+      "displayName": "Grok 4.3",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "xAI Grok 4.3 通用模型，1M 上下文，支持可配置推理、函数调用和结构化输出",
+      "contextWindow": 1000000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "vision",
+        "tool_use",
+        "structured_output",
+        "long_context"
+      ],
+      "inputPrice": 1.25,
+      "outputPrice": 2.5,
+      "defaultTemperature": 1,
+      "defaultTopP": 1,
+      "extra": {
+        "cachedInputPrice": 0.2,
+        "aliases": [
+          "grok-4.3-latest",
+          "grok-latest"
+        ],
+        "reasoning": {
+          "supportedEfforts": [
+            "none",
+            "low",
+            "medium",
+            "high"
+          ],
+          "defaultEffort": "low"
+        }
+      }
+    },
+    {
+      "modelName": "grok-build-0.1",
+      "displayName": "Grok Build 0.1",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "xAI 面向 agentic coding 工作流的专用代码模型",
+      "contextWindow": 256000,
+      "capabilities": [
+        "chat",
+        "code",
+        "agent",
+        "tool_use",
+        "structured_output"
+      ],
+      "inputPrice": 1,
+      "outputPrice": 2,
+      "defaultTemperature": 1,
+      "defaultTopP": 1,
+      "extra": {
+        "cachedInputPrice": 0.2
+      }
+    },
+    {
+      "modelName": "grok-4.20-0309-reasoning",
+      "displayName": "Grok 4.20 Reasoning (2026-03-09)",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "xAI Grok 4.20 固定版本推理模型，当前价格与 Grok 4.3 一致",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 16384,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "vision",
+        "tool_use",
+        "structured_output",
+        "long_context"
+      ],
+      "inputPrice": 1.25,
+      "outputPrice": 2.5,
+      "defaultTemperature": 1,
+      "defaultTopP": 1,
+      "extra": {
+        "cachedInputPrice": 0.2,
+        "legacy": true
+      }
+    }
+  ]
+}

compute/v2/compatibility/providers/xiaomi.json

@@ -0,0 +1,285 @@
+{
+  "id": "provider-xiaomi-001",
+  "provider": "xiaomi",
+  "label": "小米 MiMo",
+  "baseUrl": "https://api.xiaomimimo.com/v1",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "xiaomi",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "brandGroup": "xiaomi",
+  "priceCurrency": "CNY",
+  "accessMode": "api",
+  "services": [
+    "chat",
+    "reasoning",
+    "vision",
+    "tts",
+    "asr"
+  ],
+  "models": [
+    {
+      "modelName": "mimo-v2.6-pro",
+      "displayName": "MiMo-V2.6-Pro",
+      "serviceType": [
+        "chat",
+        "reasoning",
+        "vision"
+      ],
+      "description": "小米全模态旗舰推理模型，适合复杂项目与长程智能体任务",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "audio_understanding",
+        "video_understanding",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "inputPrice": 3,
+      "outputPrice": 6,
+      "extra": {
+        "cachedInputPrice": 0.025,
+        "thinkingDefault": true,
+        "pricingNotes": "CNY per 1M tokens on the Xiaomi MiMo API."
+      }
+    },
+    {
+      "modelName": "mimo-v2.6-flash",
+      "displayName": "MiMo-V2.6-Flash",
+      "serviceType": [
+        "chat",
+        "reasoning",
+        "vision"
+      ],
+      "description": "小米全模态高效推理模型，适合高频调用与规模化任务",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "audio_understanding",
+        "video_understanding",
+        "tool_use",
+        "agent",
+        "long_context",
+        "fast"
+      ],
+      "inputPrice": 1,
+      "outputPrice": 2,
+      "extra": {
+        "cachedInputPrice": 0.02,
+        "thinkingDefault": true,
+        "pricingNotes": "CNY per 1M tokens on the Xiaomi MiMo API."
+      }
+    },
+    {
+      "modelName": "mimo-v2.6-pro-ultraspeed",
+      "displayName": "MiMo-V2.6-Pro-UltraSpeed",
+      "serviceType": [
+        "chat",
+        "reasoning",
+        "vision"
+      ],
+      "description": "MiMo-V2.6-Pro 的超高速服务，需向小米申请定制配额",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "audio_understanding",
+        "video_understanding",
+        "tool_use",
+        "agent",
+        "long_context",
+        "fast"
+      ],
+      "inputPrice": 30,
+      "outputPrice": 60,
+      "extra": {
+        "cachedInputPrice": 0.25,
+        "thinkingDefault": true,
+        "pricingNotes": "CNY per 1M tokens on the Xiaomi MiMo API; UltraSpeed requires custom rate limits."
+      }
+    },
+    {
+      "modelName": "mimo-x-flash-preview",
+      "displayName": "MiMo-X-Flash Preview",
+      "serviceType": [
+        "chat",
+        "vision"
+      ],
+      "description": "小米 MiMo-X Flash 内测模型，1M 上下文，支持图片识别",
+      "contextWindow": 1000000,
+      "capabilities": [
+        "chat",
+        "vision"
+      ],
+      "extra": {}
+    },
+    {
+      "modelName": "mimo-x-pro-preview",
+      "displayName": "MiMo-X-Pro Preview",
+      "serviceType": [
+        "chat",
+        "vision"
+      ],
+      "description": "小米 MiMo-X Pro 内测模型，1M 上下文，支持图片识别",
+      "contextWindow": 1000000,
+      "capabilities": [
+        "chat",
+        "vision"
+      ],
+      "extra": {}
+    },
+    {
+      "modelName": "mimo-v2.5-pro",
+      "displayName": "MiMo-V2.5-Pro",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "小米 MiMo-V2.5-Pro，1M 上下文，1T MoE 参数，计划于 2026-10-21 下线",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "tool_use",
+        "code"
+      ],
+      "extra": {
+        "cachedInputPrice": 0.025
+      },
+      "inputPrice": 3,
+      "outputPrice": 6
+    },
+    {
+      "modelName": "mimo-v2.5-tts",
+      "displayName": "MiMo-V2.5-TTS",
+      "serviceType": [
+        "tts"
+      ],
+      "description": "小米最新高品质语音合成模型，内置多种精调音色，支持情感/语速/语调精细控制、方言和歌唱",
+      "capabilities": [
+        "tts",
+        "style_control",
+        "chinese_optimized"
+      ],
+      "extra": {
+        "voices": [
+          {
+            "id": "mimo_default",
+            "label": "MiMo 默认"
+          },
+          {
+            "id": "冰糖",
+            "label": "冰糖"
+          },
+          {
+            "id": "茉莉",
+            "label": "茉莉"
+          },
+          {
+            "id": "苏打",
+            "label": "苏打"
+          },
+          {
+            "id": "白桦",
+            "label": "白桦"
+          },
+          {
+            "id": "Mia",
+            "label": "Mia"
+          },
+          {
+            "id": "Chloe",
+            "label": "Chloe"
+          },
+          {
+            "id": "Milo",
+            "label": "Milo"
+          },
+          {
+            "id": "Dean",
+            "label": "Dean"
+          }
+        ],
+        "audioFormats": [
+          "wav",
+          "mp3",
+          "pcm16le"
+        ],
+        "sampleRate": 24000,
+        "endpoint": "/chat/completions",
+        "authHeader": "api-key"
+      }
+    },
+    {
+      "modelName": "mimo-v2.5-asr",
+      "displayName": "MiMo-V2.5-ASR",
+      "serviceType": [
+        "asr"
+      ],
+      "description": "小米 MiMo 语音识别模型，支持中英文提示与整段 WAV/MP3 识别",
+      "contextWindow": 8192,
+      "maxOutputTokens": 2048,
+      "capabilities": [
+        "asr",
+        "multilingual"
+      ],
+      "extra": {}
+    },
+    {
+      "modelName": "mimo-v2.5-tts-voicedesign",
+      "displayName": "MiMo-V2.5-TTS-VoiceDesign",
+      "serviceType": [
+        "tts"
+      ],
+      "description": "语音设计模型，通过一句话描述即可生成全新音色",
+      "capabilities": [
+        "tts",
+        "voice_design"
+      ],
+      "extra": {
+        "audioFormats": [
+          "mp3",
+          "wav"
+        ],
+        "sampleRate": 24000,
+        "endpoint": "/chat/completions",
+        "authHeader": "api-key"
+      }
+    },
+    {
+      "modelName": "mimo-v2.5-tts-voiceclone",
+      "displayName": "MiMo-V2.5-TTS-VoiceClone",
+      "serviceType": [
+        "tts"
+      ],
+      "description": "语音克隆模型，少量样本即可高保真复刻目标音色",
+      "capabilities": [
+        "tts",
+        "voice_clone"
+      ],
+      "extra": {
+        "audioFormats": [
+          "mp3",
+          "wav"
+        ],
+        "sampleRate": 24000,
+        "endpoint": "/chat/completions",
+        "authHeader": "api-key"
+      }
+    }
+  ]
+}

compute/v2/compatibility/providers/xunfei-x2.json

@@ -0,0 +1,41 @@
+{
+  "id": "provider-xunfei-x2-001",
+  "provider": "xunfei",
+  "label": "讯飞星火 X2",
+  "baseUrl": "https://spark-api-open.xf-yun.com/x2",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "xunfei",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "CNY",
+  "services": [
+    "chat",
+    "reasoning"
+  ],
+  "models": [
+    {
+      "modelName": "spark-x",
+      "displayName": "讯飞星火 X2",
+      "serviceType": [
+        "reasoning"
+      ],
+      "description": "讯飞深度推理模型（X2），输入64K/输出128K，支持思考模式与函数调用",
+      "contextWindow": 65536,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "math",
+        "code",
+        "tool_use"
+      ],
+      "defaultTemperature": 1.2,
+      "defaultTopP": 0.95,
+      "extra": {
+        "pricingNote": "官方公开文档说明模型价格以控制台及实际购买页为准，未公开固定 token 单价"
+      }
+    }
+  ]
+}

compute/v2/compatibility/providers/xunfei.json

@@ -0,0 +1,42 @@
+{
+  "id": "provider-xunfei-001",
+  "provider": "xunfei",
+  "label": "讯飞星火",
+  "baseUrl": "https://spark-api-open.xf-yun.com/v1",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "xunfei",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "CNY",
+  "services": [
+    "chat",
+    "reasoning"
+  ],
+  "tombstones": [
+    "spark-x"
+  ],
+  "models": [
+    {
+      "modelName": "4.0Ultra",
+      "displayName": "讯飞星火 4.0 Ultra",
+      "serviceType": [
+        "chat"
+      ],
+      "description": "讯飞星火最强通用模型，支持推理、代码和工具调用",
+      "contextWindow": 32768,
+      "maxOutputTokens": 32768,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "tool_use"
+      ],
+      "defaultTemperature": 1,
+      "defaultTopP": 1,
+      "extra": {
+        "pricingNote": "官方公开文档说明模型价格以控制台及实际购买页为准，未公开固定 token 单价"
+      }
+    }
+  ]
+}

compute/v2/compatibility/providers/zhipu-embedding.json

@@ -0,0 +1,39 @@
+{
+  "id": "provider-zhipu-embedding-001",
+  "provider": "zhipu-embedding",
+  "brandGroup": "zhipu",
+  "label": "智谱 Embedding",
+  "baseUrl": "https://open.bigmodel.cn/api/paas/v4",
+  "apiFormat": "openai-completions",
+  "apiKeyRef": "zhipu",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "CNY",
+  "services": [
+    "embedding"
+  ],
+  "models": [
+    {
+      "modelName": "embedding-3",
+      "displayName": "智谱 embedding-3",
+      "serviceType": [
+        "embedding"
+      ],
+      "description": "智谱嵌入模型v3，支持自定义维度；模型上下文窗口 8K，单条输入最多 3072 tokens。走 OpenAI 兼容 /embeddings 端点（Anthropic 协议不提供 embeddings）",
+      "contextWindow": 8192,
+      "capabilities": [
+        "text_embedding",
+        "semantic_search",
+        "rag",
+        "custom_dimensions"
+      ],
+      "inputPrice": 0.5,
+      "extra": {
+        "maxInputTokensPerItem": 3072,
+        "maxBatchItems": 64,
+        "vectorDimensions": "256-2048"
+      }
+    }
+  ]
+}

compute/v2/compatibility/providers/zhipu.json

@@ -0,0 +1,552 @@
+{
+  "id": "provider-zhipu-001",
+  "provider": "zhipu",
+  "label": "智谱 GLM",
+  "baseUrl": "https://open.bigmodel.cn/api/anthropic",
+  "apiFormat": "anthropic-messages",
+  "apiKeyRef": "zhipu",
+  "apiKeyVerified": false,
+  "enabled": false,
+  "status": "unconfigured",
+  "priceCurrency": "CNY",
+  "services": [
+    "chat",
+    "vision",
+    "reasoning",
+    "fast"
+  ],
+  "tombstones": [
+    "embedding-3",
+    "glm-4.7-thinking",
+    "glm-z1-air",
+    "glm-z1-airx",
+    "glm-z1-flash",
+    "glm-z1-flashx",
+    "glm-4-0520"
+  ],
+  "models": [
+    {
+      "modelName": "glm-5.3-flashx",
+      "displayName": "GLM-5.3-FlashX",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual",
+        "vision",
+        "video_understanding",
+        "tool_use",
+        "long_context",
+        "fast"
+      ],
+      "serviceType": [
+        "chat",
+        "reasoning",
+        "vision"
+      ],
+      "description": "Z.ai GLM-5.3-FlashX，高速多模态推理模型，100 万上下文，始终开启推理",
+      "extra": {
+        "thinkingOnly": true,
+        "cacheHitPrice": 0.57,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "max"
+        },
+        "pricingNotes": "国内官方标准价，人民币/百万 tokens；Flash 限时优惠以控制台实际结算为准。"
+      },
+      "inputPrice": 2,
+      "outputPrice": 7
+    },
+    {
+      "modelName": "glm-5.3-flash",
+      "displayName": "GLM-5.3-Flash",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "tool_use",
+        "agent",
+        "long_context",
+        "vision",
+        "image_understanding",
+        "video_understanding"
+      ],
+      "serviceType": [
+        "chat",
+        "reasoning",
+        "vision"
+      ],
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "description": "智谱 GLM-5.3-Flash 原生多模态模型，混合稀疏与线性注意力，面向高效编码与长程 Agent 任务，100 万上下文；强制思考，无法关闭",
+      "extra": {
+        "thinkingDefault": true,
+        "thinkingOnly": true,
+        "cacheHitPrice": 0.23,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "max"
+        },
+        "pricingNotes": "国内官方标准价，人民币/百万 tokens；Flash 限时优惠以控制台实际结算为准。"
+      },
+      "inputPrice": 0.8,
+      "outputPrice": 2.8
+    },
+    {
+      "modelName": "glm-5.3",
+      "displayName": "GLM-5.3",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "math",
+        "multilingual",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "serviceType": [
+        "chat"
+      ],
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "description": "智谱 GLM-5.3 大推理模型，面向复杂软件工程与长程 Agent 任务，100 万上下文；强制思考，无法关闭",
+      "extra": {
+        "thinkingDefault": true,
+        "thinkingOnly": true,
+        "cacheHitPrice": 2,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "max"
+        },
+        "pricingNotes": "国内官方标准价，人民币/百万 tokens；Flash 限时优惠以控制台实际结算为准。"
+      },
+      "inputPrice": 8,
+      "outputPrice": 28
+    },
+    {
+      "modelName": "glm-5.2",
+      "displayName": "GLM-5.2",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "智谱 GLM-5.2 最新旗舰模型，面向复杂长程任务与项目级工程交付，1M 上下文，128K 最大输出",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual",
+        "deep_thinking",
+        "long_context",
+        "math",
+        "tool_use",
+        "agent"
+      ],
+      "inputPrice": 8,
+      "outputPrice": 28,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cacheHitPrice": 2,
+        "cacheStoragePrice": 0,
+        "thinking": {
+          "default": "enabled"
+        },
+        "reasoning": {
+          "supportedEfforts": [
+            "max",
+            "xhigh",
+            "high",
+            "medium",
+            "low",
+            "minimal",
+            "none"
+          ],
+          "defaultEffort": "max"
+        }
+      }
+    },
+    {
+      "modelName": "glm-5.1",
+      "displayName": "GLM-5.1",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "智谱 GLM-5.1 旗舰模型，面向长程 Agentic Coding 与 Autonomous Agent 场景，200K 上下文，128K 最大输出",
+      "contextWindow": 200000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual",
+        "deep_thinking",
+        "long_context",
+        "math",
+        "tool_use",
+        "agent"
+      ],
+      "inputPrice": 6,
+      "outputPrice": 24,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cacheHitPrice": 1.3,
+        "cacheStoragePrice": 0,
+        "thinking": {
+          "default": "enabled"
+        },
+        "pricingTiers": [
+          {
+            "condition": "input_length_k_tokens: [0, 32)",
+            "inputPrice": 6,
+            "outputPrice": 24,
+            "cacheHitPrice": 1.3
+          },
+          {
+            "condition": "input_length_k_tokens: [32+)",
+            "inputPrice": 8,
+            "outputPrice": 28,
+            "cacheHitPrice": 2
+          }
+        ]
+      }
+    },
+    {
+      "modelName": "glm-5-turbo",
+      "displayName": "GLM-5-Turbo",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "智谱 GLM-5-Turbo，高阶快速模型，面向长链路任务优化，200K 上下文，128K 最大输出",
+      "contextWindow": 200000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "deep_thinking",
+        "long_context",
+        "tool_use",
+        "agent",
+        "fast"
+      ],
+      "inputPrice": 5,
+      "outputPrice": 22,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cacheHitPrice": 1.2,
+        "cacheStoragePrice": 0,
+        "decodeSpeed": "40-60 tokens/s",
+        "thinking": {
+          "default": "enabled"
+        },
+        "pricingTiers": [
+          {
+            "condition": "input_length_k_tokens: [0, 32)",
+            "inputPrice": 5,
+            "outputPrice": 22,
+            "cacheHitPrice": 1.2
+          },
+          {
+            "condition": "input_length_k_tokens: [32+)",
+            "inputPrice": 7,
+            "outputPrice": 26,
+            "cacheHitPrice": 1.8
+          }
+        ]
+      }
+    },
+    {
+      "modelName": "glm-5",
+      "displayName": "GLM-5",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "智谱 GLM-5，面向 Agentic Engineering 的旗舰基座模型，200K 上下文，128K 最大输出",
+      "contextWindow": 200000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual",
+        "deep_thinking",
+        "long_context",
+        "math",
+        "tool_use",
+        "agent"
+      ],
+      "inputPrice": 4,
+      "outputPrice": 18,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cacheHitPrice": 1,
+        "cacheStoragePrice": 0,
+        "decodeSpeed": "30-50 tokens/s",
+        "thinking": {
+          "default": "enabled"
+        },
+        "pricingTiers": [
+          {
+            "condition": "input_length_k_tokens: [0, 32)",
+            "inputPrice": 4,
+            "outputPrice": 18,
+            "cacheHitPrice": 1
+          },
+          {
+            "condition": "input_length_k_tokens: [32+)",
+            "inputPrice": 6,
+            "outputPrice": 22,
+            "cacheHitPrice": 1.5
+          }
+        ]
+      }
+    },
+    {
+      "modelName": "glm-4.7",
+      "displayName": "GLM-4.7",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "智谱 GLM-4.7，高智能模型，通用对话、推理与智能体能力全面升级，200K 上下文，128K 最大输出",
+      "contextWindow": 200000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual",
+        "deep_thinking",
+        "long_context",
+        "tool_use",
+        "agent"
+      ],
+      "inputPrice": 2,
+      "outputPrice": 8,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cacheHitPrice": 0.4,
+        "cacheStoragePrice": 0,
+        "decodeSpeed": "30-50 tokens/s",
+        "thinking": {
+          "default": "enabled",
+          "mode": "forced"
+        },
+        "pricingTiers": [
+          {
+            "condition": "input_length_k_tokens: [0, 32); output_length_k_tokens: [0, 0.2)",
+            "inputPrice": 2,
+            "outputPrice": 8,
+            "cacheHitPrice": 0.4
+          },
+          {
+            "condition": "input_length_k_tokens: [0, 32); output_length_k_tokens: [0.2+)",
+            "inputPrice": 3,
+            "outputPrice": 14,
+            "cacheHitPrice": 0.6
+          },
+          {
+            "condition": "input_length_k_tokens: [32, 200)",
+            "inputPrice": 4,
+            "outputPrice": 16,
+            "cacheHitPrice": 0.8
+          }
+        ]
+      }
+    },
+    {
+      "modelName": "glm-4.7-flashx",
+      "displayName": "GLM-4.7-FlashX",
+      "serviceType": [
+        "chat",
+        "fast"
+      ],
+      "description": "智谱 GLM-4.7-FlashX 轻量高速模型，适用于中文写作、翻译、角色扮演等通用场景，200K 上下文，128K 最大输出",
+      "contextWindow": 200000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "fast",
+        "long_context",
+        "multilingual"
+      ],
+      "inputPrice": 0.5,
+      "outputPrice": 3,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cacheHitPrice": 0.1,
+        "cacheStoragePrice": 0,
+        "decodeSpeed": "30-40 tokens/s"
+      }
+    },
+    {
+      "modelName": "glm-5v-turbo",
+      "displayName": "GLM-5V-Turbo",
+      "serviceType": [
+        "vision"
+      ],
+      "description": "智谱首个多模态 Agent 基座模型，支持视频、图像、文本和文件输入，200K 上下文，128K 最大输出",
+      "contextWindow": 200000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "vision",
+        "video_understanding",
+        "image_understanding",
+        "file_understanding",
+        "reasoning",
+        "code",
+        "deep_thinking",
+        "long_context",
+        "tool_use",
+        "agent"
+      ],
+      "inputPrice": 5,
+      "outputPrice": 22,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cacheHitPrice": 1.2,
+        "cacheStoragePrice": 0,
+        "decodeSpeed": "30-50 tokens/s",
+        "thinking": {
+          "default": "enabled"
+        },
+        "inputModalities": [
+          "image",
+          "video",
+          "file",
+          "text"
+        ],
+        "pricingTiers": [
+          {
+            "condition": "input_length_k_tokens: [0, 32)",
+            "inputPrice": 5,
+            "outputPrice": 22,
+            "cacheHitPrice": 1.2
+          },
+          {
+            "condition": "input_length_k_tokens: [32+)",
+            "inputPrice": 7,
+            "outputPrice": 26,
+            "cacheHitPrice": 1.8
+          }
+        ]
+      }
+    },
+    {
+      "modelName": "glm-4.6v",
+      "displayName": "GLM-4.6V",
+      "serviceType": [
+        "vision"
+      ],
+      "description": "智谱 GLM-4.6V 多模态模型，支持图像、视频、文本和文件理解，128K 上下文，32K 最大输出",
+      "contextWindow": 128000,
+      "maxOutputTokens": 32768,
+      "capabilities": [
+        "chat",
+        "vision",
+        "video_understanding",
+        "image_understanding",
+        "file_understanding",
+        "long_context",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 1,
+      "outputPrice": 3,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cacheHitPrice": 0.2,
+        "cacheStoragePrice": 0,
+        "decodeSpeed": "30-50 tokens/s",
+        "thinking": {
+          "default": "enabled"
+        },
+        "inputModalities": [
+          "image",
+          "video",
+          "file",
+          "text"
+        ],
+        "pricingTiers": [
+          {
+            "condition": "input_length_k_tokens: [0, 32)",
+            "inputPrice": 1,
+            "outputPrice": 3,
+            "cacheHitPrice": 0.2
+          },
+          {
+            "condition": "input_length_k_tokens: [32, 128)",
+            "inputPrice": 2,
+            "outputPrice": 6,
+            "cacheHitPrice": 0.4
+          }
+        ]
+      }
+    },
+    {
+      "modelName": "glm-4.6",
+      "displayName": "GLM-4.6",
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "智谱 GLM-4.6，擅长高级编码、复杂推理与工具调用，200K 上下文，128K 最大输出",
+      "contextWindow": 200000,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual",
+        "deep_thinking",
+        "long_context",
+        "tool_use"
+      ],
+      "inputPrice": 2,
+      "outputPrice": 8,
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "extra": {
+        "cacheHitPrice": 0.5,
+        "thinking": {
+          "default": "enabled"
+        }
+      }
+    }
+  ]
+}

compute/v2/manifest.json

@@ -0,0 +1,7 @@
+{
+  "formatVersion": 2,
+  "state": "candidate",
+  "requiredClientVersion": null,
+  "catalogPath": "compute/v2/catalog.json",
+  "digest": "f623289fb1e4f6854366094d18e619dc3a70b75e5214509b826abbe8f0557148"
+}

docs/README.md

@@ -10,3 +10,4 @@
 - [非模型来源与策略](data-sources/README.md)
 - [运行时官方来源](runtime-sources/README.md)
 - [官网缺失与来源不明独立清单](source-gaps/README.md)
+- [旧字段迁移、版本门控、域名退役与规格优化 PLAN](plans/catalog-contract-migration/PLAN.md)

docs/model-sources/MAINTENANCE.md

@@ -6,10 +6,10 @@
 
 - 核验日期：2026-10-03。
 - 配置范围：29 个 Provider、11 个套餐、23 个共享规格文件。
-- 当前模型记录：619 条（同一模型在不同接入面会分别计数）。
+- 当前模型记录：626 条（同一模型在不同接入面会分别计数）。
-- 346 条 `partial`：只核实表中列出的字段，可能仅确认官网出现该 ID；不是整条规格全通过。
+- 353 条 `partial`：只核实表中列出的字段，可能仅确认官网出现该 ID；不是整条规格全通过。
 - 272 条 `pending`：登记官网入口，但尚未取得该条模型的字段级证明。
-- 346 条 partial 中，214 条仅确认 ID，132 条还核实过其他字段；仍不代表整条通过。
+- 353 条 partial 中，216 条仅确认 ID，137 条还核实过其他字段；仍不代表整条通过。
 - 1 条 `historical`：匿名模型的历史兼容记录，不证明当前平台可用。
 
 没有把所有存量数据写成已核实。当前确认的是本轮明确核对的字段、官网公共目录以及本地配置契约；没有执行逐供应商账号的收费模型调用。

docs/model-sources/providers/alibaba/models/qwen3-coder-next.md

@@ -60,3 +60,21 @@
 }
 ```
 <!-- source-metadata:end -->
+
+## compute/model-specs/qwen.json：精确绑定
+
+仅登记既有接入精确身份；官网字段证据待核，不能把本记录当作完整能力合同。
+
+<!-- source-details: {"config":"compute/model-specs/qwen.json","id":"qwen3-coder-next"} -->
+```json
+{
+  "spec.description": "精确绑定已有接入型号，参数未核实；identity-only规格不继承旧family能力、窗口、协议或自动路由资格。"
+}
+```
+<!-- source-details:end -->
+
+<!-- source-config: compute/model-specs/qwen.json -->
+
+| 模型 ID | 上下文 / 输出 | 计价 | 已核字段→来源 | 来源入口 | 状态 | 数据指纹 |
+| --- | --- | --- | --- | --- | --- | --- |
+| `qwen3-coder-next` | 未声明 / 未声明 | 非计价主数据 | `id`→[qwen-coding](../SOURCES.md#qwen-coding) | [qwen-coding](../SOURCES.md#qwen-coding) | partial | `949ea8dd99a8c45948115ff0c61db7e27a400298c62bce1382db9cf8817eb00f` |

docs/model-sources/providers/alibaba/models/qwen3-coder-plus.md

@@ -60,3 +60,21 @@
 }
 ```
 <!-- source-metadata:end -->
+
+## compute/model-specs/qwen.json：精确绑定
+
+仅登记既有接入精确身份；官网字段证据待核，不能把本记录当作完整能力合同。
+
+<!-- source-details: {"config":"compute/model-specs/qwen.json","id":"qwen3-coder-plus"} -->
+```json
+{
+  "spec.description": "精确绑定已有接入型号，参数未核实；identity-only规格不继承旧family能力、窗口、协议或自动路由资格。"
+}
+```
+<!-- source-details:end -->
+
+<!-- source-config: compute/model-specs/qwen.json -->
+
+| 模型 ID | 上下文 / 输出 | 计价 | 已核字段→来源 | 来源入口 | 状态 | 数据指纹 |
+| --- | --- | --- | --- | --- | --- | --- |
+| `qwen3-coder-plus` | 未声明 / 未声明 | 非计价主数据 | `id`→[qwen-coding](../SOURCES.md#qwen-coding) | [qwen-coding](../SOURCES.md#qwen-coding) | partial | `ff635ac15fb8f0c970c6fb96543068b82b956d47c314031ec47d3b1b634259fa` |

docs/model-sources/providers/openai/SOURCES.md

@@ -57,6 +57,25 @@ GPT-6.1 Sol 的窗口、输出、推理档位与直连计价已再次核对。AP
 
 `extra.responsesOnly` 是为工具调用采用的保守协议收紧。官网要求工具调用走 Responses；普通无工具的 Chat Completions 仍支持，不能把此配置理解为原厂完全不支持 Chat Completions。
 
+### realtime21-spec
+
+- 入口：[realtime21-spec](https://developers.openai.com/api/docs/models/gpt-realtime-2.1.md)
+- 读取状态：`fetched`；只确认精确型号、总窗口与最大输出；不证明接入协议已验收。
+
+### realtime21mini-spec
+
+- 入口：[realtime21mini-spec](https://developers.openai.com/api/docs/models/gpt-realtime-2.1-mini.md)
+- 读取状态：`fetched`；只确认精确型号、总窗口与最大输出；不证明接入协议已验收。
+
+### realtimetranslate-spec
+
+- 入口：[realtimetranslate-spec](https://developers.openai.com/api/docs/models/gpt-realtime-translate.md)
+- 读取状态：`fetched`；只确认精确型号、总窗口与最大输出；不证明接入协议已验收。
+
+### openai-deprecations
+
+- 入口：[openai-deprecations](https://developers.openai.com/api/docs/deprecations.md)；本轮读取正文，Nano停止日2027-04-01，当前只deprecated。
+
 <!-- source-metadata:start -->
 ```json
 {
@@ -153,6 +172,46 @@ GPT-6.1 Sol 的窗口、输出、推理档位与直连计价已再次核对。AP
       "checkedAt": "2026-10-03",
       "contentSha256": "3dd20e2c95f09f3efb19c6255387a4e07d4d7c213367cca0a015e162a8f5b4e0",
       "resolvedUrl": "https://developers.openai.com/api/docs/models/gpt-5.4-nano.md"
+    },
+    {
+      "id": "realtime21-spec",
+      "url": "https://developers.openai.com/api/docs/models/gpt-realtime-2.1.md",
+      "kind": "official-doc",
+      "scope": "原厂精确模型、窗口与输出；接入协议独立验收",
+      "retrieval": "fetched",
+      "checkedAt": "2026-10-03",
+      "contentSha256": "4ab3318598ef8f29b919a3da13125b84b54c463ad71e67dee644aa6adc78a408",
+      "resolvedUrl": "https://developers.openai.com/api/docs/models/gpt-realtime-2.1.md"
+    },
+    {
+      "id": "realtime21mini-spec",
+      "url": "https://developers.openai.com/api/docs/models/gpt-realtime-2.1-mini.md",
+      "kind": "official-doc",
+      "scope": "原厂精确模型、窗口与输出；接入协议独立验收",
+      "retrieval": "fetched",
+      "checkedAt": "2026-10-03",
+      "contentSha256": "bec9dbe2d4ea92540cbc688f85a935916dbe6b0f3605405dc8ae2bc034e1094b",
+      "resolvedUrl": "https://developers.openai.com/api/docs/models/gpt-realtime-2.1-mini.md"
+    },
+    {
+      "id": "realtimetranslate-spec",
+      "url": "https://developers.openai.com/api/docs/models/gpt-realtime-translate.md",
+      "kind": "official-doc",
+      "scope": "原厂精确模型、窗口与输出；接入协议独立验收",
+      "retrieval": "fetched",
+      "checkedAt": "2026-10-03",
+      "contentSha256": "a95860e250772faa38ba78b08834abde51619679962f081c9c3b1b2f0636ab13",
+      "resolvedUrl": "https://developers.openai.com/api/docs/models/gpt-realtime-translate.md"
+    },
+    {
+      "id": "openai-deprecations",
+      "url": "https://developers.openai.com/api/docs/deprecations.md",
+      "kind": "official-doc",
+      "scope": "原厂宣布deprecated与正式停止服务日期分开；不把未来停止日当当前退役",
+      "retrieval": "fetched",
+      "checkedAt": "2026-10-03",
+      "contentSha256": "ac78c7daedd9543728f7a2c6ed453afce3bcbdcf590535fb4b65a7d2e7c43481",
+      "resolvedUrl": "https://developers.openai.com/api/docs/deprecations.md"
     }
   ]
 }

docs/model-sources/providers/openai/models/gpt-5.4-nano.md

@@ -79,3 +79,25 @@
 }
 ```
 <!-- source-metadata:end -->
+
+## compute/model-specs/openai.json：精确绑定
+
+官方规格页本轮核实窗口与输出；未扩大自动路由资格。
+
+<!-- source-details: {"config":"compute/model-specs/openai.json","id":"gpt-5.4-nano"} -->
+```json
+{
+  "spec.contextWindow": 400000,
+  "spec.maxOutputTokens": 128000,
+  "spec.description": "GPT-5.4 Nano；官网确认400000总窗口、128000最大输出。2026-10-01宣布deprecated，停止服务日2027-04-01；当前不据此退役。独立输入上限未由本页确认。"
+}
+```
+<!-- source-details:end -->
+
+<!-- source-config: compute/model-specs/openai.json -->
+
+| 模型 ID | 上下文 / 输出 | 计价 | 已核字段→来源 | 来源入口 | 状态 | 数据指纹 |
+| --- | --- | --- | --- | --- | --- | --- |
+| `gpt-5.4-nano` | 400000 / 128000 | 非计价主数据 | `id`→[gpt-5-4-nano-api](../SOURCES.md#gpt-5-4-nano-api)；`spec.contextWindow`→[gpt-5-4-nano-api](../SOURCES.md#gpt-5-4-nano-api)；`spec.maxOutputTokens`→[gpt-5-4-nano-api](../SOURCES.md#gpt-5-4-nano-api) | [gpt-5-4-nano-api](../SOURCES.md#gpt-5-4-nano-api) | partial | `f0253fa24e58c6c0ca971fa1e53fb87d5d140e843f09cfdab303b92ecf5e0664` |
+
+生命周期独立证据：[官方退役公告](../SOURCES.md#openai-deprecations)列出的停止日为2027-04-01；2026-10-01公告deprecated，不代表2026-10-03已不可调用。

docs/model-sources/providers/openai/models/gpt-5.4-pro.md

@@ -78,3 +78,23 @@
 }
 ```
 <!-- source-metadata:end -->
+
+## compute/model-specs/openai.json：精确绑定
+
+官方规格页本轮核实窗口与输出；未扩大自动路由资格。
+
+<!-- source-details: {"config":"compute/model-specs/openai.json","id":"gpt-5.4-pro"} -->
+```json
+{
+  "spec.contextWindow": 1050000,
+  "spec.maxOutputTokens": 128000,
+  "spec.description": "官方独立Pro规格；Responses协议限制由实际接入面执行，不沿用基础型号或其他代际family。"
+}
+```
+<!-- source-details:end -->
+
+<!-- source-config: compute/model-specs/openai.json -->
+
+| 模型 ID | 上下文 / 输出 | 计价 | 已核字段→来源 | 来源入口 | 状态 | 数据指纹 |
+| --- | --- | --- | --- | --- | --- | --- |
+| `gpt-5.4-pro` | 1050000 / 128000 | 非计价主数据 | `id`→[gpt-5-4-pro-api](../SOURCES.md#gpt-5-4-pro-api)；`spec.contextWindow`→[gpt-5-4-pro-api](../SOURCES.md#gpt-5-4-pro-api)；`spec.maxOutputTokens`→[gpt-5-4-pro-api](../SOURCES.md#gpt-5-4-pro-api) | [gpt-5-4-pro-api](../SOURCES.md#gpt-5-4-pro-api) | partial | `8782422b1170ce65415e84a43ff59f3668e49a4ece33e439676ef306a8c4cdbb` |

docs/model-sources/providers/openai/models/gpt-5.4.md

@@ -140,7 +140,7 @@
 
 | 模型 ID | 上下文 / 输出 | 计价 | 已核字段→来源 | 来源入口 | 状态 | 数据指纹 |
 | --- | --- | --- | --- | --- | --- | --- |
-| `gpt-5.4` | 1050000 / 128000 | 非计价主数据 | `id`→[gpt-5-4-api](../SOURCES.md#gpt-5-4-api)；`spec.contextWindow`→[gpt-5-4-api](../SOURCES.md#gpt-5-4-api)；`spec.maxOutputTokens`→[gpt-5-4-api](../SOURCES.md#gpt-5-4-api) | [gpt-5-4-api](../SOURCES.md#gpt-5-4-api) | partial | `0d0981e9f7c721fe920d732cd24fb349a81e9ef4ffec5d1c809c9bf315e09fa5` |
+| `gpt-5.4` | 1050000 / 128000 | 非计价主数据 | `id`→[gpt-5-4-api](../SOURCES.md#gpt-5-4-api)；`spec.contextWindow`→[gpt-5-4-api](../SOURCES.md#gpt-5-4-api)；`spec.maxOutputTokens`→[gpt-5-4-api](../SOURCES.md#gpt-5-4-api) | [gpt-5-4-api](../SOURCES.md#gpt-5-4-api) | partial | `cd59f54965cdd4dd1a82ca3045c38f8d9e97e6ccc20d81ba71454dfec7cab9ac` |
 
 本轮读取该模型官方明文规格页，确认原厂直连窗口、输出、标准价格与推理合同。Pro 仅限 Responses；GPT-5.4 Pro 官方默认 medium，修正旧 high。none 在产品层映射 off，schema 不接受将 none 写成 defaultEffort，故未伪造其他显式默认档。
 

docs/model-sources/providers/openai/models/gpt-realtime-2.1-mini.md

@@ -66,3 +66,23 @@
 }
 ```
 <!-- source-metadata:end -->
+
+## compute/model-specs/openai.json：精确绑定
+
+官方规格页本轮核实窗口与输出；未扩大自动路由资格。
+
+<!-- source-details: {"config":"compute/model-specs/openai.json","id":"gpt-realtime-2.1-mini"} -->
+```json
+{
+  "spec.contextWindow": 128000,
+  "spec.maxOutputTokens": 32000,
+  "spec.description": "官方精确型号与窗口/输出已核；独立实时协议适配与账号验收由接入面决定，不继承通用Chat合同。"
+}
+```
+<!-- source-details:end -->
+
+<!-- source-config: compute/model-specs/openai.json -->
+
+| 模型 ID | 上下文 / 输出 | 计价 | 已核字段→来源 | 来源入口 | 状态 | 数据指纹 |
+| --- | --- | --- | --- | --- | --- | --- |
+| `gpt-realtime-2.1-mini` | 128000 / 32000 | 非计价主数据 | `id`→[realtime21mini-spec](../SOURCES.md#realtime21mini-spec)；`spec.contextWindow`→[realtime21mini-spec](../SOURCES.md#realtime21mini-spec)；`spec.maxOutputTokens`→[realtime21mini-spec](../SOURCES.md#realtime21mini-spec) | [realtime21mini-spec](../SOURCES.md#realtime21mini-spec) | partial | `8483f11c3358f3850450c28725fdc9164005a3e17b8f9e9755a11020a7590d18` |

docs/model-sources/providers/openai/models/gpt-realtime-2.1.md

@@ -68,3 +68,23 @@
 }
 ```
 <!-- source-metadata:end -->
+
+## compute/model-specs/openai.json：精确绑定
+
+官方规格页本轮核实窗口与输出；未扩大自动路由资格。
+
+<!-- source-details: {"config":"compute/model-specs/openai.json","id":"gpt-realtime-2.1"} -->
+```json
+{
+  "spec.contextWindow": 128000,
+  "spec.maxOutputTokens": 32000,
+  "spec.description": "官方精确型号与窗口/输出已核；独立实时协议适配与账号验收由接入面决定，不继承通用Chat合同。"
+}
+```
+<!-- source-details:end -->
+
+<!-- source-config: compute/model-specs/openai.json -->
+
+| 模型 ID | 上下文 / 输出 | 计价 | 已核字段→来源 | 来源入口 | 状态 | 数据指纹 |
+| --- | --- | --- | --- | --- | --- | --- |
+| `gpt-realtime-2.1` | 128000 / 32000 | 非计价主数据 | `id`→[realtime21-spec](../SOURCES.md#realtime21-spec)；`spec.contextWindow`→[realtime21-spec](../SOURCES.md#realtime21-spec)；`spec.maxOutputTokens`→[realtime21-spec](../SOURCES.md#realtime21-spec) | [realtime21-spec](../SOURCES.md#realtime21-spec) | partial | `fe4e41e7e4fe5ac90c1d9eb640253765b7b16964a461a706e98ce12cda178de4` |

docs/model-sources/providers/openai/models/gpt-realtime-translate.md

@@ -63,3 +63,23 @@
 }
 ```
 <!-- source-metadata:end -->
+
+## compute/model-specs/openai.json：精确绑定
+
+官方规格页本轮核实窗口与输出；未扩大自动路由资格。
+
+<!-- source-details: {"config":"compute/model-specs/openai.json","id":"gpt-realtime-translate"} -->
+```json
+{
+  "spec.contextWindow": 16000,
+  "spec.maxOutputTokens": 2000,
+  "spec.description": "官方精确型号与窗口/输出已核；独立实时协议适配与账号验收由接入面决定，不继承通用Chat合同。"
+}
+```
+<!-- source-details:end -->
+
+<!-- source-config: compute/model-specs/openai.json -->
+
+| 模型 ID | 上下文 / 输出 | 计价 | 已核字段→来源 | 来源入口 | 状态 | 数据指纹 |
+| --- | --- | --- | --- | --- | --- | --- |
+| `gpt-realtime-translate` | 16000 / 2000 | 非计价主数据 | `id`→[realtimetranslate-spec](../SOURCES.md#realtimetranslate-spec)；`spec.contextWindow`→[realtimetranslate-spec](../SOURCES.md#realtimetranslate-spec)；`spec.maxOutputTokens`→[realtimetranslate-spec](../SOURCES.md#realtimetranslate-spec) | [realtimetranslate-spec](../SOURCES.md#realtimetranslate-spec) | partial | `e08efc78bf3d8f98e7a16988a3beb7f7a343c0f3e61da7bcc06914809750cd03` |

docs/plans/catalog-contract-migration/ADAPTER-DECISIONS.md

@@ -0,0 +1,62 @@
+# 原生协议适配核查决定
+
+本次是源码注册/真实派发路径核查，未作账号或媒体生成实测。机器明细：[ADAPTER-DECISIONS.json](ADAPTER-DECISIONS.json)。
+
+openai-responses/openai-completions只声明通用对话协议。Realtime WebSocket/实时翻译、原生图像/视频任务、批量ASR/TTS各有不同endpoint、请求/响应和流式状态机；同域/Bearer认证或同名apiFormat不证明模型协议适配。
+
+- 3条Realtime、21条媒体和12条ASR/TTS接入确认当前没有对应注册适配器，保留候选信息，标记protocol unsupported；不猜最低版本。
+- 图像/视频工具在findProvider后按实际provider slug查getAdapter，undefined立即返回失败，没有通用Provider.apiFormat兜底。
+- Speech Registry仅TTS/ASR；OpenAI兼容表只tts-1/tts-1-hd/whisper-1，三个Realtime型号既不在兼容表，也无独立session/WebSocket/翻译实现。
+- 已注册MiniMax/火山/百炼视频、云/NewAPI图像视频、六类语音协议未因本清单被误关。新型号是否完全符合这些已注册协议仍需其单型号协议/账号验收，当前核查没有把provider slug注册当全部型号实测。
+
+| 配置 | 型号 | 缺少适配 |
+| --- | --- | --- |
+| `compute/providers/dashscope.json` | `qwen-image-3.0` | no-registered-image_gen-provider-adapter |
+| `compute/providers/dashscope.json` | `qwen-image-3.0-pro` | no-registered-image_gen-provider-adapter |
+| `compute/providers/dashscope.json` | `paraformer-v2` | missing-speech-profile-and-compatibility |
+| `compute/providers/dashscope.json` | `wan2.7-image-pro` | no-registered-image_gen-provider-adapter |
+| `compute/providers/dashscope.json` | `wan2.7-image` | no-registered-image_gen-provider-adapter |
+| `compute/providers/dashscope.json` | `wan2.6-t2i` | no-registered-image_gen-provider-adapter |
+| `compute/providers/dashscope.json` | `wan2.2-t2i-plus` | no-registered-image_gen-provider-adapter |
+| `compute/providers/dashscope.json` | `wan2.2-t2i-flash` | no-registered-image_gen-provider-adapter |
+| `compute/providers/dashscope.json` | `qwen-image-2.0-pro` | no-registered-image_gen-provider-adapter |
+| `compute/providers/dashscope.json` | `qwen-image-2.0` | no-registered-image_gen-provider-adapter |
+| `compute/providers/kling.json` | `kling-v3` | no-registered-video_gen-provider-adapter |
+| `compute/providers/kling.json` | `kling-v2-5-turbo` | no-registered-video_gen-provider-adapter |
+| `compute/providers/kling.json` | `kling-v2-5-turbo-pro` | no-registered-video_gen-provider-adapter |
+| `compute/providers/kling.json` | `kling-v2` | no-registered-video_gen-provider-adapter |
+| `compute/providers/kling.json` | `kling-v2-master` | no-registered-video_gen-provider-adapter |
+| `compute/providers/local-whisper.json` | `whisper-large-v3` | missing-speech-profile-and-compatibility |
+| `compute/providers/minimax.json` | `speech-2.8-hd` | missing-speech-profile-and-compatibility |
+| `compute/providers/minimax.json` | `speech-2.8-turbo` | missing-speech-profile-and-compatibility |
+| `compute/providers/minimax.json` | `speech-2.6-hd` | missing-speech-profile-and-compatibility |
+| `compute/providers/minimax.json` | `speech-2.6-turbo` | missing-speech-profile-and-compatibility |
+| `compute/providers/minimax.json` | `speech-02-hd` | missing-speech-profile-and-compatibility |
+| `compute/providers/minimax.json` | `speech-02-turbo` | missing-speech-profile-and-compatibility |
+| `compute/providers/openai.json` | `gpt-image-2.5-sunburst` | no-registered-image_gen-provider-adapter |
+| `compute/providers/openai.json` | `gpt-image-2.5-flare` | no-registered-image_gen-provider-adapter |
+| `compute/providers/openai.json` | `gpt-image-2` | no-registered-image_gen-provider-adapter |
+| `compute/providers/openai.json` | `gpt-realtime-2.1` | no-model-specific-realtime-adapter |
+| `compute/providers/openai.json` | `gpt-realtime-2.1-mini` | no-model-specific-realtime-adapter |
+| `compute/providers/openai.json` | `gpt-realtime-translate` | no-model-specific-realtime-adapter |
+| `compute/providers/stability.json` | `stable-diffusion-3.5-large` | no-registered-image_gen-provider-adapter |
+| `compute/providers/xiaomi.json` | `mimo-v2.5-tts-voicedesign` | missing-speech-profile-and-compatibility |
+| `compute/providers/xiaomi.json` | `mimo-v2.5-tts-voiceclone` | missing-speech-profile-and-compatibility |
+| `compute/coding-plans/dashscope-token-plan.json` | `wan2.7-image-pro` | no-registered-image_gen-provider-adapter |
+| `compute/coding-plans/dashscope-token-plan.json` | `wan2.7-image` | no-registered-image_gen-provider-adapter |
+| `compute/coding-plans/dashscope-token-plan.json` | `qwen-image-3.0-pro` | no-registered-image_gen-provider-adapter |
+| `compute/coding-plans/minimax-coding.json` | `speech-2.8-hd` | missing-speech-profile-and-compatibility |
+| `compute/coding-plans/minimax-coding.json` | `speech-2.8-turbo` | missing-speech-profile-and-compatibility |
+
+## 真实派发证据
+
+- `packages/agent-service/src/media/adapters/registry.ts`：getImageAdapter/getVideoAdapter只查映射。
+- `packages/agent-service/src/builtin-tools/generate-image.ts`：execute调用后无adapter立即返回失败。
+- `packages/agent-service/src/builtin-tools/generate-video.ts`：同样无adapter返回失败。
+- `packages/speech-core/src/providers/index.ts`：六类已注册ASR/TTS协议。
+- `packages/speech-core/src/model-compatibility.ts`：精确OpenAI语音兼容表，不含Realtime。
+- `packages/agent-service/src/speech/speech-resolver.ts`：无speech profile/兼容项时fail，不猜协议。
+
+源文件路径相对客户端工作区 `/Users/wangyi/.codex/worktrees/catalog-contract-client/desirecore`；机器清单保存实际读取文件摘要，便于适配新增后重新核查。
+
+ASR/TTS额外12条由实际recognize/synthesize → resolveSpeechCompute → resolveProfile核实：无显式modelOrigin/speech且兼容表精确/前缀均不命中，立即报missing_model_profile。包括paraformer-v2、whisper-large-v3、MiniMax六种speech版本（含两条套餐副本）和MiMo voicedesign/voiceclone。不是据供应商slug缺失推断；也未发现local-whisper按名称另走本地执行的特殊分派。默认service-map引用whisper-large-v3仍须完成实际profile适配，不能仅凭默认映射声明ASR可用。

docs/plans/catalog-contract-migration/CLIENT-VERSIONS.md

@@ -0,0 +1,39 @@
+# 最低生效客户端版本
+
+[PLAN](PLAN.md)
+
+## 已确认与尚未确定
+
+| 项目 | 已核状态 |
+| --- | --- |
+| 最新正式release（本次读取） | v10.0.177，2026-09-27发布，非draft/prerelease，targetCommitish=dev；[发布页](https://github.com/desirecore/desirecore/releases/tag/v10.0.177) |
+| 主仓main | 412e05b043656d1f3ba2b56307ef5ceabbf8a77a，package.json仍为10.0.100，不能拿它或目录里的工作树作为已发布能力证明 |
+| 既有Provider要求 | anthropic-claude=10.0.83、github-copilot=10.0.132；其余38份Provider/套餐未声明。既有数字需在P0查当时发布包重新确认，不自动沿用 |
+| v10.0.177实际门控 | getProviderGating、前端选择器、InspectModels及Smart候选使用Provider级门控；不能据此宣布显式调用、媒体/后台任务全覆盖 |
+| Spec/模型/数据集版本门控 | 当前本仓model-specs schema没有对应字段；V2和迁移器仍需实现 |
+| V_contract | 待P1代码与正式发布确认的实际版本；本PLAN不猜10.0.178或其他未来版本 |
+
+## 版本计算
+
+P0为每种字段、协议、凭据、能力和迁移登记“实现commit、第一个包含该commit的正式tag、对应实际行为测试”。必要时以一个已验证的较新版本作保守下限，并明确这不是历史最早版本。不能用release标题、主仓版本字符串或schema存在代替实际出站行为。
+
+V_contract取支持整个V2合同和迁移器的首个已验收正式版本。每个接入模型的有效最低版本取 max(数据格式/迁移下限、Provider协议/凭据下限、模型绑定能力下限、相关退役规则实现下限)。适配器不存在时为“不可用/待适配”，不能填一个未来版本号假装支持。
+
+## 生效位置
+
+1. **加载前**：读取版本化manifest后，检查formatVersion与客户端兼容性，再解析/合并ModelSpec和Provider。遇到不支持的新格式保留最后已知可消费快照，不破坏compute.json和启动。
+2. **同步/保存**：受管条目写入、动态模型发现、构建默认包、编辑API和客户端迁移都检查；新模型自己的门槛不能因为Provider整体已支持而漏掉。
+3. **实际调用**：统一resolver在发请求、验证Key、余额检查、媒体submit/poll、后台摘要、SDK与工作流调用前再判定。调用方传明确providerId不能绕过。
+4. **UI/Agent**：同一结果派生“需更新客户端”及原因，不向Agent提供实际不可用候选；取消/升级路径不依赖隐藏整个供应商。
+5. **异常版本**：受管已发布数据的非法最低版本/未知格式拒绝；自建配置给明确错误和修复入口。当前compareSemver非法要求fail-open的行为需P1决定并测试，不直接一刀切覆盖所有用户数据。
+
+## 过渡方案
+
+- 旧客户端不认识新门控，单加requiredClientVersion不能阻止它读取新字段。因此先发布会识别V2目录和门控的客户端；V1路径保持受frozen schema约束的兼容投影/冻结快照。
+- 只支持Provider级门控的客户端如果需要精确到模型，应先升级；临时只能在不可安全拆分时提高该Provider整体门槛，清楚记录受影响的仍可用旧型号。默认方案为新客户端的逐模型门控。
+- 未铺开的客户端不能先收到已知enum扩值或删除required字段。现有frozen provider schema与manifest schema不直接改造成V2；新增版本化schema后分别校验。
+- 下发V2、停止旧写入、删除旧读取兜底和归档旧投影是四个独立动作。发布报告写明低于V_contract的行为与无法远程强制修复的旧二进制范围。
+
+## 跨版本验收
+
+覆盖门控前旧版、现有正式版177、V_contract-1、V_contract、最新版、缺失/非法版本、离线启动、不支持format、已有用户覆写，以及直接指定Provider/SDK/后台/媒体绕过反例。测试版本指构建产物有效__APP_VERSION__，不是测试源码package.json中的常量。

docs/plans/catalog-contract-migration/DOMAIN-RETIREMENT.md

@@ -0,0 +1,33 @@
+# 域名与接入下架计划
+
+[PLAN](PLAN.md)
+
+## 判定边界
+
+退役单位优先为 Provider/地域/协议/host/path，而不是整个供应商或根域。模型下架、官网换域、API路径迁移、暂时失败、停服域名是不同事件。DNS失败、证书错误、403/404、JS页面壳或公共目录缺少一个ID都不足以宣布整个域名停服。
+
+官方停止服务/迁移公告、官方文档明确替代端点与生效时间、官方支持确认是主证据；DNS/TLS/不带凭据的HTTP结果只能补充。是否已实际下架预设则用本仓Git历史证明，两类证据分开登记。
+
+## 首批清单
+
+| 对象 | 当前已知 | P0/P4动作 |
+| --- | --- | --- |
+| api.lingyiwanwu.com/v1；provider-lingyiwanwu-001 | 本仓commit a998e18（#84）已移除Provider和索引；v10.0.177退役名单仅见provider-internal-testing-001 | 查受管旧compute配置、别名/默认引用和启动迁移；补本产品接入退役记录。官方域名停止服务时间本轮未取得，不据此封锁整个lingyiwanwu.com及AgentBox子域 |
+| api.minimaxi.com 与 api.minimax.cn 等MiniMax域名 | 当前预设仍用minimaxi；来源文档出现minimax.cn。官方minimaxi平台仍可读，未取得旧API域名退役公告 | 核国内/国际和key账号范围、官方端点变更及媒体baseUrl，确认哪些是别名/迁移而非停服；不得仅按拼写统一域名。官方参考：[平台](https://platform.minimaxi.com/zh/cooperate)、[套餐入口](https://platform.minimaxi.com/subscribe/coding-plan) |
+| spark-api-open.xf-yun.com 的/v1、/v2、/x2 | X2已独立/x2，旧/v1保留Ultra，/v2同名spark-x指X1.5 | 保留按路径/型号限定的合同，不因X2推出封锁同一host上所有旧有效接入 |
+| 百川、Infini、摩尔线程、可灵、火山文档入口 | 87条记录的入口不可读；不是API域名退役证明 | 继续列为待核，获取公告/官方合同后决定，不使用读取失败自动下架 |
+| Perplexity Sonar→Agent API | 记录的是协议/接口迁移，旧同步/流式兼容与异步服务状态不同 | 核路径和请求形状，先适配Agent API；不是删除api.perplexity.ai根域 |
+
+实施P0时完整扫描baseUrl、mediaBaseUrl、coding/tokenPlan余额/额度endpoint、模型端点和taskQueryEndpoint、API-provider、云端下发端点、已发客户端内置包及受管存量配置。历史来源页与tools归档引用不计作活跃残留。
+
+## 退役注册与迁移
+
+新增版本化退役注册（建议字段）：规则ID、providerId/受管来源、exactHost、可选路径/地域/协议、产品退役理由、官方证据/状态、effectiveAt、替代接入、credential迁移边界、最低客户端版本。产品决定下架但厂家未停服时必须显式区分。
+
+- 仓库：移除活跃Provider/索引，清理services/默认映射；模型移除用tombstones。ModelSpec按原厂生命周期/历史用途处置，不因一个网关下架就删原厂规格。
+- 客户端：受管旧Provider退役/停用、缓存与动态发现过滤、任务轮询停止、默认引用修复；由统一出站resolver检查host/path，禁止透明redirect把Authorization带到未经确认的其他域。
+- 用户数据：保护自建Provider和用户覆写，不能根据label模糊删除；对确证停止服务的host阻止调用并给原因。默认保留Key引用和秘密值，不复用当前cleanupRetiredProviders的自动删Key行为去清所有厂商。
+- 替代端点：核相同主体/地域/账号与key适用范围后再给可审阅迁移；不自动把不同币种、不同套餐、不同接收域名的凭据转发过去。
+- 验收：退役规则生效前/后、离线/缓存、直接SDK和后台请求、同host其他有效路径、共享apiKeyRef、自建同名Provider、重复迁移、回滚全部覆盖。确认退役范围出站为零，但审计来源历史链接保留。
+
+当前仅能确认上述历史预置接入下架；需要P0补证之后才能给出完整“厂商已停止服务的域名”名单，本PLAN不伪造新域名退役结论。

docs/plans/catalog-contract-migration/EXECUTION.md

@@ -0,0 +1,28 @@
+# 实施任务与证据
+
+权威Plan：PLAN.md revision 3。以下为当前任务进度，不替代已核字段来源。
+
+| 阶段 | 状态 | 负责人 | 完成证据 |
+| --- | --- | --- | --- |
+| P0 字段/能力/来源清单 | 分类已完成；实际消费验收中 | 编译工作线与主工作线 | FIELD-DECISIONS、SPEC-DECISIONS，正式版本待发布验收 |
+| P1 客户端加载、迁移、版本与出站门控 | 实现及真实服务验证完成，等待合并发布 | 客户端工作线 | 340条解析一致、迁移及预算回归、版本426/退役400，客户端构建通过 |
+| P2 canonical/兼容编译与旧字段迁移 | 编译完成；存量迁移验收中 | 编译工作线 | 286 Spec/40接入/340模型，13项编译反例回归，完整V1投影一致 |
+| P3 规格精确身份/参数差异完善 | 已完成分类与有证修正 | 规格工作线 | 新增7条Spec，45差异及58spec-only均有决定；缺证保持unknown |
+| P4 受管退役接入/域名规则 | 实现及实际出站拦截验证完成 | 客户端工作线与主工作线 | 受管零一万物原host/path规则，保留凭据，自建接入不按名称删除 |
+| P5 正式客户端、数据发布和远端复查 | 等待P1验收 | 主工作线 | 不预填已发布版本 |
+
+保留原工作区logs；客户端原checkout的docs改动未触碰。客户端隔离worktree的子模块更新结果与无关gitlink将单独记录。
+
+客户端worktree子模块已按--remote更新：docs=b1394223efe77583330dbe51b961d868306b3256、drawio=88ee1f5ab4d40603a8f3aaa0d9d6dc3794c6b3f0。docs较初始dev gitlink前进，原checkout用户docs改动未触碰；本任务所需文档将在该已拉最新子模块上增量编辑后单独发布。
+
+## 截至2026-10-04的消费验证
+
+- 原始626来源记录维持353 partial/272 pending/1 historical，216仅ID，137有其他字段证明。缺证事实没有被编译伪装官方核验。
+- canonical/编译与V1完整投影16项反例通过；340条实际resolver结果与客户端比较，40接入/286规格Schema逐条通过。
+- 36条原生协议未适配（含原默认ASR）明确禁止调用但保留记录；3条API实参别名单独保留，不由metadata恢复入wire。
+- 客户端已通过changed typecheck消费闭包与完整构建；目前SDK真实正文预算、存量迁移、最后有效配置/启动、权限版本和retirement为范围回归，正式发布还未执行。
+- 实际token计数或保守文本份额检查与媒体估算区分；embedding逐item预算、ASR不把base64当文本token，多模态不是厂商精确usage验收。
+
+真实standalone重启与HTTP证明：higher-floor incoming目录不破坏合法local配置，GET config200保留2测试Provider，未来Provider changing-keyref返回426，retired managed原域changing-keyref返回400，原测试Key保留。证据仅提供fakeKeys，没有厂商账号调用。
+
+文档先行发布：desirecore-docs PR #535 已合并，客户端 docs gitlink 为 `3e600afc3b3da2686e83981f72090966b3ea0f93`。包含 ADR 207、算力模块说明及隔离真实服务验证报告。目录首批版本128为 candidate；客户端正式版本未核实前 requiredClientVersion 维持 null。

docs/plans/catalog-contract-migration/FIELD-DECISIONS.md

@@ -0,0 +1,155 @@
+# 字段归属与消费核查
+
+全部既有149条路径已分域；静态消费路径是核查候选，不宣称正式发布支持。无消费证明的值进入非运行时metadata，V1投影保留，不自动扩大能力。正式V2发布前须补实际客户端测试和版本。
+
+| 字段 | owner | 消费候选 | 发布版本 |
+| --- | --- | --- | --- |
+| `extra.adaptiveThinking` | request | packages/agent-service/src/compute/resolver.ts<br>packages/shared/src/compute/thinking-profiles.ts | 待正式发布 |
+| `extra.aliases` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.asyncHeader` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.asyncMode` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.audioCachedInputPrice` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.audioFormats` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.audioInputPrice` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.audioOutputPrice` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.audioOutputPricePerMinute` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.authHeader` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.cachedInputPrice` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.cacheHitPrice` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.cachePricing.audioInputCacheRead` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.cachePricing.inputCacheRead` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.cachePricing.inputCacheReadOver200k` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.cachePricing.read` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.cachePricing.storagePerMillionTokensPerHour` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.cachePricing.write1h` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.cachePricing.write5m` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.cacheReadPrice` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.cacheStoragePrice` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.cacheStoragePricePerMillionTokenHour` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.cacheWritePrice` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.codingPlanLongContextModelId` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.concurrencyLimit` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.decodeSpeed` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.defaultDimension` | intrinsic | 未识别；记录性隔离 | 待正式发布 |
+| `extra.deprecated` | lifecycle | packages/agent-service/src/compute/preset-merge.ts<br>packages/schemas/src/agent-service/preset-updates.ts<br>packages/shared/src/compute/model-availability.ts | 待正式发布 |
+| `extra.deprecationNotice` | lifecycle | 未识别；记录性隔离 | 待正式发布 |
+| `extra.dimensions` | intrinsic | 未识别；记录性隔离 | 待正式发布 |
+| `extra.endpoint` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.features.chatPrefixCompletion` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.features.fimCompletion` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.features.jsonOutput` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.features.toolCalls` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.forcedToolChoiceUnsupported` | intrinsic | packages/agent-service/src/compute/resolver.ts<br>packages/shared/src/compute/request-constraints.ts | 待正式发布 |
+| `extra.fps` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.imageCachedInputPrice` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.imageInputPrice` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.inputModalities` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.latencyMs` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.legacy` | lifecycle | 未识别；记录性隔离 | 待正式发布 |
+| `extra.maxAudioDuration` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.maxAudioLength` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.maxAudioSize` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.maxBatchItems` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.maxBatchSize` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.maxCloneSamples` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.maxFusionImages` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.maxInputTokensPerItem` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.maxLyricsLength` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.maxN` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.maxPromptLength` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.maxReferenceAudios` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.maxReferenceImages` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.maxReferenceVideos` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.maxTextLength` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.maxVideoDuration` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.mediaInputTypes` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.minCloneDuration` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.modelOrigin.canonicalModelId` | intrinsic | packages/agent-service/src/compute/speech-model-profile.ts<br>packages/schemas/src/agent-service/compute.ts | 待正式发布 |
+| `extra.modelOrigin.family` | intrinsic | packages/agent-service/src/compute/speech-model-profile.ts<br>packages/schemas/src/agent-service/compute.ts | 待正式发布 |
+| `extra.modelOrigin.vendor` | intrinsic | packages/agent-service/src/compute/speech-model-profile.ts<br>packages/schemas/src/agent-service/compute.ts | 待正式发布 |
+| `extra.outputFormats` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.outputSpeedTps` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.preserveThinkingDefault` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.pricePerGeneration` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.pricePerImage` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.pricePerMillionCharacters` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.pricePerSongUpToFiveMinutes` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.pricingNote` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.pricingNotes` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.pricingTiers` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.priorityPricingTiers` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.promptCacheMinTokens` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.reasoning.defaultEffort` | request | app/services/compute-api.ts<br>packages/agent-service/src/agent/query.ts<br>packages/agent-service/src/approval/ai-approve.ts | 待正式发布 |
+| `extra.reasoning.supportedEfforts` | request | app/services/compute-api.ts<br>packages/agent-service/src/agent/query.ts<br>packages/agent-service/src/approval/ai-approve.ts | 待正式发布 |
+| `extra.replacedBy` | lifecycle | 未识别；记录性隔离 | 待正式发布 |
+| `extra.requestPricingPer1k.high` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.requestPricingPer1k.low` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.requestPricingPer1k.medium` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.responseFormats` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.responsesOnly` | request | 未识别；记录性隔离 | 待正式发布 |
+| `extra.sampleRate` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.samplingParametersDeprecated` | intrinsic | packages/agent-service/src/compute/resolver.ts<br>packages/shared/src/compute/request-constraints.ts | 待正式发布 |
+| `extra.serverSideWebSearch.dialect` | request | packages/agent-service/src/auth/sapi-client.ts<br>packages/agent-service/src/compute/server-side-web-search.ts<br>packages/schemas/src/agent-service/compute.ts | 待正式发布 |
+| `extra.serverSideWebSearch.enabled` | request | packages/agent-service/src/auth/sapi-client.ts<br>packages/agent-service/src/compute/server-side-web-search.ts<br>packages/schemas/src/agent-service/compute.ts | 待正式发布 |
+| `extra.serverSideWebSearch.fallbackPriority` | request | packages/agent-service/src/auth/sapi-client.ts<br>packages/agent-service/src/compute/server-side-web-search.ts<br>packages/schemas/src/agent-service/compute.ts | 待正式发布 |
+| `extra.serverSideWebSearch.searchRequestPriceUsd` | billing | packages/agent-service/src/auth/sapi-client.ts<br>packages/agent-service/src/compute/server-side-web-search.ts<br>packages/schemas/src/agent-service/compute.ts | 待正式发布 |
+| `extra.serverSideWebSearch.toolType` | request | packages/agent-service/src/auth/sapi-client.ts<br>packages/agent-service/src/compute/server-side-web-search.ts<br>packages/schemas/src/agent-service/compute.ts | 待正式发布 |
+| `extra.sizeShortcuts.1K` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.sizeShortcuts.2K` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.sizeShortcuts.4K` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.snapshot` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.speech.defaultVoice` | intrinsic | packages/agent-service/src/compute/speech-model-profile.ts<br>packages/agent-service/src/speech/speech-resolver.ts<br>packages/speech-core/src/adapter-registry.ts | 待正式发布 |
+| `extra.speech.inputFormats` | intrinsic | packages/agent-service/src/compute/speech-model-profile.ts<br>packages/agent-service/src/speech/speech-resolver.ts<br>packages/speech-core/src/adapter-registry.ts | 待正式发布 |
+| `extra.speech.languageHints` | intrinsic | packages/agent-service/src/compute/speech-model-profile.ts<br>packages/agent-service/src/speech/speech-resolver.ts<br>packages/speech-core/src/adapter-registry.ts | 待正式发布 |
+| `extra.speech.modes` | intrinsic | packages/agent-service/src/compute/speech-model-profile.ts<br>packages/agent-service/src/speech/speech-resolver.ts<br>packages/speech-core/src/adapter-registry.ts | 待正式发布 |
+| `extra.speech.outputFormats` | intrinsic | packages/agent-service/src/compute/speech-model-profile.ts<br>packages/agent-service/src/speech/speech-resolver.ts<br>packages/speech-core/src/adapter-registry.ts | 待正式发布 |
+| `extra.speech.protocol` | intrinsic | packages/agent-service/src/compute/speech-model-profile.ts<br>packages/agent-service/src/speech/speech-resolver.ts<br>packages/speech-core/src/adapter-registry.ts | 待正式发布 |
+| `extra.speech.protocolVersion` | intrinsic | packages/agent-service/src/compute/speech-model-profile.ts<br>packages/agent-service/src/speech/speech-resolver.ts<br>packages/speech-core/src/adapter-registry.ts | 待正式发布 |
+| `extra.speech.sampleRates` | intrinsic | packages/agent-service/src/compute/speech-model-profile.ts<br>packages/agent-service/src/speech/speech-resolver.ts<br>packages/speech-core/src/adapter-registry.ts | 待正式发布 |
+| `extra.speech.voices` | intrinsic | packages/agent-service/src/compute/speech-model-profile.ts<br>packages/agent-service/src/speech/speech-resolver.ts<br>packages/speech-core/src/adapter-registry.ts | 待正式发布 |
+| `extra.streamingSupported` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportedAspectRatios` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportedAudioFormats` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportedFormats` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportedImageFormats` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportedImageSizes` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportedInputFormats` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportedLanguageCount` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportedLyricsStructures` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportedModes` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportedRatios` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportedResolutions` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportedSizes` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportedStyles` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportedVideoFormats` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportsChatApi` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportsColorPalette` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportsIncrementalStreaming` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportsIntelligentDuration` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportsNegativePrompt` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportsOutputAudio` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportsParallelToolCalls` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportsPromptExtend` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportsResponsesApi` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportsSeed` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportsThinking` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportsThinkingMode` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.supportsWatermark` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.taskQueryEndpoint` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.textCachedInputPrice` | billing | 未识别；记录性隔离 | 待正式发布 |
+| `extra.thinking.default` | request | packages/agent-service/src/approval/ai-approve.ts | 待正式发布 |
+| `extra.thinking.disabledTemperature` | request | packages/agent-service/src/approval/ai-approve.ts | 待正式发布 |
+| `extra.thinking.disableSupported` | request | packages/agent-service/src/approval/ai-approve.ts | 待正式发布 |
+| `extra.thinking.mode` | request | packages/agent-service/src/approval/ai-approve.ts | 待正式发布 |
+| `extra.thinking.preserveReasoningContent` | request | packages/agent-service/src/approval/ai-approve.ts | 待正式发布 |
+| `extra.thinkingDefault` | metadata | packages/agent-service/src/approval/constants.ts | 待正式发布 |
+| `extra.thinkingMaxTokens` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.thinkingOnly` | intrinsic | app/components/chat/ModelReasoningPill.tsx<br>packages/agent-service/src/approval/ai-approve.ts<br>packages/agent-service/src/approval/constants.ts | 待正式发布 |
+| `extra.thinkingRoundTrip.allowEmptySignature` | request | packages/agent-service/src/compute/thinking-roundtrip.ts | 待正式发布 |
+| `extra.thinkingRoundTrip.enabled` | request | packages/agent-service/src/compute/thinking-roundtrip.ts | 待正式发布 |
+| `extra.thinkingRoundTrip.protocol` | request | packages/agent-service/src/compute/thinking-roundtrip.ts | 待正式发布 |
+| `extra.thinkingRoundTrip.scope` | request | packages/agent-service/src/compute/thinking-roundtrip.ts | 待正式发布 |
+| `extra.thinkingToolTurnValidation` | request | packages/agent-service/src/compute/thinking-roundtrip.ts<br>packages/agent-service/src/runtime/thinking-tool-turn-guard.ts | 待正式发布 |
+| `extra.urlExpiry` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.vectorDimensions` | metadata | 未识别；记录性隔离 | 待正式发布 |
+| `extra.voices` | metadata | 未识别；记录性隔离 | 待正式发布 |

docs/plans/catalog-contract-migration/FIELD-INVENTORY.md

@@ -0,0 +1,157 @@
+# extra 字段全量盘点基线
+
+[字段迁移规则](FIELD-MIGRATION.md) · [PLAN](PLAN.md)
+
+2026-10-03，presetDataVersion 127。149种叶路径，233条记录包含extra。以下只是现存位置/数量；不能据此判断全部是旧字段。每项的消费点、正式版本、目标字段、语义和迁移决定在P0填写；已声明不消费的thinkingDefault已在专题中区分。数组作为一个路径，不把每个音色/价阶当新字段。
+
+| 字段路径 | Provider记录 | 套餐记录 | Spec记录 | P0处置 |
+| --- | --- | --- | --- | --- |
+| `extra.adaptiveThinking` | 10 | 1 | 8 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.aliases` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.asyncHeader` | 13 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.asyncMode` | 6 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.audioCachedInputPrice` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.audioFormats` | 3 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.audioInputPrice` | 3 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.audioOutputPrice` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.audioOutputPricePerMinute` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.authHeader` | 3 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.cachedInputPrice` | 19 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.cacheHitPrice` | 26 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.cachePricing.audioInputCacheRead` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.cachePricing.inputCacheRead` | 7 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.cachePricing.inputCacheReadOver200k` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.cachePricing.read` | 5 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.cachePricing.storagePerMillionTokensPerHour` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.cachePricing.write1h` | 5 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.cachePricing.write5m` | 5 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.cacheReadPrice` | 8 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.cacheStoragePrice` | 8 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.cacheStoragePricePerMillionTokenHour` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.cacheWritePrice` | 8 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.codingPlanLongContextModelId` | 0 | 1 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.concurrencyLimit` | 4 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.decodeSpeed` | 6 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.defaultDimension` | 4 | 0 | 4 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.deprecated` | 6 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.deprecationNotice` | 6 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.dimensions` | 1 | 0 | 4 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.endpoint` | 18 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.features.chatPrefixCompletion` | 4 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.features.fimCompletion` | 4 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.features.jsonOutput` | 4 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.features.toolCalls` | 4 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.forcedToolChoiceUnsupported` | 4 | 0 | 3 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.fps` | 2 | 0 | 2 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.imageCachedInputPrice` | 5 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.imageInputPrice` | 5 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.inputModalities` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.latencyMs` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.legacy` | 6 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.maxAudioDuration` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.maxAudioLength` | 3 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.maxAudioSize` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.maxBatchItems` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.maxBatchSize` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.maxCloneSamples` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.maxFusionImages` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.maxInputTokensPerItem` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.maxLyricsLength` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.maxN` | 7 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.maxPromptLength` | 4 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.maxReferenceAudios` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.maxReferenceImages` | 4 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.maxReferenceVideos` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.maxTextLength` | 6 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.maxVideoDuration` | 15 | 0 | 4 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.mediaInputTypes` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.minCloneDuration` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.modelOrigin.canonicalModelId` | 0 | 0 | 2 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.modelOrigin.family` | 0 | 0 | 2 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.modelOrigin.vendor` | 0 | 0 | 2 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.outputFormats` | 3 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.outputSpeedTps` | 6 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.preserveThinkingDefault` | 2 | 2 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.pricePerGeneration` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.pricePerImage` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.pricePerMillionCharacters` | 6 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.pricePerSongUpToFiveMinutes` | 3 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.pricingNote` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.pricingNotes` | 51 | 1 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.pricingTiers` | 21 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.priorityPricingTiers` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.promptCacheMinTokens` | 3 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.reasoning.defaultEffort` | 40 | 12 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.reasoning.supportedEfforts` | 44 | 12 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.replacedBy` | 6 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.requestPricingPer1k.high` | 3 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.requestPricingPer1k.low` | 3 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.requestPricingPer1k.medium` | 3 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.responseFormats` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.responsesOnly` | 3 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.sampleRate` | 3 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.samplingParametersDeprecated` | 7 | 0 | 8 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.serverSideWebSearch.dialect` | 19 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.serverSideWebSearch.enabled` | 19 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.serverSideWebSearch.fallbackPriority` | 19 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.serverSideWebSearch.searchRequestPriceUsd` | 19 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.serverSideWebSearch.toolType` | 19 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.sizeShortcuts.1K` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.sizeShortcuts.2K` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.sizeShortcuts.4K` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.snapshot` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.speech.defaultVoice` | 0 | 0 | 1 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.speech.inputFormats` | 0 | 0 | 1 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.speech.languageHints` | 0 | 0 | 1 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.speech.modes` | 0 | 0 | 2 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.speech.outputFormats` | 0 | 0 | 1 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.speech.protocol` | 0 | 0 | 2 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.speech.protocolVersion` | 0 | 0 | 2 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.speech.sampleRates` | 0 | 0 | 1 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.speech.voices` | 0 | 0 | 1 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.streamingSupported` | 4 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportedAspectRatios` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportedAudioFormats` | 3 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportedFormats` | 7 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportedImageFormats` | 7 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportedImageSizes` | 8 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportedInputFormats` | 3 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportedLanguageCount` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportedLyricsStructures` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportedModes` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportedRatios` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportedResolutions` | 15 | 0 | 2 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportedSizes` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportedStyles` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportedVideoFormats` | 4 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportsChatApi` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportsColorPalette` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportsIncrementalStreaming` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportsIntelligentDuration` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportsNegativePrompt` | 3 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportsOutputAudio` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportsParallelToolCalls` | 0 | 2 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportsPromptExtend` | 3 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportsResponsesApi` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportsSeed` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportsThinking` | 4 | 7 | 6 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportsThinkingMode` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.supportsWatermark` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.taskQueryEndpoint` | 13 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.textCachedInputPrice` | 5 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.thinking.default` | 13 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.thinking.disabledTemperature` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.thinking.disableSupported` | 3 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.thinking.mode` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.thinking.preserveReasoningContent` | 3 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.thinkingDefault` | 9 | 9 | 14 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.thinkingMaxTokens` | 6 | 2 | 1 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.thinkingOnly` | 8 | 4 | 10 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.thinkingRoundTrip.allowEmptySignature` | 10 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.thinkingRoundTrip.enabled` | 10 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.thinkingRoundTrip.protocol` | 10 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.thinkingRoundTrip.scope` | 10 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.thinkingToolTurnValidation` | 10 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.urlExpiry` | 2 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.vectorDimensions` | 1 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |
+| `extra.voices` | 4 | 0 | 0 | 查消费点、归属和语义后定keep/migrate/metadata/retire |

docs/plans/catalog-contract-migration/FIELD-MIGRATION.md

@@ -0,0 +1,40 @@
+# 字段迁移：全量清单与规则
+
+[PLAN](PLAN.md) · [149种现有extra路径清单](FIELD-INVENTORY.md)
+
+## 全量的含义
+
+覆盖 Provider、套餐、ModelSpec、索引/manifest、API-provider合同、客户端本地compute.json、构建内置包、云端同步结果、编辑API和UI保存路径。每个字段登记“来源位置 → schema → 写入点 → 运行时读取点 → 首个正式支持版本 → 新位置 → 冲突规则 → 迁移方式 → 测试”。仅rg命中注释、schema或i18n不算证明字段被运行时消费。
+
+当前149种extra叶路径是盘点起点，不是149个废弃字段。所有路径都须归类：有效规范字段保留；旧别名映射；记录性字段转metadata；接入协议专用字段有独立owner；未识别字段隔离待审。禁止为了“全迁移”把有意义的有效字段重新命名一遍。
+
+## 首批映射与禁止的转换
+
+以下目标中除已存在的extra.reasoning、ModelSpec.extra.speech等外，其余是待P1客户端/schema实现的V2结构建议，不得现在写入预设。
+
+| 当前字段/形态 | 分类/目标 | 转换约束 |
+| --- | --- | --- |
+| extra.reasoningEffort、defaultReasoningEffort、扁平defaultEffort | 已清除仓内残留；存量受管配置迁至extra.reasoning的明确支持集合/默认值 | API编辑DTO的reasoningEffort可能是有效输入名，不按同名全部删除；必须追到落盘位置。新旧同存时规范字段优先并记录冲突，不丢掉high/none |
+| thinkingDefault、supportsThinking、thinkingMaxTokens、extra.thinking.* | 分开处理官方默认说明、固有“不能关闭”、接入开关/预算及本仓默认；归入相应V2思考/请求约束或metadata | v10.0.177的thinkingDefault明确不改变行为，true不能推导effort或thinkingOnly；supportsThinking没有消费证明前不用于扩大能力 |
+| cachedInputPrice、cacheHitPrice、cacheReadPrice、cachePricing.read | 接入面V2 billing/cache-read价签，同币种/模态/计价单位才可规范 | 原价、平台价、加价后的价格分开；不混USD/CNY或每token/每百万token；有阶梯/过期条件必须保留 |
+| cacheWritePrice、cachePricing.write5m/write1h、cacheStoragePrice* | 分别规范为写入TTL价/缓存存储价，仍属接入面 | 5分钟与1小时写价不能合并；缓存存储计费必须保留时间维度 |
+| pricePerImage、pricePerGeneration、pricePerMillionCharacters、pricePerSong*、audioOutputPricePerMinute、requestPricingPer1k | 接入面统一计价条目，保留unit、currency、modality、tier、effective范围 | 不生成虚假的inputPrice/outputPrice；未知价缺省，动态API -1不变成免费 |
+| Provider.contextWindow/maxOutputTokens/defaultTemperature/defaultTopP等规格副本 | 作者数据迁到canonical规格；合法平台差异改为显式access override；旧Provider字段由编译投影生成 | 45处不同值不是45个错误；先分型。最大输入、总窗口、输出独立，不能output<=input一刀切 |
+| audioFormats、supportedAudioFormats、supportedFormats、outputFormats、voices、sampleRate、endpoint、authHeader等 | 先按ASR/TTS/图像/视频具体语义分组；稳定语音profile可复用extra.speech；端点/认证留Provider协议绑定 | 同名格式不同模态不能合并；声音ID/采样率/默认声音、batch与streaming、协议版本均须保留；不能把http auth字段移进原厂身份 |
+| dimensions、vectorDimensions、defaultDimension | 规范embedding维度与接入限制，原厂可选集合和当前端点支持集合分别记录 | Cohere原生Embed维度不能让compatibility端点发送dimensions；默认返回维度不等于可选择该参数 |
+| legacy、deprecated、replacedBy、deprecationNotice、retiredAt | V2模型/接入生命周期与本仓兼容状态分开 | legacy不等于退役；客户端软deprecated不等于原厂停服；retiredAt当前仅记录性，需P1运行时门控 |
+| ModelSpec内的reasoning/thinkingRoundTrip/thinkingToolTurnValidation/serverSideWebSearch | 不属于原厂统一规格，回到接入面合同；已有合法声明保留 | 已发布applySpec会排除这些接入键，不能通过“补Spec”来使其生效；平台搜索价同样留接入面 |
+
+## 存量迁移
+
+1. 基于正式发布包与实际读写路径生成dry-run，不仅扫描本仓JSON；受管preset、synced/cloud、自建和userOverrides分别处理。
+2. 新客户端读旧值后内存规范化，运行时只读规范结构；保存时仅写新结构，记录migrationEpoch并保留一次可恢复的本地基线。基线包含敏感配置时沿用已有保护，不提交或回显。
+3. 规范值存在优先保留；无法判断的旧值进入待处理记录，不用静默覆盖来制造幂等。用户覆写与厂商原值分别保存。
+4. 第二次迁移结果应逐字段一致；自建ID、providerId、apiModelId、凭据引用、币种/单位和用户选择保持。未覆盖的任意自定义extra不按名称猜测删除。
+5. 回滚时必须能从新结构安全恢复受支持的旧投影，且不能让退役域名或停服模型再次可调用。兼容读取退出与历史文件保留分别制定期限。
+
+## 防回流
+
+增加V2旧键拒写名单和“新字段有消费点/版本声明”校验；对受管数据的未知行为键失败，不依赖开放extra静默忽略。记录性metadata与自定义参数扩展留各自明确边界。CLI、UI保存、Provider发现/同步和云端pricing转换都必须使用同一规范化入口。
+
+V1投影继续遵守frozen契约：defaultTemperature/defaultTopP不能写null，必填字段和已知enum不能提前删除/扩展。V2明确“不支持采样”后由新客户端请求约束落实，不能把Spec里的null直接复制到旧Provider造成整份配置拒收。

docs/plans/catalog-contract-migration/MODEL-SPECS.md

@@ -0,0 +1,50 @@
+# ModelSpec 完善与 Provider 取舍
+
+[PLAN](PLAN.md) · [真实匹配/差异基线](SPEC-BASELINE.md)
+
+## 方案选择
+
+| 方案 | 收益 | 问题 | 决策 |
+| --- | --- | --- | --- |
+| Provider和Spec各自完整手工维护 | 兼容现有客户端，修改直接 | 相同事实重复、45处数值差异难追踪，不能区分正确覆盖和错误 | 退出手工双写 |
+| 只保留Provider | 端点/账号/价格可直接读取 | 原厂身份/参数重复到各套餐，动态云目录无统一规格，匹配和路由继续靠名字猜测 | 不采用 |
+| 只保留ModelSpec | 原厂规格集中 | 无法代表不同地域/套餐价格、凭据、协议、实际可用模型；spec-only不是可调用 | 不采用 |
+| ModelSpec canonical + 接入绑定/覆盖 + 自动编译投影 | 各事实只维护一次，可解释平台差异，兼顾动态发现和旧客户端 | 需要客户端resolver、版本化schema和生成工具先实现 | 推荐 |
+
+## 作者数据和运行时对象
+
+建议V2增加版本化目录（如compute/v2/model-specs、providers、coding-plans、retirements及manifest，具体路径P1确定），保留根目录V1兼容投影。新增的字段/目录都是设计提案，不是当前已实现API。
+
+- **ModelSpec**：稳定modelRef、vendor/canonicalModelId、明确版本/变体、exact与经过证实的aliases、原厂spec、逐字段证据、生命周期；routing仍为模型域中的明确产品政策，来源不能伪装成官网参数。
+- **AccessModel**：实际apiModelId/modelName、指向modelRef的绑定、是否当前可用、区域/套餐、协议/adapterVersion、认证/媒体/任务端点、billing、已证实的limit/capability/request override、所需客户端版本。
+- **EffectiveModel**：客户端/编译器按同一规则计算；identity/请求ID分开，原厂声明/平台合同/用户覆写分开，并带字段来源、覆盖理由和匹配层级用于诊断。用户配置读取和云端同步使用同一resolver。
+
+旧Provider中必须的displayName/serviceType/capabilities等并不马上删；生成器从V2对象投影到frozen V1 schema。V2作者数据不再人工复制它们。一个编译输入产生两个按版本选择的格式，兼容副本不是第二份真相源。
+
+## 有效值规则
+
+| 字段类别 | 合成规则 |
+| --- | --- |
+| 身份 | access精确绑定优先；实际请求ID原样保留。显示别名不等于原厂调用实参；不通过剥离preview/free/日期等自动宣布两个型号是同一合同 |
+| 输入/窗口/输出 | 每个限制独立表达。已证实的access限定值优先于通用原厂默认，并保留理由/证据；请求预算再受有效上限约束。平台经验证可有不同上下文、tokenizer或托管变体，不用机械min/全相等抹掉差异 |
+| 能力 | 原厂事实与已验证接入开放范围组合；接入明确不支持时为不可用，不做capabilities并集。未声明/待核是unknown，不能自动转成false或全支持；用户标签不扩大官方受管调用权限 |
+| 思考模式 | routing的标准模式与接入supportedEfforts映射/集合求交，固有不能关闭约束仍成立；原厂默认、平台默认和产品默认分别记录。none与off的产品映射不直接改成所有上游相同参数 |
+| 请求限制 | 模型拒绝采样/强制工具等固有限制按精确身份生效；adaptive/budget/Responses形状、thinking回放、平台搜索与参数名称由协议绑定落实。 family不得带高风险约束到不同代际 |
+| 价格 | 直连/套餐从各自官方价签，动态云端从实际计价平台合同；billing唯一在access，Spec无价格。不要沿用“所有价格都来自NewAPI”来描述直连预设 |
+| 用户覆写 | 保留显式合法偏好；不能突破上游硬上限/停服/客户端门槛。覆写的本地数值不冒充官方证明；冲突给可理解原因 |
+| 推断兜底 | family仅展示/保守补缺，不能新增Agent资格、未知工具/模态或强制协议转换。未命中精确合同的动态模型保持未知或授权探测，不自动注册虚假Spec |
+
+当前v10.0.177 applySpec在普通精确合并可覆盖数值，但provider-management/compute-routes强制inferred=true只补缺；capabilities却始终并集，extra整体浅合并。P1要统一这些路径，给每种字段明确合成规则，避免不同入口拿到不同能力。不能只改JSON就宣布合成语义已经统一。
+
+## 完善内容与顺序
+
+1. **身份与匹配优先**：补GPT-5.4 Nano独立规格、收窄gpt-5.4 pattern；17未匹配和7family逐条分类，见基线。普通固定型号应有精确绑定，ark-code-latest等动态套餐别名必须通过其实际目录/合同解析，不能永久绑定某个猜测原厂。
+2. **完整输入预算**：新增独立maxInputTokens与总context、maxOutput字段；窗口单位、tokenizer、独立限制/组合预算、默认max_tokens与reasoning预算均分开。讯飞不能拿65536输入证明总窗口；未知总窗口不补造。
+3. **45处差异分类**：A官方错误需修，B有证据的平台覆盖需保留，C单位换算需明确，D旧假定/没有证据需待核。每项登记决定与原值/新值，单供应商提交；不批量复制最大值。
+4. **模态/工具与协议**：明确输入/输出模态、tools/结构化结果、native/transcode路径和适配状态；Spec描述型号并不宣布客户端已适配。原生TTS/Live/转写/视频/Rerank完善规格后，仍须独立接入验收。
+5. **生命周期可执行**：preview、active、deprecated、retired、historical、unknown及其证据分开；原厂生命周期与某接入账户/地域/套餐开放状态分开。对已确认retired/未适配/缺最低版本的对象禁止自动路由和出站。
+6. **路由完整性**：仅当前access可用、版本支持、合同足够且用户策略允许的精确型号成为候选；metadata和ID存在不授权调用。pending可保留检索/手动候选与说明，不借补Spec扩大自动权限。
+7. **58条无静态接入规格**：分别标记动态云目录服务、family保守兜底、历史兼容、待协议适配、可能孤立。只有真正无使用方且原厂/别名生命周期已核才归档，不用静态Provider是否存在直接删除。
+8. **缓存更新**：Spec/绑定/生命周期的变更用catalog digest或明确版本失效；退出依赖修改_index描述来触发mtime的维护惯例。原地换文件、离线缓存和回滚均须验证。
+
+来源仍沿用每供应商SOURCES、单模型、单接入和source-gaps细分；新增字段的每一条已核声明要能指向确切接入面证据。不将当前132条部分参数已核自动标成完整Spec。

docs/plans/catalog-contract-migration/PLAN.md

@@ -0,0 +1,80 @@
+---
+plan_path: /Users/wangyi/.codex/worktrees/c0a8/config-center/docs/plans/catalog-contract-migration/PLAN.md
+revision: 3
+related_workdirs:
+  - /Users/wangyi/.codex/worktrees/c0a8/config-center
+  - /Users/wangyi/.codex/worktrees/catalog-contract-client/desirecore
+---
+# PLAN：旧字段迁移、客户端门控、域名退役与模型规格优化
+
+状态：实施中。用户已要求逐一完成；先做P0/客户端P1，并行完成编译与规格治理，客户端实际发布后再启用新数据。P0、P1、P2、P3、P4、P5的任务和证据记录见 EXECUTION.md。
+
+## 目标与基线
+
+- config-center 基线：main `4f4e5093e91ead6c9f01aacc188d23ea4b5e8917`，presetDataVersion 127。
+- 已发布客户端核查对象：[v10.0.177](https://github.com/desirecore/desirecore/releases/tag/v10.0.177)，tag commit `dd5a8c9f2fb000f7b2c589ec092068874c7dec5c`。release 标记为正式版本，发布时间 2026-09-27，targetCommitish 为 dev；主仓 main `412e05b...` 不能代替该发布基线。
+- 范围：29 份 Provider、11 份套餐、23 份规格文件；340 条接入模型记录、279 条 ModelSpec，共 619 条来源记录。
+- 当前 40 份接入配置仅 2 份有 requiredClientVersion。149 种 extra 叶路径出现在 233 条记录中；149 种不等于 149 个旧字段，全部需要完成消费/归属分类。
+- 用已发布客户端的真实匹配器核对：303 exact、3 normalized、1 stripped、9 pattern、7 family、17 未匹配；45 条记录存在 contextWindow/maxOutputTokens 差异。58 条规格未被静态接入记录命中，不能据此认定无用，可能服务云端动态目录、历史身份或待适配协议。
+- 本轮已迁移的 14 个无效推理键不再残留；下一阶段不能把“旧键告警为零”当作全部字段已规范。
+
+## 推荐取舍
+
+保留 Provider 与 model-specs。推荐“规格单处维护 + 接入面显式覆盖 + 编译得到有效配置与兼容副本”。不要合并成一个巨大 Provider 文件，也不要删掉 Provider 只留下规格。
+
+| 域 | 唯一作者数据 | 不应承担 |
+| --- | --- | --- |
+| ModelSpec | 精确模型身份、原厂已核事实、输入/输出独立限制、模态与固有限制、模型级路由策略、证据和生命周期 | endpoint、凭据、账号开放名单、平台价格、接入 effort、thinking 回放或平台 WebSearch |
+| Provider / 套餐 | 地域、端点、凭据、账号可用模型、协议/适配器、计价单位与价签、显式接入限制和 API 参数映射 | 再次手工复制所有模型事实、靠 family 推定新型号能力 |
+| 编译器 / 客户端 resolver | 按字段规则组合 ModelSpec 与接入覆盖，记录每个有效值的来源，生成旧客户端副本 | 私自补 unknown、合并不同单位或把完整能力列表取并集 |
+| 用户配置 | 用户选项、合法覆写和自建接入 | 被远端预设覆盖或被按名称猜测的迁移修改 |
+
+格式路线：新增版本化 V2 目录/manifest/schema 与客户端读取选择器，旧路径维持冻结的 V1 兼容投影。新版本原始作者数据不再保留运行时旧字段；兼容投影由生成器产出，过渡期不手工双写。完成客户端铺开后停止更新旧投影并归档，移除条件必须明确。具体见[规格与接入职责设计](MODEL-SPECS.md)。
+
+## 执行阶段
+
+| 阶段 | 工作 | 产物 | 进入下一阶段的条件 |
+| --- | --- | --- | --- |
+| P0：冻结清单与语义 | 完整盘点仓内/存量客户端字段；核149种extra路径的定义和真正读写位置；分型45条参数差异、17个无匹配和7个family；核域名下架事实 | 字段映射矩阵、版本能力矩阵、域名退役候选、来源缺口和dry-run报告 | 每个字段都有归属与处置，未知项独立登记；不能批量猜值 |
+| P1：客户端先行 | V2 schema/目录选择、全局合同版本门控、Provider/模型/适配器门控、统一有效配置resolver、配置迁移与回滚；补退役域名出站拦截 | 客户端代码PR、跨版本/存量配置测试、正式发布 V_contract | 真正正式发布且验收支持新合同的版本存在；版本号不能用未发布main或猜10.0.178填写 |
+| P2：数据与字段迁移 | 迁移受管旧字段；消除重复作者字段；保留有依据的接入覆盖；按冻结schema生成V1投影；同步来源/缺口文档 | V2 canonical、编译器、字段移除清单、来源登记、必要manifest递增 | 全部受管旧运行时键在V2写入端为零；迁移幂等、无静默丢值、价格/单位/覆写保持；V1兼容检查通过 |
+| P3：ModelSpec完善 | 先修精确身份、错误pattern、独立输入预算，再逐条解决45处数值差异；补生命周期/适配状态；逐家核官网 | 单供应商PR、已核字段、显式override及自动路由资格清单 | 每条启用模型有精确绑定或明确动态或仅手动例外；客户端原生适配未实现则保留记录并明确禁止调用，不能因为API格式名称已识别而宣称支持；差异都分类为正确覆盖/错误/待核，不能要求全相等 |
+| P4：域名退役 | 对已确认退役的接入移除活跃引用与索引；模型墓碑、Provider退役注册、存量停用、缓存/任务/默认回退同步 | 域名注册表、客户端退役迁移、活跃引用零残留与出站反例测试 | 确认范围内不再发送凭据或轮询；历史来源链接可保留；不会误伤同域其他有效API和用户自建数据 |
+| P5：发布与退出兼容 | 在客户端门控已经生效后发布V2数据，校验正式版本和全部来源；审查精确head、CI、review线程，再合并和远端复核 | 发布批次报告、各版本行为矩阵、剩余待核列表、V1退出记录 | 低版本不会应用新合同；当前版本和新版通过回归；没有把Schema通过当账号/协议实测 |
+
+P4 的实现机制可在 P1 并行开发；确认过的遗留退役接入可单独走窄范围修复，不必等所有官网字段核完。大范围字段删除和V2切换必须经过 P1。
+
+## 首批优先事项
+
+1. **P0/P1，版本门控**：40份接入及279条规格声明其能力依赖。以已发布tag检查功能，证明升级前后的读取、拒绝和缓存行为；显示“需更新”之外，还要堵住显式Provider、SDK、工作流、后台辅助和媒体调用绕过。
+2. **P3，GPT-5.4 Nano误匹配**：当前走 gpt-5.4 的 pattern，Nano 400000 窗口可能被 1050000 规格混淆。补独立exact规格、收窄pattern，并用真实匹配器验证；其他代际误吞一并巡检。
+3. **P3，独立输入限制**：Mini/Nano 的272000输入与400000总窗口、讯飞65536输入与131072输出分别表达；不再用 contextWindow 伪装最大输入。
+4. **P0/P2，参数差异**：Codex订阅272000与原厂1050000不是自动错误；MiniMax-M3共享输出512000与直连131072证据仍不完整；OpenRouter上限按其独立合同处理。见[匹配与差异基线](SPEC-BASELINE.md)。
+5. **P0/P4，历史下架接入**：零一万物已由本仓#84删除，但已发布客户端退役名单仅见internal-testing。补受管旧配置迁移；“本产品已下架”与“厂家域名已停止服务”的证明分别登记，不能直接封整个lingyiwanwu.com。
+6. **P2，定价/媒体/思考旧字段**：价格同义字段、音频格式/音色、思考默认和旧metadata优先分类；不把缓存读价、TTL写价、按字符/每秒/每search unit价格改成同一token价。
+
+## 并行实施安排
+
+本次实施采用三个 subAgent 与主工作线协作。配置中心复用当前工作区；客户端从最新dev创建隔离worktree，按要求拉取最新子模块。
+
+- A：字段归属、迁移器与兼容投影；独占迁移工具、字段schema和迁移测试。
+- B：ModelSpec精确绑定、参数与官方字段核查；按供应商分工，不同时修改同一 canonical。
+- C：域名/Provider退役清单、客户端停用与出站拦截；独占退役注册与相关测试。
+- 主工作线：客户端版本矩阵、格式门控、resolver接口、交叉review、manifest和发布；共享schema变更先定接口再分配，不能多代理同时改同一文件。
+
+P0 的只读清单可并行；P1 完成后再进行数据迁移和下架发布。客户端仓 worktree 若需建立，先同步并更新最新子模块，保留无关gitlink及用户改动。
+
+## 完成标准
+
+- 所有仓内受管字段及已识别存量旧字段都完成“保留有效规范字段/映射迁移/转记录性metadata/明确退役”的处置；运行时写入V2旧键为零，旧读取兜底有退出条件。
+- 40份接入、全部V2规格/协议的依赖都可计算最低版本；未支持版本在读取和实际调用两层被拦；离线保留最后一份可消费数据。
+- 已确认退役域名/路径/接入在活跃配置、默认映射、同步发现和后台任务中零残留；无法证实退役的域名留待核，不误删。
+- 每个接入绑定准确规格或明确动态例外；45处差异有解释、7处family没有扩大调用或路由权限；58个未命中静态接入的规格完成用途分类。
+- 原厂事实、接入限制、价格、用户覆写、产品路由政策都可追溯；未知不自动变成零、默认值或全能力。
+- 老/最低/当前客户端、受管旧配置、用户自建配置、重复迁移、回滚、停服域名和新模型调用矩阵通过。
+
+核验和发布细则见[验收与发布](VALIDATION.md)；字段、版本、域名、规格专题分别拆分在同目录，避免把证据和全部任务堆在本文件。
+
+## 实施中补充的实际边界（revision 3）
+
+真实适配器与profile执行已核36条当前未适配接入（3Realtime、21图像/视频、12ASR/TTS），原默认local-whisper/whisper-large-v3亦缺profile。此任务落实适配状态与实际调用门控，不以保留记录宣称这些接口已可用，不静默替换用户默认。原生适配器新增属于已列独立接入依赖，详见ADAPTER-DECISIONS。

docs/plans/catalog-contract-migration/RETIREMENT-DECISIONS.md

@@ -0,0 +1,7 @@
+# 退役判定与实施范围
+
+依据：本仓 #84 / commit a998e18；原配置 provider-lingyiwanwu-001，baseUrl=https://api.lingyiwanwu.com/v1。本轮核查使用Git历史，未把官网正文读取失败当作厂商域名停服。
+
+已形成候选数据规则 catalog/v2/retirements.json：受管旧Provider身份与原host/path同时匹配，2026-08-28起本产品预置接入已退役。保留凭据，不按label模糊匹配，不封AgentBox、官网或其他同厂产品。客户端实现/正式版本与验证证据将追加在EXECUTION.md。
+
+MiniMax旧/new域名、讯飞/v1/v2/x2、Perplexity路径迁移和87条不可读官网记录没有获得整域停服证明，本轮不将它们加入停服域名denylist。所有已下架模型的tombstones仍按各接入面保留。

docs/plans/catalog-contract-migration/SPEC-BASELINE.md

@@ -0,0 +1,151 @@
+# ModelSpec匹配与数值差异基线
+
+[PLAN](PLAN.md) · [职责/合成设计](MODEL-SPECS.md)
+
+使用正式发布tag v10.0.177的model-spec-matcher.ts纯函数，对本仓V127进行离线只读匹配，未调用模型API。该检查说明数据可匹配性，不证明所有运行时入口会以覆盖模式调用applySpec，也不证明差异全是错误。
+
+340接入记录/279规格：303 exact、3 normalized、1 stripped、9 pattern、7 family、17 none；45条记录有窗口或输出数值差异。
+
+## 无匹配（17条）与family（7条）
+
+| 接入配置 | 模型ID | 匹配 | 当前落入规格 | 分类与处理 |
+| --- | --- | --- | --- | --- |
+| `compute/providers/openai.json` | `gpt-realtime-2.1` | none | 无 | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/providers/openai.json` | `gpt-realtime-2.1-mini` | none | 无 | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/providers/openai.json` | `gpt-realtime-translate` | none | 无 | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/providers/dashscope.json` | `qwen-mt-plus` | none | 无 | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/providers/volcengine.json` | `doubao-seedream-4-0-250828` | none | 无 | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/providers/moonshot.json` | `moonshot-v1-8k-vision-preview` | family | moonshot-v1-8k | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/providers/moonshot.json` | `moonshot-v1-32k-vision-preview` | family | moonshot-v1-8k | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/providers/moonshot.json` | `moonshot-v1-128k-vision-preview` | family | moonshot-v1-8k | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/providers/zhipu.json` | `glm-4.7-flashx` | family | glm-4.7 | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/providers/siliconflow.json` | `BAAI/bge-m3` | none | 无 | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/providers/local-whisper.json` | `whisper-large-v3` | family | whisper-1 | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/providers/ollama.json` | `llama3.1:70b` | none | 无 | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/providers/xai.json` | `grok-build-0.1` | none | 无 | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/providers/openrouter.json` | `cohere/command-a-plus` | none | 无 | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/providers/openrouter.json` | `openrouter/auto` | none | 无 | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/coding-plans/dashscope-coding.json` | `qwen3-coder-plus` | family | qwen3-coder | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/coding-plans/dashscope-coding.json` | `qwen3-coder-next` | family | qwen3-coder | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/coding-plans/volcengine-coding.json` | `ark-code-latest` | none | 无 | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/coding-plans/tencent-token.json` | `tc-code-latest` | none | 无 | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/coding-plans/moonshot-coding.json` | `kimi-for-coding` | none | 无 | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/coding-plans/baidu-coding.json` | `qianfan-code-latest` | none | 无 | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/coding-plans/infini-coding.json` | `deepseek-v3` | none | 无 | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/coding-plans/moorethread-coding.json` | `mt-coder` | none | 无 | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+| `compute/coding-plans/kwai-coding.json` | `kwai-coder` | none | 无 | P0区分固定型号补exact、动态别名、family保守兜底或待协议适配；不据此自动补参数 |
+
+## 数值差异（45条）
+
+| 接入配置 | 模型ID | 规格/匹配 | 接入值 → Spec值 | 处置状态 |
+| --- | --- | --- | --- | --- |
+| `compute/providers/openai.json` | `gpt-5.4-nano` | gpt-5.4 (pattern) | contextWindow: 400000 → 1050000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/openai-codex.json` | `gpt-reserve` | gpt-reserve (exact) | contextWindow: 272000 → 1050000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/openai-codex.json` | `gpt-5.6-sol` | gpt-5.6-sol (exact) | contextWindow: 272000 → 1050000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/openai-codex.json` | `gpt-5.6-terra` | gpt-5.6-terra (exact) | contextWindow: 272000 → 1050000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/openai-codex.json` | `gpt-5.6-luna` | gpt-5.6-luna (exact) | contextWindow: 272000 → 1050000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/openai-codex.json` | `gpt-5.5` | gpt-5.5 (exact) | contextWindow: 272000 → 1050000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/openai-codex.json` | `gpt-5.4` | gpt-5.4 (exact) | contextWindow: 272000 → 1050000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/openai-codex.json` | `gpt-5.4-mini` | gpt-5.4-mini (exact) | contextWindow: 272000 → 400000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/volcengine.json` | `doubao-seed-2.1-turbo` | doubao-seed-2.1-turbo (exact) | maxOutputTokens: 262144 → 128000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/volcengine.json` | `doubao-seed-2.0-lite` | doubao-seed-2.0-lite (exact) | maxOutputTokens: 128000 → 32000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/volcengine.json` | `doubao-seed-2.0-mini` | doubao-seed-2.0-mini (exact) | maxOutputTokens: 128000 → 32000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/moonshot.json` | `kimi-k2.7-code` | kimi-k2.7-code (exact) | maxOutputTokens: 32768 → 16384 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/moonshot.json` | `kimi-k2.7-code-highspeed` | kimi-k2.7-code (pattern) | maxOutputTokens: 32768 → 16384 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/moonshot.json` | `kimi-k2.6` | kimi-k2.6 (exact) | maxOutputTokens: 32768 → 16384 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/moonshot.json` | `kimi-k2.5` | kimi-k2.5 (exact) | contextWindow: 262144 → 256000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/moonshot.json` | `moonshot-v1-32k-vision-preview` | moonshot-v1-8k (family) | contextWindow: 32768 → 8192 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/moonshot.json` | `moonshot-v1-128k-vision-preview` | moonshot-v1-8k (family) | contextWindow: 131072 → 8192 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/zhipu.json` | `glm-5.2` | glm-5.2 (exact) | contextWindow: 1000000 → 1048576；maxOutputTokens: 131072 → 32768 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/zhipu.json` | `glm-5.1` | glm-5.1 (exact) | maxOutputTokens: 131072 → 128000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/zhipu.json` | `glm-5-turbo` | glm-5-turbo (exact) | maxOutputTokens: 131072 → 128000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/zhipu.json` | `glm-5` | glm-5 (exact) | maxOutputTokens: 131072 → 128000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/zhipu.json` | `glm-4.7` | glm-4.7 (exact) | maxOutputTokens: 131072 → 128000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/zhipu.json` | `glm-4.7-flashx` | glm-4.7 (family) | maxOutputTokens: 131072 → 128000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/zhipu.json` | `glm-5v-turbo` | glm-5v-turbo (exact) | maxOutputTokens: 131072 → 128000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/zhipu.json` | `glm-4.6` | glm-4.6 (exact) | maxOutputTokens: 131072 → 128000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/minimax.json` | `MiniMax-M3` | MiniMax-M3 (exact) | maxOutputTokens: 131072 → 512000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/xai.json` | `grok-4.20-0309-reasoning` | grok-4.20-0309-reasoning (exact) | contextWindow: 1000000 → 2000000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/openrouter.json` | `minimax/minimax-m3` | MiniMax-M3 (normalized) | contextWindow: 1048576 → 1000000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/openrouter.json` | `deepseek/deepseek-v4.1-flash` | deepseek-flash (exact) | contextWindow: 1048576 → 1000000；maxOutputTokens: 943718 → 384000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/openrouter.json` | `z-ai/glm-5.3-flash` | glm-5.3-flash (exact) | maxOutputTokens: 943717 → 131072 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/providers/openrouter.json` | `moonshotai/kimi-k3` | kimi-k3 (exact) | maxOutputTokens: 943718 → 1048576 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/coding-plans/zhipu-coding.json` | `glm-5.2` | glm-5.2 (exact) | contextWindow: 1000000 → 1048576；maxOutputTokens: 131072 → 32768 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/coding-plans/zhipu-coding.json` | `glm-5-turbo` | glm-5-turbo (exact) | maxOutputTokens: 131072 → 128000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/coding-plans/zhipu-coding.json` | `glm-4.7` | glm-4.7 (exact) | maxOutputTokens: 131072 → 128000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/coding-plans/dashscope-coding.json` | `qwen3-coder-plus` | qwen3-coder (family) | contextWindow: 1000000 → 262144 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/coding-plans/dashscope-coding.json` | `kimi-k2.5` | kimi-k2.5 (exact) | contextWindow: 262144 → 256000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/coding-plans/dashscope-coding.json` | `glm-5` | glm-5 (exact) | contextWindow: 202752 → 200000；maxOutputTokens: 131072 → 128000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/coding-plans/dashscope-coding.json` | `glm-4.7` | glm-4.7 (exact) | contextWindow: 202752 → 200000；maxOutputTokens: 131072 → 128000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/coding-plans/dashscope-coding.json` | `MiniMax-M2.5` | MiniMax-M2.5 (exact) | contextWindow: 196608 → 204800；maxOutputTokens: 8192 → 131072 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/coding-plans/dashscope-token-plan.json` | `kimi-k2.5` | kimi-k2.5 (exact) | contextWindow: 262144 → 256000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/coding-plans/dashscope-token-plan.json` | `glm-5` | glm-5 (exact) | contextWindow: 202752 → 200000；maxOutputTokens: 131072 → 128000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/coding-plans/dashscope-token-plan.json` | `MiniMax-M2.5` | MiniMax-M2.5 (exact) | contextWindow: 196608 → 204800；maxOutputTokens: 8192 → 131072 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/coding-plans/dashscope-token-plan.json` | `glm-4.7` | glm-4.7 (exact) | maxOutputTokens: 131072 → 128000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/coding-plans/volcengine-coding.json` | `doubao-seed-2.1-turbo` | doubao-seed-2.1-turbo (exact) | maxOutputTokens: 262144 → 128000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+| `compute/coding-plans/minimax-coding.json` | `MiniMax-M3` | MiniMax-M3 (exact) | maxOutputTokens: 131072 → 512000 | 待分型：平台覆盖/错误/单位/缺证，不要求全部相等 |
+
+## 未被静态接入记录命中的规格（58条）
+
+其中4条具有routing。静态未命中不证明无云端使用方、不证明退役。分别核动态目录、家族兜底、历史兼容、原生协议待适配与真正孤立后决定保留或归档。
+
+| 规格文件 | ID | 当前routing存在 |
+| --- | --- | --- |
+| `compute/model-specs/anthropic.json` | `claude-opus-4-8` | 否 |
+| `compute/model-specs/anthropic.json` | `claude-opus-4-7` | 否 |
+| `compute/model-specs/anthropic.json` | `claude-opus` | 否 |
+| `compute/model-specs/anthropic.json` | `claude-sonnet-4-6` | 否 |
+| `compute/model-specs/anthropic.json` | `claude-sonnet-4-5` | 否 |
+| `compute/model-specs/anthropic.json` | `claude-sonnet` | 否 |
+| `compute/model-specs/openai.json` | `dall-e-3` | 否 |
+| `compute/model-specs/openai.json` | `gpt-4o-realtime` | 否 |
+| `compute/model-specs/openai.json` | `gpt-4o-realtime-preview` | 否 |
+| `compute/model-specs/openai.json` | `gpt-oss-120b` | 否 |
+| `compute/model-specs/openai.json` | `gpt-5.6-sol-pro` | 否 |
+| `compute/model-specs/openai.json` | `gpt-5.6-terra-pro` | 否 |
+| `compute/model-specs/openai.json` | `gpt-5.6-luna-pro` | 否 |
+| `compute/model-specs/google.json` | `gemini-3.5-transcribe-live` | 否 |
+| `compute/model-specs/google.json` | `gemini-3.5-transcribe` | 否 |
+| `compute/model-specs/google.json` | `gemini-3.8-live-extended-thinking` | 否 |
+| `compute/model-specs/google.json` | `gemini-3.8-live` | 否 |
+| `compute/model-specs/google.json` | `gemini-3.8-flash-lite-tts` | 否 |
+| `compute/model-specs/google.json` | `gemini-3.8-flash-tts` | 否 |
+| `compute/model-specs/google.json` | `gemini-3.1-flash-lite-image` | 否 |
+| `compute/model-specs/google.json` | `gemini-3.1-flash-image` | 否 |
+| `compute/model-specs/deepseek.json` | `deepseek-chat` | 否 |
+| `compute/model-specs/deepseek.json` | `deepseek-reasoner` | 否 |
+| `compute/model-specs/qwen.json` | `qwen3.8-max-preview` | 是 |
+| `compute/model-specs/qwen.json` | `qwen-max` | 否 |
+| `compute/model-specs/qwen.json` | `qwen-plus` | 否 |
+| `compute/model-specs/qwen.json` | `qwen-turbo` | 否 |
+| `compute/model-specs/qwen.json` | `qwen3-max-trans` | 否 |
+| `compute/model-specs/qwen.json` | `qwen3.5-35b-a3b` | 否 |
+| `compute/model-specs/qwen.json` | `qwen3.5-27b` | 否 |
+| `compute/model-specs/moonshot.json` | `kimi-k2-thinking` | 否 |
+| `compute/model-specs/moonshot.json` | `kimi-k2` | 否 |
+| `compute/model-specs/zhipu.json` | `glm-4.7-thinking` | 否 |
+| `compute/model-specs/minimax.json` | `MiniMax-H3-Max` | 否 |
+| `compute/model-specs/minimax.json` | `MiniMax-H3` | 否 |
+| `compute/model-specs/minimax.json` | `MiniMax-Text-01` | 否 |
+| `compute/model-specs/xai.json` | `grok-4.5` | 否 |
+| `compute/model-specs/xai.json` | `grok-4-1-fast-reasoning` | 否 |
+| `compute/model-specs/xiaomi.json` | `mimo-v2.5` | 是 |
+| `compute/model-specs/xiaomi.json` | `mimo-v2-pro` | 否 |
+| `compute/model-specs/xiaomi.json` | `mimo-v2-omni` | 否 |
+| `compute/model-specs/xiaomi.json` | `mimo-v2-flash` | 否 |
+| `compute/model-specs/xiaomi.json` | `mimo-v2-tts` | 否 |
+| `compute/model-specs/cohere.json` | `rerank-v4.0-fast` | 否 |
+| `compute/model-specs/cohere.json` | `rerank-v4.0-pro` | 否 |
+| `compute/model-specs/cohere.json` | `north-mini-code-1-0` | 否 |
+| `compute/model-specs/cohere.json` | `north-small-translate-1-0` | 否 |
+| `compute/model-specs/cohere.json` | `rerank-v3.5` | 否 |
+| `compute/model-specs/kling.json` | `kling-3.0-turbo` | 否 |
+| `compute/model-specs/stability.json` | `stable-audio-3.0` | 否 |
+| `compute/model-specs/tencent.json` | `hy3` | 否 |
+| `compute/model-specs/tencent.json` | `hy3-preview` | 否 |
+| `compute/model-specs/volcengine.json` | `doubao-seedance-2.0-fast` | 是 |
+| `compute/model-specs/volcengine.json` | `doubao-seedance-2.0-mini` | 是 |
+| `compute/model-specs/volcengine.json` | `doubao-seedance-1.5-pro` | 否 |
+| `compute/model-specs/volcengine.json` | `doubao-seedance-1.0-pro` | 否 |
+| `compute/model-specs/volcengine.json` | `doubao-seedance-1.0-pro-fast` | 否 |
+| `compute/model-specs/stealth.json` | `ox-alpha` | 否 |