修复算力模型采样默认值

compute/providers/cohere.json

@@ -9,15 +9,21 @@
   "enabled": false,
   "status": "unconfigured",
   "priceCurrency": "USD",
-  "services": ["chat", "embedding", "rerank"],
+  "services": [
+    "chat",
+    "embedding",
+    "rerank"
+  ],
   "models": [
     {
       "modelName": "command-a-03-2025",
       "displayName": "Command A",
-      "serviceType": ["chat"],
-      "description": "Cohere 旗舰对话模型，262K 上下文，擅长 RAG 和工具调用",
-      "contextWindow": 262144,
-      "maxOutputTokens": 8192,
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Cohere 旗舰对话模型，256K 上下文，擅长 RAG 和工具调用",
+      "contextWindow": 256000,
+      "maxOutputTokens": 8000,
       "capabilities": [
         "chat",
         "reasoning",
@@ -27,18 +33,46 @@
         "rag",
         "long_context"
       ],
-      "inputPrice": 2.50,
-      "outputPrice": 10.00,
-      "defaultTemperature": 1,
-      "defaultTopP": 1,
-      "extra": {}
+      "inputPrice": 2.5,
+      "outputPrice": 10.0,
+      "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，131K 上下文，多语言支持",
-      "contextWindow": 131072,
+      "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",
@@ -46,22 +80,35 @@
       ],
       "inputPrice": 0.12,
       "outputPrice": 0,
-      "extra": {}
+      "extra": {
+        "dimensions": [
+          256,
+          512,
+          1024,
+          1536
+        ],
+        "defaultDimension": 1536,
+        "pricingNotes": "Embedding models are priced by embedded tokens; price is per 1M tokens."
+      }
     },
     {
       "modelName": "rerank-v3.5",
       "displayName": "Rerank V3.5",
-      "serviceType": ["rerank"],
-      "description": "Cohere 语义重排序模型，用于检索结果精排",
-      "contextWindow": 0,
+      "serviceType": [
+        "rerank"
+      ],
+      "description": "Cohere 语义重排序模型，用于检索结果精排，默认单文档截断 4096 token",
+      "contextWindow": 4096,
       "maxOutputTokens": 0,
       "capabilities": [
         "rerank",
         "semantic_reranking"
       ],
-      "inputPrice": 2.00,
+      "inputPrice": 2.0,
       "outputPrice": 0,
-      "extra": {}
+      "extra": {
+        "pricingNotes": "Cohere Rerank is priced by search units, not input/output tokens; one search unit is one query with up to 100 documents. The token price fields are retained for schema compatibility."
+      }
     }
   ]
 }