feat(compute): add MiMo speech protocol metadata

compute/model-specs/xiaomi.json

@@ -156,6 +156,7 @@
       },
       "spec": {
         "contextWindow": 8192,
+        "maxOutputTokens": 8192,
         "capabilities": [
           "tts",
           "multilingual",
@@ -164,7 +165,43 @@
         "serviceType": [
           "tts"
         ],
-        "description": "小米 MiMo 语音合成，支持多语言与风格控制"
+        "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"
       }
     },
     {
@@ -177,6 +214,8 @@
         ]
       },
       "spec": {
+        "contextWindow": 8192,
+        "maxOutputTokens": 2048,
         "capabilities": [
           "asr",
           "multilingual"
@@ -184,7 +223,27 @@
         "serviceType": [
           "asr"
         ],
-        "description": "小米 MiMo 语音识别模型"
+        "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"
       }
     },
     {

compute/providers/xiaomi.json

@@ -14,7 +14,8 @@
   "services": [
     "chat",
     "reasoning",
-    "tts"
+    "tts",
+    "asr"
   ],
   "models": [
     {
@@ -49,24 +50,37 @@
       ],
       "extra": {
         "voices": [
-          {"id": "default_zh", "label": "默认中文女声"},
-          {"id": "default_en", "label": "默认英文女声"},
           {"id": "mimo_default", "label": "MiMo 默认"},
-          {"id": "Bingtang", "label": "冰糖"},
-          {"id": "Moli", "label": "茉莉"},
-          {"id": "Suda", "label": "苏打"},
-          {"id": "Baihua", "label": "白桦"},
+          {"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": ["mp3", "wav"],
+        "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",