feat(model-specs): 新增模型规格库——跨 provider 模型参数统一维护

compute/model-specs/_index.json

@@ -0,0 +1,26 @@
+{
+  "description": "模型规格库加载顺序索引。元素为对应 JSON 文件的 basename（不含 .json）。",
+  "order": [
+    "anthropic",
+    "openai",
+    "google",
+    "deepseek",
+    "qwen",
+    "moonshot",
+    "zhipu",
+    "minimax",
+    "xai",
+    "mistral",
+    "xiaomi",
+    "baichuan",
+    "baidu",
+    "cohere",
+    "kling",
+    "lingyiwanwu",
+    "perplexity",
+    "stability",
+    "tencent",
+    "volcengine",
+    "xunfei"
+  ]
+}

compute/model-specs/anthropic.json

@@ -0,0 +1,65 @@
+{
+  "description": "Anthropic Claude 系列模型规格。参数来源：config-center compute/providers/anthropic.json。",
+  "specs": [
+    {
+      "id": "claude-opus-4-7",
+      "displayName": "Claude Opus 4.7",
+      "family": "claude-opus",
+      "match": { "patterns": ["claude-opus-4-7*"] },
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 128000,
+        "capabilities": ["chat", "reasoning", "code", "vision", "tool_use"],
+        "serviceType": ["chat"],
+        "defaultTemperature": 1,
+        "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"
+      }
+    },
+    {
+      "id": "claude-sonnet-4-6",
+      "displayName": "Claude Sonnet 4.6",
+      "family": "claude-sonnet",
+      "match": { "patterns": ["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"
+      }
+    },
+    {
+      "id": "claude-sonnet-4-5",
+      "displayName": "Claude Sonnet 4.5",
+      "family": "claude-sonnet",
+      "match": { "patterns": ["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"
+      }
+    },
+    {
+      "id": "claude-haiku-4-5",
+      "displayName": "Claude Haiku 4.5",
+      "family": "claude-haiku",
+      "match": { "patterns": ["claude-haiku-4-5*"] },
+      "spec": {
+        "contextWindow": 200000,
+        "maxOutputTokens": 64000,
+        "capabilities": ["chat", "reasoning", "code", "vision", "tool_use"],
+        "serviceType": ["chat"],
+        "defaultTemperature": 1,
+        "supportsReasoning": true,
+        "description": "Anthropic 最快模型，具备接近前沿模型的智能水平"
+      }
+    }
+  ]
+}

compute/model-specs/baichuan.json

@@ -0,0 +1,61 @@
+{
+  "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": [
+    {
+      "id": "Baichuan-M3-Plus",
+      "displayName": "百川 M3-Plus",
+      "family": "baichuan-m3",
+      "match": { "exact": ["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%"
+      }
+    },
+    {
+      "id": "Baichuan-M3",
+      "displayName": "百川 M3",
+      "family": "baichuan-m3",
+      "match": { "exact": ["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"
+      }
+    },
+    {
+      "id": "Baichuan-M2-Plus",
+      "displayName": "百川 M2-Plus",
+      "family": "baichuan-m2",
+      "match": { "exact": ["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"
+      }
+    },
+    {
+      "id": "Baichuan-M2",
+      "displayName": "百川 M2",
+      "family": "baichuan-m2",
+      "match": { "exact": ["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/model-specs/baidu.json

@@ -0,0 +1,76 @@
+{
+  "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": [
+    {
+      "id": "ernie-5.0-thinking-latest",
+      "displayName": "文心 ERNIE 5.0",
+      "family": "ernie-5",
+      "match": { "exact": ["ernie-5.0-thinking-latest"], "patterns": ["ernie-5.0-thinking*"] },
+      "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"
+      }
+    },
+    {
+      "id": "ernie-5.0",
+      "displayName": "文心 ERNIE 5.0 非思考",
+      "family": "ernie-5",
+      "match": { "exact": ["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"
+      }
+    },
+    {
+      "id": "ernie-4.5-turbo-128k",
+      "displayName": "文心 ERNIE 4.5 Turbo",
+      "family": "ernie-4.5",
+      "match": { "exact": ["ernie-4.5-turbo-128k"], "patterns": ["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"
+      }
+    },
+    {
+      "id": "ernie-4.5-turbo-20260402",
+      "displayName": "文心 ERNIE 4.5 Turbo 20260402",
+      "family": "ernie-4.5",
+      "match": { "exact": ["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"
+      }
+    },
+    {
+      "id": "ernie-x1.1",
+      "displayName": "文心 ERNIE X1.1",
+      "family": "ernie-x1",
+      "match": { "exact": ["ernie-x1.1"], "patterns": ["ernie-x1*"] },
+      "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/model-specs/cohere.json

@@ -0,0 +1,57 @@
+{
+  "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": [
+    {
+      "id": "command-a-03-2025",
+      "displayName": "Command A",
+      "family": "command-a",
+      "match": { "exact": ["command-a-03-2025"], "patterns": ["command-a*"] },
+      "spec": {
+        "contextWindow": 256000,
+        "maxOutputTokens": 8000,
+        "capabilities": ["chat", "reasoning", "code", "vision", "tool_use", "rag", "long_context"],
+        "serviceType": ["chat"],
+        "defaultTemperature": 0.3,
+        "description": "Cohere 旗舰对话模型，256K 上下文，擅长 RAG 和工具调用"
+      }
+    },
+    {
+      "id": "command-r7b-12-2024",
+      "displayName": "Command R7B",
+      "family": "command-r7b",
+      "match": { "exact": ["command-r7b-12-2024"], "patterns": ["command-r7b*"] },
+      "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"
+      }
+    },
+    {
+      "id": "embed-v4.0",
+      "displayName": "Embed V4",
+      "family": "embed-v4",
+      "match": { "exact": ["embed-v4.0"], "patterns": ["embed-v4*"] },
+      "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"
+      }
+    },
+    {
+      "id": "rerank-v3.5",
+      "displayName": "Rerank V3.5",
+      "family": "rerank-v3",
+      "match": { "exact": ["rerank-v3.5"], "patterns": ["rerank-v3*"] },
+      "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/model-specs/deepseek.json

@@ -0,0 +1,58 @@
+{
+  "description": "DeepSeek 系列模型规格。参数来源：config-center compute/providers/deepseek.json。",
+  "specs": [
+    {
+      "id": "deepseek-chat",
+      "displayName": "DeepSeek Chat",
+      "family": "deepseek-chat",
+      "match": { "patterns": ["deepseek-chat*"] },
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 384000,
+        "capabilities": ["chat", "code", "reasoning", "multilingual", "tool_use"],
+        "serviceType": ["chat"],
+        "defaultTemperature": 1
+      }
+    },
+    {
+      "id": "deepseek-reasoner",
+      "displayName": "DeepSeek Reasoner",
+      "family": "deepseek-reasoner",
+      "match": { "patterns": ["deepseek-reasoner*"] },
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 384000,
+        "capabilities": ["chat", "reasoning", "deep_thinking", "code", "math", "tool_use"],
+        "serviceType": ["reasoning"],
+        "defaultTemperature": null,
+        "supportsReasoning": true
+      }
+    },
+    {
+      "id": "deepseek-v4-pro",
+      "displayName": "DeepSeek V4 Pro",
+      "family": "deepseek-v4",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 384000,
+        "capabilities": ["chat", "code", "reasoning", "deep_thinking", "multilingual", "tool_use", "math"],
+        "serviceType": ["chat", "reasoning"],
+        "defaultTemperature": 1,
+        "supportsReasoning": true
+      }
+    },
+    {
+      "id": "deepseek-v4-flash",
+      "displayName": "DeepSeek V4 Flash",
+      "family": "deepseek-v4",
+      "spec": {
+        "contextWindow": 1000000,
+        "maxOutputTokens": 384000,
+        "capabilities": ["chat", "code", "reasoning", "deep_thinking", "multilingual", "tool_use"],
+        "serviceType": ["chat", "reasoning"],
+        "defaultTemperature": 1,
+        "supportsReasoning": true
+      }
+    }
+  ]
+}