From 1a61a7bb1701ed65b8eb51f7f66309378f8c4a97 Mon Sep 17 00:00:00 2001 From: SZCJW <792430652@qq.com> Date: Mon, 30 Mar 2026 03:29:24 +0800 Subject: [PATCH] x --- README.md | 7 ++++--- app/config.py | 2 ++ app/engine.py | 2 ++ app/model_catalog.py | 7 ++++++- app/schemas.py | 1 + 5 files changed, 15 insertions(+), 4 deletions(-) diff --git a/README.md b/README.md index 5808d73..ea76160 100644 --- a/README.md +++ b/README.md @@ -30,7 +30,7 @@ - `services.openai.host` / `services.openai.port`：OpenAI 协议服务监听地址与端口（默认 `0.0.0.0:8001`） - `public_model_name`：对外固定模型名，切换底层模型时可保持调用方参数不变 -- `reasoning_enabled`：思考推理开关，默认 `false`（关闭） +- `reasoning_enabled`：是否启用推理解析器参数注入，默认 `false` - `api_key`：OpenAI 接口访问密钥 - `tensor_parallel_size`：张量并行数，双卡建议 `2` - `dtype`：推理精度，默认 `bfloat16` @@ -84,7 +84,7 @@ curl http://localhost:/v1/models curl -X POST "http://localhost:8001/v1/chat/completions" \ -H "Content-Type: application/json" \ -H "Authorization: Bearer " \ - -d "{\"model\":\"Qwen_local_model\",\"messages\":[{\"role\":\"user\",\"content\":\"你好，介绍一下你自己\"}],\"temperature\":0.7}" + -d "{\"model\":\"Qwen_local_model\",\"messages\":[{\"role\":\"user\",\"content\":\"你好，介绍一下你自己\"}],\"temperature\":0.7,\"chat_template_kwargs\":{\"enable_thinking\":false}}" ``` ## OpenClaw 调用说明 @@ -92,7 +92,8 @@ curl -X POST "http://localhost:8001/v1/chat/completions" \ - Base URL 使用 `http://<服务器IP>:8001/v1` - API Key 使用 `config.json` 中 `api_key` - 模型名固定使用 `config.json` 中 `public_model_name`（默认 `Qwen_local_model`） -- 若要启用思考推理，将 `config.json` 中 `reasoning_enabled` 设为 `true` +- 思考模式按请求控制：`chat_template_kwargs.enable_thinking=false/true` +- 仅当模型需要推理解析器时，再将 `config.json` 中 `reasoning_enabled` 设为 `true` - 若使用工具调用，`config.json` 中应配置 `tool_call_parser` 与 `enable_auto_tool_choice` - 服务强制离线模式，不会回退到 Hugging Face 远程下载 - 所有路径按 Ubuntu 规范填写，本地模型建议使用 `/opt/model/<模型目录>` diff --git a/app/config.py b/app/config.py index 392c42f..e4a9e6d 100644 --- a/app/config.py +++ b/app/config.py @@ -17,6 +17,7 @@ class Settings(BaseModel): port: int = 8000 openai_host: str = "0.0.0.0" openai_port: int = 8001 + vllm_openai_internal_url: str = "http://127.0.0.1:8001/v1" public_model_name: str = "Qwen_local_model" reasoning_enabled: bool = False model_root: str = "/opt/model" @@ -46,6 +47,7 @@ def get_settings() -> Settings: port=runtime["port"], openai_host=runtime["openai_host"], openai_port=runtime["openai_port"], + vllm_openai_internal_url=runtime["vllm_openai_internal_url"], public_model_name=runtime["public_model_name"], reasoning_enabled=runtime["reasoning_enabled"], model_root=runtime["model_root"], diff --git a/app/engine.py b/app/engine.py index 0379882..7c3e82f 100644 --- a/app/engine.py +++ b/app/engine.py @@ -25,6 +25,8 @@ class InferenceEngine: } if req.stop: payload["stop"] = req.stop + if req.enable_thinking is not None: + payload["chat_template_kwargs"] = {"enable_thinking": req.enable_thinking} response = self.client.post( f"{self.settings.vllm_openai_internal_url}/chat/completions", headers=headers, diff --git a/app/model_catalog.py b/app/model_catalog.py index 6c01468..7c89c0a 100644 --- a/app/model_catalog.py +++ b/app/model_catalog.py @@ -63,11 +63,16 @@ def resolve_runtime_settings(content: dict[str, Any]) -> dict[str, Any]: api_service = dict(services.get("api", {})) openai_service = dict(services.get("openai", {})) models = dict(content.get("models", {})) + openai_port = _to_int(openai_service.get("port"), 8001) + internal_url = _to_str(content.get("vllm_openai_internal_url")) + if not internal_url: + internal_url = f"http://127.0.0.1:{openai_port}/v1" return { "host": str(api_service.get("host", "0.0.0.0")), "port": _to_int(api_service.get("port"), 8000), "openai_host": str(openai_service.get("host", "0.0.0.0")), - "openai_port": _to_int(openai_service.get("port"), 8001), + "openai_port": openai_port, + "vllm_openai_internal_url": internal_url.rstrip("/"), "public_model_name": _to_str(content.get("public_model_name"), "Qwen_local_model"), "api_key": str(content.get("api_key", "")).strip() or None, "reasoning_enabled": _to_bool(content.get("reasoning_enabled"), False), diff --git a/app/schemas.py b/app/schemas.py index 16e0a8d..026e23a 100644 --- a/app/schemas.py +++ b/app/schemas.py @@ -10,6 +10,7 @@ class GenerateRequest(BaseModel): top_p: float = Field(default=0.95, gt=0.0, le=1.0) repetition_penalty: float = Field(default=1.0, ge=0.5, le=2.0) stop: Optional[List[str]] = None + enable_thinking: Optional[bool] = None class GenerateResponse(BaseModel):