diff --git a/apps/scheduler/call/api/api.py b/apps/scheduler/call/api/api.py index 47337d385d1b525f405af42fdb65a059ae994aa6..6a769416eafafa88b025aeb1484b56294fa67778 100644 --- a/apps/scheduler/call/api/api.py +++ b/apps/scheduler/call/api/api.py @@ -56,8 +56,11 @@ class API(CoreCall, input_model=APIInput, output_model=APIOutput): content_type: ContentType | None = Field(description="API接口的Content-Type", default=None) timeout: int = Field(description="工具超时时间", default=300, gt=30) - body: dict[str, Any] = Field(description="已知的部分请求体", default={}) - query: dict[str, Any] = Field(description="已知的部分请求参数", default={}) + headers: dict[str, Any] = Field(description="API接口的请求头", default={}) + query: dict[str, Any] = Field(description="API接口的请求参数", default={}) + json_body: Any = Field(description="JSON格式的请求体", default=None, alias="json") + formData: Any = Field(description="Form Data格式的请求体", default=None) + xWwwFormUrlencoded: Any = Field(description="x-www-form-urlencoded格式的请求体", default=None) i18n_info: ClassVar[dict[str, dict]] = { LanguageType.CHINESE: { @@ -103,8 +106,11 @@ class API(CoreCall, input_model=APIInput, output_model=APIOutput): return APIInput( url=self.url, method=self.method, + headers=self.headers, query=self.query, - body=self.body, + json_body=self.json_body, + formData=self.formData, + xWwwFormUrlencoded=self.xWwwFormUrlencoded, ) async def _exec( @@ -134,18 +140,26 @@ class API(CoreCall, input_model=APIInput, output_model=APIOutput): req_cookie = {} req_params = {} + # 合并headers字典 + if data.headers: + req_header.update(data.headers) + + # 合并query字典 + if data.query: + req_params.update(data.query) + # 创建请求工厂 request_factory = partial( self._client.request, method=self.method, url=self.url, cookies=req_cookie, + headers=req_header, ) # 根据HTTP方法创建请求 if self.method in [HTTPMethod.GET.value, HTTPMethod.DELETE.value]: # GET/DELETE 请求处理 - req_params.update(data.query) return await request_factory(params=req_params) if self.method in [HTTPMethod.POST.value, HTTPMethod.PUT.value, HTTPMethod.PATCH.value]: @@ -153,9 +167,23 @@ class API(CoreCall, input_model=APIInput, output_model=APIOutput): if not self.content_type: raise CallError(message="API接口的Content-Type未指定", data={}) - # 根据Content-Type设置请求参数 - req_body = data.body - req_header.update({"Content-Type": self.content_type}) + # 根据Content-Type获取请求体数据 + req_body = None + if self.content_type == ContentType.JSON.value: + req_body = data.json_body + # 如果json字段是字符串，需要解析成对象 + if isinstance(req_body, str): + try: + req_body = json.loads(req_body) + except Exception as e: + raise CallError( + message="API接口的JSON请求体格式错误", + data={"json": req_body}, + ) from e + elif self.content_type == ContentType.FORM_URLENCODED.value: + req_body = data.xWwwFormUrlencoded + elif self.content_type == ContentType.MULTIPART_FORM_DATA.value: + req_body = data.formData # 根据Content-Type决定如何发送请求体 content_type_handlers = { @@ -169,8 +197,9 @@ class API(CoreCall, input_model=APIInput, output_model=APIOutput): raise CallError(message="API接口的Content-Type不支持", data={}) request_kwargs = {} - request_kwargs.update(handler(req_body, files)) - return await request_factory(**request_kwargs) + if req_body is not None: + request_kwargs.update(handler(req_body, files)) + return await request_factory(params=req_params, **request_kwargs) raise CallError(message="API接口的HTTP Method不支持", data={}) diff --git a/apps/scheduler/call/api/schema.py b/apps/scheduler/call/api/schema.py index 055008a889676e4e2fbd5057912ba78421e52aa2..373f119c89ced55bfb9c28618595560ec38655d2 100644 --- a/apps/scheduler/call/api/schema.py +++ b/apps/scheduler/call/api/schema.py @@ -15,8 +15,11 @@ class APIInput(DataBase): url: SkipJsonSchema[str] = Field(description="API调用工具的URL") method: SkipJsonSchema[str] = Field(description="API调用工具的HTTP方法") + headers: dict[str, Any] = Field(description="API调用工具的请求头", default={}) query: dict[str, Any] = Field(description="API调用工具的请求参数", default={}) - body: dict[str, Any] = Field(description="API调用工具的请求体", default={}) + json_body: Any = Field(description="JSON格式的请求体", default=None, alias="json") + formData: Any = Field(description="Form Data格式的请求体", default=None) + xWwwFormUrlencoded: Any = Field(description="x-www-form-urlencoded格式的请求体", default=None) class APIOutput(DataBase): diff --git a/apps/schemas/collection.py b/apps/schemas/collection.py index eb81215c8861df7c3c5f26cdb34ec3f1e198ab7a..f316809e01adbfae77a65a0310b290cc4668684b 100644 --- a/apps/schemas/collection.py +++ b/apps/schemas/collection.py @@ -96,18 +96,45 @@ class LLM(BaseModel): type: list[str] | str = Field( default=['chat'], description="模型类型，支持单个类型或多个类型") - # 模型能力字段 + # 模型能力字段 - 基础能力 provider: str = Field(default="", description="模型提供商") - supports_thinking: bool = Field(default=False, description="是否支持思维链") - can_toggle_thinking: bool = Field( - default=False, description="是否支持开关思维链（仅当supports_thinking=True时有效）") + supports_streaming: bool = Field(default=True, description="是否支持流式输出") supports_function_calling: bool = Field( default=True, description="是否支持函数调用") supports_json_mode: bool = Field(default=True, description="是否支持JSON模式") supports_structured_output: bool = Field( default=False, description="是否支持结构化输出") + + # 推理能力 + supports_thinking: bool = Field(default=False, description="是否支持思维链") + can_toggle_thinking: bool = Field( + default=False, description="是否支持开关思维链（仅当supports_thinking=True时有效）") + supports_reasoning_content: bool = Field( + default=False, description="是否返回reasoning_content字段") + + # 参数支持 max_tokens_param: str = Field( default="max_tokens", description="最大token参数名") + supports_temperature: bool = Field(default=True, description="是否支持temperature参数") + supports_top_p: bool = Field(default=True, description="是否支持top_p参数") + supports_top_k: bool = Field(default=False, description="是否支持top_k参数") + supports_frequency_penalty: bool = Field(default=False, description="是否支持frequency_penalty参数") + supports_presence_penalty: bool = Field(default=False, description="是否支持presence_penalty参数") + supports_min_p: bool = Field(default=False, description="是否支持min_p参数") + + # 高级功能 + supports_response_format: bool = Field(default=True, description="是否支持response_format参数") + supports_tools: bool = Field(default=True, description="是否支持tools参数") + supports_tool_choice: bool = Field(default=True, description="是否支持tool_choice参数") + supports_extra_body: bool = Field(default=True, description="是否支持extra_body参数") + supports_stream_options: bool = Field(default=True, description="是否支持stream_options参数") + + # 特殊参数 + supports_enable_thinking: bool = Field(default=False, description="是否支持enable_thinking参数") + supports_thinking_budget: bool = Field(default=False, description="是否支持思维链token预算") + supports_enable_search: bool = Field(default=False, description="是否支持联网搜索") + + # 其他信息 notes: str = Field(default="", description="备注信息") created_at: float = Field(default_factory=lambda: round( diff --git a/apps/schemas/response_data.py b/apps/schemas/response_data.py index 685dacf2c17f52d9ff1f105ebb26014b245e4332..fd6912027fea007460344437d9f7138bf2a16b83 100644 --- a/apps/schemas/response_data.py +++ b/apps/schemas/response_data.py @@ -662,14 +662,40 @@ class LLMProviderInfo(BaseModel): is_editable: bool = Field(default=True, description="是否可编辑", alias="isEditable") type: list[str] = Field(default=['chat'], description="模型类型列表") - # 模型能力字段 + # 模型能力字段 - 基础能力 provider: str = Field(default="", description="模型提供商") - supports_thinking: bool = Field(default=False, description="是否支持思维链", alias="supportsThinking") - can_toggle_thinking: bool = Field(default=False, description="是否支持开关思维链", alias="canToggleThinking") + supports_streaming: bool = Field(default=True, description="是否支持流式输出", alias="supportsStreaming") supports_function_calling: bool = Field(default=True, description="是否支持函数调用", alias="supportsFunctionCalling") supports_json_mode: bool = Field(default=True, description="是否支持JSON模式", alias="supportsJsonMode") supports_structured_output: bool = Field(default=False, description="是否支持结构化输出", alias="supportsStructuredOutput") + + # 推理能力 + supports_thinking: bool = Field(default=False, description="是否支持思维链", alias="supportsThinking") + can_toggle_thinking: bool = Field(default=False, description="是否支持开关思维链", alias="canToggleThinking") + supports_reasoning_content: bool = Field(default=False, description="是否返回reasoning_content字段", alias="supportsReasoningContent") + + # 参数支持 max_tokens_param: str = Field(default="max_tokens", description="最大token参数名", alias="maxTokensParam") + supports_temperature: bool = Field(default=True, description="是否支持temperature参数", alias="supportsTemperature") + supports_top_p: bool = Field(default=True, description="是否支持top_p参数", alias="supportsTopP") + supports_top_k: bool = Field(default=False, description="是否支持top_k参数", alias="supportsTopK") + supports_frequency_penalty: bool = Field(default=False, description="是否支持frequency_penalty参数", alias="supportsFrequencyPenalty") + supports_presence_penalty: bool = Field(default=False, description="是否支持presence_penalty参数", alias="supportsPresencePenalty") + supports_min_p: bool = Field(default=False, description="是否支持min_p参数", alias="supportsMinP") + + # 高级功能 + supports_response_format: bool = Field(default=True, description="是否支持response_format参数", alias="supportsResponseFormat") + supports_tools: bool = Field(default=True, description="是否支持tools参数", alias="supportsTools") + supports_tool_choice: bool = Field(default=True, description="是否支持tool_choice参数", alias="supportsToolChoice") + supports_extra_body: bool = Field(default=True, description="是否支持extra_body参数", alias="supportsExtraBody") + supports_stream_options: bool = Field(default=True, description="是否支持stream_options参数", alias="supportsStreamOptions") + + # 特殊参数 + supports_enable_thinking: bool = Field(default=False, description="是否支持enable_thinking参数", alias="supportsEnableThinking") + supports_thinking_budget: bool = Field(default=False, description="是否支持思维链token预算", alias="supportsThinkingBudget") + supports_enable_search: bool = Field(default=False, description="是否支持联网搜索", alias="supportsEnableSearch") + + # 其他信息 notes: str = Field(default="", description="备注信息") diff --git a/apps/services/llm.py b/apps/services/llm.py index eb095b3e945647a66e1665a11578cd3f59dc2b26..a57e714e9445ad36e3d357dfe1c6f830944fd9b4 100644 --- a/apps/services/llm.py +++ b/apps/services/llm.py @@ -23,6 +23,66 @@ logger = logging.getLogger(__name__) class LLMManager: """大模型管理""" + @staticmethod + def _create_llm_provider_info(llm: dict) -> LLMProviderInfo: + """ + 从数据库 LLM 文档创建 LLMProviderInfo 对象的辅助方法 + + :param llm: 数据库中的 LLM 文档 + :return: LLMProviderInfo 对象 + """ + # 标准化type字段为列表格式 + llm_type = llm.get("type", "chat") + if isinstance(llm_type, str): + llm_type = [llm_type] + + return LLMProviderInfo( + llmId=llm["_id"], # _id已经是UUID字符串 + icon=llm["icon"], + openaiBaseUrl=llm["openai_base_url"], + openaiApiKey=llm["openai_api_key"], + modelName=llm["model_name"], + maxTokens=llm["max_tokens"], + isEditable=bool(llm.get("user_sub")), # 系统模型（user_sub=""）不可编辑 + type=llm_type, # 始终返回列表格式 + + # 模型能力字段 - 基础能力 + provider=llm.get("provider", ""), + supportsStreaming=llm.get("supports_streaming", True), + supportsFunctionCalling=llm.get("supports_function_calling", True), + supportsJsonMode=llm.get("supports_json_mode", True), + supportsStructuredOutput=llm.get("supports_structured_output", False), + + # 推理能力 + supportsThinking=llm.get("supports_thinking", False), + canToggleThinking=llm.get("can_toggle_thinking", False), + supportsReasoningContent=llm.get("supports_reasoning_content", False), + + # 参数支持 + maxTokensParam=llm.get("max_tokens_param", "max_tokens"), + supportsTemperature=llm.get("supports_temperature", True), + supportsTopP=llm.get("supports_top_p", True), + supportsTopK=llm.get("supports_top_k", False), + supportsFrequencyPenalty=llm.get("supports_frequency_penalty", False), + supportsPresencePenalty=llm.get("supports_presence_penalty", False), + supportsMinP=llm.get("supports_min_p", False), + + # 高级功能 + supportsResponseFormat=llm.get("supports_response_format", True), + supportsTools=llm.get("supports_tools", True), + supportsToolChoice=llm.get("supports_tool_choice", True), + supportsExtraBody=llm.get("supports_extra_body", True), + supportsStreamOptions=llm.get("supports_stream_options", True), + + # 特殊参数 + supportsEnableThinking=llm.get("supports_enable_thinking", False), + supportsThinkingBudget=llm.get("supports_thinking_budget", False), + supportsEnableSearch=llm.get("supports_enable_search", False), + + # 其他信息 + notes=llm.get("notes", ""), + ) + @staticmethod async def list_llm_provider() -> list[LLMProvider]: """ @@ -123,34 +183,8 @@ class LLMManager: result = await llm_collection.find(base_query).sort({"created_at": 1}).to_list(length=None) llm_list = [] - for llm in result: - # 标准化type字段为列表格式 - llm_type = llm.get("type", "chat") - if isinstance(llm_type, str): - llm_type = [llm_type] - - llm_item = LLMProviderInfo( - llmId=llm["_id"], # _id已经是UUID字符串 - icon=llm["icon"], - openaiBaseUrl=llm["openai_base_url"], - openaiApiKey=llm["openai_api_key"], - modelName=llm["model_name"], - maxTokens=llm["max_tokens"], - isEditable=bool(llm.get("user_sub")), # 系统模型（user_sub=""）不可编辑 - type=llm_type, # 始终返回列表格式 - # 模型能力字段 - provider=llm.get("provider", ""), - supportsThinking=llm.get("supports_thinking", False), - canToggleThinking=llm.get("can_toggle_thinking", False), - supportsFunctionCalling=llm.get( - "supports_function_calling", True), - supportsJsonMode=llm.get("supports_json_mode", True), - supportsStructuredOutput=llm.get( - "supports_structured_output", False), - maxTokensParam=llm.get("max_tokens_param", "max_tokens"), - notes=llm.get("notes", ""), - ) + llm_item = LLMManager._create_llm_provider_info(llm) llm_list.append(llm_item) return llm_list @@ -188,23 +222,53 @@ class LLMManager: if not req.openai_base_url: req.openai_base_url = standard_url - model_info = model_registry.get_model_info(provider, req.model_name) - # 标准化type字段为列表格式 model_type = req.type if isinstance(model_type, str): model_type = [model_type] - # 使用请求中的能力信息，如果没有则从model_registry获取，最后使用默认值 + # 从model_registry获取完整的模型能力 + from apps.llm.model_types import ModelType + capabilities_obj = model_registry.get_model_capabilities(provider, req.model_name, ModelType.CHAT) + + # 使用请求中的能力信息，如果没有则从capabilities_obj获取，最后使用默认值 capabilities = { "provider": provider, - "supports_thinking": req.supports_thinking if req.supports_thinking is not None else (model_info.supports_thinking if model_info else False), - "can_toggle_thinking": req.can_toggle_thinking if req.can_toggle_thinking is not None else (model_info.can_toggle_thinking if model_info else False), - "supports_function_calling": req.supports_function_calling if req.supports_function_calling is not None else (model_info.supports_function_calling if model_info else True), - "supports_json_mode": req.supports_json_mode if req.supports_json_mode is not None else (model_info.supports_json_mode if model_info else True), - "supports_structured_output": req.supports_structured_output if req.supports_structured_output is not None else (model_info.supports_structured_output if model_info else False), - "max_tokens_param": req.max_tokens_param or (model_info.max_tokens_param if model_info else "max_tokens"), - "notes": req.notes or (model_info.notes if model_info else ""), + + # 基础能力 + "supports_streaming": req.supports_streaming if hasattr(req, 'supports_streaming') and req.supports_streaming is not None else (capabilities_obj.supports_streaming if capabilities_obj else True), + "supports_function_calling": req.supports_function_calling if req.supports_function_calling is not None else (capabilities_obj.supports_function_calling if capabilities_obj else True), + "supports_json_mode": req.supports_json_mode if req.supports_json_mode is not None else (capabilities_obj.supports_json_mode if capabilities_obj else True), + "supports_structured_output": req.supports_structured_output if req.supports_structured_output is not None else (capabilities_obj.supports_structured_output if capabilities_obj else False), + + # 推理能力 + "supports_thinking": req.supports_thinking if req.supports_thinking is not None else (capabilities_obj.supports_thinking if capabilities_obj else False), + "can_toggle_thinking": req.can_toggle_thinking if req.can_toggle_thinking is not None else (capabilities_obj.can_toggle_thinking if capabilities_obj else False), + "supports_reasoning_content": req.supports_reasoning_content if hasattr(req, 'supports_reasoning_content') and req.supports_reasoning_content is not None else (capabilities_obj.supports_reasoning_content if capabilities_obj else False), + + # 参数支持 + "max_tokens_param": req.max_tokens_param or (capabilities_obj.max_tokens_param if capabilities_obj else "max_tokens"), + "supports_temperature": req.supports_temperature if hasattr(req, 'supports_temperature') and req.supports_temperature is not None else (capabilities_obj.supports_temperature if capabilities_obj else True), + "supports_top_p": req.supports_top_p if hasattr(req, 'supports_top_p') and req.supports_top_p is not None else (capabilities_obj.supports_top_p if capabilities_obj else True), + "supports_top_k": req.supports_top_k if hasattr(req, 'supports_top_k') and req.supports_top_k is not None else (capabilities_obj.supports_top_k if capabilities_obj else False), + "supports_frequency_penalty": req.supports_frequency_penalty if hasattr(req, 'supports_frequency_penalty') and req.supports_frequency_penalty is not None else (capabilities_obj.supports_frequency_penalty if capabilities_obj else False), + "supports_presence_penalty": req.supports_presence_penalty if hasattr(req, 'supports_presence_penalty') and req.supports_presence_penalty is not None else (capabilities_obj.supports_presence_penalty if capabilities_obj else False), + "supports_min_p": req.supports_min_p if hasattr(req, 'supports_min_p') and req.supports_min_p is not None else (capabilities_obj.supports_min_p if capabilities_obj else False), + + # 高级功能 + "supports_response_format": req.supports_response_format if hasattr(req, 'supports_response_format') and req.supports_response_format is not None else (capabilities_obj.supports_response_format if capabilities_obj else True), + "supports_tools": req.supports_tools if hasattr(req, 'supports_tools') and req.supports_tools is not None else (capabilities_obj.supports_tools if capabilities_obj else True), + "supports_tool_choice": req.supports_tool_choice if hasattr(req, 'supports_tool_choice') and req.supports_tool_choice is not None else (capabilities_obj.supports_tool_choice if capabilities_obj else True), + "supports_extra_body": req.supports_extra_body if hasattr(req, 'supports_extra_body') and req.supports_extra_body is not None else (capabilities_obj.supports_extra_body if capabilities_obj else True), + "supports_stream_options": req.supports_stream_options if hasattr(req, 'supports_stream_options') and req.supports_stream_options is not None else (capabilities_obj.supports_stream_options if capabilities_obj else True), + + # 特殊参数 + "supports_enable_thinking": req.supports_enable_thinking if hasattr(req, 'supports_enable_thinking') and req.supports_enable_thinking is not None else (capabilities_obj.supports_enable_thinking if capabilities_obj else False), + "supports_thinking_budget": req.supports_thinking_budget if hasattr(req, 'supports_thinking_budget') and req.supports_thinking_budget is not None else (capabilities_obj.supports_thinking_budget if capabilities_obj else False), + "supports_enable_search": req.supports_enable_search if hasattr(req, 'supports_enable_search') and req.supports_enable_search is not None else (capabilities_obj.supports_enable_search if capabilities_obj else False), + + # 其他信息 + "notes": req.notes or (capabilities_obj.notes if capabilities_obj and hasattr(capabilities_obj, 'notes') else ""), } if llm_id: @@ -386,21 +450,7 @@ class LLMManager: llm_list = [] for llm in result: - # 标准化type字段为列表格式 - llm_type = llm.get("type", "embedding") - if isinstance(llm_type, str): - llm_type = [llm_type] - - llm_item = LLMProviderInfo( - llmId=llm["_id"], # _id已经是UUID字符串 - icon=llm["icon"], - openaiBaseUrl=llm["openai_base_url"], - openaiApiKey=llm["openai_api_key"], - modelName=llm["model_name"], - maxTokens=llm["max_tokens"], - isEditable=bool(llm.get("user_sub")), # 有user_sub的是用户创建的，可编辑 - type=llm_type, - ) + llm_item = LLMManager._create_llm_provider_info(llm) llm_list.append(llm_item) return llm_list @@ -425,21 +475,7 @@ class LLMManager: llm_list = [] for llm in result: - # 标准化type字段为列表格式 - llm_type = llm.get("type", "embedding") - if isinstance(llm_type, str): - llm_type = [llm_type] - - llm_item = LLMProviderInfo( - llmId=llm["_id"], # _id已经是UUID字符串 - icon=llm["icon"], - openaiBaseUrl=llm["openai_base_url"], - openaiApiKey=llm["openai_api_key"], - modelName=llm["model_name"], - maxTokens=llm["max_tokens"], - isEditable=bool(llm.get("user_sub")), # 系统模型（user_sub=""）不可编辑 - type=llm_type, - ) + llm_item = LLMManager._create_llm_provider_info(llm) llm_list.append(llm_item) return llm_list @@ -460,21 +496,7 @@ class LLMManager: llm_list = [] for llm in result: - # 标准化type字段为列表格式 - llm_type = llm.get("type", "reranker") - if isinstance(llm_type, str): - llm_type = [llm_type] - - llm_item = LLMProviderInfo( - llmId=llm["_id"], # _id已经是UUID字符串 - icon=llm["icon"], - openaiBaseUrl=llm["openai_base_url"], - openaiApiKey=llm["openai_api_key"], - modelName=llm["model_name"], - maxTokens=llm["max_tokens"], - isEditable=bool(llm.get("user_sub")), # 有user_sub的是用户创建的，可编辑 - type=llm_type, - ) + llm_item = LLMManager._create_llm_provider_info(llm) llm_list.append(llm_item) return llm_list @@ -499,21 +521,7 @@ class LLMManager: llm_list = [] for llm in result: - # 标准化type字段为列表格式 - llm_type = llm.get("type", "reranker") - if isinstance(llm_type, str): - llm_type = [llm_type] - - llm_item = LLMProviderInfo( - llmId=llm["_id"], # _id已经是UUID字符串 - icon=llm["icon"], - openaiBaseUrl=llm["openai_base_url"], - openaiApiKey=llm["openai_api_key"], - modelName=llm["model_name"], - maxTokens=llm["max_tokens"], - isEditable=bool(llm.get("user_sub")), # 系统模型（user_sub=""）不可编辑 - type=llm_type, - ) + llm_item = LLMManager._create_llm_provider_info(llm) llm_list.append(llm_item) return llm_list @@ -541,14 +549,27 @@ class LLMManager: # 优先使用配置文件中明确指定的provider，如果没有则从endpoint推断 provider = getattr(model_config, 'provider', '') or getattr( model_config, 'backend', '') or get_provider_from_endpoint(model_config.endpoint) - model_info = model_registry.get_model_info( - provider, model_config.model) - - # 根据provider获取图标，如果没有则使用配置文件中的图标 + + # 根据模型类型选择正确的ModelType + from apps.llm.model_types import ModelType + model_type_map = { + "chat": ModelType.CHAT, + "embedding": ModelType.EMBEDDING, + "reranker": ModelType.RERANK, + "function_call": ModelType.FUNCTION_CALL, + } + registry_model_type = model_type_map.get(model_type, ModelType.CHAT) + + # 从model_registry获取完整的模型能力 + capabilities = model_registry.get_model_capabilities( + provider, model_config.model, registry_model_type) + + # 获取图标 provider_icon = llm_provider_dict.get(provider, {}).get( "icon", getattr(model_config, 'icon', '')) # 创建系统模型 + # 对于非chat模型，使用默认值；对于chat模型，从capabilities获取 system_llm = LLM( _id=model_id, user_sub="", # 系统级别模型 @@ -560,13 +581,41 @@ class LLMManager: max_tokens=getattr(model_config, 'max_tokens', None), type=[model_type], # 使用列表格式 provider=provider, - supports_thinking=model_info.supports_thinking if model_info else False, - can_toggle_thinking=model_info.can_toggle_thinking if model_info else False, - supports_function_calling=model_info.supports_function_calling if model_info else True, - supports_json_mode=model_info.supports_json_mode if model_info else True, - supports_structured_output=model_info.supports_structured_output if model_info else False, - max_tokens_param=model_info.max_tokens_param if model_info else "max_tokens", - notes=model_info.notes if model_info else "", + + # 基础能力 - 使用getattr安全访问，适用于不同类型的能力对象 + supports_streaming=getattr(capabilities, 'supports_streaming', True) if capabilities else True, + supports_function_calling=getattr(capabilities, 'supports_function_calling', True) if capabilities else True, + supports_json_mode=getattr(capabilities, 'supports_json_mode', True) if capabilities else True, + supports_structured_output=getattr(capabilities, 'supports_structured_output', False) if capabilities else False, + + # 推理能力 + supports_thinking=getattr(capabilities, 'supports_thinking', False) if capabilities else False, + can_toggle_thinking=getattr(capabilities, 'can_toggle_thinking', False) if capabilities else False, + supports_reasoning_content=getattr(capabilities, 'supports_reasoning_content', False) if capabilities else False, + + # 参数支持 + max_tokens_param=getattr(capabilities, 'max_tokens_param', "max_tokens") if capabilities else "max_tokens", + supports_temperature=getattr(capabilities, 'supports_temperature', True) if capabilities else True, + supports_top_p=getattr(capabilities, 'supports_top_p', True) if capabilities else True, + supports_top_k=getattr(capabilities, 'supports_top_k', False) if capabilities else False, + supports_frequency_penalty=getattr(capabilities, 'supports_frequency_penalty', False) if capabilities else False, + supports_presence_penalty=getattr(capabilities, 'supports_presence_penalty', False) if capabilities else False, + supports_min_p=getattr(capabilities, 'supports_min_p', False) if capabilities else False, + + # 高级功能 + supports_response_format=getattr(capabilities, 'supports_response_format', True) if capabilities else True, + supports_tools=getattr(capabilities, 'supports_tools', True) if capabilities else True, + supports_tool_choice=getattr(capabilities, 'supports_tool_choice', True) if capabilities else True, + supports_extra_body=getattr(capabilities, 'supports_extra_body', True) if capabilities else True, + supports_stream_options=getattr(capabilities, 'supports_stream_options', True) if capabilities else True, + + # 特殊参数 + supports_enable_thinking=getattr(capabilities, 'supports_enable_thinking', False) if capabilities else False, + supports_thinking_budget=getattr(capabilities, 'supports_thinking_budget', False) if capabilities else False, + supports_enable_search=getattr(capabilities, 'supports_enable_search', False) if capabilities else False, + + # 其他信息 + notes=getattr(capabilities, 'notes', "") if capabilities else "", ) # 使用by_alias=True将id字段作为_id插入，保持UUID字符串格式