hack/tools/scripts/llm-key-hunter/results/_cache/content/aceli12138_pubmedagent.json

{"repo": "aceli12138/pubmedagent", "files": {"src/agents/medical_agent.py": {"01c2ffe428176dd3712e0d7fa8b7fdba1e0c1e64": {"t": 1785745450.1477273, "n": 4746, "h": "1e7f820856903623", "c": "from langchain_openai import ChatOpenAI\nfrom langchain.agents import AgentExecutor, create_react_agent\nfrom langchain.memory import ConversationBufferMemory\nfrom langchain_deepseek import ChatDeepSeek\nfrom src.prompts import get_medical_prompt\n# from src.tools import create_pubmed_tool\nfrom src.tools import tools\nfrom src.utils import load_config\nimport asyncio\n\nclass LiteratureAgent:\n    def __init__(self):\n        # 加载配置\n        config = load_config()\n        \n        # 创建工具\n        # self.tools = [create_pubmed_tool()]\n        self.tools = tools\n        \n        # 创建语言模型\n        if config.get(\"use_deepseek\", False):\n            self.llm = ChatDeepSeek(\n                api_key=config[\"deepseek_api_key\"],\n                base_url=config[\"deepseek_base_url\"],\n                model_name=config[\"deepseek_model_name\"],\n                temperature=0.3,\n                streaming=True\n            )\n        else:\n            self.llm = ChatOpenAI(\n                model=config[\"openai_model_name\"],\n                temperature=0.3,\n                api_key=config[\"openai_api_key\"],\n                base_url=config.get(\"openai_base_url\"),\n                streaming=True\n            )\n        \n        # 创建记忆\n        self.memory = ConversationBufferMemory(\n            return_messages=True,\n            memory_key=\"chat_history\",\n            output_key=\"output\",\n        )\n        \n        # 创建代理\n        agent = create_react_agent(\n            llm=self.llm,\n            tools=self.tools,\n            prompt=get_medical_prompt(),\n        )\n        \n        # 创建代理执行器\n        self.agent_executor = AgentExecutor(\n            agent=agent,\n            tools=self.tools,\n            memory=self.memory,\n            verbose=True,\n            return_intermediate_steps=True,\n            handle_parsing_errors=True,\n            max_iterations=3,\n            max_execution_time=None\n        )\n    \n    def query(self, input_text: str) -> dict:\n        \"\"\"处理用户查询\"\"\"\n        return self.agent_executor.invoke({\"input\": input_text})\n\n    async def aquery(self, input_text: str):\n        \"\"\"异步处理用户查询\"\"\"\n        async for event in self.agent_executor.astream_events(\n            {\"input\": input_text},\n            version=\"v1\"\n        ):\n            event_type = event.get(\"event\")\n            event_data = event.get(\"data\", {})\n            \n            if event_type == \"on_tool_start\":\n                yield {\n                    \"type\": \"thinking\",\n                    \"content\": {\n                        \"thought\": {\n                            \"action\": event_data.get(\"name\", \"使用工具\"),\n                            \"action_input\": str(event_data.get(\"input\", \"\")),\n                            \"observation\": \"\"\n                        }\n                    }\n                }\n            elif event_type == \"on_tool_end\":\n                yield {\n                    \"type\": \"thinking\",\n                    \"content\": {\n                        \"thought\": {\n                            \"action\": \"观察结果\",\n                            \"action_input\": \"\",\n                            \"observation\": str(event_data.get(\"output\", \"\"))\n                        }\n                    }\n                }\n            elif event_type == \"on_agent_action\":\n                # 确保思考内容是字符串\n                thought = event_data.get(\"thought\", \"\")\n                if not isinstance(thought, str):\n                    thought = str(thought)\n                \n                yield {\n                    \"type\": \"thinking\",\n                    \"content\": {\n                        \"thought\": {\n                            \"action\": \"思考\",\n                            \"action_input\": thought,\n                            \"observation\": \"\"\n                        }\n                    }\n                }\n            elif event_type == \"on_agent_finish\":\n                output = event_data.get(\"output\", \"\")\n                if not isinstance(output, str):\n                    output = str(output)\n                    \n                yield {\n                    \"type\": \"final\",\n                    \"content\": output\n                }\n                \n            # 添加对链结束事件的处理\n            elif event_type == \"on_chain_end\":\n                if \"output\" in event_data:\n                    output = event_data.get(\"output\", \"\")\n                    # 确保输出是字符串\n                    if not isinstance(output, str):\n                        if isinstance(output, dict) and \"output\" in output:\n                            output = output[\"output\"]\n                        else:\n                            output = str(output)\n                    \n                    yield {\n                        \"type\": \"final\",\n                        \"content\": output\n                    } "}, "f1a97d3ceefc159df7df8d0bc74518393b306e1c": {"t": 1785753381.2779016, "n": 4746, "h": "1e7f820856903623", "c": "from langchain_openai import ChatOpenAI\nfrom langchain.agents import AgentExecutor, create_react_agent\nfrom langchain.memory import ConversationBufferMemory\nfrom langchain_deepseek import ChatDeepSeek\nfrom src.prompts import get_medical_prompt\n# from src.tools import create_pubmed_tool\nfrom src.tools import tools\nfrom src.utils import load_config\nimport asyncio\n\nclass LiteratureAgent:\n    def __init__(self):\n        # 加载配置\n        config = load_config()\n        \n        # 创建工具\n        # self.tools = [create_pubmed_tool()]\n        self.tools = tools\n        \n        # 创建语言模型\n        if config.get(\"use_deepseek\", False):\n            self.llm = ChatDeepSeek(\n                api_key=config[\"deepseek_api_key\"],\n                base_url=config[\"deepseek_base_url\"],\n                model_name=config[\"deepseek_model_name\"],\n                temperature=0.3,\n                streaming=True\n            )\n        else:\n            self.llm = ChatOpenAI(\n                model=config[\"openai_model_name\"],\n                temperature=0.3,\n                api_key=config[\"openai_api_key\"],\n                base_url=config.get(\"openai_base_url\"),\n                streaming=True\n            )\n        \n        # 创建记忆\n        self.memory = ConversationBufferMemory(\n            return_messages=True,\n            memory_key=\"chat_history\",\n            output_key=\"output\",\n        )\n        \n        # 创建代理\n        agent = create_react_agent(\n            llm=self.llm,\n            tools=self.tools,\n            prompt=get_medical_prompt(),\n        )\n        \n        # 创建代理执行器\n        self.agent_executor = AgentExecutor(\n            agent=agent,\n            tools=self.tools,\n            memory=self.memory,\n            verbose=True,\n            return_intermediate_steps=True,\n            handle_parsing_errors=True,\n            max_iterations=3,\n            max_execution_time=None\n        )\n    \n    def query(self, input_text: str) -> dict:\n        \"\"\"处理用户查询\"\"\"\n        return self.agent_executor.invoke({\"input\": input_text})\n\n    async def aquery(self, input_text: str):\n        \"\"\"异步处理用户查询\"\"\"\n        async for event in self.agent_executor.astream_events(\n            {\"input\": input_text},\n            version=\"v1\"\n        ):\n            event_type = event.get(\"event\")\n            event_data = event.get(\"data\", {})\n            \n            if event_type == \"on_tool_start\":\n                yield {\n                    \"type\": \"thinking\",\n                    \"content\": {\n                        \"thought\": {\n                            \"action\": event_data.get(\"name\", \"使用工具\"),\n                            \"action_input\": str(event_data.get(\"input\", \"\")),\n                            \"observation\": \"\"\n                        }\n                    }\n                }\n            elif event_type == \"on_tool_end\":\n                yield {\n                    \"type\": \"thinking\",\n                    \"content\": {\n                        \"thought\": {\n                            \"action\": \"观察结果\",\n                            \"action_input\": \"\",\n                            \"observation\": str(event_data.get(\"output\", \"\"))\n                        }\n                    }\n                }\n            elif event_type == \"on_agent_action\":\n                # 确保思考内容是字符串\n                thought = event_data.get(\"thought\", \"\")\n                if not isinstance(thought, str):\n                    thought = str(thought)\n                \n                yield {\n                    \"type\": \"thinking\",\n                    \"content\": {\n                        \"thought\": {\n                            \"action\": \"思考\",\n                            \"action_input\": thought,\n                            \"observation\": \"\"\n                        }\n                    }\n                }\n            elif event_type == \"on_agent_finish\":\n                output = event_data.get(\"output\", \"\")\n                if not isinstance(output, str):\n                    output = str(output)\n                    \n                yield {\n                    \"type\": \"final\",\n                    \"content\": output\n                }\n                \n            # 添加对链结束事件的处理\n            elif event_type == \"on_chain_end\":\n                if \"output\" in event_data:\n                    output = event_data.get(\"output\", \"\")\n                    # 确保输出是字符串\n                    if not isinstance(output, str):\n                        if isinstance(output, dict) and \"output\" in output:\n                            output = output[\"output\"]\n                        else:\n                            output = str(output)\n                    \n                    yield {\n                        \"type\": \"final\",\n                        \"content\": output\n                    } "}}}}