本文介绍如何将记忆库集成到Agent。 ## 前提条件 * 已使用[VeADK](https://volcengine.github.io/veadk-python/agent/agent/)开发框架构建Agent代码。 * 已创建记忆库或导入已有记忆库，详情请参见[创建记忆库](https://www.volcengine.com/docs/86681/1844843)。 ## 步骤一：配置环境变量将记忆库集成至Agent中时，您可参照以下两种方式为其配置记忆库环境变量信息。 * 方式一：在智能体运行时中关联该记忆库。请参见[更新关联记忆库](https://www.volcengine.com/docs/86681/1844839#nxOIEDxL)。 * 方式二：获取环境变量后，在代码中配置。 1. 获取配置环境。 > 您也可以通过调用[GetMemoryConnectionInfo](https://www.volcengine.com/docs/86681/1913827)接口获取。 1. 登录[AgentKit控制台](https://console.volcengine.com/agentkit)。 2. 在左侧导航树中,选择“记忆库”，进入Memory页面。 3. 单击目标记忆库ID，进入“基本信息”页签。 4. 单击“集成代码”页签，然后根据开发框架获取Memory环境变量信息。 ![图片](https://portal.volccdn.com/obj/volcfe/cloud-universal-doc/upload_280730af495dd2eea75cebc582a22612.png =493x) 2. 配置环境变量。 ```mixin-react return ( ); ``` ## 步骤二：veADK代码集成示例 ### mem0记忆库 1. 安装veADK和mem0ai包。 ```Python pip install veadk-python mem0ai==0.1.118 ``` 2. 在Agent中集成mem0记忆库。 ```Python # 参考如下 memory 示例集成代码，修改 agent 代码 import asyncio import os from veadk import Agent, Runner from veadk.memory.long_term_memory import LongTermMemory from dotenv import load_dotenv load_dotenv() if os.getenv("DATABASE_MEM0_API_KEY") is None: raise ValueError("MEM0_API_KEY is not set") if os.getenv("DATABASE_MEM0_BASE_URL") is None: raise ValueError("DATABASE_MEM0_BASE_URL is not set") # Replace the following from your own user & session management system app_name = "ltm_demo" user_id = "temp_user" teaching_session_id = "teaching_session" student_session_id = "student_session" long_term_memory = LongTermMemory(backend="mem0", app_name=app_name) agent = Agent(long_term_memory=long_term_memory) runner = Runner( agent=agent, app_name=app_name, user_id=user_id, ) teaching_prompt = "My secret is 0xabcd" asyncio.run(runner.run(messages=teaching_prompt, session_id=teaching_session_id)) # save the teaching prompt and answer in long term memory asyncio.run(runner.save_session_to_long_term_memory(session_id=teaching_session_id,user_id=user_id)) # now, let's validate this in a new session, with the same user_id configured in the Runner student_prompt = "What is my secret?" response = asyncio.run( runner.run(messages=student_prompt, session_id=student_session_id) ) print(response) ``` ### Viking记忆库 1. 安装veadk、volcengine\-python\-sdk和volcengine。 ```JSON pip install veadk-python volcengine-python-sdk volcengine ``` 2. 在Agent中集成Viking记忆库。 ```Python # 参考如下 memory 示例集成代码，修改 agent 代码 import asyncio import os from veadk import Agent, Runner from veadk.memory.long_term_memory import LongTermMemory from dotenv import load_dotenv load_dotenv() collection_name = os.getenv("DATABASE_VIKINGMEM_COLLECTION") if not collection_name: raise ValueError("DATABASE_VIKINGMEM_COLLECTION environment variable is not set") long_term_memory = LongTermMemory(backend="viking_mem", index=collection_name) # Replace the following from your own user & session management system app_name = "ltm_demo" user_id = "temp_user" teaching_session_id = "teaching_session" student_session_id = "student_session" agent = Agent(long_term_memory=long_term_memory) runner = Runner( agent=agent, app_name=app_name, user_id=user_id, ) teaching_prompt = "My secret is 0xabcd" asyncio.run(runner.run(messages=teaching_prompt, session_id=teaching_session_id)) # save the teaching prompt and answer in long term memory asyncio.run(runner.save_session_to_long_term_memory(session_id=teaching_session_id,user_id=user_id)) # now, let's validate this in a new session, with the same user_id configured in the Runner student_prompt = "What is my secret?" response = asyncio.run( runner.run(messages=student_prompt, session_id=student_session_id) ) print(response) ```