Aurora DSQL 聊天记忆

为了在聊天会话中实现更长期的持久性，您可以将默认的内存中 chatHistory 替换为无服务器的 PostgreSQL 兼容的 Amazon Aurora DSQL 数据库。

这与 PostgreSQL 集成非常相似，但有一些差异使其与 DSQL 兼容

1. PostgreSQL 中的 id 列是 SERIAL 自动递增的，而 DSQL 是使用数据库函数 gen_random_uuid 的 UUID。
2. 创建了 created_at 列来跟踪消息的顺序和历史记录。
3. PostgreSQL 中的 message 列是 JSONB，而 DSQL 是 TEXT，使用 Javascript 解析处理

设置

转到您的 AWS 控制台并创建一个 Aurora DSQL 集群，https://console.aws.amazon.com/dsql/clusters

提示

有关安装集成包的通用说明，请参阅此部分。

npm

npm install @langchain/openai @langchain/community @langchain/core pg @aws-sdk/dsql-signer

Yarn

yarn add @langchain/openai @langchain/community @langchain/core pg @aws-sdk/dsql-signer

pnpm

pnpm add @langchain/openai @langchain/community @langchain/core pg @aws-sdk/dsql-signer

用法

每个聊天历史会话都存储在 Aurora DSQL（Postgres 兼容）数据库中，并且需要会话 ID。

与 Aurora DSQL 的连接通过 PostgreSQL 连接池处理。您可以通过 pool 参数传递连接池的实例，也可以通过 poolConfig 参数传递连接池配置。有关更多信息，请参阅 pg-node 关于连接池的文档。提供的连接池优先，因此，如果同时传递了连接池实例和连接池配置，则仅使用连接池。

有关如何为 DSQL 进行身份验证和授权的选项，请查看 https://docs.aws.amazon.com/aurora-dsql/latest/userguide/authentication-authorization.html。

以下示例使用 AWS-SDK 生成身份验证令牌，该令牌将传递给连接池配置

import pg from "pg";

import { DsqlSigner } from "@aws-sdk/dsql-signer";
import { AuroraDsqlChatMessageHistory } from "@langchain/community/stores/message/aurora_dsql";
import { ChatOpenAI } from "@langchain/openai";
import { RunnableWithMessageHistory } from "@langchain/core/runnables";

import {
  ChatPromptTemplate,
  MessagesPlaceholder,
} from "@langchain/core/prompts";
import { StringOutputParser } from "@langchain/core/output_parsers";

async function getPostgresqlPool() {
  const signer = new DsqlSigner({
    hostname: process.env.DSQL_ENDPOINT!,
  });

  const token = await signer.getDbConnectAdminAuthToken();

  if (!token) throw new Error("Auth token error for DSQL");

  const poolConfig: pg.PoolConfig = {
    host: process.env.DSQL_ENDPOINT,
    port: 5432,
    user: "admin",
    password: token,
    ssl: true,
    database: "postgres",
  };

  const pool = new pg.Pool(poolConfig);
  return pool;
}

const pool = await getPostgresqlPool();

const model = new ChatOpenAI();

const prompt = ChatPromptTemplate.fromMessages([
  [
    "system",
    "You are a helpful assistant. Answer all questions to the best of your ability.",
  ],
  new MessagesPlaceholder("chat_history"),
  ["human", "{input}"],
]);

const chain = prompt.pipe(model).pipe(new StringOutputParser());

const chainWithHistory = new RunnableWithMessageHistory({
  runnable: chain,
  inputMessagesKey: "input",
  historyMessagesKey: "chat_history",
  getMessageHistory: async (sessionId) => {
    const chatHistory = new AuroraDsqlChatMessageHistory({
      sessionId,
      pool,
      // Can also pass `poolConfig` to initialize the pool internally,
      // but easier to call `.end()` at the end later.
    });
    return chatHistory;
  },
});

const res1 = await chainWithHistory.invoke(
  {
    input: "Hi! I'm MJDeligan.",
  },
  { configurable: { sessionId: "langchain-test-session" } }
);
console.log(res1);
/*
  "Hello MJDeligan! It's nice to meet you. My name is AI. How may I assist you today?"
*/

const res2 = await chainWithHistory.invoke(
  { input: "What did I just say my name was?" },
  { configurable: { sessionId: "langchain-test-session" } }
);
console.log(res2);

/*
  "You said your name was MJDeligan."
*/

// If you provided a pool config you should close the created pool when you are done
await pool.end();

API 参考

• AuroraDsqlChatMessageHistory 来自 @langchain/community/stores/message/aurora_dsql
• ChatOpenAI 来自 @langchain/openai
• RunnableWithMessageHistory 来自 @langchain/core/runnables
• ChatPromptTemplate 来自 @langchain/core/prompts
• MessagesPlaceholder 来自 @langchain/core/prompts
• StringOutputParser 来自 @langchain/core/output_parsers