MongoDB Chat Memory
For longer-term persistence across chat sessions, you can swap out the default in-memory chatHistory
that backs chat memory classes like BufferMemory
for a MongoDB instance.
Setup
You need to install Node MongoDB SDK in your project:
- npm
- Yarn
- pnpm
npm install -S mongodb
yarn add mongodb
pnpm add mongodb
You will also need a MongoDB instance to connect to.
Usage
Each chat history session stored in MongoDB must have a unique session id.
import { MongoClient, ObjectId } from "mongodb";
import { BufferMemory } from "langchain/memory";
import { ChatOpenAI } from "langchain/chat_models/openai";
import { ConversationChain } from "langchain/chains";
import { MongoDBChatMessageHistory } from "langchain/stores/message/mongodb";
const client = new MongoClient(process.env.MONGODB_ATLAS_URI || "");
await client.connect();
const collection = client.db("langchain").collection("memory");
// generate a new sessionId string
const sessionId = new ObjectId().toString();
const memory = new BufferMemory({
chatHistory: new MongoDBChatMessageHistory({
collection,
sessionId,
}),
});
const model = new ChatOpenAI({
modelName: "gpt-3.5-turbo",
temperature: 0,
});
const chain = new ConversationChain({ llm: model, memory });
const res1 = await chain.call({ input: "Hi! I'm Jim." });
console.log({ res1 });
/*
{
res1: {
text: "Hello Jim! It's nice to meet you. My name is AI. How may I assist you today?"
}
}
*/
const res2 = await chain.call({ input: "What did I just say my name was?" });
console.log({ res2 });
/*
{
res1: {
text: "You said your name was Jim."
}
}
*/
// See the chat history in the MongoDb
console.log(await memory.chatHistory.getMessages());
// clear chat history
await memory.chatHistory.clear();
API Reference:
- BufferMemory from
langchain/memory
- ChatOpenAI from
langchain/chat_models/openai
- ConversationChain from
langchain/chains
- MongoDBChatMessageHistory from
langchain/stores/message/mongodb