LangChain Integration

Overview

MetaMemory provides a drop-in BaseMemory adapter for LangChain. Replace your existing ConversationBufferMemory import with the MetaMemory adapter and your chains get persistent multi-vector memory with five-channel retrieval. Zero architecture changes required.

Installation

npm install metamemory @langchain/core

Basic Usage

import { MetaMemoryAdapter } from 'metamemory/integrations/langchain';
import { ConversationChain } from 'langchain/chains';
import { ChatOpenAI } from '@langchain/openai';

const memory = new MetaMemoryAdapter({
  userId: 'agent-1',
  // All standard MemoryEngine options are supported
  similarityThreshold: 0.55,
});

const chain = new ConversationChain({
  llm: new ChatOpenAI({ modelName: 'gpt-4' }),
  memory,
});

const response = await chain.call({
  input: 'What happened with the Redis deployment last week?',
});

What the Adapter Does

Behind the scenes, the adapter:

• Implements LangChain's BaseMemory interface
• Stores each conversation turn as a memory with automatic multi-vector encoding
• On retrieval, runs five-channel adaptive search and returns relevant context
• Tracks emotional state per turn when emotion data is available
• Groups conversation turns into episodes automatically

Migration from ConversationBufferMemory

Replace the import and initialization. No other code changes needed:

// Before
import { ConversationBufferMemory } from 'langchain/memory';
const memory = new ConversationBufferMemory();

// After
import { MetaMemoryAdapter } from 'metamemory/integrations/langchain';
const memory = new MetaMemoryAdapter({ userId: 'agent-1' });