langchain

LangChain JS example

Exercises Neo4jChatMessageHistory and Neo4jEntityRetriever directly, without depending on any specific @langchain/* package version. These classes are duck-typed against the LangChain JS interfaces, so they fit into RunnableWithMessageHistory and a retriever chain regardless of LangChain JS version skew.

Run it

cp .env.example .env       # set MEMORY_API_KEY
npm install
npm start

The script:

1. Creates a fresh conversation.
2. Persists three messages via the BaseChatMessageHistory-shaped API.
3. Reads them back.
4. Adds an entity and queries it via the retriever interface.

Using inside a real LangChain JS app

import { ChatOpenAI } from "@langchain/openai";
import { RunnableWithMessageHistory } from "@langchain/core/runnables";
import { Neo4jChatMessageHistory } from "@neo4j-labs/agent-memory/integrations/langchain";

const chain = new ChatOpenAI({ model: "gpt-4o-mini" }).pipe(...);

const memoryChain = new RunnableWithMessageHistory({
  runnable: chain,
  getMessageHistory: (sessionId) => new Neo4jChatMessageHistory(memory, sessionId),
  inputMessagesKey: "input",
  historyMessagesKey: "history",
});

See also

• How-to: LangChain JS
• LangChain JS docs