Retrieval Augmented Generation

To pass this lesson, you will need to use the Neo4jVectorStore created in the Vectors lesson to retrieve similar documents to the user’s query and add them to the ChatMessagePromptTemplate to provide context to the LLM.

import {
 } from "@langchain/core/prompts";
import { ChatOpenAI, OpenAIEmbeddings } from "@langchain/openai";
import { StringOutputParser } from "@langchain/core/output_parsers";
import { RunnableSequence, RunnablePassthrough } from "@langchain/core/runnables";
import { Neo4jVectorStore } from "@langchain/community/vectorstores/neo4j_vector";

export async function call(
  message: string,
  sessionId: string
): Promise<string> {
  const embeddings = new OpenAIEmbeddings({
    openAIApiKey: process.env.OPEN_AI_API_KEY,
  const store = await Neo4jVectorStore.fromExistingGraph(embeddings, {
    url: process.env.NEO4J_URI,
    username: process.env.NEO4J_USERNAME,
    password: process.env.NEO4J_PASSWORD,
    nodeLabel: "Talk",
    textNodeProperties: ["title", "description"],
    indexName: "talk_embeddings_openai",
    embeddingNodeProperty: "embedding",
    retrievalQuery: `
    RETURN node.description AS text, score,
    node {
      .time, .title,
      url: ''+ node.url,
      speaker: [
        (node)-[:GIVEN_BY]->(s) |
        s { .name, .company, .x_handle, .bio }
      room: [ (node)-[:IN_ROOM]->(r) | ][0],
      tags: [ (node)-[:HAS_TAG]->(t) | ]

    } AS metadata
  const retriever = store.asRetriever();

  // 1. create a prompt template
  const prompt = ChatPromptTemplate.fromMessages([
      `You are a helpful assistant helping users with queries
      about the CityJS Athens conference.
      Answer the user's question to the best of your ability.
      If you do not know the answer, just say you don't know.`
      `Here are some talks to help you answer the question.
      Don't use your pre-trained knowledge to answer the question.
      Always include a full link to the meetup.
      If the answer isn't included in the documents, say you don't know.

    HumanMessagePromptTemplate.fromTemplate(`Question: {message}`),

  // 2. choose an LLM
  const llm = new ChatOpenAI({
    openAIApiKey: process.env.OPENAI_API_KEY,
    temperature: 0.9,

  // 3. parse the response
  const parser = new StringOutputParser();

  // 4. runnable sequence (LCEL)
  const chain = RunnableSequence.from<RunInput, string>([
      documents: RunnableSequence.from([
        (input) => input.message,
        retriever.pipe((docs) => JSON.stringify(docs)),
    new StringOutputParser(),

  // 5. invoke the chain
  const output = await chain.invoke(
    { message },

  return output;


Nice work!


Hi, I am an Educational Learning Assistant for Intelligent Network Exploration. You can call me E.L.A.I.N.E.

How can I help you today?