How can I use LangChain to filter data and then implement RAG?

MongoDB Atlas Vector Search
1

For data in the pages collection, we would like to perform a RAG based on a similarity search on data where the creator userId field matches the userId of the currently logged in user and the pages data.

The LLM is using LangChain.

However, I am getting an error saying that I need index as a token.

Error during the document retrieval or generation process: MongoServerError: PlanExecutor error during aggregation :: caused by :: Path 'userId' needs to be indexed as token
    at Connection.sendCommand (/workspaces/Siki/backend/node_modules/mongodb/src/cmap/connection.ts:511:17)
    at processTicksAndRejections (node:internal/process/task_queues:95:5)
    at async Connection.command (/workspaces/Siki/backend/node_modules/mongodb/src/cmap/connection.ts:575:22)
    at async Server.command (/workspaces/Siki/backend/node_modules/mongodb/src/sdam/server.ts:322:16)
    at async executeOperation (/workspaces/Siki/backend/node_modules/mongodb/src/operations/execute_operation.ts:181:12)
    at async AggregationCursor._initialize (/workspaces/Siki/backend/node_modules/mongodb/src/cursor/aggregation_cursor.ts:71:22)
    at async AggregationCursor.[kInit] (/workspaces/Siki/backend/node_modules/mongodb/src/cursor/abstract_cursor.ts:644:21)
    at async next (/workspaces/Siki/backend/node_modules/mongodb/src/cursor/abstract_cursor.ts:717:7)
    at async AggregationCursor.[Symbol.asyncIterator] (/workspaces/Siki/backend/node_modules/mongodb/src/cursor/abstract_cursor.ts:302:26)
    at async AggregationCursor.toArray (/workspaces/Siki/backend/node_modules/mongodb/src/cursor/abstract_cursor.ts:438:22) {
  ok: 0,
  code: 8,
  codeName: 'UnknownError',
  '$clusterTime': {
    clusterTime: new Timestamp({ t: 1713417868, i: 4 }),
    signature: {
      hash: Binary.createFromBase64('WUOtgKIQnk2CBY4KhQ86oggZnZo=', 0),
      keyId: new Long('7314012942091943938')
    }
  },
  operationTime: new Timestamp({ t: 1713417868, i: 4 }),
  [Symbol(errorLabels)]: Set(0) {}
}

As a countermeasure, I created an AtlasSearch Index in the MongoDB Atlas UI as follows, but the result remains the same and the same error occurs.

{
  "mappings": {
    "dynamic": false,
    "fields": {
      "userId": {
        "analyzer": "lucene.standard",
        "type": "string"
      }
    }
  },
  "storedSource": {
    "include": [
      "userId"
    ]
  }
}

import { ChatOpenAI } from "@langchain/openai";
import { MongoDBAtlasVectorSearch } from "@langchain/mongodb";
import { MongoClient } from "mongodb";
import { OpenAIEmbeddings } from "@langchain/openai";
import { ChatPromptTemplate } from "@langchain/core/prompts";
import { StringOutputParser } from "@langchain/core/output_parsers";
import { createStuffDocumentsChain } from "langchain/chains/combine_documents";
import { sleep } from "langchain/util/time";
import { SystemPrompt, contentPrompt } from "./promptUtils"


async function retrieverGeneration(userId: string, userMessage: string) {
    let client;
    try {
        client = new MongoClient(process.env.MONGODB_URI || "");
        await client.connect();
        console.log("MongoDB connected successfully.");
    } catch (error) {
        console.error("Failed to connect to MongoDB:", error);
        throw new Error("MongoDB connection failed.");
    }

    try {
        const namespace = "Siki.pages";
        const [dbName, collectionName] = namespace.split(".");
        const collection = client.db(dbName).collection(collectionName);
        // Check if the index already exists and create it if it does not
        const indexExists = await collection.indexExists("userId_1");
        if (!indexExists) {
            await collection.createIndex({ "userId": 1 });
            console.log("Index created on 'userId'");
        }

        const vectorStore = new MongoDBAtlasVectorSearch(new OpenAIEmbeddings({
            openAIApiKey: process.env.OPENAI_API_KEY,
            batchSize: 2048,
            modelName: "text-embedding-3-small",
        }), {
            collection,
            indexName: "vector_index",
            textKey: "content",
            embeddingKey: "vector",
        });

        const retriever = vectorStore.asRetriever({
            k: 3,
            searchType: "similarity",
            filter: {
                preFilter: {
                    userId: userId // Filtering by userId
                }
            }
        });


        await sleep(2000);
        const prompt = ChatPromptTemplate.fromMessages([
            ["system", SystemPrompt],
            ["human", contentPrompt],
        ]);
        const llm = new ChatOpenAI({ modelName: "gpt-4", temperature: 0, openAIApiKey: process.env.OPENAI_API_KEY });
        const ragChain = await createStuffDocumentsChain({
            llm,
            prompt,
            outputParser: new StringOutputParser(),
        });

        const retrievedDocs = await retriever.getRelevantDocuments(userMessage);
        console.log("Documents retrieved:", retrievedDocs.length);
        const response = await ragChain.invoke({
            question: userMessage,
            context: retrievedDocs,
        });
        console.log("Response generated:", response);
        return response;
    } catch (error) {
        console.error("Error during the document retrieval or generation process:", error);
        throw error;
    } finally {
        await client.close();
        console.log("MongoDB connection closed.");
    }
}

export { retrieverGeneration };

How can I use LangChain to filter data and then implement RAG?

2

I found document.
fix it!

1 Like