How to pass data from database to Langchain like context/source

[teomikitenko]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

const ragChain = RunnableSequence.from([
    {
      context:  embedding, // how pass my own data
      question: new RunnablePassthrough(),
    },
    prompt,
    llm,
    parcer,
  ]);

I build my own RAG application but have a trouble, Logic my app is - user push button and load some document(text), after this i generate embeddings using this document and write its to database,i use Convex. And i want get this embedding from this document on database and add to llm to generate answer. But i dont find any example how do this, find only example how use ConvexVectoreStore and transform it to retriver - but in my case its dont work(i have several reason). I need just use embedding from document and pass to context and i dont know how and langchain document/discord dont give me answer. Chat bot in langchain site recomend me create CustomRetriever class - but why doc dont have answer, Pls help me

windows 10, using next,js convex and langchain

[dosubot[bot]]

To pass data from a database to LangChain as context or source, you can follow these steps:

1. Retrieve Embeddings from the Database: Fetch the embeddings from your database.
2. Create Documents with Retrieved Embeddings: Use the retrieved embeddings to create Document objects.
3. Set Up the Runnable Chain: Create a chain of runnables to process the context and question.

Here's an example of how you can achieve this:

import { ChatOpenAI } from "@langchain/openai";
import { Document } from "@langchain/core/documents";
import { ChatPromptTemplate } from "@langchain/core/prompts";
import { RunnableLambda, RunnableMap, RunnablePassthrough } from "@langchain/core/runnables";
import { StringOutputParser } from "@langchain/core/output_parsers";
import pg from "pg";

// Step 1: Retrieve embeddings from the database
async function fetchEmbeddingsFromDatabase() {
  // Implement your logic to fetch embeddings from the database
  return [
    { pageContent: "Harrison worked at Kensho" },
    { pageContent: "Bears like to eat honey." },
  ];
}

const embeddingsData = await fetchEmbeddingsFromDatabase();

// Step 2: Create Documents with Retrieved Embeddings
const documents = embeddingsData.map(data => new Document(data));

// Step 3: Set Up the Runnable Chain
const retriever = {
  invoke: async (input) => {
    // Implement your logic to retrieve the relevant document based on the input
    return documents[0].pageContent; // Example: returning the first document's content
  }
};

const prompt = ChatPromptTemplate.fromMessages([
  ["ai", `Answer the question based on only the following context:\n\n{context}`],
  ["human", "{question}"],
]);

const model = new ChatOpenAI({});
const outputParser = new StringOutputParser();

const setupAndRetrieval = RunnableMap.from({
  context: new RunnableLambda({
    func: (input) => retriever.invoke(input),
  }).withConfig({ runName: "contextRetriever" }),
  question: new RunnablePassthrough(),
});

const chain = setupAndRetrieval.pipe(prompt).pipe(model).pipe(outputParser);

// Step 4: Invoke the Chain
const response = await chain.invoke("Where did Harrison work?");
console.log(response);
/**
Harrison worked at Kensho.
 */

This code demonstrates how to retrieve embeddings from a database, create documents, set up a retriever, and pass the context to LangChain for generating answers in a RAG application ^[1][2][3].

To continue talking to Dosu, mention @dosu.

[whokc-github]

Hi @dosu, currently I have a List[Document] instead of a single document, how can I modify the code above (especially the retriever creation) in order to use LLM to produce response from a question? Thanks.

• William