This is a question not an issue.

[playground]

Hi @gregnr! First of all, thanks for sharing and step by step video. It's very helpful. Do you have a video on RAG and storing embeddings in vector db, preferably with Chromadb?

[gregnr]

Hey! Here is a video I made on RAG using similar open embedding models:

https://youtu.be/ibzlEQmgPPY

It uses Postgres + pgvector as the vector store, but it shouldn't be too hard to swap out pgvector for ChromaDB if you want. The advantage of pgvector IMO is that all your data (app and embeddings) live in the same place vs. jumping between multiple DBs.

Let me know if this helps 😃

[playground]

Hi @gregnr I have yet another question :-), I'm still trying to digest the info from open-source-embeddings as you demonstrated.

Currently, I'm using OpenAIEmbeddings for my documents and storing them in Chromadb. I would like to switch over to use open-source model instead of using OpenAI as you described but I'm a bit hung up on how to go about doing this. Please advise. Also, what is your opinion on using Neo4j KG with RAG?

const embeddings = new OpenAIEmbeddings({
    openAIApiKey: process.env.OPENAI_KEY,
  });

  const vectorStore = await Chroma.fromDocuments(
    chunks,
    new OpenAIEmbeddings({
      openAIApiKey: process.env.OPENAI_KEY,
    }),
    {
      collectionName: "test-data",
      url: "http://localhost:8000,
      collectionMetadata: {
        "hnsw:space": "cosine"
      }
    }
  );

and query against this dataset with questions.

        const embeddings = new OpenAIEmbeddings({openAIApiKey: process.env.OPENAI_API_KEY}); 
        const model = new OpenAI({ openAIApiKey: process.env.OPENAI_API_KEY });
        const data = await chromaDB.retrieveQAChain(collection, query, embeddings, model);