uncleJim21 / Mongo-Vector-Search

Playing with mongo vector search

What is This?

This project is just an example for using co-located embeddings & operational data on a MongoDB server. It uses the sample m_flix mongoDB data to:

1. Send an arbitrary search query for a list of movies
2. Get embeddings from OpenAI's Ada model
3. Do vector search on the movie data's plot description
4. Return the top 10 most similar items

Setup

1. Go through setting up an atlas instance & vector search index as described here
2. Create a user + connection string and store that as MONGO_URI in .env
3. Create an Open AI API key and store it in .env as OPEN_AI_KEY

Usage

1. Run npm i to get all the necessary modules
2. Run node start to start the server
3. In a new tab, call ./example_request.sh YOUR_SEARCH_TERM_HERE
4. Within a few seconds you will get a list of movies with the closest matches to your search term

Future Work

1. Figuring out how to run all on a local mongodb instance
2. Wrap for L402s.
3. Doing RAG