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:
- Send an arbitrary search query for a list of movies
- Get embeddings from OpenAI's Ada model
- Do vector search on the movie data's plot description
- Return the top 10 most similar items
- Go through setting up an atlas instance & vector search index as described here
- Create a user + connection string and store that as
MONGO_URI
in .env - Create an Open AI API key and store it in .env as
OPEN_AI_KEY
- Run
npm i
to get all the necessary modules - Run
node start
to start the server - In a new tab, call ./example_request.sh YOUR_SEARCH_TERM_HERE
- Within a few seconds you will get a list of movies with the closest matches to your search term
- Figuring out how to run all on a local mongodb instance
- Wrap for L402s.
- Doing RAG