Skip to content

Latest commit

 

History

History
17 lines (13 loc) · 1.15 KB

README.md

File metadata and controls

17 lines (13 loc) · 1.15 KB

recipe-rag

Find recipe that suits your needs. Built on top of Gemma-2B with RAG hosted on MongoDB. Made using the first 1k recipe from the recipe-nlg dataset.

The interface in action

How to replicate

  • The notebook was developed on Google Colab, but can be adapted for use in other environments.
  • Create MONGO_URI and HF_TOKEN environment values / secrets for your MongoDB database and Hugging Face account respectively.
  • Create a MongoDB collection for hosting the embeddings and assign a vector search index using the guide, while setting path as "embedding", similarity as "cosine" and numDimensions as 1024 to match the embedding model.
  • In a GPU environment, run the notebook and follow instructions.

TODO

  • Enrich response with the matched recipe.
  • Create a simple chat interface.

Reference