Chat-with-Github-Repo

[marcusquinn]

• https://github.com/peterw/Chat-with-Github-Repo
• https://twitter.com/pwang_szn/status/1650801868568772608
• https://twitter.com/i/status/1650801915918434304

1. Clones a git repository
2. Loading documents from the repository
3. Splitting the documents into smaller text chunks
4. We add those text chunks to a DeepLake instance with their computed embeddings using OpenAI Embeddings.
   Then have a main function that calls everything.
5. Now that we've embedded all the data, we can access the data from
   @activeloopai
   One thing with Deeplake is that you can also store metadata w/ your embeddings if you need to.

At a high-level, here's what's going on:

The DeepLake instance is loaded with embeddings from the dataset that we have in DeepLake.

The chatbot searches the dataset for relevant information and generates responses using GPT-3.5-turbo based on the user's input.