Goal is to make a Retrieval-Augmented Generator (RAG) for any github repo, currently the project uses a Small Language Model i.e. tinyllama to generate text based on the content of the README.md file.

TODO:

• Add directories and files recursively for the whole github repo.
• Experiment with different models maybe RLHF models would perform better.
• Automate the process using github action or make a pipeline which creates a RAG just by giving the repo link.
• Make a interface using streamlit.
• Manage ollama dependencies.

Demo:

demo.mp4