Github ChatBot LLM

Chatbot Powered by Time aware RAG, Reranking algorithms's & LLM --> Team MasterJackers

Comment

Inspiration

Every Software Developer team uses Github as its primary way to share code and work together. The problem is that when thins get big, commits get very difficult to get tracked, and also very difficult for each individual to know what is everyone doing inside the repository. That's why we built a Chabot able to digest all the repository data efficiently and give insights with any questions you make.

What it does

It gives insights about the commits history of the repository, knows what each code/file is about in the repo, and also is able to differentiate between what each collaborator has work on. The most interesting feature is that we make it possible to digest all this data history into time-aware data so we can choose between time frames for searching specific data.

How we built it

We used various Pipelines & technology, primary using LLama Index & TimeScaleVectorStores for making a connecting a PostgreSQL Database and then turning it into a VectorStore DataBase which powers the use of Hypertables for time-aware RAG. We used ChatGpt for powering the NLP tasks, and various pipelines for efficiently preprocessing the raw data.

Challenges we ran into

Making the RAG time-aware, connecting everything together.

Accomplishments that we're proud of

Finishing a working beta in less than 24 hours.

What we learned

The use of various libraries like llama-index and how to work with RAG.

What's next for Github ChatBot LLM

Upgrades and better responses.

Built With

  • llama-index
  • python
  • rag
  • timescalevector
Share this project:

Updates