JuggerRAG

• 

  the completion function that processes the prompt through an LLM, e.g mistral

• 

  the similar chunks function that get's N number of top similar chunks

Inspiration

The inspiration for JuggerRAG stemmed from the challenges people face when accessing legal information in Pakistan. Laws are often complex and scattered across multiple sources, making it difficult for the average pakistani to navigate them. We wanted to create a tool that bridges this gap, empowering users to find accurate legal information quickly and effortlessly. By combining Retrieval-Augmented Generation (RAG) with a user-friendly interface, we aimed to make legal assistance more accessible to everyone.

What it does

It's a chatbot for legal assistance, it handles queries according to Pakistan's law and the Islamic law.

How we built it

First we gathered the legal handbooks that complies with Pakistani law, then we moved to the second step, familiarizing ourselves with snowflake and snowsql which was a bit of a challenge, but as soon as we got the hang of it, we processed the data in the database, and then concurrently one of us was working on the streamlit frontend. which was built earlier than the backend, we also did a lot of changes to the prompt and front-end.

Challenges we ran into

The main challenge we faced was first to gather Pakistan's law compliant handbooks in pdf format, then second challenge which was not that big of a challenge was familiarize ourselves with snowflake, then another challenge was that we couldn't get the LLM's complete function to work for some reason, after many tries we moved to an alternative approach of using hugging-face's inference API for mistral LLM.

Accomplishments that we're proud of

We're proud of the fact that we completed this project, as this is our first project outside of our coursework, that we did completely on our own, and with a lot of dedication and some sleepless nights despite our final exams.

What we learned

1. How to work as a team.
2. How to solve problems ourselves by reading documentation and hit and try method mostly.
3. How to divide tasks.
4. How to stay consistent despite other commitments. ## What's next for JuggerRAG Moving it to a larger scale to cover more areas of the world to enable users from other parts of the world to have easy access to legal information.

Built With

• huggingface
• mistral
• python
• snowflake