Better Call GPT

Inspiration

Inspired by chat-GPT, we aimed to use the chat format to create an AI assistant that could better help law firms by greatly reducing the time it takes to analyze client documents. For the technologies, we were inspired by the youtuber "Fireship" and his utilization of a vector database to give GPT long term memory. For the theme we were inspired by the hit AMC drama "Better Call Saul".

What it does

Given a vector database that is populated with a lawyers documents, our app can create summaries, analyses, and can reference multiple documents to answer a single prompt. We took the utility of a vector database and reduced queries to a conversation that can be easily used and understood.

How we built it

We split our team into 2 groups; A frontend and a backend. For the backend application we created a vector database instance using Weaviate and docker to for hosting. We then created multiple API endpoints that can interact with the database, and call gpt-3.5 for the chat functionality and analysis,.

Challenges we ran into

• Parsing very verbose JSONs caused some roadblocks
• Learning how to use Docker and Weaviate for the vector database
• CORS issues when making API requests

Accomplishments that we're proud of

• Getting a working vector database
• Getting the bot to reference specific documents when asked
• Frontend design

What we learned

• How an AI can utilize a vector database to add long term memory.
• How to better split up team work

What's next for Better Call GPT

• Using GPT's Whisper API, add a speech to text function so that users can talk directly to the bot.
• Increase security by encrypting documents

Built With

• javascript
• material-ui
• node.js
• openai
• react
• weaviate