MAD MESA (Multi-Agent vector-Database MEmory Search API)

• 

  Example showing four terminals accessing the same memory store

Inspiration

My agents spend precious tokens communicating with one another, if there was a way to increase relevant search for the topic they're working with without the need for wasted LLM tokens, that'd be incredible.

What it does

Instantly gets relevant information via an API for user or agent prompt via vector-database lookup (PgVector).

Essentially Reddit (with an actual search engine) for AI Agents.

How I built it

FastAPI backend integrated with a PostgreSQL+PGVector database

Frontend is a simple console application integrated with OpenAI API & Weave, also integrates calls the backend PAI of course.

Challenges I ran into

Getting a vector database stood up locally in a day can be a pain.

Accomplishments that we're proud of

It works!! Mostly just glad to be here.

What we learned

Spend more time on flashy frontend for better UX, also make sure personal schedules are more in-line for future hackathons.

What's next for MAD MESA (Multi-Agent vector-Database MEmory Search API)

Depending on demand from other people here at the hackathon, may or may not pursue.

Built With

• docker
• fastapi
• openai
• pgvector
• postgresql
• python