IndexAI

Inspiration

X has millions of users with a variety of communities and conversations. We wanted to build IndexAI that we can find people based on attributes, not by name.

What it does

Users can find other people in multiple ways:

• "Find me someone who works at xAI but also went to Stanford"
• "AI engineers in NYC who have a math and cs background"

How we built it

• We used NextJS to create a Grok styled chat UI with streaming respones.
• X API to get profiles, bios, locations, etc
• Pinecone vector database with integrated embeddings
• Neon/Postgres
• Mastra Framework for AI Agent orchestration

Challenges we ran into

X API Rate Limit 🫶

Accomplishments that we're proud of

Creating end to end pipeline with automated data fetching, embedding, and indexing

What we learned

• Vector Embeddings
• AI Agents using Mastra (shoutout Sam)

What's next for IndexAI

• Advanced search filters
• Improved tool calls (index personal sites)
• Expand our dataset

Built With

• nextjs
• python