Gemini Beyond Prompts
AI dashboard prototype combining a Google Gemini chat with document analysis (RAG) and agent screens in a single interface.
Challenge
Explore how generative AI features go beyond a prompt box: conversations with context, semantic search over your own documents and tasks carried out by agents.
Solution
A React, Vite and TypeScript application with three modules: a Google Gemini chat that keeps the conversation context, a RAG-style document library and an assistant with research, planning and execution agents inspired by LangGraph. The chat calls the API for real; document analysis and agents are interface prototypes with simulated processing, and the database for the full version (Supabase with pgvector) is already modeled.
My contribution
Solo study project: interface, Gemini API integration and module structure.
Outcomes
- Demo published on Vercel.
- Working chat with the Gemini API, using the visitor's own key.
- Database modeled in Supabase with pgvector for the next step: semantic search.