
Mindly — AI Knowledge Workspace
AI-powered knowledge management workspace that ingests research documents, notes, and technical files into an interactive knowledge base powered by Google Gemini LLM and pgvector RAG pipeline.
- Next.js
- FastAPI
- Gemini LLM
- pgvector
- PostgreSQL
- Redis
- Docker
Behind the build
Built asynchronous background parser workers in FastAPI to chunk documents and generate vector embeddings stored in PostgreSQL via pgvector, integrated with Google Gemini LLM for context-aware response generation.
Trade-off. Initial document processing takes 2-3 seconds, but ensures sub-100ms latency during live query streaming.
Next.js App Router → FastAPI Workers → Google Gemini LLM → pgvector (PostgreSQL) → Redis Cache
