projects

bash-A: RAG Chatbot

2025 · live · github

A RAG chatbot that uses PDF documents uploaded by admins as a knowledge base. Provides contextually aware, streaming responses to user queries.

Technical Details

  • Role-Based Access: Differentiates between admin and user roles
  • Admin-Only PDF Ingestion: Admins can upload and process PDF files to build the knowledge base
  • RAG-Powered Chat: The LLM uses a vector search over the document knowledge base to generate relevant answers
  • Streaming UI: Responses are streamed token-by-token for a fluid chat experience
  • Vector Search: Uses Neon serverless Postgres with an HNSW index for efficient semantic search
  • Authentication: Clerk (handles user roles)
  • Database: Neon Serverless Postgres + pgvector
  • ORM: Drizzle ORM
  • AI & Embeddings: Google Gemini (gemini-2.5-flash-lite, text-embedding-004) via Vercel AI SDK
  • Styling: Tailwind CSS
  • UI: Shadcn UI & Vercel AI Elements