Supabot: Personalized AI Chat Bot for Supabase

Supabot transforms raw Supabase documentation into intelligent, contextual AI responses using a hybrid search approach.

Project Structure

Supabot is divided into two repositories:

1. Data Pipeline

   • HTML Scraping: Extracts raw content from Supabase documentation.
   • Text Cleaning: Processes and structures the scraped text for consistency.
   • Vector Embeddings: Generates 1536-dimensional embeddings locally (no LLM required).
   • Hybrid Search: Combines semantic vector search with traditional keyword matching for optimal results.

2. Client (RAG)

   • Fetches clean, vectorized data from the database.
   • Implements Retrieval-Augmented Generation (RAG) to deliver contextual answers based on relevant documentation.

Technology Stack

• Supabase + pgvector: Uses PostgreSQL with the pgvector extension to store and search 1536-dimensional embeddings.
• Hybrid Search Engine: Integrates semantic search and keyword matching for improved accuracy.

Features

• Personalized AI Chat Bot: Delivers contextual answers from Supabase docs.
• Efficient Data Pipeline: Local embedding and hybrid search for fast, accurate retrieval.
• Clean Separation: Data processing and client RAG are maintained in separate repositories.

Find the pipeline source code on GitHub.