Skip to main content

Supabase RAG Engine

This document describes the Supabase RAG Engine, a production-ready system for generating intelligent insights from document collections.

Overview​

The Supabase RAG Engine is a sophisticated RAG intelligence layer with hybrid search, production-ready resilience patterns, comprehensive observability and performance monitoring, and advanced quality metrics.

Key Functionalities:

  • Hybrid Search: Combines vector similarity search with keyword-based full-text search using Reciprocal Rank Fusion (RRF) for optimal result quality and performance.
  • Resilience: Implements a fallback search strategy to ensure users always get relevant results, even when precision-focused searches fail.
  • Quality Metrics: Calculates relevance density to distinguish between documents that are about a topic versus just mention it.
  • Insight Generation: Uses a multi-stage process to generate document summaries, direct answers, and related questions.

Technology Stack:

  • Supabase: The core platform, providing the PostgreSQL database, Edge Functions, and other backend services.
  • PostgreSQL: The database, with the pgvector and pg_trgm extensions.
  • Deno: The runtime for the Edge Functions.
  • OpenAI: The LLM provider for insight generation.

Edge Functions​

query-knowledge-base​

Performs a hybrid search (semantic + keyword) with a fallback strategy.

generate-rag-insights​

Generates insights from the search results using a multi-stage process.

Database Schema​

The database schema is defined in the supabase/bootstrap.sql file. It includes tables for documents, document chunks, insight cache, and search history.