AgentSkillsCN

supabase-search

使用 PostgreSQL 全文搜索在 Supabase 中搜索知识库。使用“/supabase-search <查询>”进行搜索。

SKILL.md
--- frontmatter
name: supabase-search
description: Search knowledge base in Supabase using PostgreSQL full-text search. Use "/supabase-search <query>" to search.
user-invocable: true

Supabase Knowledge Search Skill

Purpose

Search the knowledge base stored in Supabase PostgreSQL database using full-text search and vector similarity (when pgvector is enabled).

Features

  • •Full-text search across knowledge_base table
  • •Metadata filtering (file_path, file_name, etc.)
  • •Results ranked by relevance
  • •Source attribution with GitHub links

Usage

bash
/supabase-search "RTL multiplication optimization"
/supabase-search "NoiseComputer architecture"
/supabase-search --limit 10 "debate results"

Search Methods

  1. •Full-text search: PostgreSQL tsvector and tsquery
  2. •Metadata filtering: JSON field searches
  3. •Vector similarity: pgvector cosine similarity (if configured)

Implementation

Python script: supabase_search.py

  • •Connects to Supabase PostgreSQL
  • •Executes search queries
  • •Formats results with context
  • •Includes source attribution

Configuration

Requires environment variables:

  • •SUPABASE_URL: Your Supabase project URL
  • •SUPABASE_SERVICE_ROLE_KEY: Service role key for database access

Output Format

code
Search Results for "RTL optimization"
=====================================

1. NoiseComputer RTL Implementation (Score: 0.95)
   Source: docs/brain/DECISIONS.md
   Content: RTL 곱셈 회피를 위해 256x256 lookup table 사용...

2. Previous Debate on Multiplication (Score: 0.87)
   Source: debate_results/2026-01-19T20:30:00Z
   Content: Claude and Gemini discussed alternatives to multiplication...

Integration with Multi-AI System

  • •Automatically searches before debates (context gathering)
  • •Used by Claude/Gemini to reference previous decisions
  • •Provides evidence for arguments
  • •Links to source files in GitHub

Future Enhancements

  • •Vector similarity search with embeddings
  • •Semantic search using Supabase AI
  • •Multi-language support
  • •Relevance ranking improvements