MCP Jina Supabase RAG

# Slash Command for mcp-jina-supabase-rag Create this file: `~/.claude/commands/jina.md` ```markdown --- allowed-tools: mcp__jina-supabase-rag__* argument-hint: <url_pattern> <project_name> [options] description: Crawl documentation and index to Supabase RAG --- # Index Documentation to Supabase Crawl and index documentation websites to Supabase for RAG (Retrieval-Augmented Generation). ## Arguments - **$1** - URL pattern (required) - Single URL: `https://docs.example.com/guide` - Pattern with wildcard: `https://docs.example.com/*` - **$2** - Project name (required) - Alphanumeric identifier for this documentation set - Used to filter searches later - Examples: `supabase-docs`, `anthropic-claude`, `nextjs-docs` - **$3-$N** - Optional arguments (parsed from natural language) - Discovery method: `sitemap` or `crawl` - Extraction method: `jina` or `crawl4ai` ## How It Works 1. **URL Discovery**: Tries sitemap.xml first, falls back to recursive crawling 2. **Content Extraction**: Uses Jina AI for fast extraction, Crawl4AI as fallback 3. **Chunking**: Intelligently splits content by headers and paragraphs 4. **Embedding**: Generates OpenAI embeddings for semantic search 5. **Storage**: Stores in Supabase with pgvector for similarity search ## Instructions Parse the user's command to extract: 1. URL pattern (first positional argument) 2. Project name (second positional argument) 3. Optional preferences from remaining arguments Call the `crawl_and_index` tool with: - `url_pattern`: The URL or pattern - `project_name`: The project identifier - `discovery_method`: "auto" (default), "sitemap", or "crawl" - `extraction_method`: "auto" (default), "jina", or "crawl4ai" After indexing completes, inform the user they can search with: ``` /search "query text" project-name ``` ## Examples ### Basic Usage ```bash # Index Supabase documentation /jina https://supabase.com/docs/* supabase-docs # Index Anthropic Claude docs /jina https://docs.anthropic.com/claude/* anthropic-claude # Index Next.js documentation /jina https://nextjs.org/docs/* nextjs-docs ``` ### With Options ```bash # Force sitemap discovery only /jina https://docs.example.com/* example-docs sitemap # Force Jina extraction /jina https://docs.example.com/* example-docs jina # Force Crawl4AI for both discovery and extraction /jina https://docs.example.com/* example-docs crawl crawl4ai ``` ### Single Page ```bash # Index a single guide page /jina https://docs.example.com/getting-started example-docs ``` ## Task Based on the provided URL pattern and project name: 1. Parse the command arguments 2. Call `mcp__jina-supabase-rag__crawl_and_index` with appropriate parameters 3. Report progress and final statistics 4. Remind user how to search the indexed content ## Searching Indexed Content After indexing, users can search with: ```bash # Search specific project "Search the supabase-docs for authentication examples" # This should call: mcp__jina-supabase-rag__search_documents # Search all projects "Search all indexed docs for API rate limiting" ``` Or create a companion command `/search` for convenience. ``` ## Installation ```bash # Create the command file mkdir -p ~/.claude/commands cat > ~/.claude/commands/jina.md << 'EOF' [paste the markdown above] EOF ``` ## Testing ```bash # List available commands /help # Try the jina command /jina https://docs.anthropic.com/claude/docs/* anthropic-test ```