inspecting-rag-catalog

Deterministic process for answering "What's in our RAG?" by listing all indexed sources.

When to Use

• When asked what documents are in the RAG system.
• When verifying ingestion success.
• When planning which sources to query.
• Before ingesting new content (to avoid duplicates).

Prerequisites

• Active Qdrant vector store with at least one indexed document (running via Docker).
• FastMCP RAG Server must be actively running via scripts/mcp/servers/rag/start_rag_server.bat (exposing SSE on port 8000).

Process

1. Fetch Catalog Data:
   • Execute the native catalog script: python "scripts/list_ebooks.py" locally from this skill directory.
   • Wait for the JSON response containing the list of sources.
2. Present results: Parse the JSON string and display the list of unique PDF sources with their paths.
3. Summarize: Report the total count and topic areas covered.

Important Rules

• No semantic search needed: This is a metadata-only operation, not a vector search.
• Exact tool: Always use list_library_sources, never search_library for catalog queries.

Best Practices

• Run this before ingesting new content to check for duplicates.
• Use the source list to inform which topics are queryable via retrieving-rag-context.