Library RAG: Semantic Search

Script paths below are relative to this skill's base directory ($SKILL_DIR), which Claude Code reports when the skill loads.

Semantic search over your local library of markdown-converted papers using sentence-transformers embeddings and ChromaDB.

Prerequisites

• uv installed (standard in this project)
• Papers ingested via ingest.py (which converts to markdown, organizes files, and adds metadata to references.bib)
• references.bib with md_path fields linking citation keys to markdown files

Important: Only files registered in references.bib are indexed. Loose markdown files in library/markdown/ without a bib entry will be flagged as "unlinked" during indexing. Run ingest.py to register them.

Quick Start

# Index your library (first time or after adding papers)
uv run "$SKILL_DIR"/scripts/rag.py index