skill-rag-indexer

To index local markdown files for semantic search, you configure an embeddings provider and run a build command to generate dense vector representations stored in SQLite. This creates a queryable local index for offline-first workflows.

The best way to search local skill documents offline is using a SQLite-based index that combines dense vector embeddings with keyword matching. This local retrieval approach provides fast semantic search without requiring external network calls.

Yes, you need a configured embeddings provider to build a local RAG index. The indexing process relies on this provider to generate dense vector representations from your SKILL.md documents before storing them in SQLite.

You can update your semantic search index incrementally without a full rebuild. The indexer supports incremental updates to reflect changes in your skills, ensuring your local repository stays current efficiently.

The local skill indexer supports both vector search and keyword matching. It operates over a local SQLite index, allowing you to query your skill repository using dense semantic retrieval and traditional keyword-based matching.