Build or incrementally update a LanceDB semantic search index over markdown KB files.
npx skills add https://github.com/mathiasbourgoin/roster --skill kb-reindex
Please help me install this Agent Skill. Skill: kb-reindex Source: https://github.com/mathiasbourgoin/roster/tree/main/.opencode/skills/kb-reindex Command: npx skills add https://github.com/mathiasbourgoin/roster --skill kb-reindex
This Skill eliminates the tedious, error-prone manual work of building and maintaining semantic search indexes for project knowledge bases (KB), enabling fast, accurate semantic search over markdown KB files without constant manual intervention.
Use the kb-reindex skill to build a full semantic search index for all files in your project's kb/ directory, or update the existing index with only the files you've changed recently.
High-intent search queries and answers about installing and using this skill.
To build a semantic search index for your project knowledge base, use this Skill to scan your kb/ directory and generate a LanceDB vector index from all markdown documentation files in one go.
Yes, you can incrementally update a vector index. The Skill detects changed or deleted markdown files in your knowledge base and only re-indexes those specific files to save time.
No, semantic search indexing does not override your original markdown documentation. The LanceDB index acts as a non-intrusive search cache, keeping markdown files as the source of truth.
Yes, you can configure the embedding model and chunking strategy for your knowledge base index. The Skill supports tailoring chunking by section or fixed character length to fit your search needs.
The best way to exclude operational files from a knowledge base vector index is using the Skill's built-in exclusion rules, which automatically skip operational report files and the index directory itself during cold start indexing.
mathiasbourgoinSearch LanceDB knowledge base chunks using hybrid vector and BM25 rankings.
jtucker9Retrieve past CC session details via semantic vector search with LanceDB and SQLite fallback.
ataschzEmbed text into Gemini vector representations for retrieval and semantic search.
softsparkIndex project source files and documentation into a vector store for semantic search.
