Tags

Tags give the ability to mark specific points in history as being important

  • v0.1.0

    19694d31 · feat: library API for custom schemas and filtered queries · 
    v0.1.0: Initial release

Local-first vector search for structure-aware plain text documents.
Extracted from a production second-brain knowledge base where it replaced
a 1200-line monolith, validated through real consumption as a library
dependency before tagging.

Features:
- Structure-aware chunking that respects meeting-date section boundaries
- Hybrid search (vector + FTS) with RRF reranking via LanceDB
- Automatic date extraction from natural language queries
- Section expansion: retrieves sibling chunks for full meeting context
- Library API with custom LanceModel schemas and filtered queries
- CLI for index, query, and status operations

Tested: 58 tests across chunking, date parsing, query integration,
and custom schema library usage patterns.

Built on: LanceDB, sentence-transformers (BGE), tantivy FTS.