semtree v0.3.0

New features

Hybrid search (semtree-rag)

• HybridSearcher fuses vector similarity (SearchEngine) with BM25 keyword matching (LexicalIndex) via reciprocal rank fusion - catches concepts a grep misses while keeping exact-identifier precision
• SearchMode selects Hybrid (default), Semantic, or Lexical
• LexicalIndex::from_chunks builds an in-memory BM25 index from a chunk registry

CLI (semtree-cli)

• semtree search --mode hybrid|semantic|lexical (defaults to hybrid)
• semtree search --lang / --kind / --path filters, applied after the vector lookup

Documentation

• Crate-level docs (//!) on all six library crates - each now has a pitch, a components table, and an example rendered on docs.rs
• New runnable example semtree-rag/examples/build_your_own.rs - assemble the pipeline from its building blocks
• New example crate examples/mcp-server - expose semtree as a Model Context Protocol server (tool search_code) for AI agents, using the rmcp SDK over stdio
• README: composable-library positioning, hybrid-search callout, MCP-server section, docs.rs/downloads badges