v0.2.19

What's New

Features

• Semantic Selection: New selection_mode="semantic" using embedding providers (OpenAI, Voyage, Cohere, Sentence-Transformers, Ollama)
• Cross-Document Pattern Cache: PatternCache with bloom filter for efficient pattern reuse across documents
• Streaming Compression: Memory-bounded streaming with ~190K tokens/sec throughput
• Parallel Suffix Array: Rayon-based parallel SA in Rust (1.4-1.8x speedup)
• Template Extraction: Infrastructure for parameterized pattern compression
• Quality Monitoring: Production-ready monitoring with ASCII exports (Prometheus, JSONL, summaries)

Documentation

• Added Warp Terminal MCP setup instructions
• Updated API reference with all new features
• Added semantic selection guide to ML_INTEGRATION.md

Fixes

• Resolved ruff and mypy compliance issues
• Applied code formatting across codebase