Highlight
SST (Simplified Syntax Tree) Pipeline: The heuristic auto-source engine now parses pages into an immutable, typed AST (SST::Document → Segmenter → Scoring → Extractor) rather than running heuristic scoring directly over mutable Nokogiri DOM walks. This architecture delivers ~6× faster execution with roughly half the allocations on heuristic benchmarks while improving article extraction precision and filtering noise.
Selectors: Removed legacy :link selector alias in favor of standardized :url selector configuration (5cc2a34).
What's Changed
- feat(auto_source): SST heuristic pipeline and Nokogiri boundary pushback by @gildesmarais in #435
- feat(auto_source): tiered extract, lazy SST, and --limit by @gildesmarais in #436
- refactor(html): share article rules and raise test flight level by @gildesmarais in #437
- fix: close SSRF gaps, strip feed style, and emit channel author/image by @gildesmarais in #438
- refactor!: decompose JsonState scraper, optimize semantic traversal, and remove legacy :link alias by @gildesmarais in #439
- chore(release): bump version to v0.25.0 by @gildesmarais in #440
Full Changelog: v0.24.0...v0.25.0
Contributors
gildesmarais