Files
AstrBot/astrbot
Loagaeth e26fe1c3f5 feat(kb): add Markdown-aware chunker for structured documents (#8151)
* feat(kb): add Markdown-aware chunker for structured documents

* fix: address review feedback from sourcery-ai and gemini

- Clamp max_heading_depth to 1-6 to prevent regex errors
- Deduct prefix length when splitting oversized sections
- Replace hardcoded "[续]" with configurable continuation_prefix
- Skip fenced code blocks when detecting headings
- Cap pending size to prevent chunks exceeding chunk_size
- Refactor into dataclass + helper methods

* fix: handle unmatched fenced code block at EOF

* fix: prevent chunks exceeding chunk_size with long heading prefixes

* chore: rf

---------

Co-authored-by: Soulter <905617992@qq.com>
2026-05-30 17:43:13 +08:00
..