Spaces:
Running
Running
Commit History
-
chore: update .gitignore, clean up junk, finalize V2.3.0 arch doc
.gitignore: - Add *.log to prevent log files from being tracked - Add scripts/*.db for accidental DB copies in scripts dir Deleted: - scripts/preindex_abbrevs.log (659 KB log, no longer needed) TIPITAKA_WEB_ARCHITECTURE_V2.md: - Update to V2.3.0 with all changes: * Adjacent <B> merge for heading+fn ref * Marker normalization exclusions (], ., Thai digits) * Dialogue markers อนุ. วิ. ปฏิ. * Pre-indexed abbrev system (1,742 pages, DB-only) * Direct SQLite abbrev lookup (no RAG/LLM import)5d89832 -
refactor: disable LLM fallback for (ย่อ), fix JSON string array parsing
All 1,742 pages with (ย่อ) markers are now pre-indexed in DB. Remove expensive DeepSeek API call on every (ย่อ) click. Fix expand_abbreviation to parse plain string arrays from preindex_abbrevs.efc32dd
-
fix: add อนุ. วิ. ปฏิ. dialogue markers to split formatting
Add อนุ. (อนุเคราะห์), วิ. (วิสัชนา), and ปฏิ. (ปฏิปุจฉา) to both marker_normalization (split onto new lines) and num_pattern (format as numbered-line) so dialogue in Abhidhamma/Paṭṭhāna pages renders each speaker on its own line.9babb2f -
fix: exclude '.' from marker normalization preceding char
Prevents ๕๘. (๑) from being split into ๕๘. + \n(๑) by the marker normalization step, which was causing section markers [๙๔] ๕๘. and their sub-items (๑) to render on separate paragraphs.70f46ee -
fix: merge adjacent <B> headings with <I> footnote refs, fix marker normalization exclusion list
- Merge <B>text</B><I>fn-</I><B>text</B> into one heading before B→h4 conversion to prevent footnote refs from splitting into multiple h4. - Add ']' and Thai digits (\u0E50-\u0E59) to marker normalization preceding-char exclusion list to prevent: • [๓๒] ๒. → [๓๒] + \n๒. • ๑๖. → ๑ + \n๖.6b67cae