v0.4.3 v1.23.0
high risk
airweave-ai/airweaveSep 21, 2026 → Sep 22, 20261 day between releases
What's New
Major overhaul of PDF text extraction with RTL support and CMap interpolation (v1.23.0) vs generic fixes (v0.4.3)
New Features
- Fixed extraction of text from malformed Form XObjects
- Fixed right-to-left text reading order
- CMap gaps field in API
- New OCR reason: invisible_text_layer
- Built-in CMaps for standalone binaries
Improvements
- Improved handling of re-encoded fonts and CMap gaps
- Improved font decoding for missing glyphs
- Interpolated missing ToUnicode CMap entries
- Interpolated single-glyph gaps in ToUnicode CMap
Bug Fixes
- Fixed extraction of text from malformed Form XObjects
- Fixed right-to-left text reading order
- Fixed Form XObject parsing for large BBox
- Fixed re-encoded simple font decoding
Upgrade Notes
The API now includes a new 'CMap gaps' field and a new 'invisible_text_layer' OCR reason. Users should verify their OCR integration handles these new fields and that standalone binaries now include built-in CMaps.
The Two Releases
v1.23.0
1.23.0 — Text-layer repairs
Sep 22, 2026· 1 month ago
Improved PDF text-layer extraction by fixing right-to-left reading order, handling malformed Form XObjects, and interpolating missing ToUnicode CMap entries.
Full release notes