- Bugfix: Fix broken tagged release pipeline
- Bugfix: Fix broken ESM output in 4.0.0 release
- Feature: Support both x64 and ARM64 architectures across macOS, Linux and Windows
- Enhancement: Improved image binarization for dark/light theme support
- Enhancement: Ship both CommonJS and ESM
- Maintenance: Removed
cross-fetch dependency in favor of Node.js built-in fetch (requires Node >= 22)
- Maintenance: Replaced Jest with Vitest for testing
- Maintenance: Updated to latest nut.js core API
- BREAKING: Disable auto-registration and require explicit call to
useOcrPlugin
- BREAKING: Replaced jimp with image-js for image processing
- Improvement: Don't binarize images by default as it might lead to lower quality results
- Since this is a breaking change, you can still enable binarization by setting
binarize to true in preprocessConfig of the provider config
- Bugfix: Endless loop in findPartialMatches (#610)
- Feature: Added support for RegExp text search queries
- Bugfix: Drop alpha channel on input images to prevent issues with image processing
- Bugfix: Fix case-sensitivity for comparisons when searching for text
- BREAKING: Updated processing of textLine queries to only return bounding box around match, not the whole line
- Enhancement: Added auto wrapping for single language values in provider config
- Bugfix: Respect custom confidence values from match requests
- Enhancement: Return info about next best match in case of a miss
- Enhancement: Remove language data file from plugin package
- Enhancement: Default to configured LanguageModelType when preloading languages
- Enhancement: Improve typing of readText
- Updated loading of language data from cache
- Disabled sourcemaps
- Removed image debug output in preprocess step