v0.8.0

What's Changed

• fix: remove text recognition length limit by @GreatV in #150
• refactor(predictors): centralize auto-download path resolution by @GreatV in #152
• feat: support loading models from memory by @GreatV in #153
• docs: add FAQ, document in-memory model loading; bump version to 0.8.0 by @GreatV in #154

Full Changelog: v0.7.4...v0.8.0

v0.7.4

What's Changed

• fix(vl): avoid OOM in PaddleOCR-VL vision attention on low-VRAM GPUs by @GreatV in #148
• chore: bump version to 0.7.4, document environment variables, trim verbose comments by @GreatV in #149

Full Changelog: v0.7.3...v0.7.4

v0.7.3

What's Changed

• fix(vl): avoid BF16 on unsupported CUDA devices by @GreatV in #144
• ci: pin macOS runner to macos-26 to avoid image drift breaking cached builds by @GreatV in #145
• bump version to 0.7.3 by @GreatV in #146

Full Changelog: v0.7.2...v0.7.3

v0.7.2

What's Changed

• build(deps): update candle-transformers requirement from 0.10.2 to 0.11.0 by @dependabot[bot] in #141
• build(deps): update candle-core requirement from 0.10.2 to 0.11.0 by @dependabot[bot] in #140
• build(deps): update itertools requirement from 0.14 to 0.15 by @dependabot[bot] in #137
• refactor: replace crop_rotated_bounding_box with get_rotate_crop_image in BBoxCrop by @GreatV in #142

Full Changelog: v0.7.1...v0.7.2

v0.7.1

What's Changed

• feat: add PP-OCRv6 model support and optimize inference performance by @GreatV in #135
• build(deps): update imageproc requirement from 0.26.0 to 0.27.0 by @dependabot[bot] in #134
• chore(release): bump version to 0.7.1 by @GreatV in #136

Full Changelog: v0.7.0...v0.7.1

v0.7.0

What's Changed

• Add Windows CI for tests by @Copilot in #111
• perf: reduce image copies in OCR pipeline by @GreatV in #112
• build(deps): update tokenizers requirement from 0.22 to 0.23 by @dependabot[bot] in #113
• Support hierarchical speculative decoding and align inference outputs by @GreatV in #116
• build(deps): replace clipper2 with clipper2-rust by @GreatV in #117
• feat: auto-download ONNX models from ModelScope by @GreatV in #118
• Update deps and adapt Hayro 0.6 rendering by @GreatV in #119
• chore: update hayro dependency to version 0.7 and adjust related depe… by @GreatV in #121
• fix: improve clarity and formatting in HSD documentation by @GreatV in #122
• Fix math formatting in hsd.md by @GreatV in #123
• Fix link to ModelScope in README by @GreatV in #124
• Enhance image processing and OCR capabilities by @GreatV in #126
• build(deps): update nalgebra requirement from 0.34 to 0.35 by @dependabot[bot] in #125
• feat: update dependencies to use workspace features for improved management by @GreatV in #127
• refactor(oar-ocr-vl): dedupe weight loading, config, and vision RoPE helpers by @GreatV in #128
• fix: stabilize formula/layout inference, dedup VL masks, and tidy code & docs by @GreatV in #129
• refactor(oar-ocr-vl)!: remove Hierarchical Speculative Decoding (HSD) by @GreatV in #130
• Optimize DB layout postprocessing and align detection with official implementations by @GreatV in #131
• feat: add PP-OCRv6 models with direct links from PaddlePaddle's official channels by @GreatV in #132
• feat(core): SIMD-accelerate hot CPU normalization and CTC-decode kernels by @GreatV in #133

Full Changelog: v0.6.3...v0.7.0

v0.6.3

What's Changed

• build(deps): update candle dependencies from 0.9.2 to 0.10.2 by @GreatV in #104
• build(deps): update hayro requirement from 0.5 to 0.6 by @dependabot[bot] in #108
• build: dont pull in ureq on default-features = false by @sftse in #109
• bump(package): update version to 0.6.3 by @GreatV in #110

New Contributors

• @sftse made their first contribution in #109

Full Changelog: v0.6.2...v0.6.3

v0.6.2

What's Changed

• feat(vl): add GLM-OCR support with native Candle inference by @GreatV in #89
• feat(vl): add MinerU2.5 support for document parsing by @GreatV in #90
• feat(vl): add Metal GPU acceleration support for Apple devices by @GreatV in #92
• build(deps): update toml requirement from 0.9 to 1.0 by @dependabot[bot] in #94
• refactor(inference): unify OrtInfer to single multi-input/multi-output interface by @GreatV in #93
• build(deps): update rand requirement from 0.8 to 0.10 by @dependabot[bot] in #95
• Add TensorRT execution provider options for timing and engine cache by @aisinker in #97
• Fix hull angle sort using atan2 for total ordering by @inzanez in #98
• fix(structure): improve layout sorting, title levels, and inline formula handling by @GreatV in #99
• fix(structure): fix table cell matching, batch formula inference, and improve markdown output by @GreatV in #100

New Contributors

• @aisinker made their first contribution in #97
• @inzanez made their first contribution in #98

Full Changelog: v0.6.1...v0.6.2

v0.6.1

What's Changed

• refactor(derive): use darling for cleaner attribute parsing by @GreatV in #82
• build(deps): update hayro requirement from 0.4 to 0.5 by @dependabot[bot] in #83
• build(deps): update darling requirement from 0.20 to 0.23 by @dependabot[bot] in #84
• refactor(vl): add unified attention module and batched inference support by @GreatV in #85
• build(deps): update hayro requirement from 0.4 to 0.5 by @dependabot[bot] in #87
• build(deps): update criterion requirement from 0.5 to 0.8 by @dependabot[bot] in #86
• feat(vl): add PaddleOCR-VL-1.5 support with text spotting and seal recognition by @GreatV in #88

Full Changelog: v0.6.0...v0.6.1

v0.6.0

What's Changed

• Refactor oarocr module to use oar_ocr_core crate by @GreatV in #73
• Refactor adapter builders by @GreatV in #74
• refactor: Improve pattern matching and error handling by @GreatV in #75
• Add HunyuanOCR vision model implementation by @GreatV in #76
• build(deps): update ndarray requirement from 0.16 to 0.17 by @dependabot[bot] in #78
• Refactoring and code simplification by @GreatV in #79
• feat: Implement cross-cell OCR box splitting for improved table recognition by @GreatV in #80
• feat: implement LightOnOCR model by @GreatV in #81

Full Changelog: v0.5.1...v0.6.0