DocLayout v3.1.0

DocLayout 3.1.0 completes the pinned PP-DocLayoutV3 ONNX decode and carries its usable geometry and reading order through conversion. Sol still reads the whole page and supplies text and HTML.

Highlights

• Retains V3 masks, classes, scores, boxes, and raw order keys; derives page-pixel contours and draws them in annotated images/PDFs. Invalid contours fall back to V3 rectangles, and unavailable or unmatched V3 geometry falls back to Sol without losing content.
• Uses a compact region guide and deterministic one-to-one reconciliation, with source-footprint lineage and explicit matching/fallback diagnostics. V3-only detections remain diagnostic evidence.
• Keeps automatic CUDA detection with CPU retry and visible whole-page Sol fallback on V3 runtime failure.
• Adds GUI authorization-field extraction from saved raw Markdown, local JSON/SQLite results, and optional classification that remains off by default.

Compatibility

• The pinned official PaddlePaddle/PP-DocLayoutV3_onnx artifact remains unchanged. Direct ONNX Runtime and its dependencies are now part of the base package; the former optional layout extra is not needed. First use may download the hash-verified weights.
• New conversions use pipeline sol-layout-v3/v4; accepted V3 matches can change geometry and reading order. Saved conversions retain their historical provenance, and field-only retries do not reconvert PDFs.
• Python >=3.11 is required. PyPI publication remains deferred.

Verification and limits

• 443 offline tests passed, 1 skipped; Ruff F,E9, uv lock, and dependency checks passed.
• Wheel and source archive built. An isolated Windows Python 3.13 wheel install passed CLI, packaged resource, GUI/API import, dependency, and uv-tool smoke checks.
• These checks do not establish model accuracy or fresh-machine GPU behavior. See the dated verification record in docs/layout-v3-plan.md for live-validation limits.

Install the base wheel with uv:
uv pip install "https://github.com/pypi-ahmad/DocLayout/releases/download/v3.1.0/doclayout-3.1.0-py3-none-any.whl"

DocLayout v3.0.0

DocLayout 3.0.0 adds optional local PP-DocLayoutV3 guidance to the existing GPT-6 Sol transcription pipeline. Sol still reads the full page and writes HTML. Confident, policy-qualified matches can use V3 rectangles and reading order; unmatched or ambiguous Sol content is retained. If layout preparation or inference fails, conversion continues with Sol and records the fallback.

Highlights

• Revision-pinned, hash-checked ONNX weights downloaded to an ignored local cache when absent; the layout extra requires Python 3.11+.
• GUI, CLI, API, PDF, and OCR conversion paths share layout guidance and diagnostics. Auto device mode can fall back to CPU; explicit CUDA does not silently switch devices.
• JSON, chunks, Markdown, OCR JSON, and annotations consume the resulting document. Annotations draw rectangles, not segmentation masks.
• Updated installation, configuration, architecture, OpenWiki, and interactive diagrams.

Install
The wheel and source archive are attached. PyPI publication remains deferred. For the base CLI without Git:

uv pip install "https://github.com/pypi-ahmad/DocLayout/releases/download/v3.0.0/doclayout-3.0.0-py3-none-any.whl"

To use V3 guidance, install the downloaded wheel with the layout extra, optionally adding gui or server. Configure Sol credentials before conversion.

Verification and limits

• 514 offline tests passed; 3 skipped. The uv lock check, Ruff correctness rules, and whitespace check passed.
• Both distributions built. The source archive excludes local cache content and includes the license, notice, and four prompt resources.
• An isolated Windows Python 3.13 wheel install with gui, server, and layout extras passed dependency and import checks.
• No live Sol, CPU-layout, or GPU-layout conversion was run for this release. Model accuracy and fresh-machine GPU behavior remain unverified. Repository-wide style Ruff and ty are not clean.

DocLayout v2.1.1

DocLayout v2.1.1 packages the changes on main since v2.1.0, including bounded document processing, the authenticated HTTP API, safer GUI previews, and updated exports.

Compatibility

HTTP conversion now requires a separate DOCLAYOUT_API_TOKEN sent as Authorization: Bearer <token>. Filepath requests also require an operator configured DOCLAYOUT_INPUT_ROOT; uploads do not. Failed requests use HTTP error statuses and a detail field instead of a successful status with success: false. Existing API clients must update their authentication and error handling. See the API setup guide.

Changes

• Bound document, archive, image, worksheet, embedded resource, upload, page selection, and model generated table expansion.
• Release provider resources deterministically and restrict referenced HTML and EPUB resources to embedded data.
• Sanitize Markdown previews and keep GUI launchers on loopback; the Windows launcher no longer terminates an existing port listener.
• Add launch folder credential loading, estimated Sol/Luna API costs, timestamped exports, interactive architecture diagrams, and updated guides.

Install

The wheel and source archive are attached below. For the base CLI without Git:

uv pip install "https://github.com/pypi-ahmad/DocLayout/releases/download/v2.1.1/doclayout-2.1.1-py3-none-any.whl"

The gui, server, and full extras add the browser app, HTTP API, and office/HTML/EPUB converters respectively. PyPI publication is deferred.

Verification

214 offline tests passed; 1 skipped. Ruff correctness and uv lock checks passed. A clean checkout produced both distributions. Fresh Windows Python 3.14 installs of the wheel with base, gui, and server dependencies passed consistency checks; CLI help, GUI/API imports, all four prompt resources, and an isolated uv tool install passed. No live model requests or Modal deployment checks were performed.

DocLayout v2.1.0

DocLayout 2.1.0 adds direct file conversion with selectable exports and GitHub installation through uv tool, uv pip, or pip.

Install

uv tool install "git+https://github.com/pypi-ahmad/DocLayout.git@v2.1.0"

For the browser workbench, install doclayout[gui] @ git+https://github.com/pypi-ahmad/DocLayout.git@v2.1.0 instead. Optional server and full extras add the HTTP API and document-format converters.

Set OPENAI_API_KEY and, for a custom compatible endpoint, OPENAI_BASE_URL. Then run:

doclayout input.pdf output --all
doclayout input.pdf output --markdown --html

The default is Markdown. Other flags select JSON, chunks, metadata, crops, annotated PDF, annotated PNGs, or a complete ZIP. Extraction runs once per selected page. Generated files with matching names are overwritten; use separate output directories to retain multiple documents.

Manual cloning and existing folder conversion remain supported. See the README and usage guide for setup and options.

Verification

• 130 offline tests passed; 3 billable integration tests skipped.
• Wheel and source builds, isolated pip/uv pip/uv tool installation, CLI exports with mocked extraction, GUI startup, and API startup passed.
• Runtime prompt files are unchanged. No live model calls were made for this release.

PyPI publication is deferred. Download the wheel or source archive below, or install from the GitHub tag.