Fullbleed 2.6.0a6: component pagination and authored showcases

Fullbleed's component preview now includes the table pagination and compiled-reflow fixes from stable 2.5.32. Oversized top- and baseline-aligned cells continue across pages with their content, columns and source diagnostics preserved. The added fixtures cover baseline-aligned text and top-aligned lists; rich cells with middle or bottom alignment still have documented limits.

Install this experimental Python API explicitly:

python -m pip install --pre "fullbleed==2.6.0a6"

The API remains under fullbleed.ui.experimental and may change. Normal installation continues to select the stable release.

Author a document

• Bootstrap components and styling: reusable cards, typed tables, named-part styles, scoped CSS and a static component catalog. Export a complete HTML document or fragment, with or without styles. Bootstrap JavaScript is not included.
• Northstar and Common Ground, two authoring paths: the same invoice and report authored in HTML/CSS and Python components, with four matching rendered pages and a local PDF editing preview.
• An authored accessible guide: a two-page component document with headings, image text, lists and table headers, plus styled and unstyled HTML exports. This exact PDF passes the retained veraPDF PDF/UA-1 machine checks; human assistive-technology testing has not been performed.
• Try the component notebook: edit order data, rich table cells and styles, then download a PDF and both HTML exports.

Verification

The exact-tag release passed. All 15 public PyPI wheels and the source archive match the qualified artifacts. A fresh public installation passes the runtime, source-kit, table and line-height checks; the notebook's four execution cases also pass. The Windows/Linux docs jobs retain 120 PDFs matching the reviewed outputs. Website deployment and real browser downloads pass at three widths. Template version 0.1.2 works through the default CLI registry, with 12 local installs and 48 CI installs covering current and historical versions. Normal pip installation still selects stable 2.5.32.

Source qualification includes 938 Python tests with no skips, 1,425 Rust tests with Python/SVG features, 1,424 default-feature Rust tests and 11 doctests in each configuration. Linux and Windows CI also cover Python 3.10–3.14 and the 3.15 release candidate. On 3.15, each platform passes 937 tests and retains the known _xxsubinterpreters skip.

The table gates retain 180 native table cases, 240 pagination cases and 240 component/export cases. The 420 shared native fixtures match stable 2.5.32 byte for byte. The full independent corpus retains 1,620 passes, 24 failures and 18 disputed references; all 1,662 results and 5,489 saved files match the qualified alpha 5 and stable 2.5.32 baselines.

The external validator checks cover the retained specimens: all 120 pagination PDFs and all 60 component-table PDF/UA-1 PDFs pass their requested profiles. Existing caption findings in the other PDF/UA-2 fixtures remain tracked in #98. The dedicated PDF/VT job is unconfigured and skipped. These checks do not establish general Bootstrap/CSS, accessibility, PDF/VT or human usability conformance.

The original preview timeout and its successful isolated retry are retained. The framework sample now has its own alpha 6 preview filename; the image is byte-identical to the reviewed alpha 5 sample. Known Linux reader-image differences match the retained alpha 5 Linux images, and the corresponding PDFs match the qualified PDFs exactly.

The public-package evidence also retains the initial PyPI install-index propagation delay and the subsequent ordinary-index and fresh-install checks.

Download the source qualification evidence and public package evidence, website and CLI delivery evidence, and complete corpus artifact. The attached source kits and PDFs are versioned; prior releases remain available.