This file records the notable changes in each Slide Alchemy release.
Compared with v0.1.0, this release focuses on safer inputs, reproducible asset slicing, stricter QA, and clearer failure recovery.
- Validate required fields, types, positive dimensions, bounding boxes, and component references in
element_analysis.json. - Reject unsafe or empty crop IDs and prevent crop output paths from escaping the selected output directory.
- Reject invalid contact-sheet dimensions and column counts.
- Make the default
unittestdiscovery command run the bundled regression suite.
- Correctly map SVGs whose
viewBoxstarts at a non-zero origin. - Reject non-positive SVG and slide dimensions.
- Fail clearly on unsupported visible SVG elements instead of silently producing incomplete PPTX output.
- Allow crop specs to store their key color while remaining compatible with the original crop-array format and CLI overrides.
- Store the magenta key color in all bundled strict-run crop specs, so the example sheets can be reproduced without extra arguments.
- Recursively inspect PNG assets and build contact sheets from nested directories.
- Fail asset QA when no PNG files are found instead of reporting a false success.
- Expand the failure decision tree with concrete recovery steps for base generation, element classification, asset sheets, SVG conversion, text layout, PPTX composition, and preview QA.
- Add regression coverage for SVG coordinate mapping, unsupported SVG elements, element-analysis validation, nested and empty asset directories, and crop-spec key colors.
- Initial public workflow for rebuilding image-based slides into editable, component-based PPTX files.