Repository navigation
PDF Reader 0.1.5
- Stop native processor descendants on POSIX timeout/cancellation and promptly reject cancelled queued work.
- Track complete text delivery across contiguous read offsets without combining different text representations; failed calls and cancelled image saves no longer claim delivery.
- Validate OCR language arguments consistently and distinguish unknown image capability from explicitly text-only routes.
- Exercise actual DeepSeek model metadata, arbitrary model IDs, route/capability changes, and real images without network or model calls.
- Add Harness
0.2.1-rc.5and0.2.2-alpha.2package lanes with exact Cordis families, alpha native-executor fixtures, and a current rc.5 CLI lifecycle lane.
Install the attached .tgz using the correct DSH profile's Plugins page or CLI. Verify it against SHA256SUMS. This GitHub release does not update an installed profile or confer Store approval.
Validation covers real Harness services and extracted package contents, plus controlled model-adapter behavior. Paid provider output quality and packaged desktop GUIs are not certified. PDF native processing is tested on Linux/macOS; Windows native execution remains unverified. Poppler and optional Tesseract must be installed separately. The unscoped npm name belongs to another author; use this repository archive.
Validation: 46 extracted-package tests and 12 actual-adapter cases passed on each current runtime. All ten package/profile CI lanes passed on the merged release commit, including real rc.5 installation, startup, tool execution and uninstall. Functional changes are in PRs #8, #9, and #10.