Repository navigation
ThreadShelf Local Suite 1.4.0
Reader, Analytics and Alternative are now 1.4.0.
- Reader and Alternative add speech playback with bundled offline Pocket TTS and Kokoro CPU voices, optional provider connections, streaming, per-message controls and an audio cache.
- TTS settings retain fixed navigation above scrollable content, including after cache deletion.
- The illustrated manual and application screenshots are refreshed; an updated offline manual ZIP is included.
- Existing MathJax 4.1.3 rendering and readable ChatGPT/Codex file citations are preserved.
- Analytics retains its read-only behaviour. Profiles, archive/schema identifiers and the Electron/Node/npm/ONNX pins are unchanged.
- Windows portable ZIPs contain the application EXE at their root. Linux TAR.GZ packages contain a root native binary; DEB packages install a real native executable in
/usr/bin.
Choose one of the nine native packages for your application and OS. The Reader Preview and Alternative Demo HTML files contain synthetic examples. Native packages include their runtime and do not require Node.js or npm.
All six Windows/Linux package jobs passed their archive, native runtime and GUI gates before publication. See ThreadShelf-Local-Suite-1.4.0-Verification.md and the matching evidence ZIP for exact checks and limits. Independent Checks workflow results are reported separately and are not inferred here. Verify downloads using the single SHA256SUM.txt.
The release includes the exact tagged suite source as ZIP and TAR.GZ. Unchanged dependency source companions retain their original filenames and checksums:
- Electron/ONNX source parts and metadata (v1.3.0), with the v1.3.0 checksum manifest.
- Native sharp/libvips sources (v1.0.0), with the v1.0.0 checksum manifest.
Keep every numbered source part with its matching metadata/assembler ZIP. Speech payloads include their own notices, voice attribution and corresponding GPL source archives. Windows packages remain unsigned. DEB payloads are validated and exercised after extraction; system-wide installation and every Linux distribution are not claimed tested. Remote speech services require a configured endpoint and any credentials required by that provider; their availability and output are not verified by the offline package gates.