Repository navigation
WTM 0.4.3
WTM 0.4.3 · build 14 · macOS 15+ · Apple Silicon
Fixed
- Enable and validate Homebrew-installed llama.cpp without rejecting the standard
admin-group-writable Cellar directory. - Validate symlinked GGUF weights in Hugging Face caches and revalidate the approved
executable symlink before launching its canonical target. - Preserve relative shared-library lookup when staging llama.cpp for execution.
- Explain how to start Ollama when its local API is unavailable for model cleanup.
Changed
- Clarified inventory terminology and refreshed release documentation and screenshot tooling.
- Added regression coverage for llama.cpp executable approval, GGUF symlinks, and real
inference, plus an opt-in real Ollama deletion test.
Includes a signed and notarized DMG, SHA-256 checksums, SPDX SBOM, build metadata, and GitHub artifact attestations.