- Downgrade LLamaSharp packages to 0.25.0 to match Vulkan backend availability
- Add LLamaSharp.Backend.Vulkan for AMD/Intel/NVIDIA GPU acceleration
- Fix _gpuLayers bug: was reading LlamaCppTimeout instead of a dedicated field
- Add GpuLayerCount to JournalConfig, sourced from JOURNAL_GPU_LAYERS env var
- Document AI/LLM notes in README (version pinning, known vulkaninfo issue)
Co-Authored-By: Oz <oz-agent@warp.dev>
638e5d0b8a