v0.47.0

Mesh-LLM/mesh-llmv0.47.0Mar 26, 2026by michaelneale

AI Summary

Expands Linux support and enhances hardware detection capabilities with automatic build scripts and improved RAM/VRAM reporting.

Key Highlights

Linux support for CUDA, CPU-only, and Jetson/Tegra builds
Auto-detecting `just build` scripts for different platforms
Improved RAM detection logic for CPU-only and GPU machines
CUDA architecture auto-detection via nvidia-smi

New Features

Linux build scripts and install instructions
RAM and VRAM offload capacity reporting
Tegrastats fallback for VRAM detection on Jetson
CI workflow with cargo test and CLI smoke tests

Full Release Notes

- Linux support: build from source with CUDA, CPU-only, or Jetson/Tegra
- Linux RAM detection: CPU-only machines detect system RAM, GPU machines report VRAM + RAM offload capacity
- Jetson/Tegra: tegrastats fallback for VRAM detection (nvidia-smi broken on Orin AGX)
- Linux build scripts: `just build` auto-detects platform (macOS→Metal, Linux→CUDA)
- CUDA arch auto-detection: nvidia-smi → deviceQuery → GPU model lookup
- CI: adds `cargo build --release` + CLI smoke test
- Docs: Linux install instructions on docs site and README
- Install: `mkdir -p ~/.local/bin` in curl command for fresh systems