Troubleshooting
MLX is not installed
Run Bootstrap in Setup, or Install MLX for Mac on the same page. aituner installs MLX into its own private Python environment; it needs Python 3 from Homebrew.
A model will not start
Setup shows the model server's log. The most common cause is memory: another app or model is using the GPU memory the model needs. Quit other models (including Ollama, from its entry in the sidebar), or choose a smaller model; Downloads shows what fits.
Replies are slow
Check Monitor: if the GPU is not busy, the prompt may be very long. With Claude Code, use the aituner-claude launcher, which sends about 1,800 tokens per request instead of about 27,700. Measure speeds in Downloads, On this Mac, and prefer the recommended model.
An editor cannot connect
A model must be running: editors talk to aituner's gateway, which is up only while a model is served. Check the base URL and key on the Setup page, and set up the tool again if you changed models.
A download stopped
Open Downloads, Queue: failed items can be retried and resume where they stopped. Check the free space shown for your models folder; a network share must be mounted.
An update did not install
aituner refuses any download that fails its checks, and says why. Check your connection and try Check for Updates again, or download the release from the download page. Homebrew installs update with brew upgrade --cask aituner.
Still stuck?
aituner's log is at ~/Library/Application Support/aituner/aituner.log. Open an issue on GitHub with what you tried and the relevant lines, or write to [email protected].