Support

Support

Astronomical is actively developed and used daily. If something does not work, we want to hear about it — most issues are answered or fixed quickly.

Before you report

• System requirements. Astronomical needs an Apple silicon Mac (M1 or later) running macOS 14 (Sonoma) or later. Support across that spectrum is declared best-effort: hardware features the GPU cannot prove fall back to public MLX paths with no loss of declared model precision. Memory needs depend on the model you load; the app shows a recommended ceiling before a model is admitted.
• Check your version. The native menu shows the app version and build number. Mention both when reporting.
• Latest Stable release. The current signed Mac release is available on GitHub Releases — download it or update through the in-app Sparkle updater.
• Try the obvious fix. A stuck model download recovers by retrying the download; a wedged server responds to "Restart server" in the popover.
• Ask the command line. astronomical status reports the worker state, the resident and default models, and any active download. astronomical doctor shows which instance this command line is bound to and whether each channel's daemon answers, the first thing to check when a client cannot connect.

What's New

Release notes are derived from the signed public update feed the app itself checks, so this list stays current without manual edits.

What to include in a report

1. The app version and build number from the native menu.
2. Your Mac model and installed memory.
3. What you did, what you expected, and what happened.
4. For serving problems: the response you received and the relevant lines from the app's log file, stored in the app's data location under logs/.
5. For coding-agent problems: which agent and version, the astronomical launch command you ran, and the output of astronomical doctor.
6. For crashes on Mac App Store builds: whether you have diagnostics sharing enabled, and the crash report time if Apple showed you one.

Please do not paste model outputs, documents, or images you meant to keep private — a description of the behavior is enough, and logs can be reviewed locally before you attach anything.

Common questions

• A model will not download. Model files come from public model repositories. Check your connection, then retry the download from the model library. Corporate and school networks sometimes block these repositories.
• The app says there is not enough memory. The memory ceiling is a safety boundary, not a bug. Pick a smaller quantization of the same model, or free wired memory and retry. The recommendation in the native menu accounts for your installed memory.
• How do I use a coding agent with my local models? Run astronomical launch claude, astronomical launch codex, or astronomical launch opencode. The launcher finds the running app, picks a chat model that speaks the agent's protocol (or takes --model), and starts the agent for one session without changing its saved settings. If it answers "Start Astronomical first.", the app for this command line's channel is not running: astronomical doctor shows which channel that is.
• How do I use the local API? The app serves an OpenAI-compatible interface and an Anthropic-compatible Messages interface on this Mac's loopback address. The menu popover shows the port of the installed channel — 6732 for Stable and 6733 for Development — and Observatory's Connect page shows the full base URL. Point a local client at it; the local API page carries the model identifier, the credential field, and configuration ready to paste into a coding agent. Requests never leave this machine.
• Where are my models stored? In the app's designated application-support location. App Store builds keep everything inside their sandbox container; you can remove all of it by deleting the app's data.

Contact

The fastest route is a GitHub issue — it reaches the developer directly and keeps the answer searchable for everyone.

Astronomical is open source under its repository license. Privacy policy: privacy.html.