Status, logs and metrics

Project and host health

amicu doctor
amicu doctor --output json

doctor checks the host (OS, memory, free disk), whether llama-server is found, whether the project config parses, and whether the state directory is writable. JSON output has the shape { "version": "v1", "ok": bool, "checks": [...] }. A warn check does not fail the command; a fail check does.

Local deployments

Local deployments are user services.

• macOS: logs are ~/Library/Logs/amicu/<label>.out.log and <label>.err.log. Inspect the service with launchctl list and the plist in ~/Library/LaunchAgents/.
• Linux: the service is a transient systemd user unit. Use systemctl --user status <label> and journalctl --user --unit=<label>.

The label is io.amicu.llama.<deployment id>, for both llama.cpp and vLLM deployments. Both runtimes expose GET /health and GET /v1/models on the port you set. llama.cpp's /health is open by design even when an API key is set; /v1/models needs the key.

llama-server serves Prometheus metrics at /metrics because amicu adds --metrics.

Remote provider deployments

Status, logs, and metrics for a remote deployment come from the provider. Today amicu ships these commands for Modal only (amicu modal status, logs, metrics, stop); the full list, the metrics shown, and their limits are on the Modal page. Other providers will document theirs on their own page when they are wired into apply.