run

ownllm run

Boots Ollama, cloudflared, and the gateway as supervised subprocesses, then blocks the terminal. Logs from all three are merged, prefixed with their source, and tailed to stdout. Ctrl+C triggers a graceful shutdown.

This is the debug entry point — it's the same code path as the service-managed daemon, just attached to your terminal. For long-lived deployments, register a service instead so the agent restarts on boot.

When to use it

• You just paired and want to see the boot sequence end-to-end.
• A previous service install fails and you need raw logs.
• You're iterating on a config change and want a quick restart loop.

When not to use it

• Long-lived servers — use ownllm service install + start.
• CI: run blocks; use start + stop instead.

Output

[ownllm] paired tenant=acme-prod
[ollama] loading model llama-3.3-70b
[cloudflared] tunnel UP, hostname=acme-agent.ownllm.app
[gateway] listening on 127.0.0.1:11435
[ownllm] heartbeat ok

Restart and reload

While run is in the foreground, send SIGHUP to reload the runtime manifest from the site without restarting the process:

kill -HUP $(pgrep -f "ownllm run")

SIGINT (Ctrl+C) triggers a clean shutdown: drains the gateway request queue, signals Ollama to unload models, then closes the tunnel.