Commands / kosa8 model serve
kosa8 model serve
Serve a model on the local gateway until interrupted
Usage
kosa8 serve NAME
Options
| Flag | Type | What it does | Default |
|---|---|---|---|
--backend | string | inference runtime: auto (llama-server if installed, else kosa8's own engine), llama, or engine | auto |
--sandbox | string | the sandbox whose agent this gateway serves (clients may also send X-Kosa8-Sandbox) | — |
--session | string | AI session id the audit records belong to (default: a new one) | — |