Referencegemmapod CLI
gemmapod run
Register a pod with the Host (starting one if needed). Lets you override the Ollama model at runtime.
Synopsis
gemmapod run <dir-or-toml> [--model <name>] [--port <n>] [--once]Description
gemmapod run registers a pod with the running Host — and
starts the Host in the foreground if none is running. The pod
directory (or path to a pod.toml) is required.
By default the model comes from pod.toml. You can override it for
this run via --model, or interactively pick one from the locally
installed Ollama models when stdin is a terminal.
Options
| Flag | Effect |
|---|---|
--model <name> | Use this Ollama model instead of the one in pod.toml. Skips the interactive picker, so this is the right flag for scripts and CI. |
--port <n> | Host port to use when starting a new one. Default: try 57447..57456. Overrides GEMMAPOD_PORT. |
--once | Exit non-zero if signaling doesn't connect within 10s. Intended for CI smoke tests. |
Model selection precedence
--model <name>— explicit, no prompt- Interactive picker (TTY only, when Ollama is reachable and has ≥1 model installed). Pod's
pod.tomlmodel is preselected. pod.toml'smodel = "..."field- Fallback:
gemma4:e4b
The chosen model is logged as pod started — model: <name> so you
can confirm what's actually in use.
Examples
# Use whatever pod.toml says, prompt on TTY
gemmapod run ./examples/raj-card
# Pin a model explicitly (good for scripts)
gemmapod run ./examples/raj-card --model llama3:8b
# CI smoke: fail fast if signaling can't connect
gemmapod run ./examples/raj-card --model gemma4:e4b --onceSee also
gemmapod status— runtime health checkpod.tomlreference — where the default model lives