edgible ai

Higher-level helpers for running a local LLM (Ollama) and exposing it as an Edgible application. Convenient for getting a private API endpoint to a model you’re hosting yourself.

edgible ai setup

End-to-end setup: install Ollama, pull a model, optionally install Open WebUI, optionally expose either as Edgible applications.

edgible ai setup --model llama3 --auto-install
edgible ai setup --model llama3 --expose-ollama --setup-webui

Flag	Description
--model <name>	Model to pull (e.g. llama3, mistral).
--auto-install	Install Ollama if missing.
--local-only	Don’t create Edgible applications.
--expose-ollama	Publish the Ollama API as an application.
--setup-webui	Install Open WebUI on the device.
--device-id <id>	Device to place on (defaults to local).
--ollama-device-id <id>	Override the placement of just the Ollama workload.
--webui-device-id <id>	Override the placement of just the WebUI workload.
--gateway-ids <ids>	Pin to specific gateways.
--webui-deployment <type>	How WebUI runs (docker or managed-process).
--non-interactive	Fail rather than prompt.

edgible ai stop

Stop Ollama and any UI you started with setup.

edgible ai stop

edgible ai serve

Start Open WebUI locally pointing at a (possibly different) Ollama URL.

edgible ai serve --port 3200 --ollama-url http://localhost:11434

Flag	Description
--port <n>	Port WebUI listens on. Default 3200.
--ollama-url <url>	Where to reach Ollama.
-d, --detached	Run in the background. Default true.

edgible ai status

Show whether Ollama and WebUI are running.

edgible ai status

edgible ai test

Run a one-shot prompt against Ollama and print the response.

edgible ai test --model llama3

edgible ai teardown

Stop services and remove Edgible applications.

edgible ai teardown --stop-ollama --remove-volumes

Flag	Description
--stop-ollama	Also stop Ollama on the device.
--remove-volumes	Also delete model storage volumes.