llm command
Manage the local LLM harness.
fobis llm <subcommand> [options]How to use it, with examples: Models, knowledge and agents.
All options
fobis llm ask
Ask the configured model a question (smoke test, scripting); the answer streams to standard output.
Arguments: PROMPT
| Option | Value | Default | Description |
|---|---|---|---|
--schema | TEXT | JSON Schema file: the answer is constrained to it. | |
--system | TEXT | System prompt. | |
--provider | TEXT | managed, ollama or openai. | |
--model, -m | TEXT | Model (alias or spec for managed). |
fobis llm doctor
Check the local LLM setup: platform, runtime, model, and (managed) a timed sample answer.
| Option | Value | Default | Description |
|---|---|---|---|
--sample, --no-sample | Start the model and time a short answer. |
fobis llm list
List the models in the store.
| Option | Value | Default | Description |
|---|---|---|---|
--json | Machine-readable output. |
fobis llm pull
Download a model into the store (resumable; verified against the Hub's sha256).
Arguments: MODEL
No options.
fobis llm recommended
List curated models (alias, size, licence): nothing is downloaded.
No options.
fobis llm rm
Remove a model from the store.
Arguments: MODEL
No options.
fobis llm runtime
Install, list and remove the llama.cpp runtime (llama-server).
fobis llm runtime install
Download, verify and install llama-server (the pinned build by default).
| Option | Value | Default | Description |
|---|---|---|---|
--variant | TEXT | auto | auto, cpu, metal, cuda12, cuda13, vulkan, rocm or sycl. |
--build | TEXT | A llama.cpp build tag other than the pinned one. | |
--sha256 | TEXT | Expected sha256 of a non-pinned build. | |
--trust | Accept a non-pinned build without a known sha256. |
fobis llm runtime list
List the installed llama.cpp builds.
No options.
fobis llm runtime remove
Remove an installed llama.cpp build.
Arguments: NAME
No options.
fobis llm serve
Keep a llama-server running between commands (start, stop, status).
fobis llm serve start
Start a persistent llama-server; later commands with the same model use it instead of starting their own.
| Option | Value | Default | Description |
|---|---|---|---|
--model, -m | TEXT | Model to serve (default: the configured one). |
fobis llm serve status
Show the persistent llama-server, if one is running.
| Option | Value | Default | Description |
|---|---|---|---|
--json | Machine-readable output (without the API key). |
fobis llm serve stop
Stop the persistent llama-server.
No options.