Skip to main content

Command Line Interface (CLI) Reference

Use the merry command line tool to manage, build, run, and inspect local models.

Global Commands​

merry serve​

Start the GoingMerry background server daemon.

  • Environment variables: Binds to MERRY_HOST (defaults to 127.0.0.1:11434).

merry run​

Retrieve a model configuration and open an interactive terminal prompt session.

  • Usage: merry run <model_name> [prompt]
  • Flags:
    • --format json: Format response outputs in strict JSON format.
    • --keepalive <duration>: Specify how long the model weights remain cached in memory (e.g. 10m, 2h, 0 for instant unload).
    • --system <prompt>: Override the system instruction prompt.
    • --verbose: Print benchmark statistics (load duration, tokens/s, prompt evaluations) after completions.

merry pull​

Retrieve model weights from the registry without launching execution.

  • Usage: merry pull <model_name>
  • Flags:
    • --insecure: Allow connecting to custom registries using unencrypted HTTP requests.

merry push​

Upload a customized local model tag configuration to your developer registry.

  • Usage: merry push <username>/<model_name>

merry list​

List all model configurations stored in your local cache.

  • Usage: merry list

merry ps​

Inspect loaded model configurations currently active in memory.

  • Usage: merry ps

merry create​

Compile a new model tag from a Modelfile.

  • Usage: merry create <model_name> -f <path_to_modelfile>
  • Flags:
    • -f, --file: Path to the Modelfile.

merry show​

Inspect the metadata parameters, system prompt, and templates for a local model.

  • Usage: merry show <model_name>
  • Flags:
    • --license: Print the model's licensing agreement.
    • --system: Print the model's system prompt instructions.
    • --template: Print the model's prompt templating structure.

merry cp​

Clones an existing model tag.

  • Usage: merry cp <source_model> <target_name>

merry rm​

Purge a model tag and its cached weights.

  • Usage: merry rm <model_name>