NAME

MODEL — backend-agnostic model management and inline switching

SYNOPSIS

/model [list|use NAME|show NAME|status]

@NAME [prompt...]

DESCRIPTION

The /model command manages models across all backends (llamafile and Ollama) using a single consistent interface. Use it when you don’t want to remember which backend is active.

The @NAME prefix switches the active model inline, within the current prompt. History is preserved — the switch is like a new character entering the scene.

SUBCOMMANDS

/model /model show NAME Print the currently active model and backend. If NAME is given, show details for that model.

/model list List all registered models across llamafile and Ollama, marking the active entry with an arrow.

/model use NAME Switch to the named model. Harvey checks llamafile models first, then Ollama models. Equivalent to /llamafile use or /ollama use depending on where NAME is registered.

/model status Show whether the active backend is reachable.

AT-MENTION SWITCHING

Prefix any prompt with @NAME to switch to that model for this turn and all subsequent turns:

@phi-mini summarise this in under 100 words

If NAME is not recognised, the @ prefix is treated as part of the prompt and no switch occurs. History is never cleared on a switch — the new model sees the full conversation so far.

Switch notes are written to the session file: [[model switch: phi-mini (llamafile) at 2026-06-20 14:32:11]]

MODEL ALIASES

/model alias list List all defined short-name aliases.

/model alias add ALIAS FULL_NAME Define an alias (also accepts: set).

/model alias remove ALIAS Remove an alias (also accepts: rm, delete).