model

name	model
description	Inspect or set OmA model-selection strategy (balanced, auto, custom) for default task routing.

Run OmA model-strategy manager.

Input: $ARGUMENTS

Protocol:

1. Parse requested strategy and optional lane overrides.
   • examples:
     • auto
     • balanced
     • custom planning=gemini-3.1-pro-preview execution=gemini-3-flash-preview quick=gemini-3.1-flash-lite-preview review_verify=gemini-3.1-pro-preview
2. Supported strategies:
   • balanced (default): role-aware lane split (gemini-3.1-pro-preview for planning/review, gemini-3-flash-preview for execution, gemini-3.1-flash-lite-preview for quick low-risk edits)
   • auto: defer lane model selection to Gemini runtime auto-model policy across all lanes
   • custom: operator-provided lane map with concrete model names or Gemini CLI aliases
3. If no strategy is provided, show current policy from .omg/state/model.json when available.
4. Produce an effective lane map for:
   • planning
   • execution
   • quick_edit
   • review_verify
5. If filesystem tools are available, write/update .omg/state/model.json.
6. Clarify boundary:
   • this command sets OmA orchestration defaults
   • runtime model forcing still depends on Gemini CLI runtime controls (for example /model or launch-time --model when supported)

Suggested model.json shape: { "strategy": "balanced|auto|custom", "lane_models": { "planning": "gemini-3.1-pro-preview|auto|", "execution": "gemini-3-flash-preview|auto|", "quick_edit": "gemini-3.1-flash-lite-preview|auto|", "review_verify": "gemini-3.1-pro-preview|auto|" }, "updated_at": "", "source": "oma:model" }

Response:

• Keep it concise and operator-facing.
• Include Model Strategy, Effective Allocation, Runtime Notes, and Next Command.