CrunTeam Crun Model Router
Use ../../runtime/crun_cli.py and ../../catalog/models.json. Treat the catalog as the source of truth for routing labels and priority, and the authenticated Models endpoint as the source of truth for the selected model's current input schema.
Build routing intent
Normalize the request into one JSON object:
{
"modality": "image|video|audio",
"operation": "text-to-image|image-edit|text-to-video|image-to-video|reference-to-video|text-to-speech|music-generate|...",
"quality": "balanced|best",
"speed": "balanced|fast",
"native_audio": false,
"reference_media": []
}
Preserve explicit constraints. Infer the operation from supplied media only when unambiguous.
Route and inspect
Prefer a JSON file or stdin for portable invocation:
python <runtime>/crun_cli.py models route --intent-file <intent.json>
python <runtime>/crun_cli.py models describe --model <selected-model>
Use selected by default. Present alternatives only when their tradeoffs matter or the user asks. Read required_input_fields and input_schema, then ask only for required information that cannot be derived. Never invent fields or copy a sibling model's schema.
Use models list when the static candidate is rejected or missing. Never infer capability or priority from a model name or schema; update models.json explicitly. Report a hard-constraint conflict instead of weakening it silently.
After routing, hand off affordability checks to ../crun-account-credits/SKILL.md and all creation or monitoring work to ../crun-task-runner/SKILL.md.