Reasoning: effort to model suffix

What it does

When the request carries a supported reasoning effort (none, minimum, low, medium, high, xhigh, max), this transform appends the first matching rule's suffix to the upstream model name. {effort} in the suffix is replaced by the effort value.

Requests without a supported effort pass through unchanged.

Reference

When to use

• An upstream encodes reasoning strength in the model name, for example model-high.

Configuration

Example rule

{
  "transform": "reasoning_effort_to_model_suffix",
  "enabled": true,
  "phase": "request",
  "models": [
    "*"
  ],
  "config": {
    "rules": [
      {
        "pattern": "grok-*",
        "suffix": "-{effort}"
      }
    ]
  }
}

This transform is Provider-scope only, because it changes upstream model selection.