Feature request: opt-in to emit Anthropic output_config.effort for providers that require it

What feature would you like to see?

An opt-in to make the Anthropic provider emit the output_config.effort field on the wire.

Some Anthropic-compatible providers require output_config.effort in the request to control reasoning effort, but kimi-code never emits this field unless the model is recognized as supporting it. Users routing such providers through the Anthropic transport currently have no way to opt in.

Proposed shape: an effort_param boolean on the model configuration, plus a KIMI_MODEL_EFFORT_PARAM environment variable, mirroring how adaptive_thinking / KIMI_MODEL_ADAPTIVE_THINKING work today. Default off, so existing behavior is unchanged.

Additional information

I searched existing issues and PRs for "effortParam", "output_config", and "anthropic effort" and found no prior request for this.

I have a working implementation of this proposal (v1 and v2 engines, with tests and docs) that I am currently testing on my fork: https://github.com/kevingatera/kimi-code/pull/1 — happy to open it as a PR here once maintainers signal the approach is acceptable.

Start by reviewing the existing adaptive_thinking / KIMI_MODEL_ADAPTIVE_THINKING configuration path, then inspect the v1 and v2 engine implementation, tests, and docs referenced in the proposal. Done means an opt-in effort parameter emits Anthropic output_config.effort for compatible providers while the default remains unchanged, with coverage for the environment variable and configuration behavior.