Function: mapThinkingToEffort()

function mapThinkingToEffort(thinking, model): OpenAIEffort | undefined;

Defined in: providers/openai-models.ts:147

Our ThinkingConfig → the reasoning.effort we send. undefined means omit reasoning entirely and let the model use its own default; tools work at any effort on this endpoint.

⚠️ 'none' is NEVER returned, and that is deliberate rather than tidy. 'none' is the one value measured to be rejected (Astra, 400); 'low' is measured to work on all three GPT-6 models. Mapping disabled → 'low' means the string 'none' cannot leave this process, so the single per-model exception we know about can never fire and allowedEfforts() never has to be right for us to be correct. The cost is honest and small: a caller asking for reasoning off gets minimal reasoning instead.

Parameters

Parameter	Type
thinking	ThinkingConfig | undefined
model	string

Returns

OpenAIEffort | undefined