Expand description
Anthropic extended-thinking configuration.
Setting crate::ChatRequest::thinking to a Some(...) value
enables Anthropic’s extended-thinking mode — the model produces a
private chain-of-thought (surfaced as
crate::ChatStreamEvent::ThinkingToken on the streaming path)
before the user-facing reply. Available only on the
Anthropic-direct path; silently ignored on other providers.
Enums§
- Thinking
Mode - How much the model is allowed to “think” before replying.