Whether to use the Responses (Agent) API instead of the Chat Completions API.
If not specified then will be inferred based on invocation params. Specifically,
requests will be routed to the Responses API when the payload includes a built-in
tool (any tools[*] whose type is not "function") or any of the
Responses-only fields: previous_response_id, instructions, input, include.
Set explicitly to True to always use the Responses API, or False to always
use Chat Completions.
The Perplexity Agent API does not accept Chat-Completions-only knobs. When routing through Responses (whether explicitly or by inference):
temperature, top_p, top_k, stop, and metadata are dropped
at the boundary with a WARNING log so the behavior change is
discoverable. The class default temperature is dropped silently
(it would otherwise spam every call), but a user-supplied
temperature (init, invoke(temperature=...), or .bind) still
warns.tool_choice raises ValueError rather than being dropped, since
downstream agent loops cannot recover from a silently-disabled
forced tool call.preset causes model to be dropped because the Agent
API rejects bare Chat-Completions model names when model is
provided. If model was explicitly set by the user, a WARNING is
logged so the override is discoverable.Use use_responses_api=False if you need any of these parameters to
take effect.