Creates a new ResponseGenerator bound to the given model settings.
AI backend model ID and sampling parameters.
Generates a tutor response for the given context.
Session state, student transcript, and conversation history.
A TutorResponse ready to hand to TTS.
Streams a tutor response token-by-token, invoking onToken for each
partial output. Useful for low-latency TTS streaming pipelines.
Session state, student transcript, and conversation history.
Called with each text token as it arrives from the model.
The fully assembled TutorResponse once streaming completes.
Produces spoken tutor responses from student transcripts and session context.