OptionalchoiceNumber of response choices requested from the model.
OptionaldataIdentifier of the external data source used to ground the request.
OptionalfrequencyFrequency penalty applied during token sampling.
OptionalmaxMaximum number of tokens requested for generation.
OptionalmodelName of the requested model.
OptionaloutputRequested output type (for example, json).
OptionalpresencePresence penalty applied during token sampling.
OptionalseedSeed used to make sampling reproducible.
OptionalstopStop sequences supplied with the request.
OptionalsystemStructured system instructions provided to the model.
OptionaltemperatureSampling temperature for the request.
OptionaltopNucleus-sampling top-p value for the request.
Request-side GenAI parameters captured for agent invocation telemetry.