RealtimeSession class

Configuration for a Realtime session.

The Realtime API enables real-time audio conversations with the model using WebSockets. Audio configuration is nested under audio (audio.input.* / audio.output.*), output modalities are configured via outputModalities, and the maximum output token cap is set on maxOutputTokens.

Note: the prompt field from the spec is intentionally not modelled in this PR — the spec helper schema Prompt has no Dart class yet. Tracked as a follow-up for prompt-template support.

Annotations
  • @immutable

Constructors

RealtimeSession({String? id, String? object, String? type, String? model, int? expiresAt, RealtimeAudioConfig? audio, List<String>? outputModalities, String? instructions, List<RealtimeTool>? tools, RealtimeToolChoice? toolChoice, InfOrInt? maxOutputTokens, bool? parallelToolCalls, RealtimeReasoning? reasoning, RealtimeTracingConfig? tracing, RealtimeTruncation? truncation, List<String>? include})
Creates a RealtimeSession.
const
RealtimeSession.fromJson(Map<String, dynamic> json)
Creates a RealtimeSession from JSON.
factory

Properties

audio → RealtimeAudioConfig?
Nested audio configuration.
final
expiresAt → int?
Expiration timestamp (Unix epoch seconds). May be null on partial frames before the session is fully provisioned.
final
hashCode → int
The hash code for this object.
no setteroverride
id → String?
The session identifier (sess_…). Optional on partial frames.
final
include → List<String>?
Additional fields to include in server outputs (e.g. 'item.input_audio_transcription.logprobs').
final
instructions → String?
System instructions.
final
maxOutputTokens → InfOrInt?
Maximum output tokens ('inf' or a specific integer).
final
model → String?
The model identifier. Always present for realtime sessions; omitted for transcription sessions (which don't carry a top-level model).
final
object → String?
The object type (e.g. 'realtime.session'). Optional on partial frames.
final
outputModalities → List<String>?
Output modalities (e.g. ['audio'] or ['text']). Defaults to ['audio']. The server only accepts a single modality at a time.
final
parallelToolCalls → bool?
Whether the model may call multiple tools in parallel. Only supported by reasoning Realtime models such as gpt-realtime-2.
final
reasoning → RealtimeReasoning?
Reasoning configuration. Only supported by reasoning models.
final
runtimeType → Type
A representation of the runtime type of the object.
no setterinherited
toolChoice → RealtimeToolChoice?
Tool choice setting.
final
tools → List<RealtimeTool>?
Tools available to the model.
final
tracing → RealtimeTracingConfig?
Tracing configuration.
final
truncation → RealtimeTruncation?
Truncation configuration.
final
type → String?
The session type ('realtime' or 'transcription'). Optional on partial frames.
final

Methods

copyWith({Object? id = unsetCopyWithValue, Object? object = unsetCopyWithValue, Object? type = unsetCopyWithValue, Object? model = unsetCopyWithValue, Object? expiresAt = unsetCopyWithValue, Object? audio = unsetCopyWithValue, Object? outputModalities = unsetCopyWithValue, Object? instructions = unsetCopyWithValue, Object? tools = unsetCopyWithValue, Object? toolChoice = unsetCopyWithValue, Object? maxOutputTokens = unsetCopyWithValue, Object? parallelToolCalls = unsetCopyWithValue, Object? reasoning = unsetCopyWithValue, Object? tracing = unsetCopyWithValue, Object? truncation = unsetCopyWithValue, Object? include = unsetCopyWithValue}) → RealtimeSession
Returns a copy of this RealtimeSession with the given fields replaced.
noSuchMethod(Invocation invocation) → dynamic
Invoked when a nonexistent method or property is accessed.
inherited
toJson() → Map<String, dynamic>
Converts to JSON.
toString() → String
A string representation of this object.
override

Operators

operator ==(Object other) → bool
The equality operator.
override