Maximum time in milliseconds to wait for remaining tool results before flushing (default 15s).
Returns true if there are no outstanding tool calls.
Number of tool calls still awaiting results in the current batch.
Arms or re-arms the timeout with a custom duration and callback.
Clears all pending tool calls and cancels any active timeout (e.g. on user interruption, error, or session close).
Returns true when THIS result drains the batch and the caller should send response.create. If the call was not in the outstanding set (unknown or duplicate), returns false.
The tool call identifier.
Records an outstanding tool call ID that must be answered.
The tool call identifier.
OptionalonTimeoutFlush: () => voidOptional callback to invoke if the batch times out waiting for remaining results.
Alias for Closed.
Resets the barrier state and cancels any pending timeout (alias for Clear).
Alias for Opened.
OptionalonTimeoutFlush: () => void
Coordinates batched tool-call completions for realtime sessions that require a single turn-continuation trigger (e.g. OpenAI Live
response.create) per batch of parallel tool calls, rather than firing per tool result.Requirements (§4.2 / §4.3 of plans/realtime/gpt-live-1.md):
response.item.create).response.createmust be sent when the outstanding set drains.