runAgentLoop function

Future<List<Message>> runAgentLoop({ required List<Message> prompts, required Context context, required AgentLoopConfig config, required StreamFunction streamFunction, required ToolExecutor toolExecutor, required AgentEventSink emit, CancelToken? cancelToken, CancelTokenSource? cancelTokenSource, })

Starts an agent loop with new prompt messages, delivering events to emit and resolving with the messages produced by the run.

This is the emit-based core beneath agentLoop (pi's runAgentLoop); use it when events must be awaited as part of the run (the stateful Agent does this for its listeners). Unlike agentLoop, exceptions from emit or a TransformContextHook propagate to the caller instead of being swallowed into a closed stream.

Implementation

Future<List<Message>> runAgentLoop({
  required List<Message> prompts,
  required Context context,
  required AgentLoopConfig config,
  required StreamFunction streamFunction,
  required ToolExecutor toolExecutor,
  required AgentEventSink emit,
  CancelToken? cancelToken,

  /// The writable side of [cancelToken] (issue #1923 AC1): the provider-
  /// phase steering interrupt cancels the run through it and re-arms the
  /// latch before the resubmit. Null keeps the interrupt inert (the
  /// pre-#1923 boundary-only steering semantics).
  CancelTokenSource? cancelTokenSource,
}) {
  return _runAgentLoop(
    prompts: prompts,
    context: context,
    config: config,
    emit: emit,
    streamFunction: streamFunction,
    toolExecutor: toolExecutor,
    cancelToken: cancelToken,
    cancelTokenSource: cancelTokenSource,
  );
}