generate method

@override Future<ChatResponse> generate(ChatRequest request, { AgenticContext? context, })

Generates a complete answer.

Throws an AgenticException subclass on failure — never a transport exception, and never a provider-specific type.

Implementation

@override
Future<ChatResponse> generate(
  ChatRequest request, {
  AgenticContext? context,
}) async {
  _requests.add(request);
  if (latency > Duration.zero) {
    await (context?.clock ?? const SystemClock()).delay(latency);
  }
  context?.throwIfCancelled();

  final turn = _nextTurn();
  if (turn.error case final error?) throw error;
  if (turn.response case final response?) {
    return response.copyWith(metadata: request.metadata);
  }
  // A chunk-scripted turn asked for by a non-streaming caller: assemble it,
  // so the same script serves both modes.
  final builder = ChatResponseBuilder(modelId: info.id)
    ..addAll(turn.chunks ?? const <ChatChunk>[]);
  return builder.build();
}