Skip to main content

​
Parameters

ParameterTypeDefaultDescription
_agentAnyNoneReference to the agent instance for streaming operations
_taskOptional[Task]NoneThe task being executed
_accumulated_textstr""Text content accumulated during streaming
_streaming_eventsList[ModelResponseStreamEvent][]All streaming events received during execution
_final_outputOptional[OutputDataT]NoneThe final output after streaming completes
_all_messagesList[ModelMessage][]Internal storage for all messages during streaming
_run_boundariesList[int][0]Indices marking where each run starts in the message list
_is_completeboolFalseWhether the streaming operation has completed
_context_enteredboolFalseWhether we’re currently inside the async context manager
_start_timeOptional[float]NoneTimestamp when streaming started
_end_timeOptional[float]NoneTimestamp when streaming completed
_first_token_timeOptional[float]NoneTimestamp when first token was received
_modelAnyNoneModel override for streaming
_debugboolFalseDebug mode for streaming
_retryint1Number of retries for streaming

​
Functions

​
stream_output

Stream text content from the agent response. Returns:
  • AsyncIterator[str]: Incremental text content as it becomes available
Example: 
async with agent.stream(task) as result:
    async for text_chunk in result.stream_output():
        print(text_chunk, end='', flush=True)
    print()  # New line after streaming

​
stream_events

Stream raw events from the agent response. This provides access to the full stream events including tool calls, thinking events, and other non-text content. Returns:
  • AsyncIterator[ModelResponseStreamEvent]: Raw streaming events
Example: 
async with agent.run_stream(task) as result:
    async for event in result.stream_events():
        if isinstance(event, PartStartEvent):
            print(f"New part: {type(event.part).__name__}")
        elif isinstance(event, FinalResultEvent):
            print("Final result available")

​
get_final_output

Get the final accumulated output after streaming completes. Returns:
  • Optional[OutputDataT]: The final output, or None if streaming hasn’t completed

​
is_complete

Check if streaming has completed. Returns:
  • bool: Whether streaming has completed

​
get_accumulated_text

Get all text accumulated so far. Returns:
  • str: All text accumulated during streaming

​
get_streaming_events

Get all streaming events received so far. Returns:
  • List[ModelResponseStreamEvent]: Copy of all streaming events

​
get_text_events

Get only text-related streaming events. Returns:
  • List[ModelResponseStreamEvent]: Only text-related streaming events

​
get_tool_events

Get only tool-related streaming events. Returns:
  • List[ModelResponseStreamEvent]: Only tool-related streaming events

​
get_streaming_stats

Get detailed statistics about the streaming session. Returns:
  • Dict[str, Any]: Statistics including:
    • total_events: Total number of streaming events
    • text_events: Number of text events
    • tool_events: Number of tool events
    • accumulated_chars: Number of accumulated characters
    • is_complete: Whether streaming is complete
    • has_final_output: Whether final output is available
    • event_types: Breakdown of event types

​
get_performance_metrics

Get detailed performance metrics for the streaming session. Returns:
  • Dict[str, Optional[float]]: Performance metrics including:
    • start_time: When streaming started
    • end_time: When streaming ended
    • first_token_time: When first token was received
    • total_duration: Total streaming duration
    • time_to_first_token: Time to receive first token
    • tokens_per_second: Estimated tokens per second
    • characters_per_second: Characters per second

​
all_messages

Get all messages from the streaming session. Returns:
  • List[ModelMessage]: List of all ModelMessage objects from the streaming session

​
new_messages

Get messages from the last run only. Returns:
  • List[ModelMessage]: List of all ModelMessage objects from the most recent run

​
add_messages

Add messages to the internal message store. Parameters:
  • messages (List[ModelMessage]): List of ModelMessage objects to add

​
add_message

Add a single message to the internal message store. Parameters:
  • message (ModelMessage): A ModelMessage object to add

​
start_new_run

Mark the start of a new run in the message history. This should be called before adding messages from a new streaming run to properly track run boundaries for the new_messages() method.

​
__str__

String representation returns the accumulated text. Returns:
  • str: String representation of accumulated text

​
__repr__

Detailed representation for debugging. Returns:
  • str: Detailed representation including status, character count, and event count

​
Features

  • Async Context Manager: Full async context manager support for streaming operations
  • Real-time Streaming: Stream text content and events as they become available
  • Event Tracking: Comprehensive tracking of all streaming events
  • Performance Metrics: Detailed timing and performance statistics
  • Message Management: Full message history tracking during streaming
  • Tool Integration: Support for tool calls and complex event sequences
  • Error Handling: Robust error handling with completion state management
  • Debug Support: Extensive debugging and monitoring capabilities
  • Generic Support: Generic output data type support
  • State Management: Complete state tracking for streaming operations
⌘I