| Name | Type | Description |
|---|---|---|
user_input* | str | The user's input message |
agent* | Any | The LangGraph agent to execute |
assistant_id* | str | None | The agent identifier |
session_state* | Any | Session state with auto_approve flag |
adapter* | TextualUIAdapter | |
backend | Any | Default: None |
image_tracker | MediaTracker | None | Default: None |
context | CLIContext | None | Default: None |
sandbox_type | str | None | Default: None |
message_kwargs | dict[str, Any] | None | Default: None |
rubric | str | None | Default: None |
blocked_goal_retry_context | str | None | Default: None |
turn_stats | SessionStats | None | Default: None |
Execute a task with output directed to Textual UI.
This is the Textual-compatible version of execute_task() that uses the TextualUIAdapter for all UI operations.
The TextualUIAdapter for UI operations
Optional backend for file operations
Optional tracker for images
Optional CLIContext with model override and params. The
current approval mode (session_state.auto_approve) is written
into context["auto_approve"] on every stream iteration before it
is passed to the graph via context=, so the interrupt_on when
predicate can suppress interrupts at the source.
Sandbox provider name for trace metadata, or None
if no sandbox is active.
Extra fields merged into the stream input message
dict (e.g., additional_kwargs for persisting skill metadata
in the checkpoint).
Acceptance criteria supplied to RubricMiddleware via graph
input state.
One-turn model context for retrying a previously blocked goal. This is carried via runtime context so it is not parsed for file mentions or checkpointed as human input.
Pre-created SessionStats to accumulate into.
When the caller holds a reference to the same object, stats are available even if this coroutine is cancelled before it can return.
If None, a new instance is created internally.
