airflow.providers.anthropic.triggers.agent¶

Classes¶

AnthropicAgentSessionTrigger

Poll a Managed Agents session until it reaches a terminal status.

Module Contents¶

class airflow.providers.anthropic.triggers.agent.AnthropicAgentSessionTrigger(conn_id, session_id, poll_interval, end_time, expect_outcome=False, kickoff_event_id=None)[source]¶

Bases: airflow.triggers.base.BaseTrigger

Poll a Managed Agents session until it reaches a terminal status.

Parameters:

conn_id (str) – The Anthropic connection ID.
session_id (str) – The session to poll.
poll_interval (float) – Seconds to sleep between polls.
end_time (float) – Wall-clock deadline (time.time() epoch seconds). Wall-clock is used deliberately: the trigger is serialized and may resume in a different triggerer process, where a per-process time.monotonic() value is meaningless.
expect_outcome (bool) – Whether the session is running a user.define_outcome loop (completion is then judged from outcome_evaluations, not from idle alone).
kickoff_event_id (str | None) – ID of the kickoff event, used to correlate the terminal idle event on a message run (defeats the start race).

conn_id[source]¶

session_id[source]¶

poll_interval[source]¶

end_time[source]¶

expect_outcome = False[source]¶

kickoff_event_id = None[source]¶

serialize()[source]¶

Serialize AnthropicAgentSessionTrigger arguments and class path.

async on_kill()[source]¶

Archive the session when a user kills the deferred task.

Runs in the triggerer event loop on Airflow 3.3+ (a no-op override on older versions, which never call it). Closes the gap the operator’s on_kill leaves for deferred tasks, which have released their worker slot.

async run()[source]¶

Poll the session and yield exactly one terminal event.