Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype to production without rebuilding the stack.
Python agentagentic-aiagentic-frameworkagentic-workflowaiautomationclaudedeepseekgenaigptllmlow-codemcpnextjsno-codeopenaipythonskillsworkflow
"""Keep delayed execution history consistent with durable workflow cancellation."""
from sqlalchemy import select
from sqlalchemy.orm import Session
from graphon.entities import WorkflowNodeExecution
from graphon.enums import WorkflowExecutionStatus, WorkflowNodeExecutionStatus
from libs.datetime_utils import ensure_naive_utc
from models.workflow import WorkflowNodeExecutionModel, WorkflowRun
def get_stopped_workflow_run(
session: Session, *, tenant_id: str, workflow_run_id: str | None, lock: bool = False
) -> WorkflowRun | None:
if workflow_run_id is None:
return None
statement = select(WorkflowRun).where(WorkflowRun.tenant_id == tenant_id, WorkflowRun.id == workflow_run_id)
if lock:
statement = statement.with_for_update()
run = session.scalar(statement)
if run is not None and run.status == WorkflowExecutionStatus.STOPPED and run.stop_requested_at is not None:
return run
return None
def apply_workflow_stop_to_node(
execution: WorkflowNodeExecution | WorkflowNodeExecutionModel, stopped_run: WorkflowRun | None
) -> None:
if stopped_run is None or execution.status not in (
WorkflowNodeExecutionStatus.RUNNING,
WorkflowNodeExecutionStatus.PAUSED,
WorkflowNodeExecutionStatus.RETRY,
):
return
execution.status = WorkflowNodeExecutionStatus.FAILED
execution.error = stopped_run.error or "User requested stop"
execution.finished_at = stopped_run.finished_at
if stopped_run.finished_at is not None:
execution.elapsed_time = max(
(stopped_run.finished_at - ensure_naive_utc(execution.created_at)).total_seconds(), 0.0
)