Here are all keyword arguments accepted by Prefect 3.x’s

@flow

decorator (keyword-only). Defaults shown as in the API:

@flow(
  name=None,
  description=None,
  version=None,
  flow_run_name=None,
  task_runner=None,
  timeout_seconds=None,
  validate_parameters=True,
  persist_result=None,
  result_storage=None,
  result_serializer=None,
  cache_policy=None,
  cache_expiration=None,
  cache_key_fn=None,
  on_completion=None,
  on_failure=None,
  on_cancellation=None,
  on_crashed=None,
  on_running=None,
  retries=None,
  retry_delay_seconds=None,
  retry_jitter_factor=None,
  log_prints=None,
)

Brief descriptions: - name: Optional custom name for the flow (defaults to function name) - description: Text description (defaults from function docstring if present) - version: Optional version string for this flow - flow_run_name: String or callable to name individual flow runs - task_runner: Task runner to execute tasks (e.g., thread/process-based) - timeout_seconds: Max runtime in seconds for the flow - validate_parameters: Whether to validate/coerce parameters (default True) - persist_result: Whether to persist the flow’s result (inherits setting if None) - result_storage: Storage location/block for persisted results - result_serializer: Serializer to use for results - cache_policy: Policy controlling flow-level result caching - cache_expiration: How long a cached result is valid - cache_key_fn: Function to compute a cache key from inputs - on_completion: List of hooks to run when the flow completes - on_failure: List of hooks to run when the flow fails (after retries) - on_cancellation: List of hooks to run when the flow is cancelled - on_crashed: List of hooks to run when the flow crashes - on_running: List of hooks to run when the flow enters Running - retries: Number of times to retry the flow on failure - retry_delay_seconds: Delay between retries (can be a number or sequence) - retry_jitter_factor: Adds jitter to retry delays - log_prints: Capture print() as logs (inherits setting if None) References: - API reference for flows: prefect.flows API - State change hooks: State change hooks - Results and persistence: Results and persistence - Retries: Retries

@Marvin is there a way to set a global “on failure” for every flow of my repo ?

Short answer: Yes—but not by flipping a single code switch in Prefect itself. You have two good options: 1) Prefect Cloud Automations (recommended for “apply to everything”) - Create an Automation that listens for the

prefect.flow-run.Failed

event and sends a notification or runs an action. You can scope it to all flows in the workspace or filter by things like deployment, work pool, etc. - No code changes to your flows are required. - Manage via the Cloud UI (create/enable/disable) or API. The CLI can list/inspect/enable/disable, but not create. - Docs: - Creating automations - Event triggers - State change hooks (concepts behind flow hooks) 2) Repo-level wrapper around @flow (code-based, works anywhere) - Define your failure handler once and wrap Prefect’s

@flow

to inject

on_failure

for every flow you define in your repo. Example:

# global_flow.py
from prefect import flow
from prefect.flows import Flow

async def notify_failure(flow: Flow, flow_run, state):
    # Your global handler logic here (send Slack, log, etc.)
    print(f"[GLOBAL FAILURE] {flow.name} -> {state.type}: {state.message}")

def global_flow(*, on_failure=None, **kwargs):
    # Ensure our global hook is always included (and preserve any flow-specific hooks)
    hooks = [notify_failure, *(on_failure or [])]
    return flow(on_failure=hooks, **kwargs)

# usage in your flows
from .global_flow import global_flow

@global_flow(name="my_flow")
def my_flow():
    ...

Notes: -

on_failure

hooks run only after all retries are exhausted and the flow enters a final Failed state. - If you’re on Prefect OSS (no Cloud), use the wrapper approach; if you’re on Cloud, Automations are the simplest “global” solution with no code changes. - CLI reference for Automations (manage-only):

prefect automation --help

If you tell me whether you’re on Prefect Cloud or OSS, and whether you want notifications (Slack/Email) or to run custom logic, I can give you a tailored setup.