NoClick Docs
Sign in

A health check for your automations

A morning check of every automation in your account, built in the app or from code - what isn't set up, which step broke in its last failed run and why - that pauses the ones failing over and over until someone fixes them.

pip install noclick
python 08_automations/automation_health_check.py
EnvironmentNOCLICK_API_KEY
View the code

Automations break quietly: a token expires, a channel is renamed, an API starts refusing a field, and a schedule keeps firing into the same error every hour. This example reads every workflow your account owns the way the app shows it, says what stands in each one's way, finds the step that broke in its last failed run, and pauses the ones that keep failing, so they stop spending until someone looks. It changes nothing unless you pass --pause.

pip install noclick
export NOCLICK_API_KEY=nk_live_...
python 08_automations/automation_health_check.py            # report only
python 08_automations/automation_health_check.py --pause    # also pause the ones failing over and over

Run it from a scheduler every morning, or hand the same calls to an agent: the account's own agent uses exactly these routes.

import sys

import noclick

RECENT = 5    # how many recent runs to judge a workflow by
FAILING = 3   # paused when at least this many of them failed

What stands in its way

One workflow reads in full: its steps (each with its settings, its account and, for a trigger, what keeps it from firing), its readiness and whether it's paused. readiness.issues are what the app's own check finds, each naming its step: one with no account connected, an account that stopped working, a trigger that couldn't register.

def report_setup(workflow: noclick.Workflow) -> None:
    for issue in workflow.readiness.get("issues", []):
        print(f"  needs: {issue['message']}")
    for step in workflow.steps:
        if step.connection and step.connection["healthy"] is False:
            print(f"  {step.name}: its account stopped working. {step.connection['hint']}")
        if step.trigger and step.trigger["error"]:
            print(f"  {step.name}: {step.trigger['error']}")

Why the last failed run failed

runs() lists recent runs however they started: a schedule, a webhook, the app, your code. get_run reads one with what each step did, so the broken step and its own error are right there.

def report_failures(workflow: noclick.Workflow) -> int:
    recent = workflow.runs(limit=RECENT)
    failed = [run for run in recent if run.status == "failed"]
    if failed:
        broken = next((s for s in workflow.get_run(failed[0].id).steps if s.status == "failed"), None)
        where = f"step '{broken.name}': {broken.error}" if broken else failed[0].error
        print(f"  {len(failed)} of its last {len(recent)} runs failed; the latest at {where}")
    return len(failed)

Pause the ones that keep failing

A pause switches the workflow's triggers off with the reason beside them in the app; resume() later switches back on only what the pause switched off. Runs you start from code still run.

def main() -> None:
    pause = "--pause" in sys.argv
    # Every workflow the account can open comes back, shared ones too; `access` says which are yours.
    owned = [w for w in noclick.Workflow.list() if w.access == "owner"]
    print(f"{len(owned)} workflows of your own\n")
    for listed in owned:
        workflow = noclick.Workflow.get(listed.id)  # a list is light: read each one in full
        print(f"{workflow.name}{'  (paused)' if workflow.paused else ''}")
        report_setup(workflow)
        failures = report_failures(workflow)
        if pause and failures >= FAILING and not workflow.paused:
            workflow.pause(f"{failures} of its last {RECENT} runs failed; paused by the morning check.")
            print("  paused it")
        print()


if __name__ == "__main__":
    main()

Clean up

Nothing here creates anything, and reading costs nothing. To switch a paused workflow back on once it's fixed: noclick.Workflow.from_name("Hourly digest").resume().