watchdock-errors

Python SDK for application-level error tracking on the Watchdock platform.

Installation

pip install watchdock-errors

With framework extras:

pip install "watchdock-errors[django]"
pip install "watchdock-errors[fastapi]"
pip install "watchdock-errors[celery]"

Quickstart

import watchdock_errors

watchdock_errors.init(
    api_key="wdk_xxx",
    environment="production",
    release="1.0.0",
)

Django

# settings.py
MIDDLEWARE = [
    ...
    "watchdock_errors.integrations.django.DjangoErrorMiddleware",
]

Or via INSTALLED_APPS for automatic registration:

INSTALLED_APPS = [
    ...
    "watchdock_errors.integrations.django",
]

FastAPI

from watchdock_errors.integrations.fastapi import setup_watchdock

setup_watchdock(app)

Celery

from celery import Celery
from watchdock_errors.integrations.celery import register

app = Celery("myproject")
app.config_from_object("django.conf:settings", namespace="CELERY")
app.autodiscover_tasks()

register(app)

Call register(app) once, after autodiscover_tasks(). It reports, automatically:

• Task registry — every registered task (and, for periodic tasks, their beat schedule), once per worker on startup.
• Failures, retries, and revocations — via task_failure, task_retry, and task_revoked. Successful runs are not reported.
• Heartbeats for periodic tasks — a lightweight ping on every run of a beat-scheduled task, so Watchdock can flag one that's gone silent. Non-periodic tasks are never pinged.

watchdock_errors.init() must be called before these signals fire — if you're on Django, a worker_process_init handler that calls init() covers every forked worker process.

Manual capture

# Capture the current exception
try:
    process_payment()
except Exception:
    watchdock_errors.capture_exception()

# Capture a specific exception
watchdock_errors.capture_exception(exc)

# Capture a message
watchdock_errors.capture_message("Stripe webhook signature invalid")

# Capture a message with a custom level
watchdock_errors.capture_message("Queue depth high", level="warning")

Event levels

Every event carries a level. Exceptions default to "error"; messages default to "info" unless you pass level explicitly to capture_message.

Correlating with nginx requests

If your app is behind nginx and you've added $request_id to your access log format (see the nginx log collection docs) and forwarded it to your app via proxy_set_header X-Request-Id $request_id;, this SDK automatically reads that header off every captured request and attaches it as trace_id — no code changes needed. This lets WatchDock link a failed request in your nginx access logs directly to the exception it produced.

You can also pass trace_id explicitly, which takes priority over the auto-extracted value:

watchdock_errors.capture_exception(exc, trace_id=my_trace_id)
watchdock_errors.capture_message("Queue depth high", level="warning", trace_id=my_trace_id)

SDK initialization

When init() is called, the SDK sends a one-time, fire-and-forget ping to the platform (with the SDK version and environment) to register that it started up. This never blocks application startup and any failure is silently ignored.

PII scrubbing

By default, Authorization, Cookie, Set-Cookie, and X-Api-Key headers are stripped, the request body is not sent, and sensitive-looking query-param values (names containing token, secret, password, auth, key, session, credential, otp, pin, ssn, etc.) are redacted in both query_params and the captured url itself. Set send_pii=True to disable scrubbing.

Use the before_send hook for custom scrubbing:

def scrub(event):
    event["request"]["headers"].pop("X-Internal-Token", None)
    return event  # return None to drop the event entirely

watchdock_errors.init(api_key="wdk_xxx", before_send=scrub)