Flask Error Monitoring: Setup Guide

Same story as Django, different framework: Flask intercepts exceptions raised in a view so it can return an error response, which means the exception's journey ends inside Flask and never reaches the process-level exception hook. If your monitoring setup relies purely on sys.excepthook, a Flask app can 500 all day without a single event showing up.

Why doesn't Flask report exceptions the normal way?

Flask's dispatch machinery wraps every view call in its own error handling so it can render a proper HTTP error response instead of crashing the WSGI worker. Before it does, it sends the got_request_exception signal — powered by the blinker library — with the live exception still accessible. That signal is the integration point.

How do you capture Flask exceptions with Ravn?

import ravn
from ravn.integrations.flask import init_app
from flask import Flask

app = Flask(__name__)

ravn.init(api_key="your_api_key")
init_app(app)

# Every unhandled exception in any view is now captured,
# with the request path and method attached automatically.

Under the hood, it's a thin signal receiver:

def init_app(app) -> None:
    from flask import got_request_exception, request

    def _handle(sender, exception, **extra):
        capture_exception(exception, metadata={
            "mechanism": "flask",
            "path": request.path,
            "method": request.method,
        })

    got_request_exception.connect(_handle, app, weak=False)

Why does the handler need weak=False?

Flask's signals run on blinker, which by default holds only a weak reference to connected receivers. A local closure with no other strong reference gets garbage-collected almost immediately after the connecting function returns — the signal silently stops firing, with nothing in your logs to tell you it happened. It looks wired up. It isn't. Passing weak=False makes blinker hold a strong reference, so the handler survives for the life of the process.

This is the same failure mode as the Django integration, and it's worth checking for in any hand-rolled Flask signal handler — it's an easy thing to get wrong silently.

Do I need to install blinker separately?

No, if you're on a reasonably current Flask. Blinker has been a required dependency since Flask 2.3, so it's already installed alongside Flask itself in any modern project.

Does this work with blueprints?

got_request_exception is application-wide — it fires for unhandled exceptions regardless of whether the view is registered directly on the app or through a blueprint. No blueprint-specific configuration is needed.

For the framework-agnostic version of this setup, see the Python error monitoring guide, or the equivalent walkthroughs for Django and FastAPI.

Frequently asked questions