aioprometheus

23.12.0 verified Mon Apr 27 auth: no python

A Prometheus Python client library for asyncio-based applications. Version 23.12.0. Supports creating metrics (Counter, Gauge, Histogram, Summary) and exposing them via an async HTTP server or pushing to Pushgateway. Compatible with Python 3.8+. Released regularly on PyPI.

pip install aioprometheus

Common errors

error ModuleNotFoundError: No module named 'aioprometheus.metrics' ↓

cause Importing from old submodules removed in version 21.

fix

Use from aioprometheus import Counter instead.

error ValueError: The 'labels' argument must be a dict ↓

cause Passing labels as keyword arguments or non-dict.

fix

Pass labels as a dictionary: counter.inc({"endpoint": "/"}).

Warnings

breaking Version 21.x removed the old submodule import paths. Use top-level imports: `from aioprometheus import Counter`, not `from aioprometheus.metrics import Counter`. ↓

fix Update all imports to `from aioprometheus import ...`.

deprecated The `render` function signature changed in 22.x. Previously `render(metrics, accept_header)` now `render(metrics, accept=None)`. Defaults to text format. ↓

fix Pass `accept` as keyword argument if needed.

gotcha Metrics must be registered with a collector registry before they are exposed. Not doing so results in empty /metrics output. ↓

fix Always add the metric to the service's collector: `service.prometheus_metrics.add_collector(my_metric)`.

Imports

Counter
```
from aioprometheus import Counter
```
Gauge
```
from aioprometheus import Gauge
```
Histogram
```
from aioprometheus import Histogram
```
Summary
```
from aioprometheus import Summary
```

MetricsMiddleware

wrong

from aioprometheus.middleware import MetricsMiddleware

correct

from aioprometheus import MetricsMiddleware

Old import path removed in 21.x

Service
```
from aioprometheus import Service
```

render

wrong

from aioprometheus.renderer import render

correct

from aioprometheus import render

Renderer was refactored; use top-level render

Quickstart

Create a Counter, register it with a Service, and expose metrics on port 8080.

from aioprometheus import Counter, Service
import asyncio

async def main():
    counter = Counter("requests_total", "Total requests")
    counter.inc({"endpoint": "/"})
    service = Service()
    service.prometheus_metrics.add_collector(counter)
    await service.start(addr="0.0.0.0", port=8080)
    try:
        await asyncio.Event().wait()
    finally:
        await service.stop()

asyncio.run(main())