larzmetrics

Application metrics with Prometheus export. Pure Python, zero dependencies.

Instrument your code and expose the numbers in the Prometheus text format every scraper and dashboard understands — without pulling in the official client (and its protobuf).

from larzmetrics import Registry

reg = Registry()
reqs = reg.counter("http_requests_total", "requests", labelnames=["method"])
reqs.labels(method="GET").inc()

reg.gauge("temperature", "temp").set(21.5)

lat = reg.histogram("request_seconds", "latency")
with lat.time():
    handle_request()

print(reg.render())          # Prometheus text exposition format

Why

• Counters, gauges, histograms — monotonic counters, settable gauges, and histograms with configurable buckets, _sum/_count, and a time() context manager for latencies.
• Labels — metric.labels(method="GET").inc(), exactly like the Prometheus client.
• Scrape-ready — render() produces the standard text exposition format (# HELP/# TYPE, _bucket{le=...}, +Inf), so any Prometheus/Grafana/ OpenMetrics tool reads it. Serve it at /metrics and you're done.
• Zero dependencies. No prometheus_client, no protobuf.

Install

pip install larzmetrics

Usage

reg = Registry()
c = reg.counter("name", "help", labelnames=["a", "b"])
c.labels(a="1", b="2").inc(5)

g = reg.gauge("temp", "help"); g.set(20); g.inc(); g.dec(2)

h = reg.histogram("lat", "help", buckets=[0.1, 0.5, 1, 2, 5])
h.observe(0.3)
with h.time():
    work()

reg.render()      # str, Prometheus format

Pairs with larzlog for observability and any WSGI app (serve render() at /metrics).

Tests

python -m unittest discover -s tests -v   # 10 tests

The Larz stack

One of 30+ pure-Python, zero-dependency libraries at github.com/larz-scripter.

License

MIT © larz-scripter