dorm.contrib.prometheus

Stdlib-only Prometheus text-exposition exporter — no prometheus_client dependency. Connects to post_query to emit:

Metric	Type	Labels
dorm_queries_total	counter	vendor, outcome
dorm_query_duration_seconds	histogram	vendor
dorm_pool_size	gauge	alias
dorm_pool_in_use	gauge	alias
dorm_cache_hits_total	counter	alias
dorm_cache_misses_total	counter	alias

Quick start (FastAPI / any ASGI)

from fastapi import FastAPI
from fastapi.responses import PlainTextResponse
from dorm.contrib.prometheus import install, metrics_response

app = FastAPI()

@app.on_event("startup")
def startup():
    install()  # connect counters / histograms to dorm signals

@app.get("/metrics")
def metrics():
    return PlainTextResponse(
        metrics_response(),
        media_type="text/plain; version=0.0.4",
    )

API

dorm.contrib.prometheus.install() -> None

Connect the receiver. Idempotent.

Called explicitly so projects that don't expose metrics pay no overhead from the post-query timing fan-out.

dorm.contrib.prometheus.uninstall() -> None

Disconnect the receiver and reset every counter / histogram.

dorm.contrib.prometheus.metrics_response() -> str

Return the Prometheus text-exposition payload as str.

Wrap in your framework's Response helper. Output is a string; the canonical content type is text/plain; version=0.0.4.

dorm.contrib.prometheus.record_cache_hit(alias: str = 'default') -> None

Optional helper for cache backends — call from your custom BaseCache.get when the read landed a hit. RedisCache / LocMemCache don't call this automatically yet so apps that don't care pay zero overhead.

dorm.contrib.prometheus.record_cache_miss(alias: str = 'default') -> None

Histogram buckets

Fixed layout (1 ms → 5 s, doubling): 0.001, 0.002, 0.005, 0.01, 0.02, 0.05, 0.1, 0.25, 0.5, 1.0, 2.5, 5.0. Apps that need richer / configurable buckets should swap to prometheus_client and translate from the same dorm signals.