Prometheus Exporter¶

The Prometheus exporter exposes profiling events as Prometheus metrics: counters and histograms for HTTP requests, function calls, and database queries. Optionally, collector health metrics (queue depth, events dropped) can be registered for the same registry.

Installation¶

pip install profilis[prometheus]

Metrics¶

Metric	Type	Labels	Description
`profilis_http_requests_total`	Counter	service, instance, worker, route, status	Total HTTP requests
`profilis_http_request_duration_seconds`	Histogram	service, instance, worker, route, status	HTTP request duration
`profilis_function_calls_total`	Counter	service, instance, worker, function	Total function calls (decorated)
`profilis_function_duration_seconds`	Histogram	service, instance, worker, function	Function call duration
`profilis_db_queries_total`	Counter	service, instance, worker, db_vendor	Total DB queries
`profilis_db_query_duration_seconds`	Histogram	service, instance, worker, db_vendor	DB query duration
`profilis_events_dropped_total`	Counter	—	Events dropped (queue full; optional health)
`profilis_queue_depth`	Gauge	—	Current collector buffer size (optional health)

Default histogram buckets (seconds): [0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 2, 5, 10]. You can override them when creating the exporter.

Quick Start (ASGI / FastAPI)¶

from prometheus_client import CollectorRegistry
from profilis.exporters.prometheus import (
    PrometheusExporter,
    make_asgi_app,
    register_collector_health_metrics,
)
from profilis.core.async_collector import AsyncCollector

registry = CollectorRegistry()
prom = PrometheusExporter(registry, service="myapp")
collector = AsyncCollector(prom, queue_size=2048, batch_max=128)
register_collector_health_metrics(registry, collector)

# Wire collector into your app (e.g. instrument_fastapi with same collector)
# Then expose /metrics:
app.mount("/metrics", make_asgi_app(registry))

Quick Start (Flask)¶

from prometheus_client import CollectorRegistry
from profilis.exporters.prometheus import (
    PrometheusExporter,
    make_metrics_blueprint,
    register_collector_health_metrics,
)
from profilis.core.async_collector import AsyncCollector

registry = CollectorRegistry()
prom = PrometheusExporter(registry, service="myapp")
collector = AsyncCollector(prom, queue_size=2048, batch_max=128)
register_collector_health_metrics(registry, collector)

# Wire collector into ProfilisFlask(app, collector=collector, ...)
app.register_blueprint(make_metrics_blueprint(registry))  # GET /metrics
# Or with prefix: make_metrics_blueprint(registry, url_prefix="/custom")

Configuration¶

PrometheusExporter(registry, service="", instance="", worker="", buckets=DEFAULT_BUCKETS) Use a dedicated CollectorRegistry() so scrapes only see Profilis metrics. Set service, instance, worker for label values.
register_collector_health_metrics(registry, collector) Registers profilis_events_dropped_total and profilis_queue_depth for the given AsyncCollector. Call after creating the collector and use the same registry as the exporter.
make_asgi_app(registry) Returns an ASGI app that serves GET /metrics in Prometheus text format. Mount at /metrics (or another path).
make_metrics_blueprint(registry, url_prefix="") Returns a Flask blueprint that serves GET /metrics. Register with app.register_blueprint(...).

Composing with Other Exporters¶

You can use the Prometheus exporter alongside JSONL or console by using a sink that fans out to multiple exporters:

def sink(batch):
    jsonl_exporter(batch)  # or your JSONL writer
    prom(batch)

collector = AsyncCollector(sink, queue_size=2048, batch_max=128)