Prometheus Tornado Instrumentator

A Tornado-focused Prometheus instrumentation library inspired by trallnag/prometheus-fastapi-instrumentator.

The goal is to provide a Tornado-native user experience:

from prometheus_tornado_instrumentator import Instrumentator

Instrumentator().instrument(app).expose(app)

Installation

From PyPI:

pip install prometheus-tornado-instrumentator

From a local checkout:

uv pip install -e .

Quick Start

from tornado.web import Application, RequestHandler

from prometheus_tornado_instrumentator import Instrumentator


class HelloHandler(RequestHandler):
    def get(self):
        self.write({"message": "hello"})


app = Application([(r"/hello", HelloHandler)])
Instrumentator().instrument(app).expose(app)

This instruments registered Tornado handlers and exposes Prometheus scrape output at /metrics by default.

Features

• Default Prometheus metrics for request count, request size, response size, and latency.
• Custom sync or async instrumentation functions.
• Custom metric namespace and subsystem.
• Custom Prometheus CollectorRegistry.
• Gzip support for metrics output.
• Regex-based handler exclusions.
• Dynamic route instrumentation for handlers added with Application.add_handlers().
• Prometheus client multiprocess support through PROMETHEUS_MULTIPROC_DIR.

Scope

This package instruments Tornado HTTP RequestHandler routes.

It does not provide FastAPI, Starlette, ASGI mounted-app, included-router, or websocket route semantics. See docs/parity.md for the upstream semantic-parity notes.

Documentation

• Usage
• API reference
• Development
• Release process
• Upstream parity
• Documentation index

Development

Use uv for local development:

uv sync --extra dev
uv run --extra dev python -m pytest -q

License

This project is licensed under the ISC License. See LICENSE.

This project is inspired by trallnag/prometheus-fastapi-instrumentator, which is also licensed under the ISC License. See NOTICE for the initial upstream baseline.