Demo: Prometheus plugin example

[savathoon]

Using this as a sanity check for our generic metrics plugin interface for compatibility with other metric reporting libraries

[Copilot]

Pull Request Overview

This PR introduces a demonstration Prometheus metrics reporter plugin to validate compatibility with the generic metrics plugin interface. The implementation provides a complete Prometheus integration with support for counters, gauges, histograms, and summaries.

Key changes include:

• Complete Prometheus metrics reporter plugin implementation with thread-safe batching and automatic label management
• Plugin packaging configuration with setuptools entry points and dependency management
• Comprehensive documentation covering installation, configuration, and usage examples

Reviewed Changes

Copilot reviewed 4 out of 5 changed files in this pull request and generated 4 comments.

File	Description
setup.py	Plugin packaging configuration with dependencies and entry points
requirements.txt	Runtime dependencies for pluggy and prometheus_client
metrics_reporter.py	Core Prometheus metrics reporter implementation with batching and label support
README.md	Complete documentation for installation, configuration, and usage

[Copilot]

MultiProcessValue() is being called as a function, but it should be assigned as a class. This should be prometheus_client.values.ValueClass = prometheus_client.values.MultiProcessValue (without parentheses).

Suggested change

	prometheus_client.values.ValueClass = prometheus_client.values.MultiProcessValue()
	prometheus_client.values.ValueClass = prometheus_client.values.MultiProcessValue

[Copilot]

There is trailing whitespace at the end of this line that should be removed.

Suggested change

	prometheus_client>=0.17.0
	prometheus_client>=0.17.0

[Copilot]

The labelnames parameter should be consistent across all instances of the same metric. If a metric is created with certain labelnames, all subsequent uses must have the same labelnames. Consider determining all possible labels upfront or using a different approach to handle dynamic labels.

Suggested change

	labelnames=list(labels.keys()) if labels else [],
	labelnames=self.metric_labelnames.setdefault(
	prometheus_name, list(labels.keys()) if labels else []
	),

[Copilot]

[nitpick] The hardcoded service name 'access' should be configurable or derived from environment variables to make the plugin more reusable across different services.

Suggested change

	merged_tags["service"] = "access"
	merged_tags["service"] = os.environ.get("SERVICE_NAME", "access")

[somethingnew2-0]

Prometheus metrics work by having an external process scrape an exposed http endpoint (eg. /metrics) that outputs the metrics. For this plugin to work, it would require the Access Flask app to expose some sort of metrics route for the Prometheus server to scrape.

The Prometheus client_python documentation sorta shows how this works in their example code with start_http_server https://prometheus.github.io/client_python/

[savathoon]

Prometheus metrics work by having an external process scrape an exposed http endpoint (eg. /metrics) that outputs the metrics. For this plugin to work, it would require the Access Flask app to expose some sort of metrics route for the Prometheus server to scrape.

The Prometheus client_python documentation sorta shows how this works in their example code with start_http_server https://prometheus.github.io/client_python/

🤔 we could set an env variable to configure that endpoint if prometheus. I mostly used this + the otel client lib to make sure the functions I was exposing via pluggy hook impl were generic enough to work for most metrics library options. Happy to reconfigure as just an endpoint that can be scraped by an external client.