Instrument Python Apps with Azure Monitor

This skill covers azure-monitor-opentelemetry, a Python distro that wires up Application Insights with a single configure_azure_monitor() call, reading the connection string from APPLICATIONINSIGHTS_CONNECTION_STRING by default or accepting it explicitly. It auto-instruments eleven common libraries for tracing - Flask, Django, FastAPI, Requests, urllib3, httpx, aiohttp, psycopg2, pymysql, pymongo, and redis - so calling configure_azure_monitor() before your app code picks up traces from any of them without manual instrumentation. The instrumentations parameter can narrow this down to a specific subset, for example only Flask and Requests, when auto-instrumentation should be limited to the libraries a given service actually uses. On top of auto-instrumentation, the skill covers adding custom traces via trace.get_tracer(__name__).start_as_current_span(...), custom metrics via metrics.get_meter(__name__).create_counter(...), and custom logs through the standard logging module, all of which flow into Application Insights once configured.

Use it when standing up Application Insights telemetry for a Python service (Flask, Django, or FastAPI) with minimal setup code, or when tuning an existing setup's sampling rate, cloud role name, enabled instrumentations, or live metrics. Live Metrics (enabled via enable_live_metrics=True) gives a near-real-time view of requests and dependencies, which is a separate concern from the batched telemetry the rest of the distro exports. It's a one-line auto-instrumentation distro, not a replacement for the lower-level OpenTelemetry SDK when you need fine-grained control the configuration options don't expose.

configure_azure_monitor() takes six documented parameters: connection_string (defaults to the env var), credential (an Azure credential such as DefaultAzureCredential for AAD auth, default None), sampling_ratio (0.0-1.0, default 1.0), resource (an OpenTelemetry Resource, auto-detected by default), instrumentations (a list to restrict which libraries are enabled, default all), and enable_live_metrics (default False). A cloud role name for Application Map is set via a Resource:

configure_azure_monitor(
    resource=Resource.create({SERVICE_NAME: "my-service-name"})
)

There's no return value from configure_azure_monitor() itself - it mutates global OpenTelemetry state, after which trace/metrics/logging calls anywhere in the app are automatically exported to Application Insights.

Builds on the OpenTelemetry Python SDK and Azure Identity's DefaultAzureCredential for AAD auth, auto-instruments the eleven libraries listed above, and is distributed as the azure-monitor-opentelemetry PyPI package.

Python developers adding Application Insights telemetry to a Flask, Django, or FastAPI service who want auto-instrumentation over hand-wiring OpenTelemetry, and who follow the SDK's guidance: call configure_azure_monitor() before importing instrumented libraries, use environment variables for the connection string in production, set a cloud role name for multi-service apps, enable sampling under high traffic, use structured logging, add custom span attributes for debugging, and prefer AAD authentication for production workloads.