Repository navigation
Python: setup_observability does not support Application Insights QuickPulse aka Live Metrics #2649
Description
In the Python SDK, the setup_observability() path does not take advantage of the Azure Monitor OTel distro's configure_azure_monitor call to wire up Application Insights, instead relying on creating the exporters manually. As such, it doesn't have the enable_live_metrics functionality, and so Live Metric (QuickPulse) telemetry is not emitted which breaks important functionality. It'd be great to get all the Agent insights without losing Live Metrics.
I've not yet found a workaround for this, but am continuing to hack away at it.
Metadata
Metadata
Labels
Type
Projects
- Status
Activity
[-]setup_observability does not support Application Insights QuickPulse aka Live Metrics[/-][+]Python: setup_observability does not support Application Insights QuickPulse aka Live Metrics[/+]TaoChenOSU commentedon Dec 10, 2025
Hi Brandon H (@brandonh-msft),
Have you tried directly using
configure_azure_monitorinstead of relying onsetup_observability?brandonh-msft commentedon Dec 11, 2025
Yes, in fact we were already using that approach in our codebase and then added AF, but we were not seeing Agent/Tool/etc-level telemetry until we added the call to
setup_observability; these agent calls are what we're after in our telemetry.eavanvalkenburg commentedon Dec 11, 2025
I have opened a PR that makes that pattern the norm, including in samples. Since there is so much going on inside
configure_azure_monitorthat it isn't feasible to bring all that into our setup.TaoChenOSU commentedon Dec 11, 2025
I think there is a gap we have to fill, which is enabling instrumentation on the code without setting up the Monitoring backends. Left comments in the PR.
eavanvalkenburg commentedon Dec 15, 2025
Tao Chen (@TaoChenOSU) Brandon H (@brandonh-msft) the changes I made in #2782 make this explicit, we now have two different commands, one to enable the instrumentation, which would be used when you setup your otel providers in some other way (like with configure_azure_monitor), while the other configures providers for you, as well as enable instrumentation.