Arize AI Community Icon

Home
Events

Arize AX Releases
Arize AX Support
Arize News
Arize Observe
Discussions
Introductions
Phoenix Support

Powered by Tightknit

Phoenix Support

Request for Built-In Instrumentation in Arize Phoenix and MS Semantic Kernel | Arize AI Community

Phoenix Support

Request for Built-In Instrumentation in Arize Phoenix and MS Semantic Kernel

·Feb 13, 2025 10:45 AM

Feature request: Is it possible to add built-in instrumentation for Arize Phoenix and Microsoft Semantic Kernel? Also, keen if they are any existing samples using client side OTEL?

3 comments

· Sorted by Oldest

RunLLM
APP
·
Currently, there is no specific mention of built-in instrumentation for Microsoft Semantic Kernel in Arize Phoenix. However, Phoenix supports OpenTelemetry (OTEL) for tracing, which can be used to manually instrument applications, including those using Microsoft Semantic Kernel, by following the manual instrumentation guide.For client-side OTEL samples, you can refer to the Phoenix documentation which provides guidance on setting up tracing for JavaScript applications using OTEL. This includes setting up the OpenTelemetry SDK and using it to instrument your code.Would you like more detail on how to set up manual instrumentation or client-side OTEL? Sources: - manual instrumentation guide - Phoenix documentation
💡 Hint: Mention RunLLM in the thread for followups.
Anthony P.
·
Hey there Farzad S. thanks for reaching out! We are always accepting feedback on what instrumentations we build! Leave us a feature request here https://github.com/Arize-ai/openinference/issues/new?template=feature_request.md with relevant links and the team can go over and prioritize it. As for client-side OTEL, it exists but it is experimental on the OTEL side https://opentelemetry.io/docs/languages/js/getting-started/browser/ and not quite supported on our side. I have personally been able to get it working by modifying instrumentation code and following our tutorials, swapping out otel node for otel browser packages where applicable, but it is not particularly easy. Would love another feature request to help us prioritize this use case, would also be interested in hearing how it may be useful for you!
Farzad S.
·
Yes I’m happy to partner here!