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

Phoenix Support

Comparing Phoenix with Evidently for ML Observability Solutions

·Apr 19, 2024 05:18 PM

I’m considering phoenix for some proof of concept work for ML observability.

1.
How does it compare to Evidently and other open source ML Observability solutions?
2.
How does the LLM eval in Phoenix compare to Guardrails AI evals?

3 comments

· Sorted by Oldest

Mikyo
·
Hey Amy thanks for checking out phoenix!
1.
How does it compare to Evidently and other open source ML Observability solutions?
There are some overlaps in terms of the fact that they both can be used to surface up data drift and data quality problems. I'd say Phoenix is more focused on unstructured data and LLM tracing where as Evidently is more focused on structured data. I have to be honest however that it's been some time since I've tried Evidently so let us know if there's any gaps that you'd like to see filled by phoenix!
Mikyo
·
1.
How does the LLM eval in Phoenix compare to Guardrails AI evals?
The distinction is probably that Guardrails is designed for model evals and inference time checks against things like PII and Jailbreaks vs Phoenix Evals is designed for task evaluations - validating that your application is working as designed based on application traces and datasets. Hope that helps.
Tammy L.
·
Amy To Mikyo's note, Phoenix OSS is more tailored to deep learning and LLM use cases, but Arize does have a free plan if you want to evaluate what's possible on the ML monitoring front