[ENHANCEMENT]: Support for OTel GenAI SemConv?

[dedeswim]

Is your feature request related to a problem? Please describe.

Hi, thank you so much for maintaining Phoenix!

As I understand it, currently Phoenix shows LLM conversations in a nice UI if OTel telemetry is sent following OpenInference's semantic conventions. Are you planning to add support for OTel's own GenAI semantic conventions?

Describe the solution you'd like

I would like to be able to use directly OTel's GenAI semantic conventions instead of using OpenInference's, as I expect OTel ones to be more widely supported in the future.

[axiomofjoy]

Thanks for filing this issue @dedeswim. We have discussed adding support for ingesting OTel gen AI traces directly into Phoenix and will track upvotes on this issue to gauge demand. In the meantime, we do offer span processors for OTel gen AI native libraries such as Pydantic that convert Pydantic + OTel gen AI attributes into OpenInference.

[dedeswim]

Hi @axiomofjoy, thank you for your reply! I spotted a few bugs in the span processor for PydanticAI. Would you be open to a PR to fix them?

[axiomofjoy]

Hey @dedeswim, definitely!