OTEL, again

[rumpl]

https://opentelemetry.io/blog/2025/ai-agent-observability/

[krissetto]

do we need this issue? seems like this is all still early in the definiton phase on their side, i think this issue will end up rotting here for quite a while potentially :)

The GenAI Special Interest Group (SIG) in OpenTelemetry is actively defining GenAI semantic conventions

[rumpl]

I think we do, at least it reminds us that one day we will need to change our otel

[elgreco247]

Maybe we could start with the gen_ai attibutes listed here?
https://mlflow.org/docs/latest/genai/tracing/opentelemetry/attribute-mapping/#opentelemetry-genai-semantic-conventions

[tdabasinskas]

Hi,

Any plans on this? It seems that current OTEL implementation is not covering everything, as well as it's not following the semantics. Would be great to the official conventions from https://opentelemetry.io/docs/specs/semconv/gen-ai.

If you're open to it, I could draft a PR (we actually have most of this internally already).

[krissetto]

Hey! We don't have any specific plans for this at the moment.

Since some time has passed since i last checked the otel spec and things hopefully are in a better place now, feel free to open a PR!

We definitely appreciate the contribution

[tdabasinskas]

Hey! We don't have any specific plans for this at the moment.

Since some time has passed since i last checked the otel spec and things hopefully are in a better place now, feel free to open a PR!

We definitely appreciate the contribution

I've opened #2620. I know it's quite large, but I believe it might make sense to wire full OTEL support while it's still pre-v1.