Hello, is there a way to add an evaluation directly with the span while tracing? Currently what we do is add a unique ID to the span metadata, and then fetch the span id from the unique ID to add the eval afterward. But it feels really clunky, did we miss something in the docs? (We are using manual instrumentation)
Hi Remi, that鈥檚 a good question, our evaluation ingestion isn鈥檛 really set up well for sending in one evaluation at a time, so even if you handled the evals on your end probably the best way is to make a REST request with a zipped protobuf as the payload
This is the route you'd need to post to: https://github.com/Arize-ai/phoenix/blob/main/src/phoenix/server/api/routers/v1/evaluations.py#L42-L100 And this is the proto: https://github.com/Arize-ai/phoenix/blob/main/src/phoenix/trace/v1/evaluation_pb2.pyi
Thank you Jason & Dustin 馃檪
R茅mi C. Feel free to follow this ticket: https://github.com/Arize-ai/phoenix/pull/3240
It will need the OSS team to take the PR in / do changes and can't commit on timing but please follow.
This will be very helpful 馃檪
https://github.com/Arize-ai/phoenix/pull/3251 is that the one that is missing for Jason's PR to work? 馃檪
yep!
Oh wait
it's actually a refactor
Does that mean this is actuall available 馃槃 ??
