Hey, do you plan to support message content as list in request payloads? This is part of the standard Open AI format when using VLMs and i.e. Smolagents defaults to it. As of now the Arize UI just shows empty blocks for system/user prompt
馃挕聽Hint: Mention RunLLM in the thread for followups.
Yea, we can support it. Can you let us know which instrumentor you鈥檙e currently using? I can look into it
Thanks! I鈥檓 using:
from phoenix.otel import register
from openinference.instrumentation.smolagents import SmolagentsInstrumentor
tracer_provider = register(
project_name="default"
)
SmolagentsInstrumentor().instrument(tracer_provider=tracer_provider)And in smolagents I鈥檓 using their standard OpenAIServerModel
I think there's an active fix out to account for the changes in smolagents! https://github.com/Arize-ai/openinference/pull/1238
