Description
The Python observability samples README (python/samples/02-agents/observability/README.md) shows how to send Agent Framework telemetry to Langfuse, Comet Opik, and MLflow, but not to Arize AX or Phoenix.
Arize AX and Phoenix use the OpenInference format. The openinference-instrumentation-agent-framework span processor converts Agent Framework's GenAI spans to it. Because the processor has to run before the exporter, the setup differs from the existing examples: it builds the tracer provider directly instead of calling configure_otel_providers().
Proposal: add an "Or with Arize AX or Phoenix" example alongside the existing third-party examples, and add the packages to the Dependencies list.
Code Sample
tracer_provider = TracerProvider(resource=Resource.create({"openinference.project.name": "<project_name>"}))
tracer_provider.add_span_processor(AgentFrameworkToOpenInferenceProcessor())
tracer_provider.add_span_processor(BatchSpanProcessor(exporter)) # OTLP exporter for Arize AX or Phoenix
trace.set_tracer_provider(tracer_provider)
enable_sensitive_telemetry()Language/SDK
Python
Description
The Python observability samples README (
python/samples/02-agents/observability/README.md) shows how to send Agent Framework telemetry to Langfuse, Comet Opik, and MLflow, but not to Arize AX or Phoenix.Arize AX and Phoenix use the OpenInference format. The
openinference-instrumentation-agent-frameworkspan processor converts Agent Framework's GenAI spans to it. Because the processor has to run before the exporter, the setup differs from the existing examples: it builds the tracer provider directly instead of callingconfigure_otel_providers().Proposal: add an "Or with Arize AX or Phoenix" example alongside the existing third-party examples, and add the packages to the Dependencies list.
Code Sample
Language/SDK
Python