datarobot-external-agent-monitoring

You instrument the agent code to export OpenTelemetry traces, logs, and metrics into DataRobot via OTLP over HTTP. This requires creating a DataRobot shell deployment to act as the telemetry sink and configuring the appropriate exporters.

Yes, the monitoring framework detects common agent frameworks including LangChain, LangGraph, CrewAI, LlamaIndex, PydanticAI, and Google ADK, applying the appropriate instrumentation to route their telemetry into DataRobot.

Configuring OTLP telemetry export requires environment-variable-driven endpoint and authentication wiring. You must set up additive tracer, log, and metric exporters without conflicting OTEL_EXPORTER_OTLP_* variables to ensure proper routing.

OpenTelemetry traces may fail to show up in DataRobot if conflicting OTEL_EXPORTER_OTLP_* environment variables override the additive exporter configuration, preventing proper routing to the DataRobot shell deployment telemetry sink.

Yes, creating a DataRobot shell deployment is required to act as the telemetry sink that receives external agent traces, logs, and metrics before the instrumentation can successfully route observability data.