airflow.providers.common.ai.observability¶

OpenTelemetry tracing for pydantic-ai agents created by this provider.

pydantic-ai ships native OpenTelemetry instrumentation that emits GenAI spans (agent run, model call, tool call) following the OTel GenAI semantic conventions. This module turns it on and points it at Airflow’s existing OpenTelemetry exporter so agent spans flow to whatever OTLP backend the deployment already runs, nested under the worker’s task span.

It deliberately does not configure an exporter or a TracerProvider of its own: it reuses the global SDK provider that core tracing ([traces] otel_on) installs in the worker process. When that provider is absent (core tracing off, or not configured in this process) no spans are emitted.

Attributes¶

SECTION

Functions¶

`genai_instrumentation_settings`()	Build pydantic-ai `InstrumentationSettings` for an agent run.
`build_run_identity_attributes`(ti)	Build the Airflow identity attributes to stamp on a run's GenAI spans.
`stamp_identity_on_agent_spans`(agent, attributes)	Stamp attributes on every GenAI span agent emits during its run.

Module Contents¶

airflow.providers.common.ai.observability.SECTION = 'common.ai'[source]¶

airflow.providers.common.ai.observability.genai_instrumentation_settings()[source]¶

Build pydantic-ai InstrumentationSettings for an agent run.

Returns None (leave the agent un-instrumented, zero overhead) when export is disabled or no live OTLP TracerProvider is configured in this worker process. include_content is off by default so prompts, completions, and tool IO are never emitted unless explicitly opted in via [common.ai] capture_content.

airflow.providers.common.ai.observability.build_run_identity_attributes(ti)[source]¶

Build the Airflow identity attributes to stamp on a run’s GenAI spans.

Reuses core’s task-span attribute keys (see _make_task_span) so agent spans filter identically to the task span they nest under, plus the per-attempt task-instance id as the run join key carried on every span.

airflow.providers.common.ai.observability.stamp_identity_on_agent_spans(agent, attributes)[source]¶

Stamp attributes on every GenAI span agent emits during its run.

pydantic-ai opens all of a run’s agent/model/tool spans from InstrumentationSettings.tracer, so wrapping that one tracer reaches them all without touching the shared core TracerProvider. No-op when the agent is not instrumented with an InstrumentationSettings (tracing off, or the caller supplied its own non-settings instrument value).

The settings object may be caller-owned and shared across agents (a single module-level InstrumentationSettings handed to several tasks) or reused across HITL re-runs. Mutating it in place would leak one run’s identity into another and nest _IdentityTracer wrappers on each stamp, so we wrap a copy and swap it onto this agent, leaving the original untouched.