chore(telemetry): remove gen_ai.agent.workflow.steps metric

Semconv PR for this metric in:
open-telemetry/semantic-conventions-genai#203
Was replaced by `gen_ai.invoke_agent.{inference,tool}_calls` metrics in:
open-telemetry/semantic-conventions-genai#336

We're hence removing the non-standard and less defined
`gen_ai.agent.workflow.steps` metric.

Co-authored-by: Max Ind <maxind@google.com>
PiperOrigin-RevId: 943154162

Author: RKest

src/google/adk/telemetry/_instrumentation.py

@@ -125,7 +125,6 @@ def record_llm_response(
 def _record_agent_metrics(
     agent_name: str,
     elapsed_s: float,
-    events: object,
     caught_error: Exception | None,
 ) -> None:
   try:
@@ -134,7 +133,6 @@ def _record_agent_metrics(
         elapsed_s,
         caught_error,
     )
-    _metrics.record_agent_workflow_steps(agent_name, events)
   except Exception:  # pylint: disable=broad-exception-caught
     logger.exception("Failed to record agent metrics for agent %s", agent_name)
 
@@ -196,7 +194,6 @@ async def record_agent_invocation(
     _record_agent_metrics(
         agent.name,
         _metrics.get_elapsed_s(span, start_time),
-        getattr(getattr(ctx, "session", None), "events", []),
         caught_error,
     )
     _flush_invoke_agent_metrics(tel_ctx, agent.name)

src/google/adk/telemetry/_metrics.py

@@ -27,7 +27,6 @@
 from opentelemetry.semconv.attributes import error_attributes
 
 if TYPE_CHECKING:
-  from google.adk.events.event import Event
   from google.adk.models.llm_request import LlmRequest
   from google.adk.models.llm_response import LlmResponse
   from opentelemetry.trace import Span
@@ -91,24 +90,6 @@
         81.92,
     ],
 )
-_agent_workflow_steps = meter.create_histogram(
-    "gen_ai.agent.workflow.steps",
-    unit="1",
-    description="Length of agentic workflow (# of events).",
-    explicit_bucket_boundaries_advisory=[
-        1,
-        2,
-        4,
-        8,
-        16,
-        32,
-        64,
-        128,
-        256,
-        512,
-        1024,
-    ],
-)
 _client_operation_duration = (
     gen_ai_metrics.create_gen_ai_client_operation_duration(meter)
 )
@@ -198,13 +179,6 @@ def record_invoke_agent_tool_calls(agent_name: str, count: int) -> None:
   _invoke_agent_tool_calls.record(count, attributes=attrs)
 
 
-def record_agent_workflow_steps(agent_name: str, events: list[Event]):
-  """Records the number of steps in the agent workflow by counting the number of events."""
-  attrs = {gen_ai_attributes.GEN_AI_AGENT_NAME: agent_name}
-  count = sum(1 for event in events if event.author == agent_name)
-  _agent_workflow_steps.record(count, attributes=attrs)
-
-
 def record_tool_execution_duration(
     tool_name: str,
     tool_type: str,

tests/unittests/telemetry/functional_node_test_cases.py

@@ -2742,9 +2742,6 @@
             value=NON_DETERMINISTIC,
         ),
     }),
-    "gen_ai.agent.workflow.steps": frozenset({
-        MetricPoint(attributes={"gen_ai.agent.name": AGENT_NAME}, value=3),
-    }),
     "gen_ai.client.operation.duration": frozenset({
         MetricPoint(
             attributes={
@@ -2802,9 +2799,6 @@
             value=NON_DETERMINISTIC,
         ),
     }),
-    "gen_ai.agent.workflow.steps": frozenset({
-        MetricPoint(attributes={"gen_ai.agent.name": AGENT_NAME}, value=3),
-    }),
     "gen_ai.client.operation.duration": frozenset({
         MetricPoint(
             attributes={

tests/unittests/telemetry/functional_test_cases.py

@@ -2291,9 +2291,6 @@
             value=NON_DETERMINISTIC,
         ),
     }),
-    "gen_ai.agent.workflow.steps": frozenset({
-        MetricPoint(attributes={"gen_ai.agent.name": AGENT_NAME}, value=3),
-    }),
     "gen_ai.client.operation.duration": frozenset({
         MetricPoint(
             attributes={
@@ -2332,9 +2329,6 @@
             value=NON_DETERMINISTIC,
         ),
     }),
-    "gen_ai.agent.workflow.steps": frozenset({
-        MetricPoint(attributes={"gen_ai.agent.name": AGENT_NAME}, value=3),
-    }),
     "gen_ai.client.operation.duration": frozenset({
         MetricPoint(
             attributes={

tests/unittests/telemetry/functional_test_helpers.py

@@ -314,11 +314,6 @@ class HistogramSpec(NamedTuple):
         attr="_tool_execution_duration",
         metric_name="gen_ai.execute_tool.duration",
     ),
-    HistogramSpec(
-        module=_metrics,
-        attr="_agent_workflow_steps",
-        metric_name="gen_ai.agent.workflow.steps",
-    ),
     HistogramSpec(
         module=_metrics,
         attr="_client_operation_duration",

tests/unittests/telemetry/test_instrumentation.py

@@ -17,10 +17,8 @@
 import time
 from unittest import mock
 
-from google.adk.telemetry import _instrumentation
 from google.adk.telemetry import _metrics
 from opentelemetry import trace
-import pytest
 
 
 def test_get_elapsed_s_span_none():
@@ -82,36 +80,3 @@ def test_get_elapsed_s_span_non_int_end():
   with mock.patch("time.monotonic", return_value=12.0):
     elapsed = _metrics.get_elapsed_s(mock_span, start_time)
   assert elapsed == 2.0
-
-
-@pytest.mark.asyncio
-async def test_record_agent_invocation_tolerates_minimal_context():
-  """Tolerates context-likes that lack session.
-
-  Test doubles, partial migrations, and external embedders can pass an
-  InvocationContext-like object with a `session` that has no `events`
-  attribute. The telemetry path must not raise AttributeError on the
-  metrics call in those cases.
-  """
-  agent = mock.MagicMock()
-  agent.name = "test_agent"
-  # Bare object without `session`.
-  bare_ctx = object()
-
-  with (
-      mock.patch.object(
-          _instrumentation, "_record_agent_metrics"
-      ) as mock_record,
-      mock.patch.object(_instrumentation, "tracing") as mock_tracing,
-  ):
-    mock_tracing.tracer.start_as_current_span.return_value.__enter__.return_value = mock.MagicMock(
-        spec=trace.Span
-    )
-    async with _instrumentation.record_agent_invocation(bare_ctx, agent):
-      pass
-
-  mock_record.assert_called_once()
-  call_args = mock_record.call_args
-  # positional: (agent_name, elapsed_s, events, caught_error)
-  assert call_args.args[0] == "test_agent"
-  assert call_args.args[2] == []  # events default

tests/unittests/telemetry/test_metrics.py

@@ -29,14 +29,12 @@ def _mock_meter_setup(monkeypatch):
   agent_duration_hist = mock.MagicMock(spec=metrics.Histogram)
   workflow_duration_hist = mock.MagicMock(spec=metrics.Histogram)
   tool_duration_hist = mock.MagicMock(spec=metrics.Histogram)
-  steps_hist = mock.MagicMock(spec=metrics.Histogram)
   client_duration_hist = mock.MagicMock(spec=metrics.Histogram)
   client_token_usage_hist = mock.MagicMock(spec=metrics.Histogram)
 
   agent_duration_hist.name = "agent_invocation_duration"
   workflow_duration_hist.name = "workflow_invocation_duration"
   tool_duration_hist.name = "tool_execution_duration"
-  steps_hist.name = "agent_workflow_steps"
   client_duration_hist.name = "client_operation_duration"
   client_token_usage_hist.name = "client_token_usage"
 
@@ -47,8 +45,6 @@ def create_histogram_side_effect(name, **_kwargs):
       return workflow_duration_hist
     elif name == "gen_ai.execute_tool.duration":
       return tool_duration_hist
-    elif name == "gen_ai.agent.workflow.steps":
-      return steps_hist
     elif name == "gen_ai.client.operation.duration":
       return client_duration_hist
     elif name == "gen_ai.client.token.usage":
@@ -66,7 +62,6 @@ def create_histogram_side_effect(name, **_kwargs):
       _metrics, "_workflow_invocation_duration", workflow_duration_hist
   )
   monkeypatch.setattr(_metrics, "_tool_execution_duration", tool_duration_hist)
-  monkeypatch.setattr(_metrics, "_agent_workflow_steps", steps_hist)
   monkeypatch.setattr(
       _metrics, "_client_operation_duration", client_duration_hist
   )
@@ -77,7 +72,6 @@ def create_histogram_side_effect(name, **_kwargs):
       "agent_duration": agent_duration_hist,
       "workflow_duration": workflow_duration_hist,
       "tool_duration": tool_duration_hist,
-      "steps": steps_hist,
       "client_duration": client_duration_hist,
       "client_token_usage": client_token_usage_hist,
   }
@@ -145,24 +139,6 @@ def test_record_workflow_invocation_duration_nested_with_error(
   assert kwargs["attributes"]["error.type"] == "ValueError"
 
 
-def test_record_agent_workflow_steps(mock_meter_setup):
-  """Tests record_agent_workflow_steps records correctly."""
-  _metrics.record_agent_workflow_steps(
-      "test_agent",
-      [
-          mock.MagicMock(author="test_agent"),
-          mock.MagicMock(author="test_agent"),
-          mock.MagicMock(author="other_agent"),
-      ],
-  )
-  steps_hist = mock_meter_setup["steps"]
-  steps_hist.record.assert_called_once()
-  args, kwargs = steps_hist.record.call_args
-  assert args[0] == 2
-  want_attributes = {"gen_ai.agent.name": "test_agent"}
-  assert kwargs["attributes"] == want_attributes
-
-
 def test_record_tool_execution_duration(mock_meter_setup):
   """Tests record_tool_execution_duration records correctly."""
   _metrics.record_tool_execution_duration(