test(gooddata-eval): cover the turn and step counts on every skill

metric_skill and alert_skill had no test for the new counters at all, and no
skill asserted that the counts reach Langfuse — the dataclass fields alone would
pass with the score writes deleted.

- counter tests for metric_skill and alert_skill
- score tests for all four skills, driven through submit_trace_scoring
- conversation pins `turns` as fixture turns plus clarification rounds, which is
  neither half on its own

Author: myhoai

packages/gooddata-eval/tests/test_agentic_alert_skill.py

@@ -996,3 +996,67 @@ def test_every_gen_ai_interval_is_accepted():
         assert AnomalyDetectionGranularity.parse(value.lower()) is AnomalyDetectionGranularity(value)
     assert AnomalyDetectionGranularity.parse(None) is None
     assert AnomalyDetectionGranularity.parse("  ") is None
+
+
+def test_run_agentic_alert_skill_counts_the_turns_and_reasoning_steps_it_used():
+    """QA-29110: the effort comparison reads these. A refusal still took a turn, and the turn
+    count is what separates a wrong answer from a run max_iterations cut short."""
+    mock_client = MagicMock()
+    mock_client.create_conversation.return_value = "conv-1"
+    mock_client.send_message.return_value = _no_alert_chat_result()
+    mock_client._base = "http://host/api/v1/actions/workspaces/ws1/ai"
+    mock_client._auth = {"Authorization": "Bearer tok"}
+
+    with _patched(mock_client, simulated_reply="Yes please"):
+        summary = run_agentic_alert_skill(
+            host="http://host/api/v1/actions/workspaces/ws1/ai",
+            token="tok",
+            workspace_id="ws1",
+            question="Create alert",
+            expected_output={"operator": "GREATER_THAN", "threshold": 100},
+            k=1,
+            max_iterations=2,
+        )
+
+    # _no_alert_chat_result has no tool calls and non-empty text, so the run replies once and
+    # stops at max_iterations: 2 turns, 1 reasoning step each.
+    assert summary.best.total_turns == 2.0
+    assert summary.best.total_steps == 2.0
+
+
+def test_alert_skill_writes_the_turn_and_step_counts_to_langfuse():
+    """The counters exist to reach Langfuse; asserting only the dataclass would pass even if
+    the scores were never written."""
+    mock_client = MagicMock()
+    mock_client.create_conversation.return_value = "conv-1"
+    mock_client.send_message.return_value = _no_alert_chat_result()
+    mock_client._base = "http://host/api/v1/actions/workspaces/ws1/ai"
+    mock_client._auth = {"Authorization": "Bearer tok"}
+    captured = {}
+
+    def _capture(_submit, _identity, **kwargs):
+        captured["write_scores"] = kwargs["write_scores"]
+
+    with (
+        _patched(mock_client),
+        patch("gooddata_eval.core.agentic.alert_skill.submit_trace_scoring", _capture),
+        pytest.raises(AlertSkillAssertionError),
+    ):
+        evaluate_agentic_alert_skill(
+            host="http://host/api/v1/actions/workspaces/ws1/ai",
+            token="tok",
+            workspace_id="ws1",
+            question="Create alert",
+            expected_output={"operator": "GREATER_THAN", "threshold": 100},
+            k=1,
+            max_iterations=1,
+            langfuse=MagicMock(),
+            dataset_item_id="item-1",
+        )
+
+    ctx = MagicMock()
+    captured["write_scores"](ctx)
+    scores = {c.kwargs["name"]: c.kwargs["value"] for c in ctx.score.call_args_list}
+
+    assert scores["turns"] == 1.0
+    assert scores["steps"] == 1.0

packages/gooddata-eval/tests/test_agentic_conversation.py

@@ -1071,3 +1071,61 @@ def test_run_agentic_conversation_sums_the_reasoning_steps_of_every_turn():
 
     assert result.total_steps == 5.0
     assert result.total_clarification_turns == 1
+
+
+def test_conversation_writes_the_turn_step_and_clarification_counts_to_langfuse():
+    """`turns` is not the clarification count: it is one per fixture turn plus every
+    simulated-user round, so a test has to pin the sum rather than either half."""
+    proposal_turn = ChatResult.model_validate(
+        {
+            "text_response": None,
+            "alertProposals": [{"cta": "Should I create this alert?", "recipients": [{"email": "a@b.com"}]}],
+            "reasoningStepCount": 2,
+            "toolCallEvents": [
+                {"functionName": "set_skills", "functionArguments": '{"skills": ["alert"]}', "result": None},
+                {"functionName": "prepare_metric_alert_proposal", "functionArguments": "{}", "result": None},
+            ],
+        }
+    )
+    created_turn = ChatResult.model_validate(
+        {
+            "text_response": "Alert created.",
+            "reasoningStepCount": 3,
+            "toolCallEvents": [
+                {"functionName": "create_metric_alert", "functionArguments": "{}", "result": '{"id": "alert-1"}'}
+            ],
+        }
+    )
+    mock_client = MagicMock()
+    mock_client.create_conversation.return_value = "conv-1"
+    mock_client.send_message.side_effect = [proposal_turn, created_turn]
+    captured = {}
+
+    def _capture(_submit, _identity, **kwargs):
+        captured["write_scores"] = kwargs["write_scores"]
+
+    with (
+        patch("gooddata_eval.core.agentic.conversation.ChatClient", return_value=mock_client),
+        patch("gooddata_eval.core.agentic.conversation.GoodDataSdk"),
+        patch("gooddata_eval.core.agentic.conversation.submit_trace_scoring", _capture),
+        patch(
+            "gooddata_eval.core.agentic.conversation._get_sim_user_response",
+            return_value="Yes, please create it.",
+        ),
+    ):
+        evaluate_agentic_conversation(
+            host="http://host/api/v1/actions/workspaces/ws1/ai",
+            token="tok",
+            workspace_id="ws1",
+            fixture=_alert_turn_fixture(),
+            langfuse=MagicMock(),
+            dataset_item_id="item-1",
+        )
+
+    ctx = MagicMock()
+    captured["write_scores"](ctx)
+    scores = {c.kwargs["name"]: c.kwargs["value"] for c in ctx.score.call_args_list}
+
+    assert scores["clarification_turns"] == 1.0
+    assert scores["turns"] == 2.0  # 1 fixture turn + 1 clarification round
+    assert scores["steps"] == 5.0

packages/gooddata-eval/tests/test_agentic_kda_skill.py

@@ -1245,3 +1245,38 @@ def test_run_agentic_kda_skill_reports_no_turns_when_the_first_send_fails():
 
     assert summary.best.total_turns == 0.0
     assert summary.best.total_steps == 0.0
+
+
+def test_kda_skill_writes_the_turn_and_step_counts_to_langfuse():
+    """The counters exist to reach Langfuse; asserting only the dataclass would pass even if
+    the scores were never written."""
+    mock_client = MagicMock()
+    mock_client.create_conversation.return_value = "conv-1"
+    mock_client.send_message.return_value = _kda_chat_result(success=True)
+    captured = {}
+
+    def _capture(_submit, _identity, **kwargs):
+        captured["write_scores"] = kwargs["write_scores"]
+
+    with (
+        patch("gooddata_eval.core.agentic.kda_skill.ChatClient", return_value=mock_client),
+        patch("gooddata_eval.core.agentic.kda_skill.submit_trace_scoring", _capture),
+    ):
+        evaluate_agentic_kda_skill(
+            host="http://host/api/v1/actions/workspaces/ws1/ai",
+            token="tok",
+            workspace_id="ws1",
+            question="What drove the change?",
+            expected_output=_EXPECTED,
+            k=1,
+            max_iterations=1,
+            langfuse=MagicMock(),
+            dataset_item_id="item-1",
+        )
+
+    ctx = MagicMock()
+    captured["write_scores"](ctx)
+    scores = {c.kwargs["name"]: c.kwargs["value"] for c in ctx.score.call_args_list}
+
+    assert scores["turns"] == 1.0
+    assert scores["steps"] == 1.0

packages/gooddata-eval/tests/test_agentic_metric_skill.py

@@ -764,3 +764,86 @@ def test_no_timer_output_by_default(monkeypatch, capsys):
     assert "[timer]" not in capsys.readouterr().out
     # Silenced, not un-measured.
     assert summary.run_results[0].timings.agent_s == 3.0
+
+
+def test_run_agentic_metric_skill_counts_the_turns_and_reasoning_steps_it_used():
+    """QA-29110: the effort comparison reads these. A clarification round is part of the work
+    the effort setting changes, so its steps count with the rest."""
+    clarify_turn = ChatResult.model_validate(
+        {"textResponse": "Which foo?", "toolCallEvents": [], "reasoningStepCount": 2}
+    )
+    created_turn = ChatResult.model_validate(
+        {
+            "textResponse": "done",
+            "reasoningStepCount": 3,
+            "toolCallEvents": [
+                {
+                    "functionName": "create_metric",
+                    "functionArguments": "{}",
+                    "result": '{"data": {"maql": "SELECT {metric/foo}"}}',
+                }
+            ],
+        }
+    )
+    mock_client = _client()
+    mock_client.send_message.side_effect = [clarify_turn, created_turn]
+
+    with _patched(mock_client, simulated_reply="It's foo"):
+        summary = run_agentic_metric_skill(
+            host="http://host/api/v1/actions/workspaces/ws1/ai",
+            token="tok",
+            workspace_id="ws1",
+            question="Create metric foo",
+            expected_output={"maql": "SELECT {metric/foo}"},
+            k=1,
+            max_iterations=2,
+        )
+
+    assert summary.best.total_turns == 2.0
+    assert summary.best.total_steps == 5.0
+
+
+def test_metric_skill_writes_the_turn_and_step_counts_to_langfuse():
+    """The counters exist to reach Langfuse; asserting only the dataclass would pass even if
+    the scores were never written."""
+    mock_client = _client()
+    mock_client.send_message.return_value = ChatResult.model_validate(
+        {
+            "textResponse": "done",
+            "reasoningStepCount": 4,
+            "toolCallEvents": [
+                {
+                    "functionName": "create_metric",
+                    "functionArguments": "{}",
+                    "result": '{"data": {"maql": "SELECT {metric/foo}"}}',
+                }
+            ],
+        }
+    )
+    captured = {}
+
+    def _capture(_submit, _identity, **kwargs):
+        captured["write_scores"] = kwargs["write_scores"]
+
+    with (
+        _patched(mock_client),
+        patch("gooddata_eval.core.agentic.metric_skill.submit_trace_scoring", _capture),
+    ):
+        evaluate_agentic_metric_skill(
+            host="http://host/api/v1/actions/workspaces/ws1/ai",
+            token="tok",
+            workspace_id="ws1",
+            question="Create metric foo",
+            expected_output={"maql": "SELECT {metric/foo}"},
+            k=1,
+            max_iterations=1,
+            langfuse=MagicMock(),
+            dataset_item_id="item-1",
+        )
+
+    ctx = MagicMock()
+    captured["write_scores"](ctx)
+    scores = {c.kwargs["name"]: c.kwargs["value"] for c in ctx.score.call_args_list}
+
+    assert scores["turns"] == 1.0
+    assert scores["steps"] == 4.0