test(decisions): audit #47508 (#48777)

* test(decisions): audit #47508

* test(decisions): price the Databricks decision cell from its cost-map row

---------

Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>

tests/integration/cost_calculation/cases/decisions.py

@@ -162,6 +162,37 @@ TYPESAFE_JEV_LATEST_SYSTEMONE: Final = CostTrackingTestCase(
         completion_tokens=22,
     ),
 )
+TYPESAFE_JEV_PREVIEW_ANSWERED_AS_PRICED_VERSION_SYSTEMONE: Final = CostTrackingTestCase(
+    name="typesafe/jev-preview-answered-as-priced-version-systemone",
+    covers="quota_management.spend_tracking.systemone.logs_cost",
+    model="typesafe/jev-preview",
+    endpoint="/v1/systemone",
+    request={
+        "model": "$MODEL",
+        "state": "The replacement part arrived but the wrong size.",
+        "questions": {"is_refund": {"type": "noul", "instructions": "Is the customer asking for a refund?"}},
+    },
+    upstream_body={
+        "model": "jev-preview",
+        "state": "The replacement part arrived but the wrong size.",
+        "questions": {"is_refund": {"type": "noul", "instructions": "Is the customer asking for a refund?"}},
+    },
+    response=JsonResponse(
+        content_type="application/json",
+        body={
+            "model": "jev-latest",
+            "answers": {"is_refund": {"type": "noul", "noul": 0.31}},
+            "usage": {"input_tokens": 311, "output_tokens": 9},
+        },
+    ),
+    expected=ExactExpected(
+        spend=311 * 4.2e-08,
+        input_cost=311 * 4.2e-08,
+        output_cost=0.0,
+        prompt_tokens=311,
+        completion_tokens=9,
+    ),
+)
 
 # perplexity/pplx-decider-v1-27b
 PERPLEXITY_PPLX_DECIDER_V1_27B_DECISIONS: Final = CostTrackingTestCase(
@@ -514,6 +545,7 @@ CASES: Final[tuple[CostTrackingTestCase, ...]] = (
     OPENAI_GPT_6_LUNA_SYSTEMONE,
     TYPESAFE_JEV_LATEST_DECISIONS,
     TYPESAFE_JEV_LATEST_SYSTEMONE,
+    TYPESAFE_JEV_PREVIEW_ANSWERED_AS_PRICED_VERSION_SYSTEMONE,
     PERPLEXITY_PPLX_DECIDER_V1_27B_DECISIONS,
     PERPLEXITY_PPLX_DECIDER_V1_27B_SYSTEMONE,
     CLOUDFLARE_CLEF_FLASH_DECISIONS,

tests/integration/cost_calculation/cost_map.py

@@ -121,6 +121,12 @@ COST_MAP_ENTRIES: Final[dict[str, dict[str, JsonValue]]] = {
         "input_cost_per_token": 4.2e-08,
         "output_cost_per_token": 0.0,
     },
+    "typesafe/jev-preview": {
+        "litellm_provider": "typesafe",
+        "mode": "evaluation",
+        "input_cost_per_token": 8.4e-08,
+        "output_cost_per_token": 0.0,
+    },
     "fireworks_ai/accounts/fireworks/routers/rune": {
         "litellm_provider": "fireworks_ai",
         "mode": "evaluation",

tests/integration/providers/test_decisions_wire.py

@@ -137,7 +137,7 @@ _PROVIDERS: Final = (
         "databricks-openjev-qwen35-4b",
         _API_KEY,
         False,
-        None,
+        "databricks/databricks-openjev-qwen35-4b",
     ),
 )
 _PERPLEXITY: Final = _PROVIDERS[0]
@@ -576,13 +576,26 @@ def test_upstream_errors_keep_their_status_and_log_an_unbilled_failure(gateway:
         )
 
 
-def test_upstream_success_without_answers_is_a_gateway_side_server_error(gateway: Gateway) -> None:
+@pytest.mark.parametrize(
+    ("reply", "field"),
+    (
+        pytest.param({"model": _PERPLEXITY.body_model, "usage": _USAGE}, "answers", id="missing_answers"),
+        pytest.param(
+            {"model": _PERPLEXITY.body_model, "answers": {"defect": {"type": "Request timed out"}}, "usage": _USAGE},
+            "answers.defect",
+            id="answer_type_timeout_text",
+        ),
+    ),
+)
+def test_a_malformed_upstream_success_is_a_server_error_naming_the_field(
+    gateway: Gateway, reply: dict[str, JsonValue], field: str
+) -> None:
     with gateway.scenario() as scenario:
-        handle: Final = _register(scenario, {"model": _PERPLEXITY.body_model, "usage": _USAGE})
+        handle: Final = _register(scenario, reply)
         model: Final = _deployment(scenario, handle, _PERPLEXITY)
         response: Final = _decide(gateway, model, num_retries=0)
-        assert 500 <= response.status_code < 600, response.text
-        assert "answers" in response.text
+        assert response.status_code == 500, response.text
+        assert field in response.text
         assert _spend_row(response.headers["x-litellm-call-id"])["status"] == "failure"