test(interactions): cover usage transformation, cost calculation, and success-handler billing

tests/test_litellm/litellm_core_utils/llm_cost_calc/test_usage_object_transformation.py

@@ -0,0 +1,73 @@
+from litellm.litellm_core_utils.llm_cost_calc.usage_object_transformation import (
+    InteractionsUsageObjectTransformation,
+)
+
+
+def _omni_flash_video_usage() -> dict:
+    return {
+        "total_tokens": 18247,
+        "total_input_tokens": 16,
+        "input_tokens_by_modality": [{"modality": "text", "tokens": 16}],
+        "total_cached_tokens": 0,
+        "total_output_tokens": 17937,
+        "output_tokens_by_modality": [
+            {"modality": "text", "tokens": 561},
+            {"modality": "video", "tokens": 17376},
+        ],
+        "total_tool_use_tokens": 0,
+        "total_thought_tokens": 294,
+    }
+
+
+def test_is_interactions_usage_dict():
+    assert InteractionsUsageObjectTransformation.is_interactions_usage_dict(_omni_flash_video_usage())
+    assert InteractionsUsageObjectTransformation.is_interactions_usage_dict({"total_output_tokens": 5})
+    assert not InteractionsUsageObjectTransformation.is_interactions_usage_dict(
+        {"prompt_tokens": 1, "completion_tokens": 2}
+    )
+    assert not InteractionsUsageObjectTransformation.is_interactions_usage_dict(
+        {"input_tokens": 1, "output_tokens": 2}
+    )
+
+
+def test_transform_interactions_usage_maps_modalities():
+    usage = InteractionsUsageObjectTransformation.transform_interactions_usage_to_chat_usage(
+        _omni_flash_video_usage()
+    )
+    assert usage.prompt_tokens == 16
+    assert usage.completion_tokens == 17937 + 294
+    assert usage.total_tokens == 18247
+    assert usage.prompt_tokens_details is not None
+    assert usage.prompt_tokens_details.text_tokens == 16
+    assert usage.prompt_tokens_details.cached_tokens == 0
+    assert usage.completion_tokens_details is not None
+    assert usage.completion_tokens_details.reasoning_tokens == 294
+    assert usage.completion_tokens_details.text_tokens == 561
+    assert usage.completion_tokens_details.video_tokens == 17376
+
+
+def test_transform_interactions_usage_reasoning_key_fallback():
+    usage = InteractionsUsageObjectTransformation.transform_interactions_usage_to_chat_usage(
+        {"total_input_tokens": 10, "total_output_tokens": 20, "total_reasoning_tokens": 5}
+    )
+    assert usage.prompt_tokens == 10
+    assert usage.completion_tokens == 25
+    assert usage.total_tokens == 35
+    assert usage.completion_tokens_details is not None
+    assert usage.completion_tokens_details.reasoning_tokens == 5
+
+
+def test_transform_interactions_usage_cached_tokens():
+    usage = InteractionsUsageObjectTransformation.transform_interactions_usage_to_chat_usage(
+        {"total_input_tokens": 100, "total_cached_tokens": 40, "total_output_tokens": 10}
+    )
+    assert usage.prompt_tokens == 100
+    assert usage.prompt_tokens_details is not None
+    assert usage.prompt_tokens_details.cached_tokens == 40
+
+
+def test_transform_interactions_usage_empty():
+    usage = InteractionsUsageObjectTransformation.transform_interactions_usage_to_chat_usage(None)
+    assert usage.prompt_tokens == 0
+    assert usage.completion_tokens == 0
+    assert usage.total_tokens == 0

tests/test_litellm/litellm_core_utils/test_litellm_logging.py

@@ -3773,3 +3773,140 @@ def test_zero_token_video_usage_preserves_duration_seconds(logging_obj):
     assert payload["metadata"]["usage_object"]["duration_seconds"] == 4.0
     assert payload["total_tokens"] == 0
     assert payload["completion_tokens"] == 0
+
+
+def _register_fake_interactions_model():
+    litellm.register_model(
+        model_cost={
+            "gemini/fake-interactions-model": {
+                "litellm_provider": "gemini",
+                "mode": "chat",
+                "input_cost_per_token": 3e-7,
+                "output_cost_per_token": 2.5e-6,
+            }
+        }
+    )
+
+
+def _interactions_usage_dict():
+    return {
+        "total_tokens": 871,
+        "total_input_tokens": 16,
+        "input_tokens_by_modality": [{"modality": "text", "tokens": 16}],
+        "total_cached_tokens": 0,
+        "total_output_tokens": 561,
+        "output_tokens_by_modality": [{"modality": "text", "tokens": 561}],
+        "total_tool_use_tokens": 0,
+        "total_thought_tokens": 294,
+    }
+
+
+_EXPECTED_INTERACTIONS_COST = 16 * 3e-7 + (561 + 294) * 2.5e-6
+
+
+def _interactions_logging_obj(call_type: str, stream: bool) -> LitellmLogging:
+    import datetime
+
+    logging_obj = LitellmLogging(
+        model="gemini/fake-interactions-model",
+        messages=[{"role": "user", "content": "hi"}],
+        stream=stream,
+        call_type=call_type,
+        start_time=datetime.datetime.now(),
+        litellm_call_id="interactions-cost-test",
+        function_id="interactions-cost-test",
+    )
+    logging_obj.update_from_kwargs(
+        kwargs={},
+        model="gemini/fake-interactions-model",
+        optional_params={},
+        litellm_params={"litellm_call_id": "interactions-cost-test"},
+        custom_llm_provider="gemini",
+    )
+    return logging_obj
+
+
+def test_interactions_create_success_tracks_response_cost():
+    import datetime
+
+    from litellm.types.interactions import InteractionsAPIResponse
+
+    _register_fake_interactions_model()
+    logging_obj = _interactions_logging_obj(call_type="acreate_interaction", stream=False)
+    result = InteractionsAPIResponse(
+        id="interaction_123",
+        model="fake-interactions-model",
+        status="completed",
+        steps=[],
+        usage=_interactions_usage_dict(),
+    )
+
+    logging_obj._success_handler_helper_fn(
+        result=result,
+        start_time=datetime.datetime.now(),
+        end_time=datetime.datetime.now(),
+        cache_hit=False,
+    )
+
+    assert logging_obj.model_call_details["response_cost"] == pytest.approx(_EXPECTED_INTERACTIONS_COST)
+    standard_logging_object = logging_obj.model_call_details["standard_logging_object"]
+    assert standard_logging_object["prompt_tokens"] == 16
+    assert standard_logging_object["completion_tokens"] == 561 + 294
+    assert standard_logging_object["total_tokens"] == 871
+    assert standard_logging_object["response_cost"] == pytest.approx(_EXPECTED_INTERACTIONS_COST)
+
+
+def test_interactions_streaming_completed_event_tracks_response_cost():
+    import datetime
+
+    from litellm.types.interactions import InteractionsAPIStreamingResponse
+
+    _register_fake_interactions_model()
+    logging_obj = _interactions_logging_obj(call_type="acreate_interaction", stream=True)
+    result = InteractionsAPIStreamingResponse(
+        event_type="interaction.completed",
+        interaction={
+            "id": "interaction_123",
+            "model": "fake-interactions-model",
+            "status": "completed",
+            "steps": [],
+            "usage": _interactions_usage_dict(),
+        },
+    )
+
+    logging_obj._success_handler_helper_fn(
+        result=result,
+        start_time=datetime.datetime.now(),
+        end_time=datetime.datetime.now(),
+        cache_hit=False,
+    )
+
+    assert logging_obj.model_call_details["response_cost"] == pytest.approx(_EXPECTED_INTERACTIONS_COST)
+    standard_logging_object = logging_obj.model_call_details["standard_logging_object"]
+    assert standard_logging_object["prompt_tokens"] == 16
+    assert standard_logging_object["completion_tokens"] == 561 + 294
+
+
+def test_interactions_get_poll_does_not_bill():
+    import datetime
+
+    from litellm.types.interactions import InteractionsAPIResponse
+
+    _register_fake_interactions_model()
+    logging_obj = _interactions_logging_obj(call_type="aget_interaction", stream=False)
+    result = InteractionsAPIResponse(
+        id="interaction_123",
+        model="fake-interactions-model",
+        status="completed",
+        steps=[],
+        usage=_interactions_usage_dict(),
+    )
+
+    logging_obj._success_handler_helper_fn(
+        result=result,
+        start_time=datetime.datetime.now(),
+        end_time=datetime.datetime.now(),
+        cache_hit=False,
+    )
+
+    assert logging_obj.model_call_details.get("response_cost") is None

tests/test_litellm/test_cost_calculator.py

@@ -3479,3 +3479,44 @@ def test_batch_cost_calculator_cache_creation_falls_back_to_input_rate():
     )
 
     assert prompt_cost == pytest.approx((1000 * 3e-6 + 8000 * 3e-7 + 2000 * 3e-6) / 2)
+
+
+def test_completion_cost_interactions_api_response():
+    from litellm.types.interactions import InteractionsAPIResponse
+
+    litellm.register_model(
+        model_cost={
+            "gemini/fake-interactions-model": {
+                "litellm_provider": "gemini",
+                "mode": "chat",
+                "input_cost_per_token": 3e-7,
+                "output_cost_per_token": 2.5e-6,
+            }
+        }
+    )
+
+    response = InteractionsAPIResponse(
+        id="interaction_123",
+        model="fake-interactions-model",
+        status="completed",
+        steps=[],
+        usage={
+            "total_tokens": 871,
+            "total_input_tokens": 16,
+            "input_tokens_by_modality": [{"modality": "text", "tokens": 16}],
+            "total_cached_tokens": 0,
+            "total_output_tokens": 561,
+            "output_tokens_by_modality": [{"modality": "text", "tokens": 561}],
+            "total_tool_use_tokens": 0,
+            "total_thought_tokens": 294,
+        },
+    )
+
+    cost = completion_cost(
+        completion_response=response,
+        model="gemini/fake-interactions-model",
+        custom_llm_provider="gemini",
+        call_type="acreate_interaction",
+    )
+
+    assert cost == pytest.approx(16 * 3e-7 + (561 + 294) * 2.5e-6)