litellm/tests/unit/integrations/test_langfuse_otel.py

import base64
import json
import os
from datetime import datetime, timezone
from typing import Final
from unittest.mock import MagicMock, patch

import pytest
from opentelemetry.sdk.trace import TracerProvider

from litellm.integrations.langfuse.langfuse_otel import LangfuseOtelLogger
from litellm.integrations.opentelemetry import OpenTelemetry, OpenTelemetryConfig
from litellm.types.llms.openai import ResponsesAPIResponse
from litellm.types.utils import StandardCallbackDynamicParams


class TestLangfuseOtelIntegration:
    def test_get_langfuse_otel_config_with_required_env_vars(self):
        """Test that config is created correctly with required environment variables."""
        # Clean environment of any Langfuse-related variables
        env_vars_to_clean = [
            "LANGFUSE_HOST",
            "OTEL_EXPORTER_OTLP_ENDPOINT",
            "OTEL_EXPORTER_OTLP_HEADERS",
        ]
        with patch.dict(
            os.environ,
            {
                "LANGFUSE_PUBLIC_KEY": "test_public_key",
                "LANGFUSE_SECRET_KEY": "test_secret_key",
            },
            clear=False,
        ):
            # Remove any existing Langfuse variables
            for var in env_vars_to_clean:
                if var in os.environ:
                    del os.environ[var]

            config = LangfuseOtelLogger.get_langfuse_otel_config()

            assert isinstance(config, OpenTelemetryConfig)
            assert config.exporter == "otlp_http"
            assert "Authorization=Basic" in config.headers
            # Note: We no longer set os.environ explicitly to avoid leakage
            # assert os.environ.get("OTEL_EXPORTER_OTLP_ENDPOINT") == "https://us.cloud.langfuse.com/api/public/otel"
            # assert "Authorization=Basic" in os.environ.get("OTEL_EXPORTER_OTLP_HEADERS", "")

    def test_get_langfuse_otel_config_missing_keys(self):
        """Test that ValueError is raised when required keys are missing."""
        with patch.dict(os.environ, {}, clear=True):
            with pytest.raises(
                ValueError,
                match="LANGFUSE_PUBLIC_KEY and LANGFUSE_SECRET_KEY must be set",
            ):
                LangfuseOtelLogger.get_langfuse_otel_config()

    def test_get_langfuse_otel_config_with_eu_host(self):
        """Test config with EU host."""
        with patch.dict(
            os.environ,
            {
                "LANGFUSE_PUBLIC_KEY": "test_public_key",
                "LANGFUSE_SECRET_KEY": "test_secret_key",
                "LANGFUSE_HOST": "https://cloud.langfuse.com",
            },
            clear=False,
        ):
            config = LangfuseOtelLogger.get_langfuse_otel_config()
            # Endpoint assertion removed as side effect is gone
            assert isinstance(config, OpenTelemetryConfig)

    def test_get_langfuse_otel_config_with_custom_host(self):
        """Test config with custom host."""
        with patch.dict(
            os.environ,
            {
                "LANGFUSE_PUBLIC_KEY": "test_public_key",
                "LANGFUSE_SECRET_KEY": "test_secret_key",
                "LANGFUSE_HOST": "https://my-langfuse.com",
            },
            clear=False,
        ):
            config = LangfuseOtelLogger.get_langfuse_otel_config()
            # Endpoint assertion removed as side effect is gone
            assert isinstance(config, OpenTelemetryConfig)

    def test_get_langfuse_otel_config_with_host_no_protocol(self):
        """Test config with custom host without protocol."""
        with patch.dict(
            os.environ,
            {
                "LANGFUSE_PUBLIC_KEY": "test_public_key",
                "LANGFUSE_SECRET_KEY": "test_secret_key",
                "LANGFUSE_HOST": "my-langfuse.com",
            },
            clear=False,
        ):
            config = LangfuseOtelLogger.get_langfuse_otel_config()
            # Endpoint assertion removed as side effect is gone
            assert isinstance(config, OpenTelemetryConfig)

    def test_set_langfuse_otel_attributes(self):
        """Test that set_langfuse_otel_attributes calls the Arize utils function."""
        from litellm.integrations.langfuse.langfuse_otel_attributes import (
            LangfuseLLMObsOTELAttributes,
        )

        mock_span = MagicMock()
        mock_kwargs = {"test": "kwargs"}
        mock_response = {"test": "response"}

        with patch(
            "litellm.integrations.arize._utils.set_attributes"
        ) as mock_set_attributes:
            LangfuseOtelLogger.set_langfuse_otel_attributes(
                mock_span, mock_kwargs, mock_response
            )

            mock_set_attributes.assert_called_once_with(
                mock_span, mock_kwargs, mock_response, LangfuseLLMObsOTELAttributes
            )
            mock_span.set_attribute.assert_any_call(
                "langfuse.observation.type", "generation"
            )

    def test_set_langfuse_environment_attribute(self):
        """Test that Langfuse environment is set correctly when environment variable is present."""
        mock_span = MagicMock()
        mock_kwargs = {"test": "kwargs"}
        test_env = "staging"

        with patch.dict(os.environ, {"LANGFUSE_TRACING_ENVIRONMENT": test_env}):
            with patch(
                "litellm.integrations.arize._utils.safe_set_attribute"
            ) as mock_safe_set_attribute:
                LangfuseOtelLogger._set_langfuse_specific_attributes(
                    mock_span, mock_kwargs, {}
                )

                # safe_set_attribute(span, key, value) → positional args
                mock_safe_set_attribute.assert_called_once_with(
                    mock_span, "langfuse.environment", test_env
                )

    def test_set_langfuse_environment_attribute_prefers_dynamic_param(self):
        """Per-key/team langfuse_environment beats the deployment env var."""

        class _RecordingSpan:
            def __init__(self):
                self.attributes = {}

            def set_attribute(self, key, value):
                self.attributes[key] = value

        span = _RecordingSpan()
        mock_kwargs = {
            "standard_callback_dynamic_params": {
                "langfuse_environment": "team-a-env"
            }
        }

        with patch.dict(
            os.environ, {"LANGFUSE_TRACING_ENVIRONMENT": "deployment-wide"}
        ):
            LangfuseOtelLogger._set_langfuse_specific_attributes(
                span, mock_kwargs, {}
            )

        assert span.attributes["langfuse.environment"] == "team-a-env"

    @pytest.mark.parametrize(
        "completion_start_time",
        [
            datetime(2026, 10, 9, 12, 0, 0, 250000, tzinfo=timezone.utc),
            datetime(2026, 10, 9, 12, 0, 0, 250000),
        ],
        ids=["aware", "naive_local_time"],
    )
    def test_set_langfuse_specific_attributes_completion_start_time(self, completion_start_time: datetime):
        """The first-token time goes out the way the Langfuse SDK sends it: a JSON string, with an explicit offset."""
        span: Final = TracerProvider().get_tracer(__name__).start_span("litellm_request")

        LangfuseOtelLogger._set_langfuse_specific_attributes(span, {"completion_start_time": completion_start_time}, {})

        sent: Final = datetime.fromisoformat(json.loads(span.attributes["langfuse.observation.completion_start_time"]))
        assert sent.tzinfo is not None
        assert sent.timestamp() == completion_start_time.timestamp()

    def test_set_langfuse_specific_attributes_without_completion_start_time(self):
        span: Final = TracerProvider().get_tracer(__name__).start_span("litellm_request")

        LangfuseOtelLogger._set_langfuse_specific_attributes(span, {}, {})

        assert "langfuse.observation.completion_start_time" not in span.attributes

    def test_extract_langfuse_metadata_basic(self):
        """Ensure metadata is correctly pulled from litellm_params."""
        metadata_in = {"generation_name": "my-gen", "custom": "data"}
        kwargs = {"litellm_params": {"metadata": metadata_in}}
        extracted = LangfuseOtelLogger._extract_langfuse_metadata(kwargs)
        assert extracted == metadata_in

    def test_extract_langfuse_metadata_with_header_enrichment(self, monkeypatch):
        """_extract_langfuse_metadata should call LangFuseLogger.add_metadata_from_header when available."""
        import sys
        import types

        # Build a stub module + class on-the-fly
        stub_module = types.ModuleType("litellm.integrations.langfuse.langfuse")

        class StubLFLogger:
            @staticmethod
            def add_metadata_from_header(litellm_params, metadata):
                # Echo back existing metadata plus a marker
                return {**metadata, "enriched": True}

        stub_module.LangFuseLogger = StubLFLogger  # type: ignore

        # Register stub in sys.modules so import inside method succeeds.
        # Use monkeypatch so the real module is restored after the test runs,
        # preventing sys.modules corruption that would break patch() targets in
        # later tests (the patch would hit the stub while the real module's
        # globals remain unpatched).
        monkeypatch.setitem(sys.modules, "litellm.integrations.langfuse.langfuse", stub_module)  # type: ignore

        kwargs = {"litellm_params": {"metadata": {"foo": "bar"}}}
        extracted = LangfuseOtelLogger._extract_langfuse_metadata(kwargs)
        assert extracted.get("foo") == "bar"
        assert extracted.get("enriched") is True

    def test_set_langfuse_specific_attributes_metadata(self):
        """Verify every supported metadata key maps to the correct OTEL attribute and complex types are JSON-serialised."""
        # Build a sample metadata payload covering all mappings
        metadata = {
            "generation_name": "gen-name",
            "generation_id": "gen-id",
            "parent_observation_id": "parent-id",
            "version": "v1",
            "mask_input": True,
            "mask_output": False,
            "trace_user_id": "user-123",
            "session_id": "sess-456",
            "tags": ["tagA", "tagB"],
            "trace_name": "trace-name",
            "trace_id": "trace-id",
            "trace_metadata": {"k": "v"},
            "trace_version": "t-ver",
            "trace_release": "rel-1",
            "existing_trace_id": "existing-id",
            "update_trace_keys": ["key1", "key2"],
            "debug_langfuse": True,
        }
        kwargs = {"litellm_params": {"metadata": metadata}}

        # Capture calls to safe_set_attribute
        with patch(
            "litellm.integrations.arize._utils.safe_set_attribute"
        ) as mock_safe_set_attribute:
            LangfuseOtelLogger._set_langfuse_specific_attributes(
                MagicMock(), kwargs, None
            )

            # Build expected calls manually for clarity
            from litellm.types.integrations.langfuse_otel import LangfuseSpanAttributes

            expected = {
                LangfuseSpanAttributes.GENERATION_NAME.value: "gen-name",
                LangfuseSpanAttributes.GENERATION_ID.value: "gen-id",
                LangfuseSpanAttributes.PARENT_OBSERVATION_ID.value: "parent-id",
                LangfuseSpanAttributes.VERSION.value: "t-ver",
                LangfuseSpanAttributes.MASK_INPUT.value: True,
                LangfuseSpanAttributes.MASK_OUTPUT.value: False,
                LangfuseSpanAttributes.TRACE_USER_ID.value: "user-123",
                LangfuseSpanAttributes.SESSION_ID.value: "sess-456",
                # Lists / dicts should be JSON strings
                LangfuseSpanAttributes.TAGS.value: json.dumps(["tagA", "tagB"]),
                LangfuseSpanAttributes.TRACE_NAME.value: "trace-name",
                LangfuseSpanAttributes.TRACE_ID.value: "traceid",  # stripped dashes
                LangfuseSpanAttributes.TRACE_METADATA.value: json.dumps({"k": "v"}),
                LangfuseSpanAttributes.RELEASE.value: "rel-1",
                LangfuseSpanAttributes.EXISTING_TRACE_ID.value: "existing-id",
                LangfuseSpanAttributes.UPDATE_TRACE_KEYS.value: json.dumps(
                    ["key1", "key2"]
                ),
                LangfuseSpanAttributes.DEBUG_LANGFUSE.value: True,
            }

            # Flatten the actual calls into {key: value}
            actual = {
                call.args[1]: call.args[2]  # (span, key, value)
                for call in mock_safe_set_attribute.call_args_list
            }

            assert (
                actual == expected
            ), "Mismatch between expected and actual OTEL attribute mapping."

    @pytest.mark.parametrize(
        "metadata, expected_version",
        [
            (
                {"version": "v-observation", "trace_version": "v-trace"},
                "v-trace",
            ),
            ({"trace_version": "v-trace"}, "v-trace"),
            ({"version": "v-observation"}, "v-observation"),
            ({"version": "v-observation", "trace_version": ""}, ""),
            ({}, None),
        ],
        ids=[
            "trace-version-wins-as-documented",
            "trace-only",
            "observation-version-is-the-fallback",
            "empty-trace-version-is-not-absent",
            "neither-key-emits-nothing",
        ],
    )
    def test_version_emitted_on_langfuse_v4_key(self, metadata, expected_version):
        kwargs = {"litellm_params": {"metadata": {"trace_release": "rel-9", **metadata}}}

        with patch(
            "litellm.integrations.arize._utils.safe_set_attribute"
        ) as mock_safe_set_attribute:
            LangfuseOtelLogger._set_langfuse_specific_attributes(
                MagicMock(), kwargs, None
            )

        emitted = {
            call.args[1]: call.args[2] for call in mock_safe_set_attribute.call_args_list
        }

        if expected_version is None:
            assert "langfuse.version" not in emitted
        else:
            assert emitted["langfuse.version"] == expected_version
        assert emitted["langfuse.release"] == "rel-9"
        for retired_key in (
            "langfuse.generation.version",
            "langfuse.trace.version",
            "langfuse.trace.release",
        ):
            assert retired_key not in emitted

    def test_set_langfuse_specific_attributes_with_content(self):
        """Test that _set_langfuse_specific_attributes correctly sets observation.output with regular content response."""
        from litellm.types.integrations.langfuse_otel import LangfuseSpanAttributes
        from litellm.types.utils import Choices, ModelResponse

        # Create response with content
        response_obj = ModelResponse(
            id="chatcmpl-test",
            model="gpt-4o",
            choices=[
                Choices(
                    finish_reason="stop",
                    message={
                        "role": "assistant",
                        "content": "The weather in Tokyo is sunny.",
                    },
                )
            ],
        )

        kwargs = {
            "messages": [{"role": "user", "content": "What's the weather in Tokyo?"}],
        }

        with patch(
            "litellm.integrations.arize._utils.safe_set_attribute"
        ) as mock_safe_set_attribute:
            LangfuseOtelLogger._set_langfuse_specific_attributes(
                MagicMock(), kwargs, response_obj
            )

            expect_output = {
                LangfuseSpanAttributes.OBSERVATION_INPUT.value: [
                    {"role": "user", "content": "What's the weather in Tokyo?"}
                ],
                LangfuseSpanAttributes.OBSERVATION_OUTPUT.value: {
                    "role": "assistant",
                    "content": "The weather in Tokyo is sunny.",
                },
            }

            # Flatten the actual calls into {key: value}
            actual = {
                call.args[1]: json.loads(call.args[2])
                for call in mock_safe_set_attribute.call_args_list
            }

            assert (
                actual == expect_output
            ), "Mismatch in observation input/output OTEL attributes."

    def test_set_langfuse_specific_attributes_with_tool_calls(self):
        """Test that _set_langfuse_specific_attributes correctly sets observation.output with tool calls in Langfuse format."""
        from litellm.types.integrations.langfuse_otel import LangfuseSpanAttributes
        from litellm.types.utils import (
            ChatCompletionMessageToolCall,
            Choices,
            Function,
            ModelResponse,
        )

        # Create response with tool calls
        response_obj = ModelResponse(
            id="chatcmpl-test",
            model="gpt-4o",
            choices=[
                Choices(
                    finish_reason="tool_calls",
                    message={
                        "role": "assistant",
                        "content": None,
                        "tool_calls": [
                            ChatCompletionMessageToolCall(
                                function=Function(
                                    arguments='{"location":"Tokyo"}', name="get_weather"
                                ),
                                id="call_123",
                                type="function",
                            )
                        ],
                    },
                )
            ],
        )

        with patch(
            "litellm.integrations.arize._utils.safe_set_attribute"
        ) as mock_safe_set_attribute:
            LangfuseOtelLogger._set_langfuse_specific_attributes(
                MagicMock(), {}, response_obj
            )

            expected = {
                LangfuseSpanAttributes.OBSERVATION_OUTPUT.value: [
                    {
                        "id": "chatcmpl-test",
                        "name": "get_weather",
                        "arguments": {"location": "Tokyo"},
                        "call_id": "call_123",
                        "type": "function_call",
                    }
                ]
            }

            # Flatten the actual calls into {key: value}
            actual = {
                call.args[1]: json.loads(call.args[2])
                for call in mock_safe_set_attribute.call_args_list
            }
            assert (
                actual == expected
            ), "Mismatch in observation output OTEL attribute for tool calls."

    def test_construct_dynamic_otel_headers_with_langfuse_keys(self):
        """Test that construct_dynamic_otel_headers creates proper auth headers when langfuse keys are provided."""
        from litellm.types.utils import StandardCallbackDynamicParams

        # Create dynamic params with langfuse keys
        dynamic_params = StandardCallbackDynamicParams(
            langfuse_public_key="test_public_key", langfuse_secret_key="test_secret_key"
        )

        logger = LangfuseOtelLogger()
        result = logger.construct_dynamic_otel_headers(dynamic_params)

        # Should return a dict with otlp_auth_headers
        assert result is not None
        assert "Authorization" in result

        # The auth header should contain the basic auth format
        auth_header = result["Authorization"]
        assert auth_header.startswith("Basic ")

        # Verify the header format by decoding
        import base64

        # Extract the base64 part from "Authorization=Basic <base64>"
        base64_part = auth_header.replace("Basic ", "")
        decoded = base64.b64decode(base64_part).decode()

        assert decoded == "test_public_key:test_secret_key"

    def test_construct_dynamic_otel_headers_empty_params(self):
        """Test that construct_dynamic_otel_headers returns empty dict when no langfuse keys are provided."""
        from litellm.types.utils import StandardCallbackDynamicParams

        # Create dynamic params without langfuse keys
        dynamic_params = StandardCallbackDynamicParams()

        logger = LangfuseOtelLogger()
        result = logger.construct_dynamic_otel_headers(dynamic_params)

        # Should return an empty dict
        assert result == {}

    def test_get_langfuse_otel_config_with_otel_host_priority(self):
        """LANGFUSE_OTEL_HOST should take priority over LANGFUSE_HOST."""
        with patch.dict(
            os.environ,
            {
                "LANGFUSE_PUBLIC_KEY": "test_public_key",
                "LANGFUSE_SECRET_KEY": "test_secret_key",
                "LANGFUSE_HOST": "https://should-not-be-used.com",
                "LANGFUSE_OTEL_HOST": "https://otel-host.com",
            },
            clear=False,
        ):
            config = LangfuseOtelLogger.get_langfuse_otel_config()
            assert isinstance(config, OpenTelemetryConfig)
            # Endpoint assertion removed as side effect is gone


def _exported_auth_headers(exporter):
    import requests_mock
    from opentelemetry.sdk.trace.export import SpanExportResult

    with requests_mock.Mocker() as collector:
        collector.post(requests_mock.ANY, status_code=200)
        assert exporter.export(()) is SpanExportResult.SUCCESS
    return [
        (
            request.url,
            {name: request.headers[name] for name in ("Authorization", "x-langfuse-ingestion-version")},
        )
        for request in collector.request_history
    ]


class TestLangfuseOtelKeyDynamicConfig:
    """Key/team-scoped Langfuse credentials must define the full export target
    (OTLP endpoint + auth), not just auth headers on the init-time exporter."""

    CLEAN_ENV_VARS = [
        "LANGFUSE_PUBLIC_KEY",
        "LANGFUSE_SECRET_KEY",
        "LANGFUSE_HOST",
        "LANGFUSE_OTEL_HOST",
        "OTEL_EXPORTER",
        "OTEL_EXPORTER_OTLP_PROTOCOL",
        "OTEL_ENDPOINT",
        "OTEL_EXPORTER_OTLP_ENDPOINT",
        "OTEL_HEADERS",
        "OTEL_EXPORTER_OTLP_HEADERS",
    ]

    def _clean_env(self):
        cleaned = {k: v for k, v in os.environ.items() if k not in self.CLEAN_ENV_VARS}
        return patch.dict(os.environ, cleaned, clear=True)

    def _dynamic_params(self, **overrides):
        from litellm.types.utils import StandardCallbackDynamicParams

        params = {
            "langfuse_public_key": "key_public",
            "langfuse_secret_key": "key_secret",
            "langfuse_host": "https://langfuse.example.com",
        }
        params.update(overrides)
        return StandardCallbackDynamicParams(**{k: v for k, v in params.items() if v is not None})

    def test_construct_dynamic_otel_config_with_key_credentials(self):
        with self._clean_env():
            logger = LangfuseOtelLogger()
            config = logger.construct_dynamic_otel_config(self._dynamic_params())

        assert config is not None
        assert config.exporter == "otlp_http"
        assert config.endpoint == "https://langfuse.example.com/api/public/otel"

        import base64

        expected_auth = base64.b64encode(b"key_public:key_secret").decode()
        assert config.headers == f"Authorization=Basic {expected_auth},x-langfuse-ingestion-version=4"

    def test_construct_dynamic_otel_config_host_without_protocol(self):
        with self._clean_env():
            logger = LangfuseOtelLogger()
            config = logger.construct_dynamic_otel_config(self._dynamic_params(langfuse_host="langfuse.example.com"))

        assert config is not None
        assert config.endpoint == "https://langfuse.example.com/api/public/otel"

    def test_construct_dynamic_otel_config_defaults_to_us_cloud(self):
        with self._clean_env():
            logger = LangfuseOtelLogger()
            config = logger.construct_dynamic_otel_config(self._dynamic_params(langfuse_host=None))

        assert config is not None
        assert config.endpoint == "https://us.cloud.langfuse.com/api/public/otel"

    def test_construct_dynamic_otel_config_falls_back_to_env_host(self):
        with self._clean_env():
            with patch.dict(os.environ, {"LANGFUSE_HOST": "https://env-host.example.com"}):
                logger = LangfuseOtelLogger()
                config = logger.construct_dynamic_otel_config(self._dynamic_params(langfuse_host=None))

        assert config is not None
        assert config.endpoint == "https://env-host.example.com/api/public/otel"

    def test_construct_dynamic_otel_config_requires_both_keys(self):
        with self._clean_env():
            logger = LangfuseOtelLogger()

            assert logger.construct_dynamic_otel_config(self._dynamic_params(langfuse_secret_key=None)) is None
            assert logger.construct_dynamic_otel_config(self._dynamic_params(langfuse_public_key=None)) is None

    def test_key_dynamic_params_create_otlp_exporter_without_global_env(self):
        """Without global LANGFUSE_* env vars, a request carrying key-scoped Langfuse
        credentials must get a tracer exporting via OTLP HTTP to that key's host."""
        from opentelemetry.exporter.otlp.proto.http.trace_exporter import (
            OTLPSpanExporter,
        )
        from opentelemetry.sdk.trace.export import BatchSpanProcessor

        with self._clean_env():
            logger = LangfuseOtelLogger()
            assert logger.OTEL_EXPORTER == "console"

            tracer = logger.get_tracer_to_use_for_request(
                {"standard_callback_dynamic_params": self._dynamic_params()}
            )

        assert tracer is not logger.tracer
        assert len(logger._tracer_provider_cache) == 1

        provider = next(iter(logger._tracer_provider_cache.values())).provider
        span_processors = provider._active_span_processor._span_processors
        assert len(span_processors) == 1
        assert isinstance(span_processors[0], BatchSpanProcessor)

        exporter = span_processors[0].span_exporter
        assert isinstance(exporter, OTLPSpanExporter)
        import base64

        expected_auth = base64.b64encode(b"key_public:key_secret").decode()
        assert _exported_auth_headers(exporter) == [
            (
                "https://langfuse.example.com/api/public/otel/v1/traces",
                {"Authorization": f"Basic {expected_auth}", "x-langfuse-ingestion-version": "4"},
            )
        ]

    def test_key_dynamic_params_reuse_cached_provider(self):
        with self._clean_env():
            logger = LangfuseOtelLogger()
            kwargs = {"standard_callback_dynamic_params": self._dynamic_params()}
            logger.get_tracer_to_use_for_request(kwargs)
            logger.get_tracer_to_use_for_request(kwargs)

        assert len(logger._tracer_provider_cache) == 1

    def test_no_dynamic_params_keeps_default_tracer(self):
        with self._clean_env():
            logger = LangfuseOtelLogger()
            tracer = logger.get_tracer_to_use_for_request({})

        assert tracer is logger.tracer
        assert logger._tracer_provider_cache == {}

    def test_key_credentials_never_passed_to_debug_logger(self):
        """The span-processor debug logs must receive a redacted header value, so the
        key-scoped Langfuse secret never enters a log record regardless of downstream
        handler configuration, while the exporter still gets the real header."""
        import base64

        from opentelemetry.exporter.otlp.proto.http.trace_exporter import (
            OTLPSpanExporter,
        )

        from litellm.integrations import opentelemetry as otel_module

        secret = base64.b64encode(b"key_public:key_secret").decode()

        recorded_arguments = []

        def _spy(message, *args, **kwargs):
            recorded_arguments.append(" ".join(str(part) for part in (message, *args)))

        with self._clean_env():
            logger = LangfuseOtelLogger()
            with patch.object(otel_module.verbose_logger, "debug", side_effect=_spy):
                logger.get_tracer_to_use_for_request(
                    {"standard_callback_dynamic_params": self._dynamic_params()}
                )

        logged = "\n".join(recorded_arguments)
        assert "initializing span processor" in logged
        assert secret not in logged
        assert f"Basic {secret}" not in logged

        provider = next(iter(logger._tracer_provider_cache.values())).provider
        exporter = provider._active_span_processor._span_processors[0].span_exporter
        assert isinstance(exporter, OTLPSpanExporter)
        assert _exported_auth_headers(exporter) == [
            (
                "https://langfuse.example.com/api/public/otel/v1/traces",
                {"Authorization": f"Basic {secret}", "x-langfuse-ingestion-version": "4"},
            )
        ]


class TestLangfuseOtelResponsesAPI:
    """Test suite for Langfuse OTEL integration with ResponsesAPI"""

    def test_langfuse_otel_with_responses_api(self):
        """Test that Langfuse OTEL logger works with ResponsesAPI responses and logs metadata."""
        # Create a mock ResponsesAPIResponse
        mock_response = ResponsesAPIResponse(
            id="response-123",
            created_at=1234567890,
            output=[
                {
                    "type": "message",
                    "content": [{"type": "text", "text": "Hello from responses API"}],
                }
            ],
            parallel_tool_calls=False,
            tool_choice="auto",
            tools=[],
            top_p=1.0,
        )

        # Create kwargs with metadata that should be logged
        test_metadata = {
            "user_id": "test123",
            "session_id": "abc456",
            "custom_field": "test_value",
            "generation_name": "responses_test_generation",
            "trace_name": "responses_api_trace",
        }

        kwargs = {
            "call_type": "responses",
            "messages": [{"role": "user", "content": "Hello"}],
            "model": "gpt-4o",
            "optional_params": {},
            "litellm_params": {"metadata": test_metadata},
        }

        mock_span = MagicMock()

        from litellm.integrations.langfuse.langfuse_otel_attributes import (
            LangfuseLLMObsOTELAttributes,
        )

        with patch(
            "litellm.integrations.arize._utils.set_attributes"
        ) as mock_set_attributes:
            with patch(
                "litellm.integrations.arize._utils.safe_set_attribute"
            ) as mock_safe_set_attribute:
                logger = LangfuseOtelLogger()
                logger.set_langfuse_otel_attributes(mock_span, kwargs, mock_response)

                # Verify that set_attributes was called for general attributes
                mock_set_attributes.assert_called_once_with(
                    mock_span, kwargs, mock_response, LangfuseLLMObsOTELAttributes
                )

                # Verify that Langfuse-specific attributes were set
                mock_safe_set_attribute.assert_any_call(
                    mock_span, "langfuse.generation.name", "responses_test_generation"
                )
                mock_safe_set_attribute.assert_any_call(
                    mock_span, "langfuse.trace.name", "responses_api_trace"
                )

    def test_responses_api_metadata_extraction(self):
        """Test that metadata is correctly extracted from ResponsesAPI kwargs."""
        # Clean up any existing module mocks
        import sys

        if "litellm.integrations.langfuse.langfuse" in sys.modules:
            sys.modules["litellm.integrations.langfuse.langfuse"]

        test_metadata = {
            "user_id": "responses_user_123",
            "session_id": "responses_session_456",
            "custom_metadata": {"key": "value"},
            "generation_name": "responses_generation",
            "trace_id": "custom_trace_id",
        }

        kwargs = {
            "call_type": "responses",
            "model": "gpt-4o",
            "litellm_params": {"metadata": test_metadata},
        }

        extracted_metadata = LangfuseOtelLogger._extract_langfuse_metadata(kwargs)

        # Verify all expected metadata was extracted (may have additional fields from header enrichment)
        for key, value in test_metadata.items():
            assert extracted_metadata[key] == value

        assert extracted_metadata["user_id"] == "responses_user_123"
        assert extracted_metadata["generation_name"] == "responses_generation"
        assert extracted_metadata["trace_id"] == "custom_trace_id"

    def test_responses_api_langfuse_specific_attributes(self):
        """Test that ResponsesAPI metadata maps correctly to Langfuse OTEL attributes."""
        metadata = {
            "generation_name": "responses_gen",
            "generation_id": "resp_gen_123",
            "trace_name": "responses_trace",
            "trace_user_id": "resp_user_456",
            "session_id": "resp_session_789",
            "tags": ["responses", "api", "test"],
            "trace_metadata": {"source": "responses_api", "version": "1.0"},
        }

        kwargs = {"call_type": "responses", "litellm_params": {"metadata": metadata}}

        mock_span = MagicMock()

        with patch(
            "litellm.integrations.arize._utils.safe_set_attribute"
        ) as mock_safe_set_attribute:
            LangfuseOtelLogger._set_langfuse_specific_attributes(mock_span, kwargs, {})

            # Verify specific attributes were set
            from litellm.types.integrations.langfuse_otel import LangfuseSpanAttributes

            expected_calls = [
                (
                    mock_span,
                    LangfuseSpanAttributes.GENERATION_NAME.value,
                    "responses_gen",
                ),
                (mock_span, LangfuseSpanAttributes.GENERATION_ID.value, "resp_gen_123"),
                (mock_span, LangfuseSpanAttributes.TRACE_NAME.value, "responses_trace"),
                (
                    mock_span,
                    LangfuseSpanAttributes.TRACE_USER_ID.value,
                    "resp_user_456",
                ),
                (
                    mock_span,
                    LangfuseSpanAttributes.SESSION_ID.value,
                    "resp_session_789",
                ),
                (
                    mock_span,
                    LangfuseSpanAttributes.TAGS.value,
                    json.dumps(["responses", "api", "test"]),
                ),
                (
                    mock_span,
                    LangfuseSpanAttributes.TRACE_METADATA.value,
                    json.dumps({"source": "responses_api", "version": "1.0"}),
                ),
            ]

            for expected_call in expected_calls:
                mock_safe_set_attribute.assert_any_call(*expected_call)

    def test_responses_api_with_output(self):
        """Test Langfuse OTEL logger with Responses API output (reasoning + message)."""
        from openai.types.responses import (
            ResponseReasoningItem,
            ResponseOutputMessage,
            ResponseOutputText,
        )
        from openai.types.responses.response_reasoning_item import Summary
        from litellm.types.integrations.langfuse_otel import LangfuseSpanAttributes

        # Create Responses API response with reasoning and message
        response_obj = ResponsesAPIResponse(
            id="response-456",
            created_at=1625247600,
            output=[
                ResponseReasoningItem(
                    id="reasoning-001",
                    type="reasoning",
                    summary=[
                        Summary(
                            text="Let me analyze this problem step by step...",
                            type="summary_text",
                        )
                    ],
                ),
                ResponseOutputMessage(
                    id="msg-001",
                    type="message",
                    role="assistant",
                    status="completed",
                    content=[
                        ResponseOutputText(
                            annotations=[],
                            text="The weather in San Francisco is sunny, 20°C.",
                            type="output_text",
                        )
                    ],
                ),
            ],
        )

        kwargs = {
            "call_type": "responses",
            "messages": [
                {"role": "user", "content": "What's the weather in San Francisco?"}
            ],
            "model": "gpt-4o",
            "optional_params": {},
        }

        mock_span = MagicMock()

        with patch(
            "litellm.integrations.arize._utils.safe_set_attribute"
        ) as mock_safe_set_attribute:
            LangfuseOtelLogger._set_langfuse_specific_attributes(
                mock_span, kwargs, response_obj
            )

            # Verify observation output was set
            output_calls = [
                call
                for call in mock_safe_set_attribute.call_args_list
                if call.args[1] == LangfuseSpanAttributes.OBSERVATION_OUTPUT.value
            ]

            assert len(output_calls) > 0, "observation.output should be set"
            output_json = output_calls[0].args[2]
            output_data = json.loads(output_json)

            # Verify output contains reasoning and message
            assert isinstance(output_data, list)
            assert len(output_data) == 2

            # Verify reasoning summary
            assert output_data[0]["role"] == "reasoning_summary"
            assert (
                output_data[0]["content"]
                == "Let me analyze this problem step by step..."
            )

            # Verify message
            assert output_data[1]["role"] == "assistant"
            assert (
                output_data[1]["content"]
                == "The weather in San Francisco is sunny, 20°C."
            )

    def test_responses_api_with_function_calls(self):
        """Test Langfuse OTEL logger with Responses API function_call output."""
        from litellm.types.integrations.langfuse_otel import LangfuseSpanAttributes
        from openai.types.responses import ResponseFunctionToolCall

        # Create Responses API response with function call
        response_obj = ResponsesAPIResponse(
            id="response-789",
            created_at=1625247700,
            output=[
                ResponseFunctionToolCall(
                    id="fc-123",
                    type="function_call",
                    name="get_weather",
                    call_id="call-abc",
                    arguments='{"location": "San Francisco", "unit": "celsius"}',
                    status="completed",
                )
            ],
        )

        kwargs = {
            "call_type": "responses",
            "messages": [
                {"role": "user", "content": "What's the weather in San Francisco?"}
            ],
            "model": "gpt-4o",
            "optional_params": {},
        }

        mock_span = MagicMock()

        with patch(
            "litellm.integrations.arize._utils.safe_set_attribute"
        ) as mock_safe_set_attribute:
            LangfuseOtelLogger._set_langfuse_specific_attributes(
                mock_span, kwargs, response_obj
            )

            # Verify observation output was set
            output_calls = [
                call
                for call in mock_safe_set_attribute.call_args_list
                if call.args[1] == LangfuseSpanAttributes.OBSERVATION_OUTPUT.value
            ]

            assert len(output_calls) > 0, "observation.output should be set"
            output_json = output_calls[0].args[2]
            output_data = json.loads(output_json)

            # Verify output contains function call
            assert isinstance(output_data, list)
            assert len(output_data) == 1

            # Verify function call details
            assert output_data[0]["type"] == "function_call"
            assert output_data[0]["id"] == "fc-123"
            assert output_data[0]["name"] == "get_weather"
            assert output_data[0]["call_id"] == "call-abc"
            assert output_data[0]["arguments"]["location"] == "San Francisco"
            assert output_data[0]["arguments"]["unit"] == "celsius"

    def test_responses_api_function_call_with_redacted_arguments(self):
        """Sentinel arguments (invalid JSON) must not kill the whole observation output."""
        from openai.types.responses import ResponseFunctionToolCall

        from litellm.types.integrations.langfuse_otel import LangfuseSpanAttributes

        response_obj = ResponsesAPIResponse(
            id="response-redacted",
            created_at=1625247700,
            output=[
                ResponseFunctionToolCall(
                    id="fc-redacted",
                    type="function_call",
                    name="get_weather",
                    call_id="call-redacted",
                    arguments="redacted-by-litellm",
                    status="completed",
                )
            ],
        )

        kwargs = {
            "call_type": "responses",
            "messages": [{"role": "user", "content": "What's the weather?"}],
            "model": "gpt-4o",
            "optional_params": {},
        }

        mock_span = MagicMock()

        with patch(  # test-quality-ok: the span attribute sink is the observable boundary; sibling tests in this class stub the same seam
            "litellm.integrations.arize._utils.safe_set_attribute"
        ) as mock_safe_set_attribute:
            LangfuseOtelLogger._set_langfuse_specific_attributes(mock_span, kwargs, response_obj)

            output_calls = [
                call
                for call in mock_safe_set_attribute.call_args_list
                if call.args[1] == LangfuseSpanAttributes.OBSERVATION_OUTPUT.value
            ]

            assert len(output_calls) > 0, "observation.output should still be set"
            output_data = json.loads(output_calls[0].args[2])
            assert output_data[0]["name"] == "get_weather"
            assert output_data[0]["arguments"] == {}


LANGFUSE_ENV_ONLY: Final = {
    key: value
    for key, value in os.environ.items()
    if not key.startswith(("LANGFUSE_", "OTEL_"))
}


def _decoded_basic_auth(header: str) -> str:
    assert header.startswith("Basic ")
    return base64.b64decode(header.removeprefix("Basic ")).decode()


def test_langfuse_otel_env_config_headers_carry_v4_ingestion_and_basic_auth() -> None:
    with patch.dict(
        os.environ, {**LANGFUSE_ENV_ONLY, "LANGFUSE_PUBLIC_KEY": "pk-lf-123", "LANGFUSE_SECRET_KEY": "sk-lf-123"}, clear=True
    ):
        logger: Final = LangfuseOtelLogger()
    headers: Final = OpenTelemetry._get_headers_dictionary(logger.config.headers)
    assert headers["x-langfuse-ingestion-version"] == "4"
    assert _decoded_basic_auth(headers["Authorization"]) == "pk-lf-123:sk-lf-123"


def test_langfuse_otel_dynamic_headers_carry_v4_ingestion_and_basic_auth() -> None:
    with patch.dict(os.environ, LANGFUSE_ENV_ONLY, clear=True):
        logger: Final = LangfuseOtelLogger()
    headers: Final = logger.construct_dynamic_otel_headers(
        StandardCallbackDynamicParams(langfuse_public_key="pk-lf-dynamic", langfuse_secret_key="sk-lf-dynamic")
    )
    assert headers is not None
    assert headers["x-langfuse-ingestion-version"] == "4"
    assert _decoded_basic_auth(headers["Authorization"]) == "pk-lf-dynamic:sk-lf-dynamic"


def test_langfuse_otel_does_not_start_proxy_request_span() -> None:
    langfuse_provider: Final = TracerProvider()
    generic_provider: Final = TracerProvider()
    with patch.dict(os.environ, LANGFUSE_ENV_ONLY, clear=True):
        langfuse_logger: Final = LangfuseOtelLogger(tracer_provider=langfuse_provider)
        generic_logger: Final = OpenTelemetry(
            config=OpenTelemetryConfig(exporter="console", skip_set_global=True), tracer_provider=generic_provider
        )
    started_at: Final = datetime(2026, 1, 1, tzinfo=timezone.utc)
    request_headers: Final = {"Authorization": "Bearer test"}
    try:
        assert (
            langfuse_logger.create_litellm_proxy_request_started_span(start_time=started_at, headers=request_headers)
            is None
        )
        assert (
            generic_logger.create_litellm_proxy_request_started_span(start_time=started_at, headers=request_headers)
            is not None
        )
    finally:
        langfuse_provider.shutdown()
        generic_provider.shutdown()