fix(langfuse): rebuild the cached bundle when mock mode or sample rate changes

The SDK keys resource bundles on the public key alone, so a bundle built with the discarding exporter for LANGFUSE_MOCK, or with an earlier LANGFUSE_SAMPLE_RATE, was handed back to a client that asked for a live exporter or a different rate. Compare both when deciding whether the cached bundle is still valid

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

litellm/integrations/langfuse/langfuse_sdk.py

@@ -487,7 +487,12 @@ def _run_teardowns(state: _LangfuseLifecycleState, clients: tuple[Langfuse, ...]
 
 
 def _evict_if_stale_locked(
-    *, public_key: object, secret_key: object, base_url: object
+    *,
+    public_key: object,
+    secret_key: object,
+    base_url: object,
+    mock_mode: bool | None = None,
+    sample_rate: float | None = None,
 ) -> LangfuseResourceManager | None:
     """Assumes ``LangfuseResourceManager._lock`` is held; returns the still-valid bundle, evicting a stale one."""
     if not public_key:
@@ -495,7 +500,16 @@ def _evict_if_stale_locked(
     cached: Final = LangfuseResourceManager._instances.get(public_key)  # pyright: ignore[reportPrivateUsage]  # registry has no public accessor
     if cached is None:
         return None
-    if getattr(cached, "secret_key", None) == secret_key and getattr(cached, "base_url", None) == base_url:
+    same_exporter_kind: Final = mock_mode is None or (
+        isinstance(getattr(cached, "span_exporter", None), DiscardingSpanExporter) == mock_mode
+    )
+    same_sample_rate: Final = sample_rate is None or getattr(cached, "sample_rate", None) == sample_rate
+    if (
+        getattr(cached, "secret_key", None) == secret_key
+        and getattr(cached, "base_url", None) == base_url
+        and same_exporter_kind
+        and same_sample_rate
+    ):
         return cached
     LangfuseResourceManager._instances.pop(public_key, None)  # pyright: ignore[reportPrivateUsage]  # registry has no public accessor
     return None
@@ -611,6 +625,8 @@ def acquire_langfuse_client(
             public_key=public_key,
             secret_key=parameters.get("secret_key"),
             base_url=parameters.get("base_url"),
+            mock_mode=mock_mode,
+            sample_rate=sample_rate,
         )
         client: Final = Langfuse(
             **parameters,  # pyright: ignore[reportArgumentType]  # kwargs-ok: dict mirrors the typed ctor, values resolved by the callers

tests/test_litellm/integrations/langfuse/test_langfuse_sdk.py

@@ -1230,3 +1230,38 @@ def test_second_client_on_the_same_key_does_not_build_another_provider():
         assert len(_litellm_built_providers) == providers_after_first
     finally:
         LangfuseResourceManager._instances.pop(pk, None)
+
+
+def test_leaving_mock_mode_on_the_same_key_stops_using_the_discarding_exporter():
+    from litellm.integrations.langfuse.langfuse_sdk import DiscardingSpanExporter
+
+    pk = "pk-mock-to-live-test"
+    LangfuseResourceManager._instances.pop(pk, None)
+    parameters = {"public_key": pk, "secret_key": "sk-live", "base_url": "http://127.0.0.1:1"}
+    try:
+        mocked = acquire_langfuse_client(parameters=parameters, environment=None, release=None, mock_mode=True)
+        assert isinstance(mocked._resources.span_exporter, DiscardingSpanExporter)
+
+        live = acquire_langfuse_client(parameters=parameters, environment=None, release=None, mock_mode=False)
+        assert live._resources is not mocked._resources
+        assert not isinstance(live._resources.span_exporter, DiscardingSpanExporter)
+        assert LangfuseResourceManager._instances.get(pk) is live._resources
+    finally:
+        LangfuseResourceManager._instances.pop(pk, None)
+
+
+def test_a_changed_sample_rate_on_the_same_key_rebuilds_the_bundle(monkeypatch: pytest.MonkeyPatch):
+    pk = "pk-resample-test"
+    LangfuseResourceManager._instances.pop(pk, None)
+    parameters = {"public_key": pk, "secret_key": "sk-resample", "base_url": "http://127.0.0.1:1"}
+    try:
+        monkeypatch.setenv("LANGFUSE_SAMPLE_RATE", "0.25")
+        quarter = acquire_langfuse_client(parameters=parameters, environment=None, release=None, mock_mode=True)
+        monkeypatch.setenv("LANGFUSE_SAMPLE_RATE", "1")
+        full = acquire_langfuse_client(parameters=parameters, environment=None, release=None, mock_mode=True)
+
+        assert quarter._resources.sample_rate == 0.25
+        assert full._resources is not quarter._resources
+        assert full._resources.sample_rate == 1.0
+    finally:
+        LangfuseResourceManager._instances.pop(pk, None)