push

Author: hassiebp

langfuse/_client/span.py

@@ -861,7 +861,7 @@ def start_observation(
         self,
         *,
         name: str,
-        as_type: ObservationTypeLiteral,
+        as_type: ObservationTypeLiteral = "span",
         input: Optional[Any] = None,
         output: Optional[Any] = None,
         metadata: Optional[Any] = None,
@@ -1109,7 +1109,7 @@ def start_as_current_observation(  # type: ignore[misc]
         self,
         *,
         name: str,
-        as_type: ObservationTypeLiteralNoEvent,
+        as_type: ObservationTypeLiteralNoEvent = "span",
         input: Optional[Any] = None,
         output: Optional[Any] = None,
         metadata: Optional[Any] = None,

tests/test_core_sdk.py

@@ -211,10 +211,15 @@ def test_create_boolean_score():
     # Retrieve and verify
     trace = api_wrapper.get_trace(trace_id)
 
-    assert trace["scores"][0]["id"] == score_id
-    assert trace["scores"][0]["dataType"] == "BOOLEAN"
-    assert trace["scores"][0]["value"] == 1
-    assert trace["scores"][0]["stringValue"] == "True"
+    # Find the score we created by name
+    created_score = next(
+        (s for s in trace["scores"] if s["name"] == "this-is-a-score"), None
+    )
+    assert created_score is not None, "Score not found in trace"
+    assert created_score["id"] == score_id
+    assert created_score["dataType"] == "BOOLEAN"
+    assert created_score["value"] == 1
+    assert created_score["stringValue"] == "True"
 
 
 def test_create_categorical_score():
@@ -260,10 +265,15 @@ def test_create_categorical_score():
     # Retrieve and verify
     trace = api_wrapper.get_trace(trace_id)
 
-    assert trace["scores"][0]["id"] == score_id
-    assert trace["scores"][0]["dataType"] == "CATEGORICAL"
-    assert trace["scores"][0]["value"] == 0
-    assert trace["scores"][0]["stringValue"] == "high score"
+    # Find the score we created by name
+    created_score = next(
+        (s for s in trace["scores"] if s["name"] == "this-is-a-score"), None
+    )
+    assert created_score is not None, "Score not found in trace"
+    assert created_score["id"] == score_id
+    assert created_score["dataType"] == "CATEGORICAL"
+    assert created_score["value"] == 0
+    assert created_score["stringValue"] == "high score"
 
 
 def test_create_score_with_custom_timestamp():
@@ -302,14 +312,19 @@ def test_create_score_with_custom_timestamp():
     # Retrieve and verify
     trace = api_wrapper.get_trace(trace_id)
 
-    assert trace["scores"][0]["id"] == score_id
-    assert trace["scores"][0]["dataType"] == "NUMERIC"
-    assert trace["scores"][0]["value"] == 0.85
+    # Find the score we created by name
+    created_score = next(
+        (s for s in trace["scores"] if s["name"] == "custom-timestamp-score"), None
+    )
+    assert created_score is not None, "Score not found in trace"
+    assert created_score["id"] == score_id
+    assert created_score["dataType"] == "NUMERIC"
+    assert created_score["value"] == 0.85
 
     # Verify timestamp is close to our custom timestamp
     # Parse the timestamp from the API response
     response_timestamp = datetime.fromisoformat(
-        trace["scores"][0]["timestamp"].replace("Z", "+00:00")
+        created_score["timestamp"].replace("Z", "+00:00")
     )
 
     # Check that the timestamps are within 1 second of each other
@@ -1995,7 +2010,7 @@ def test_generate_trace_id():
 def test_generate_trace_url_client_disabled():
     langfuse = Langfuse(tracing_enabled=False)
 
-    with langfuse.start_as_current_span(
+    with langfuse.start_as_current_observation(
         name="test-span",
     ):
         # The trace URL should be None because the client is disabled

tests/test_langchain.py

@@ -176,7 +176,7 @@ def test_callback_simple_openai():
         llm.invoke(text, config={"callbacks": [handler], "run_name": test_name})
 
         # Ensure data is flushed to API
-    handler.client.flush()
+    handler._langfuse_client.flush()
     sleep(2)
 
     # Retrieve trace
@@ -222,7 +222,7 @@ def test_callback_multiple_invocations_on_different_traces():
         handler2 = CallbackHandler()
         llm.invoke(text, config={"callbacks": [handler2], "run_name": test_name_2})
 
-    handler1.client.flush()
+    handler1._langfuse_client.flush()
 
     # Ensure data is flushed to API
     sleep(2)
@@ -281,7 +281,7 @@ def test_openai_instruct_usage():
         ]
         runnable_chain.batch(input_list)
 
-    lf_handler.client.flush()
+    lf_handler._langfuse_client.flush()
 
     observations = get_api().trace.get(trace_id).observations
 
@@ -451,7 +451,7 @@ def test_link_langfuse_prompts_invoke():
             },
         )
 
-    langfuse_handler.client.flush()
+    langfuse_handler._langfuse_client.flush()
     sleep(2)
 
     trace = get_api().trace.get(trace_id=trace_id)
@@ -538,7 +538,7 @@ def test_link_langfuse_prompts_stream():
         for chunk in stream:
             output += chunk
 
-    langfuse_handler.client.flush()
+    langfuse_handler._langfuse_client.flush()
     sleep(2)
 
     trace = get_api().trace.get(trace_id=trace_id)
@@ -624,7 +624,7 @@ def test_link_langfuse_prompts_batch():
             },
         )
 
-    langfuse_handler.client.flush()
+    langfuse_handler._langfuse_client.flush()
 
     traces = get_api().trace.list(name=trace_name).data
 
@@ -747,13 +747,13 @@ class GetWeather(BaseModel):
         }
     ]
 
-    with handler.client.start_as_current_observation(
+    with handler._langfuse_client.start_as_current_observation(
         name="test_callback_openai_functions_with_tools"
     ) as span:
         trace_id = span.trace_id
         llm.bind_tools([address_tool, weather_tool]).invoke(messages)
 
-    handler.client.flush()
+    handler._langfuse_client.flush()
 
     trace = get_api().trace.get(trace_id=trace_id)
 
@@ -846,11 +846,13 @@ def test_multimodal():
         ],
     )
 
-    with handler.client.start_as_current_observation(name="test_multimodal") as span:
+    with handler._langfuse_client.start_as_current_observation(
+        name="test_multimodal"
+    ) as span:
         trace_id = span.trace_id
         model.invoke([message], config={"callbacks": [handler]})
 
-    handler.client.flush()
+    handler._langfuse_client.flush()
 
     trace = get_api().trace.get(trace_id=trace_id)
 
@@ -935,14 +937,16 @@ def call_model(state: MessagesState):
     handler = CallbackHandler()
 
     # Use the Runnable
-    with handler.client.start_as_current_observation(name="test_langgraph") as span:
+    with handler._langfuse_client.start_as_current_observation(
+        name="test_langgraph"
+    ) as span:
         trace_id = span.trace_id
         final_state = app.invoke(
             {"messages": [HumanMessage(content="what is the weather in sf")]},
             config={"configurable": {"thread_id": 42}, "callbacks": [handler]},
         )
     print(final_state["messages"][-1].content)
-    handler.client.flush()
+    handler._langfuse_client.flush()
 
     trace = get_api().trace.get(trace_id=trace_id)
 
@@ -975,7 +979,7 @@ def test_cached_token_usage():
     # invoke again to force cached token usage
     chain.invoke({"test_param": "in a funny way"}, config)
 
-    handler.client.flush()
+    handler._langfuse_client.flush()
 
     trace = get_api().trace.get(handler.get_trace_id())
 

tests/test_otel.py

@@ -2452,16 +2452,16 @@ def test_blocked_instrumentation_scopes_export_filtering(
         allowed_tracer = tracer_provider.get_tracer("allowed-library")
 
         # Create spans from each tracer
-        langfuse_span = langfuse_tracer.start_observation("langfuse-span")
+        langfuse_span = langfuse_tracer.start_span("langfuse-span")
         langfuse_span.end()
 
-        openai_span = openai_tracer.start_observation("openai-span")
+        openai_span = openai_tracer.start_span("openai-span")
         openai_span.end()
 
-        anthropic_span = anthropic_tracer.start_observation("anthropic-span")
+        anthropic_span = anthropic_tracer.start_span("anthropic-span")
         anthropic_span.end()
 
-        allowed_span = allowed_tracer.start_observation("allowed-span")
+        allowed_span = allowed_tracer.start_span("allowed-span")
         allowed_span.end()
 
         # Force flush to ensure all spans are processed
@@ -2515,13 +2515,13 @@ def test_no_blocked_scopes_allows_all_exports(
         anthropic_tracer = tracer_provider.get_tracer("anthropic")
 
         # Create spans from each tracer
-        langfuse_span = langfuse_tracer.start_observation("langfuse-span")
+        langfuse_span = langfuse_tracer.start_span("langfuse-span")
         langfuse_span.end()
 
-        openai_span = openai_tracer.start_observation("openai-span")
+        openai_span = openai_tracer.start_span("openai-span")
         openai_span.end()
 
-        anthropic_span = anthropic_tracer.start_observation("anthropic-span")
+        anthropic_span = anthropic_tracer.start_span("anthropic-span")
         anthropic_span.end()
 
         # Force flush
@@ -2559,10 +2559,10 @@ def test_none_blocked_scopes_allows_all_exports(
         openai_tracer = tracer_provider.get_tracer("openai")
 
         # Create spans from each tracer
-        langfuse_span = langfuse_tracer.start_observation("langfuse-span")
+        langfuse_span = langfuse_tracer.start_span("langfuse-span")
         langfuse_span.end()
 
-        openai_span = openai_tracer.start_observation("openai-span")
+        openai_span = openai_tracer.start_span("openai-span")
         openai_span.end()
 
         # Force flush
@@ -2597,10 +2597,10 @@ def test_blocking_langfuse_sdk_scope_export(self, instrumentation_filtering_setu
         other_tracer = tracer_provider.get_tracer("other-library")
 
         # Create spans
-        langfuse_span = langfuse_tracer.start_observation("langfuse-span")
+        langfuse_span = langfuse_tracer.start_span("langfuse-span")
         langfuse_span.end()
 
-        other_span = other_tracer.start_observation("other-span")
+        other_span = other_tracer.start_span("other-span")
         other_span.end()
 
         # Force flush

tests/test_propagate_attributes.py

@@ -1180,9 +1180,7 @@ def test_different_tracer_spans_get_attributes(
                 langfuse_span.end()
 
                 # Create span with different tracer
-                with other_tracer.start_as_current_observation(
-                    name="other-library-span"
-                ):
+                with other_tracer.start_as_current_span(name="other-library-span"):
                     pass
 
         # Verify both spans have the propagated attributes
@@ -1222,8 +1220,8 @@ def test_nested_spans_from_multiple_tracers(
                 user_id="user_123", metadata={"experiment": "cross_tracer"}
             ):
                 # Create nested spans from different tracers
-                with tracer_a.start_as_current_observation(name="library-a-span"):
-                    with tracer_b.start_as_current_observation(name="library-b-span"):
+                with tracer_a.start_as_current_span(name="library-a-span"):
+                    with tracer_b.start_as_current_span(name="library-b-span"):
                         langfuse_leaf = langfuse_client.start_observation(
                             name="langfuse-leaf"
                         )
@@ -1251,13 +1249,13 @@ def test_other_tracer_span_before_propagate_context(
 
         with langfuse_client.start_as_current_observation(name="root"):
             # Create span BEFORE propagate_attributes
-            with other_tracer.start_as_current_observation(name="span-before"):
+            with other_tracer.start_as_current_span(name="span-before"):
                 pass
 
             # NOW set attributes
             with propagate_attributes(user_id="user_123"):
                 # Create span AFTER propagate_attributes
-                with other_tracer.start_as_current_observation(name="span-after"):
+                with other_tracer.start_as_current_span(name="span-after"):
                     pass
 
         # Verify: span-before does NOT have user_id, span-after DOES
@@ -1291,9 +1289,7 @@ def test_mixed_tracers_with_metadata(
                 )
                 langfuse_span.end()
 
-                with other_tracer.start_as_current_observation(
-                    name="library-operation"
-                ):
+                with other_tracer.start_as_current_span(name="library-operation"):
                     pass
 
         # Verify both spans have all metadata
@@ -1322,10 +1318,10 @@ def test_propagate_without_langfuse_parent(
         other_tracer = tracer_provider.get_tracer("other-library", "1.0.0")
 
         # Parent span is from different tracer
-        with other_tracer.start_as_current_observation(name="other-parent"):
+        with other_tracer.start_as_current_span(name="other-parent"):
             with propagate_attributes(user_id="user_123", session_id="session_xyz"):
                 # Create children from both tracers
-                with other_tracer.start_as_current_observation(name="other-child"):
+                with other_tracer.start_as_current_span(name="other-child"):
                     pass
 
                 langfuse_child = langfuse_client.start_observation(
@@ -1358,11 +1354,11 @@ def test_attributes_persist_across_tracer_changes(
         with langfuse_client.start_as_current_observation(name="root"):
             with propagate_attributes(user_id="persistent_user"):
                 # Bounce between different tracers
-                with tracer_1.start_as_current_observation(name="step-1"):
+                with tracer_1.start_as_current_span(name="step-1"):
                     pass
 
-                with tracer_2.start_as_current_observation(name="step-2"):
-                    with tracer_3.start_as_current_observation(name="step-3"):
+                with tracer_2.start_as_current_span(name="step-2"):
+                    with tracer_3.start_as_current_span(name="step-3"):
                         pass
 
                 langfuse_span = langfuse_client.start_observation(name="step-4")