langfuse-haystack 0.8.0__tar.gz → 0.9.0__tar.gz

{langfuse_haystack-0.8.0 → langfuse_haystack-0.9.0}/CHANGELOG.md

@@ -1,5 +1,12 @@
 # Changelog
 
+## [integrations/langfuse-v0.8.0] - 2025-01-28
+
+### 🚀 Features
+
+- Add custom Langfuse span handling support (#1313)
+
+
 ## [integrations/langfuse-v0.7.0] - 2025-01-21
 
 ### 🚀 Features
@@ -22,7 +29,7 @@
 
 ### 🌀 Miscellaneous
 
-- Chore: Langfuse - pin `haystack-ai>=2.9.0` and simplify message conversion (#1292)
+- Pin haystack-ai>=2.9.0 and simplify (#1292)
 
 ## [integrations/langfuse-v0.6.2] - 2025-01-02
 
@@ -36,7 +43,7 @@
 
 ### 🌀 Miscellaneous
 
-- Chore: Fix tracing_context_var lint errors (#1220)
+- Fix tracing_context_var lint errors (#1220)
 - Fix messages conversion to OpenAI format (#1272)
 
 ## [integrations/langfuse-v0.6.0] - 2024-11-18
@@ -53,6 +60,25 @@
 
 - Fixed TypeError in LangfuseTrace (#1184)
 
+* Added parent_span functionality in trace method
+
+* solved PR comments
+
+* Readded "end()" for solving Latency issues
+
+* chore: fix ruff linting
+
+* Handle multiple runs
+
+* Fix indentation and span closing
+
+* Fix tests
+
+---------
+
+Co-authored-by: Vladimir Blagojevic <dovlex@gmail.com>
+Co-authored-by: Silvano Cerza <silvanocerza@gmail.com>
+
 ## [integrations/langfuse-v0.5.0] - 2024-10-01
 
 ### 🧹 Chores
@@ -61,7 +87,7 @@
 
 ### 🌀 Miscellaneous
 
-- Fix: Add delay to flush the Langfuse traces (#1091)
+- Add delay to flush the Langfuse traces (#1091)
 - Add invocation_context to identify traces (#1089)
 
 ## [integrations/langfuse-v0.4.0] - 2024-09-17
@@ -93,15 +119,15 @@
 
 ### 🌀 Miscellaneous
 
-- Ci: install `pytest-rerunfailures` where needed; add retry config to `test-cov` script (#845)
-- Chore: Update Langfuse README to avoid common initialization issues (#952)
-- Chore: langfuse - ruff update, don't ruff tests (#992)
+- Install pytest-rerunfailures; change test-cov script (#845)
+- Update Langfuse README to avoid common initialization issues (#952)
+- Ruff update, don't ruff tests (#992)
 
 ## [integrations/langfuse-v0.2.0] - 2024-06-18
 
 ### 🌀 Miscellaneous
 
-- Feat: add support for Azure generators (#815)
+- Add support for Azure generators (#815)
 
 ## [integrations/langfuse-v0.1.0] - 2024-06-13
 
@@ -119,7 +145,7 @@
 
 ### 🌀 Miscellaneous
 
-- Chore: change the pydoc renderer class (#718)
-- Docs: add missing api references (#728)
+- Change the pydoc renderer class (#718)
+- Missing api references (#728)
 
 <!-- generated by git-cliff -->

{langfuse_haystack-0.8.0 → langfuse_haystack-0.9.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: langfuse-haystack
-Version: 0.8.0
+Version: 0.9.0
 Project-URL: Documentation, https://github.com/deepset-ai/haystack-core-integrations/tree/main/integrations/langfuse#readme
 Project-URL: Issues, https://github.com/deepset-ai/haystack-core-integrations/issues
 Project-URL: Source, https://github.com/deepset-ai/haystack-core-integrations/tree/main/integrations/langfuse
@@ -9,14 +9,14 @@ License-Expression: Apache-2.0
 License-File: LICENSE.txt
 Classifier: Development Status :: 4 - Beta
 Classifier: Programming Language :: Python
-Classifier: Programming Language :: Python :: 3.8
 Classifier: Programming Language :: Python :: 3.9
 Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
 Classifier: Programming Language :: Python :: Implementation :: CPython
 Classifier: Programming Language :: Python :: Implementation :: PyPy
-Requires-Python: >=3.8
+Requires-Python: >=3.9
 Requires-Dist: haystack-ai>=2.9.0
 Requires-Dist: langfuse
 Description-Content-Type: text/markdown

{langfuse_haystack-0.8.0 → langfuse_haystack-0.9.0}/pyproject.toml

@@ -7,18 +7,18 @@ name = "langfuse-haystack"
 dynamic = ["version"]
 description = ''
 readme = "README.md"
-requires-python = ">=3.8"
+requires-python = ">=3.9"
 license = "Apache-2.0"
 keywords = []
 authors = [{ name = "deepset GmbH", email = "info@deepset.ai" }]
 classifiers = [
   "Development Status :: 4 - Beta",
   "Programming Language :: Python",
-  "Programming Language :: Python :: 3.8",
   "Programming Language :: Python :: 3.9",
   "Programming Language :: Python :: 3.10",
   "Programming Language :: Python :: 3.11",
   "Programming Language :: Python :: 3.12",
+  "Programming Language :: Python :: 3.13",
   "Programming Language :: Python :: Implementation :: CPython",
   "Programming Language :: Python :: Implementation :: PyPy",
 ]
@@ -60,10 +60,6 @@ cov = ["test-cov", "cov-report"]
 cov-retry = ["test-cov-retry", "cov-report"]
 docs = ["pydoc-markdown pydoc/config.yml"]
 
-
-[[tool.hatch.envs.all.matrix]]
-python = ["3.8", "3.9", "3.10", "3.11", "3.12"]
-
 [tool.hatch.envs.lint]
 installer = "uv"
 detached = true

langfuse_haystack-0.9.0/src/haystack_integrations/components/connectors/__init__.py

@@ -0,0 +1,3 @@
+# SPDX-FileCopyrightText: 2023-present deepset GmbH <info@deepset.ai>
+#
+# SPDX-License-Identifier: Apache-2.0

{langfuse_haystack-0.8.0 → langfuse_haystack-0.9.0}/src/haystack_integrations/components/connectors/langfuse/__init__.py

@@ -1,6 +1,7 @@
 # SPDX-FileCopyrightText: 2023-present deepset GmbH <info@deepset.ai>
 #
 # SPDX-License-Identifier: Apache-2.0
+
 from .langfuse_connector import LangfuseConnector
 
 __all__ = ["LangfuseConnector"]

{langfuse_haystack-0.8.0 → langfuse_haystack-0.9.0}/src/haystack_integrations/components/connectors/langfuse/langfuse_connector.py

@@ -1,12 +1,16 @@
+# SPDX-FileCopyrightText: 2023-present deepset GmbH <info@deepset.ai>
+#
+# SPDX-License-Identifier: Apache-2.0
+
 from typing import Any, Dict, Optional
 
 import httpx
 from haystack import component, default_from_dict, default_to_dict, logging, tracing
 from haystack.utils import Secret, deserialize_secrets_inplace
 from haystack.utils.base_serialization import deserialize_class_instance, serialize_class_instance
+from langfuse import Langfuse
 
 from haystack_integrations.tracing.langfuse import LangfuseTracer, SpanHandler
-from langfuse import Langfuse
 
 logger = logging.getLogger(__name__)
 
@@ -28,7 +32,7 @@ class LangfuseConnector:
     enable Haystack tracing in your pipeline.
 
     Lastly, you may disable flushing the data after each component by setting the `HAYSTACK_LANGFUSE_ENFORCE_FLUSH`
-    environent variable to `false`. By default, the data is flushed after each component and blocks the thread until
+    environment variable to `false`. By default, the data is flushed after each component and blocks the thread until
     the data is sent to Langfuse. **Caution**: Disabling this feature may result in data loss if the program crashes
     before the data is sent to Langfuse. Make sure you will call langfuse.flush() explicitly before the program exits.
     E.g. by using tracer.actual_tracer.flush():
@@ -159,7 +163,7 @@ class LangfuseConnector:
         )
         tracing.enable_tracing(self.tracer)
 
-    @component.output_types(name=str, trace_url=str)
+    @component.output_types(name=str, trace_url=str, trace_id=str)
     def run(self, invocation_context: Optional[Dict[str, Any]] = None):
         """
         Runs the LangfuseConnector component.
@@ -176,7 +180,7 @@ class LangfuseConnector:
             "Langfuse tracer invoked with the following context: '{invocation_context}'",
             invocation_context=invocation_context,
         )
-        return {"name": self.name, "trace_url": self.tracer.get_trace_url()}
+        return {"name": self.name, "trace_url": self.tracer.get_trace_url(), "trace_id": self.tracer.get_trace_id()}
 
     def to_dict(self) -> Dict[str, Any]:
         """

{langfuse_haystack-0.8.0 → langfuse_haystack-0.9.0}/src/haystack_integrations/tracing/langfuse/__init__.py

@@ -1,6 +1,7 @@
 # SPDX-FileCopyrightText: 2024-present deepset GmbH <info@deepset.ai>
 #
 # SPDX-License-Identifier: Apache-2.0
+
 from .tracer import DefaultSpanHandler, LangfuseSpan, LangfuseTracer, SpanContext, SpanHandler
 
 __all__ = ["DefaultSpanHandler", "LangfuseSpan", "LangfuseTracer", "SpanContext", "SpanHandler"]

{langfuse_haystack-0.8.0 → langfuse_haystack-0.9.0}/src/haystack_integrations/tracing/langfuse/tracer.py

@@ -1,3 +1,7 @@
+# SPDX-FileCopyrightText: 2023-present deepset GmbH <info@deepset.ai>
+#
+# SPDX-License-Identifier: Apache-2.0
+
 import contextlib
 import os
 from abc import ABC, abstractmethod
@@ -43,6 +47,8 @@ _ALL_SUPPORTED_GENERATORS = _SUPPORTED_GENERATORS + _SUPPORTED_CHAT_GENERATORS
 
 # These are the keys used by Haystack for traces and span.
 # We keep them here to avoid making typos when using them.
+_PIPELINE_INPUT_KEY = "haystack.pipeline.input_data"
+_PIPELINE_OUTPUT_KEY = "haystack.pipeline.output_data"
 _PIPELINE_RUN_KEY = "haystack.pipeline.run"
 _COMPONENT_NAME_KEY = "haystack.component.name"
 _COMPONENT_TYPE_KEY = "haystack.component.type"
@@ -114,6 +120,14 @@ class LangfuseSpan(Span):
         """
         return self._span
 
+    def get_data(self) -> Dict[str, Any]:
+        """
+        Return the data associated with the span.
+
+        :return: The data associated with the span.
+        """
+        return self._data
+
     def get_correlation_data_for_logs(self) -> Dict[str, Any]:
         return {}
 
@@ -209,7 +223,7 @@ class SpanHandler(ABC):
         """
         Process a span after component execution by attaching metadata and metrics.
 
-        This method is called after the component yields its span, allowing you to:
+        This method is called after the component or pipeline yields its span, allowing you to:
         - Extract and attach token usage statistics
         - Add model information
         - Record timing data (e.g., time-to-first-token)
@@ -234,9 +248,10 @@ class DefaultSpanHandler(SpanHandler):
     """DefaultSpanHandler provides the default Langfuse tracing behavior for Haystack."""
 
     def create_span(self, context: SpanContext) -> LangfuseSpan:
-        message = "Tracer is not initialized"
         if self.tracer is None:
+            message = "Tracer is not initialized"
             raise RuntimeError(message)
+
         tracing_ctx = tracing_context_var.get({})
         if not context.parent_span:
             if context.operation_name != _PIPELINE_RUN_KEY:
@@ -262,13 +277,20 @@ class DefaultSpanHandler(SpanHandler):
             return LangfuseSpan(context.parent_span.raw_span().span(name=context.name))
 
     def handle(self, span: LangfuseSpan, component_type: Optional[str]) -> None:
+        # If the span is at the pipeline level, we add input and output keys to the span
+        at_pipeline_level = span.get_data().get(_PIPELINE_INPUT_KEY) is not None
+        if at_pipeline_level:
+            span.raw_span().update(
+                input=span.get_data().get(_PIPELINE_INPUT_KEY), output=span.get_data().get(_PIPELINE_OUTPUT_KEY)
+            )
+
         if component_type in _SUPPORTED_GENERATORS:
-            meta = span._data.get(_COMPONENT_OUTPUT_KEY, {}).get("meta")
+            meta = span.get_data().get(_COMPONENT_OUTPUT_KEY, {}).get("meta")
             if meta:
-                m = meta[0]
-                span._span.update(usage=m.get("usage") or None, model=m.get("model"))
-        elif component_type in _SUPPORTED_CHAT_GENERATORS:
-            replies = span._data.get(_COMPONENT_OUTPUT_KEY, {}).get("replies")
+                span.raw_span().update(usage=meta[0].get("usage") or None, model=meta[0].get("model"))
+
+        if component_type in _SUPPORTED_CHAT_GENERATORS:
+            replies = span.get_data().get(_COMPONENT_OUTPUT_KEY, {}).get("replies")
             if replies:
                 meta = replies[0].meta
                 completion_start_time = meta.get("completion_start_time")
@@ -278,7 +300,7 @@ class DefaultSpanHandler(SpanHandler):
                     except ValueError:
                         logger.error(f"Failed to parse completion_start_time: {completion_start_time}")
                         completion_start_time = None
-                span._span.update(
+                span.raw_span().update(
                     usage=meta.get("usage") or None,
                     model=meta.get("model"),
                     completion_start_time=completion_start_time,
@@ -380,3 +402,10 @@ class LangfuseTracer(Tracer):
         :return: The URL to the tracing data.
         """
         return self._tracer.get_trace_url()
+
+    def get_trace_id(self) -> str:
+        """
+        Return the trace ID.
+        :return: The trace ID.
+        """
+        return self._tracer.get_trace_id()

{langfuse_haystack-0.8.0 → langfuse_haystack-0.9.0}/tests/test_langfuse_span.py

@@ -1,3 +1,7 @@
+# SPDX-FileCopyrightText: 2023-present deepset GmbH <info@deepset.ai>
+#
+# SPDX-License-Identifier: Apache-2.0
+
 import os
 
 os.environ["HAYSTACK_CONTENT_TRACING_ENABLED"] = "true"

{langfuse_haystack-0.8.0 → langfuse_haystack-0.9.0}/tests/test_tracer.py

@@ -1,3 +1,7 @@
+# SPDX-FileCopyrightText: 2023-present deepset GmbH <info@deepset.ai>
+#
+# SPDX-License-Identifier: Apache-2.0
+
 import datetime
 import logging
 import sys

{langfuse_haystack-0.8.0 → langfuse_haystack-0.9.0}/tests/test_tracing.py

@@ -1,3 +1,7 @@
+# SPDX-FileCopyrightText: 2023-present deepset GmbH <info@deepset.ai>
+#
+# SPDX-License-Identifier: Apache-2.0
+
 import os
 import time
 from urllib.parse import urlparse
@@ -11,7 +15,6 @@ from haystack.components.generators.chat import OpenAIChatGenerator
 from haystack.dataclasses import ChatMessage
 from haystack.utils import Secret
 from requests.auth import HTTPBasicAuth
-import httpx
 
 from haystack_integrations.components.connectors.langfuse import LangfuseConnector
 from haystack_integrations.components.generators.anthropic import AnthropicChatGenerator
@@ -23,51 +26,34 @@ from haystack_integrations.tracing.langfuse.tracer import _COMPONENT_OUTPUT_KEY
 os.environ["HAYSTACK_CONTENT_TRACING_ENABLED"] = "true"
 
 
-@pytest.fixture
-def pipeline_with_env_vars(llm_class, expected_trace):
-    """Pipeline factory using environment variables for Langfuse authentication"""
-    pipe = Pipeline()
-    pipe.add_component("tracer", LangfuseConnector(name=f"Chat example - {expected_trace}", public=True))
-    pipe.add_component("prompt_builder", ChatPromptBuilder())
-    pipe.add_component("llm", llm_class())
-    pipe.connect("prompt_builder.prompt", "llm.messages")
-    return pipe
+def poll_langfuse(url: str):
+    """Utility function to poll Langfuse API until the trace is ready"""
+    # Initial wait for trace creation
+    time.sleep(10)
 
+    auth = HTTPBasicAuth(os.environ["LANGFUSE_PUBLIC_KEY"], os.environ["LANGFUSE_SECRET_KEY"])
 
-@pytest.fixture
-def pipeline_with_secrets(llm_class, expected_trace):
-    """Pipeline factory using Secret objects for Langfuse authentication"""
-    pipe = Pipeline()
-    pipe.add_component(
-        "tracer",
-        LangfuseConnector(
-            name=f"Chat example - {expected_trace}",
-            public=True,
-            secret_key=Secret.from_env_var("LANGFUSE_SECRET_KEY"),
-            public_key=Secret.from_env_var("LANGFUSE_PUBLIC_KEY"),
-        ),
-    )
-    pipe.add_component("prompt_builder", ChatPromptBuilder())
-    pipe.add_component("llm", llm_class())
-    pipe.connect("prompt_builder.prompt", "llm.messages")
-    return pipe
+    attempts = 5
+    delay = 1
+
+    res = None
+    while attempts > 0:
+        res = requests.get(url, auth=auth)
+        if res.status_code == 200:
+            return res
+
+        attempts -= 1
+        if attempts > 0:
+            time.sleep(delay)
+            delay *= 2
+
+    return res
 
 
 @pytest.fixture
-def pipeline_with_custom_client(llm_class, expected_trace):
-    """Pipeline factory using custom httpx client for Langfuse"""
+def pipeline_with_env_vars(llm_class, expected_trace):
     pipe = Pipeline()
-    custom_client = httpx.Client(timeout=30.0)  # Custom timeout of 30 seconds
-    pipe.add_component(
-        "tracer",
-        LangfuseConnector(
-            name=f"Chat example - {expected_trace}",
-            public=True,
-            secret_key=Secret.from_env_var("LANGFUSE_SECRET_KEY"),
-            public_key=Secret.from_env_var("LANGFUSE_PUBLIC_KEY"),
-            httpx_client=custom_client,
-        ),
-    )
+    pipe.add_component("tracer", LangfuseConnector(name=f"Chat example - {expected_trace}", public=True))
     pipe.add_component("prompt_builder", ChatPromptBuilder())
     pipe.add_component("llm", llm_class())
     pipe.connect("prompt_builder.prompt", "llm.messages")
@@ -83,9 +69,7 @@ def pipeline_with_custom_client(llm_class, expected_trace):
         (CohereChatGenerator, "COHERE_API_KEY", "Cohere"),
     ],
 )
-@pytest.mark.parametrize(
-    "pipeline_fixture", ["pipeline_with_env_vars", "pipeline_with_secrets", "pipeline_with_custom_client"]
-)
+@pytest.mark.parametrize("pipeline_fixture", ["pipeline_with_env_vars"])
 def test_tracing_integration(llm_class, env_var, expected_trace, pipeline_fixture, request):
     if not all([os.environ.get("LANGFUSE_SECRET_KEY"), os.environ.get("LANGFUSE_PUBLIC_KEY"), os.environ.get(env_var)]):
         pytest.skip(f"Missing required environment variables: LANGFUSE_SECRET_KEY, LANGFUSE_PUBLIC_KEY, or {env_var}")
@@ -104,32 +88,23 @@ def test_tracing_integration(llm_class, env_var, expected_trace, pipeline_fixtur
     )
     assert "Berlin" in response["llm"]["replies"][0].text
     assert response["tracer"]["trace_url"]
+    assert response["tracer"]["trace_id"]
 
     trace_url = response["tracer"]["trace_url"]
     uuid = os.path.basename(urlparse(trace_url).path)
     url = f"https://cloud.langfuse.com/api/public/traces/{uuid}"
 
-    # Poll the Langfuse API a bit as the trace might not be ready right away
-    attempts = 5
-    delay = 1
-    while attempts >= 0:
-        res = requests.get(
-            url, auth=HTTPBasicAuth(os.environ["LANGFUSE_PUBLIC_KEY"], os.environ["LANGFUSE_SECRET_KEY"])
-        )
-        if attempts > 0 and res.status_code != 200:
-            attempts -= 1
-            time.sleep(delay)
-            delay *= 2
-            continue
-        assert res.status_code == 200, f"Failed to retrieve data from Langfuse API: {res.status_code}"
+    res = poll_langfuse(url)
+    assert res.status_code == 200, f"Failed to retrieve data from Langfuse API: {res.status_code}"
 
-        # check if the trace contains the expected LLM name
-        assert expected_trace in str(res.content)
-        # check if the trace contains the expected generation span
-        assert "GENERATION" in str(res.content)
-        # check if the trace contains the expected user_id
-        assert "user_42" in str(res.content)
-        break
+    res_json = res.json()
+    assert res_json["name"] == f"Chat example - {expected_trace}"
+    assert isinstance(res_json["input"], dict)
+    assert res_json["input"]["tracer"]["invocation_context"]["user_id"] == "user_42"
+    assert isinstance(res_json["output"], dict)
+    assert isinstance(res_json["metadata"], dict)
+    assert isinstance(res_json["observations"], list)
+    assert res_json["observations"][0]["type"] == "GENERATION"
 
 
 def test_pipeline_serialization(monkeypatch):
@@ -237,21 +212,9 @@ def test_custom_span_handler():
     uuid = os.path.basename(urlparse(trace_url).path)
     url = f"https://cloud.langfuse.com/api/public/traces/{uuid}"
 
-    # Poll the Langfuse API
-    attempts = 5
-    delay = 1
-    while attempts >= 0:
-        res = requests.get(
-            url, auth=HTTPBasicAuth(os.environ["LANGFUSE_PUBLIC_KEY"], os.environ["LANGFUSE_SECRET_KEY"])
-        )
-        if attempts > 0 and res.status_code != 200:
-            attempts -= 1
-            time.sleep(delay)
-            delay *= 2
-            continue
+    res = poll_langfuse(url)
+    assert res.status_code == 200, f"Failed to retrieve data from Langfuse API: {res.status_code}"
 
-        assert res.status_code == 200
-        content = str(res.content)
-        assert "WARNING" in content
-        assert "Response too long" in content
-        break
+    content = str(res.content)
+    assert "WARNING" in content
+    assert "Response too long" in content