vellum-workflow-server 0.14.73.post2__tar.gz → 0.14.73.post5__tar.gz

{vellum_workflow_server-0.14.73.post2 → vellum_workflow_server-0.14.73.post5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: vellum-workflow-server
-Version: 0.14.73.post2
+Version: 0.14.73.post5
 Summary: 
 License: AGPL
 Requires-Python: >=3.9.0,<4

{vellum_workflow_server-0.14.73.post2 → vellum_workflow_server-0.14.73.post5}/pyproject.toml

@@ -3,7 +3,7 @@ name = "vellum-workflow-server"
 
 [tool.poetry]
 name = "vellum-workflow-server"
-version = "0.14.73.post2"
+version = "0.14.73.post5"
 description = ""
 readme = "README.md"
 authors = []
@@ -32,6 +32,7 @@ packages = [
 
 [tool.poetry.group.dev.dependencies]
 requests-mock = "^1.12.1"
+pytest-mock = "^3.14.1"
 
 [project.urls]
 Repository = 'https://github.com/vellum-ai/vembda-service'

{vellum_workflow_server-0.14.73.post2 → vellum_workflow_server-0.14.73.post5}/src/workflow_server/api/workflow_view.py

@@ -15,7 +15,6 @@ from typing import Any, Dict, Generator, Iterator, Union, cast
 
 from flask import Blueprint, Response, current_app as app, request, stream_with_context
 from pydantic import ValidationError
-import sentry_sdk
 from vellum_ee.workflows.display.nodes.get_node_display_class import get_node_display_class
 from vellum_ee.workflows.display.types import WorkflowDisplayContext
 
@@ -115,7 +114,6 @@ def stream_workflow_route() -> Response:
         )
         increment_process_count(1)
     except Exception as e:
-        sentry_sdk.set_tag("vellum_trace_id", str(context.trace_id))
         logger.exception(e)
 
         process_output_queue.put(create_vembda_rejected_event(context, traceback.format_exc()))
@@ -202,7 +200,6 @@ def stream_workflow_route() -> Response:
 
                 continue
             except Exception as e:
-                sentry_sdk.set_tag("vellum_trace_id", str(context.trace_id))
                 logger.exception(e)
                 break
 
@@ -237,6 +234,9 @@ def stream_workflow_route() -> Response:
                 f"process count: {get_active_process_count()}"
             )
         except GeneratorExit:
+            # These can happen either from Vembda disconnects (possibily from predict disconnects) or
+            # from knative activator gateway timeouts which are caused by idleTimeout or responseStartSeconds
+            # being exceeded.
             app.logger.error("Client disconnected in the middle of the Workflow Stream")
             return
         finally:
@@ -300,7 +300,6 @@ def stream_node_route() -> Response:
                     # This happens when theres a problem with the stream function call
                     # itself not the workflow runner
                     yield create_vembda_rejected_event(context, "Internal Server Error")
-                    sentry_sdk.set_tag("vellum_trace_id", str(context.trace_id))
                     app.logger.exception(stream_future.exception())
                     break
                 else:
@@ -377,7 +376,6 @@ def get_version_route() -> tuple[dict, int]:
 
         resp["nodes"] = nodes
     except Exception as e:
-        sentry_sdk.set_tag("vellum_trace_id", "unknown")
         logger.exception(f"Failed to discover nodes: {str(e)}")
         resp["nodes"] = []
 

{vellum_workflow_server-0.14.73.post2 → vellum_workflow_server-0.14.73.post5}/src/workflow_server/code_exec_runner.py

@@ -5,8 +5,6 @@ import os
 from uuid import uuid4
 from typing import Optional
 
-import sentry_sdk
-
 from vellum.workflows.exceptions import WorkflowInitializationException
 from workflow_server.api.workflow_view import get_workflow_request_context
 from workflow_server.core.events import VembdaExecutionInitiatedBody, VembdaExecutionInitiatedEvent
@@ -56,7 +54,6 @@ def run_code_exec_stream() -> None:
             fulfilled_event = serialize_vembda_rejected_event(context, str(e))
             print(f"{_EVENT_LINE}{fulfilled_event}")  # noqa: T201
     except Exception as e:
-        sentry_sdk.set_tag("vellum_trace_id", str(context.trace_id) if context else "unknown")
         logger.exception(e)
 
         event = serialize_vembda_rejected_event(context, "Internal Server Error")

{vellum_workflow_server-0.14.73.post2 → vellum_workflow_server-0.14.73.post5}/src/workflow_server/core/executor.py

@@ -15,7 +15,6 @@ from uuid import UUID, uuid4
 from typing import Any, Callable, Generator, Iterator, Optional, Tuple, Type
 
 from pebble import concurrent
-import sentry_sdk
 from vellum_ee.workflows.display.workflows import BaseWorkflowDisplay
 from vellum_ee.workflows.server.virtual_file_loader import VirtualFileFinder
 
@@ -68,7 +67,6 @@ def _stream_node_wrapper(executor_context: NodeExecutorContext, queue: Queue) ->
         for event in stream_node(executor_context=executor_context):
             queue.put(event)
     except Exception as e:
-        sentry_sdk.set_tag("vellum_trace_id", str(executor_context.trace_id))
         logger.exception(e)
         queue.put(
             VembdaExecutionFulfilledEvent(
@@ -106,7 +104,6 @@ def _stream_workflow_wrapper(executor_context: WorkflowExecutorContext, queue: Q
         if not span_id_emitted:
             queue.put(f"{SPAN_ID_EVENT}:{uuid4()}")
 
-        sentry_sdk.set_tag("vellum_trace_id", str(executor_context.trace_id))
         logger.exception(e)
         queue.put(serialize_vembda_rejected_event(executor_context, "Internal Server Error"))
     queue.put(STREAM_FINISHED_EVENT)
@@ -456,7 +453,6 @@ def _gather_workflow(context: WorkflowExecutorContext) -> Tuple[BaseWorkflow, st
         )
         return workflow, namespace
     except Exception as e:
-        sentry_sdk.set_tag("vellum_trace_id", str(context.trace_id))
         logger.exception("Failed to initialize Workflow")
         raise WorkflowInitializationException(f"Failed to initialize workflow: {e}") from e
 
@@ -465,6 +461,5 @@ def _gather_display_context(workflow: BaseWorkflow, namespace: str) -> Optional[
     try:
         return BaseWorkflowDisplay.gather_event_display_context(namespace, workflow.__class__)
     except Exception:
-        sentry_sdk.set_tag("vellum_trace_id", "unknown")
         logger.exception("Unable to Parse Workflow Display Context")
         return None

{vellum_workflow_server-0.14.73.post2 → vellum_workflow_server-0.14.73.post5}/src/workflow_server/utils/sentry.py

@@ -5,10 +5,44 @@ import sentry_sdk
 from sentry_sdk.integrations.logging import LoggingIntegration
 
 
+def _tag_trace_id(event: dict) -> None:
+    if "request" not in event:
+        return
+
+    if not isinstance(event["request"], dict):
+        return
+
+    url = event["request"].get("url")
+    if not isinstance(url, str):
+        return
+
+    if not url.endswith("/workflow/stream"):
+        return
+
+    body = event["request"].get("data")
+    if not isinstance(body, dict):
+        return
+
+    execution_context = body.get("execution_context")
+    if not isinstance(execution_context, dict):
+        return
+
+    trace_id = execution_context.get("trace_id")
+    if not isinstance(trace_id, str):
+        return
+
+    if "tags" not in event:
+        event["tags"] = {}
+
+    event["tags"]["vellum_trace_id"] = trace_id
+
+
 def before_send(event: dict, hint: dict) -> Optional[dict]:
     if "exc_info" in hint:
         _, _, _ = hint["exc_info"]
 
+    _tag_trace_id(event)
+
     return event
 
 

{vellum_workflow_server-0.14.73.post2 → vellum_workflow_server-0.14.73.post5}/src/workflow_server/utils/system_utils.py

@@ -12,8 +12,9 @@ WARN_MEMORY_PERCENT = 0.90
 FORCE_GC_MEMORY_PERCENT = 0.75
 
 _MAX_PROCESS_COUNT = math.ceil(CONCURRENCY * 1.7)
-_MEMORY_CHECK_INTERVAL_SECONDS = 3
-_MAX_MEMORY_CHECK_ATTEMPTS = 5
+# Keep these under the 15s knative response start timeout or it may trigger odd behavior
+_MEMORY_CHECK_INTERVAL_SECONDS = 2
+_MAX_MEMORY_CHECK_ATTEMPTS = 3
 _ACTIVE_PROCESS_COUNT = multiprocessing.Value("i", 0)
 _ACTIVE_PROCESS_LOCK = multiprocessing.Lock()
 

vellum_workflow_server-0.14.73.post5/src/workflow_server/utils/tests/test_sentry_integration.py

@@ -0,0 +1,69 @@
+import pytest
+from uuid import uuid4
+
+from workflow_server.server import create_app
+
+
+@pytest.fixture
+def mock_sentry_capture_envelope(mocker):
+    mock_transport = mocker.patch("sentry_sdk.client.make_transport")
+    return mock_transport.return_value.capture_envelope
+
+
+def test_sentry_integration_with_workflow_endpoints(monkeypatch, mock_sentry_capture_envelope):
+    # GIVEN sentry is configured
+    monkeypatch.setenv("SENTRY_DSN", "https://test-dsn@sentry.io/1234567890")
+
+    # AND our /workflow/stream endpoint raises an exception
+    def mock_get_version():
+        raise Exception("Test exception")
+
+    monkeypatch.setattr("workflow_server.api.workflow_view.get_version", mock_get_version)
+
+    # AND we have a mock trace_id
+    trace_id = str(uuid4())
+
+    # AND we have a mock request body
+    body = {
+        "execution_id": uuid4(),
+        "inputs": [],
+        "environment_api_key": "test",
+        "module": "workflow",
+        "timeout": 360,
+        "files": {
+            "__init__.py": "",
+            "workflow.py": """\
+from vellum.workflows import BaseWorkflow
+
+class Workflow(BaseWorkflow):
+    pass
+""",
+        },
+        "execution_context": {
+            "trace_id": trace_id,
+            "parent_context": {
+                "type": "API_REQUEST",
+                "span_id": str(uuid4()),
+                "parent": None,
+            },
+        },
+    }
+
+    # WHEN we call the /workflow/version endpoint
+    flask_app = create_app()
+
+    with flask_app.test_client() as test_client:
+        response = test_client.post("/workflow/stream", json=body)
+
+        # THEN we get a 500 error
+        assert response.status_code == 500
+
+        # AND sentry captures the error with the correct data
+        assert mock_sentry_capture_envelope.call_count == 1
+        envelope = mock_sentry_capture_envelope.call_args[0][0]
+        event = envelope.get_event()
+        assert event["level"] == "error"
+        assert "Test exception" in event["exception"]["values"][0]["value"]
+
+        # AND the trace_id is tagged
+        assert event["tags"]["vellum_trace_id"] == trace_id

{vellum_workflow_server-0.14.73.post2 → vellum_workflow_server-0.14.73.post5}/src/workflow_server/utils/tests/test_system_utils.py

@@ -85,7 +85,7 @@ def test_wait_for_available_process_never_available(mock_get_active_process_coun
     assert result is False
 
     # Should sleep for each attempt
-    assert mock_sleep.call_count == 5
+    assert mock_sleep.call_count == 3
 
 
 @patch("workflow_server.utils.system_utils.time.sleep")