vellum-ai 0.14.10__tar.gz → 0.14.11__tar.gz

{vellum_ai-0.14.10 → vellum_ai-0.14.11}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: vellum-ai
-Version: 0.14.10
+Version: 0.14.11
 Summary: 
 License: MIT
 Requires-Python: >=3.9,<4.0

{vellum_ai-0.14.10 → vellum_ai-0.14.11}/ee/vellum_cli/image_push.py

@@ -1,5 +1,6 @@
 import json
 import logging
+import re
 import subprocess
 from typing import List, Optional
 
@@ -24,9 +25,11 @@ def image_push_command(image: str, tags: Optional[List[str]] = None) -> None:
     # listing all of the architectures of the image instead of just the one that matches the machine. We can fall back
     # to using normal inspect which returns the machine image for this case though. And in the future we could figure
     # out how to call the docker host directly to do this.
+    logger.info("Pre-validating image...")
     docker_client = docker.from_env()
     check_architecture(docker_client, image, logger)
 
+    logger.info("Authenticating...")
     auth = vellum_client.container_images.docker_service_token()
 
     docker_client.login(
@@ -70,9 +73,29 @@ def image_push_command(image: str, tags: Optional[List[str]] = None) -> None:
             except Exception:
                 continue
 
-    logger.info("Updating Vellum metadata and enforcing the first law of robotics...")
-    image_details = docker_client.api.inspect_image(image)
-    sha = image_details["Id"]
+    result = subprocess.run(
+        ["docker", "inspect", "--format='{{index .RepoDigests 0}}'", image],
+        stdout=subprocess.PIPE,
+        stderr=subprocess.PIPE,
+    )
+
+    sha = ""
+    if result.returncode == 0:
+        match = re.search(r"sha256[^']*", result.stdout.decode("utf-8"))
+        if match and match.group(0):
+            sha = match.group(0)
+
+    if not sha:
+        # Fallback to using docker client if command line fails, at least on some systems
+        # this appears to give a bad sha.
+        logger.warning(
+            "Could not determine image hash with command line docker falling back to python docker client..."
+        )
+
+        image_details = docker_client.api.inspect_image(image)
+        sha = image_details["Id"]
+
+    logger.info(f"Updating Vellum metadata and validating image works in our system with image digest: {sha}...")
 
     vellum_client.container_images.push_container_image(
         name=image_name,
@@ -95,12 +118,12 @@ def check_architecture(docker_client: DockerClient, image: str, logger: logging.
         manifest = json.loads(result.stdout)
         architectures = [manifest_item["platform"]["architecture"] for manifest_item in manifest["manifests"]]
     except Exception:
-        logger.warning("Error parsing manifest response")
+        logger.debug("Error parsing manifest response")
         manifest_parse_failed = True
 
     # Fall back to inspect image if we errored out using docker command line
     if result.returncode != 0 or manifest_parse_failed:
-        logger.warning(f"Error inspecting manifest: {result.stderr.decode('utf-8').strip()}")
+        logger.debug(f"Error inspecting manifest: {result.stderr.decode('utf-8').strip()}")
         image_details = docker_client.api.inspect_image(image)
 
         if image_details["Architecture"] != _SUPPORTED_ARCHITECTURE:

vellum_ai-0.14.11/ee/vellum_ee/workflows/display/nodes/vellum/retry_node.py

@@ -0,0 +1,132 @@
+import inspect
+from typing import Any, Callable, Generic, Optional, Tuple, Type, TypeVar, cast
+
+from vellum.workflows.descriptors.base import BaseDescriptor
+from vellum.workflows.errors.types import WorkflowErrorCode
+from vellum.workflows.nodes.bases.base import BaseNode
+from vellum.workflows.nodes.core.retry_node.node import RetryNode
+from vellum.workflows.nodes.utils import ADORNMENT_MODULE_NAME
+from vellum.workflows.references.output import OutputReference
+from vellum.workflows.types.core import JsonArray, JsonObject
+from vellum.workflows.types.utils import get_original_base
+from vellum.workflows.utils.uuids import uuid4_from_hash
+from vellum.workflows.workflows.base import BaseWorkflow
+from vellum_ee.workflows.display.nodes.base_node_display import BaseNodeDisplay
+from vellum_ee.workflows.display.nodes.get_node_display_class import get_node_display_class
+from vellum_ee.workflows.display.nodes.types import NodeOutputDisplay
+from vellum_ee.workflows.display.nodes.vellum.base_adornment_node import BaseAdornmentNodeDisplay
+from vellum_ee.workflows.display.types import WorkflowDisplayContext
+
+_RetryNodeType = TypeVar("_RetryNodeType", bound=RetryNode)
+
+
+class BaseRetryNodeDisplay(BaseAdornmentNodeDisplay[_RetryNodeType], Generic[_RetryNodeType]):
+    def serialize(self, display_context: WorkflowDisplayContext, **kwargs: Any) -> JsonObject:
+        node = self._node
+        node_id = self.node_id
+
+        attributes: JsonArray = []
+        for attribute in node:
+            if inspect.isclass(attribute.instance) and issubclass(attribute.instance, BaseWorkflow):
+                # We don't need to serialize attributes that are workflows
+                continue
+
+            id = str(uuid4_from_hash(f"{node_id}|{attribute.name}"))
+            attributes.append(
+                {
+                    "id": id,
+                    "name": attribute.name,
+                    "value": self.serialize_value(display_context, cast(BaseDescriptor, attribute.instance)),
+                }
+            )
+
+        adornment: JsonObject = {
+            "id": str(node_id),
+            "label": node.__qualname__,
+            "base": self.get_base().dict(),
+            "attributes": attributes,
+        }
+
+        serialized_node = super().serialize(
+            display_context,
+            adornment=adornment,
+        )
+
+        if serialized_node["type"] == "GENERIC":
+            return serialized_node
+
+        serialized_node_definition = serialized_node.get("definition")
+        if isinstance(serialized_node_definition, dict):
+            serialized_node_definition_module = serialized_node_definition.get("module")
+            if isinstance(serialized_node_definition_module, list):
+                serialized_node_definition_module.extend(
+                    [
+                        serialized_node_definition["name"],
+                        ADORNMENT_MODULE_NAME,
+                    ]
+                )
+                serialized_node_definition["name"] = node.__name__
+
+        return serialized_node
+
+    def get_node_output_display(self, output: OutputReference) -> Tuple[Type[BaseNode], NodeOutputDisplay]:
+        inner_node = self._node.__wrapped_node__
+        if not inner_node:
+            return super().get_node_output_display(output)
+
+        node_display_class = get_node_display_class(BaseNodeDisplay, inner_node)
+        node_display = node_display_class()
+
+        inner_output = getattr(inner_node.Outputs, output.name)
+        return node_display.get_node_output_display(inner_output)
+
+    @classmethod
+    def wrap(
+        cls,
+        max_attempts: int,
+        delay: Optional[float] = None,
+        retry_on_error_code: Optional[WorkflowErrorCode] = None,
+        retry_on_condition: Optional[BaseDescriptor] = None,
+    ) -> Callable[..., Type["BaseRetryNodeDisplay"]]:
+        _max_attempts = max_attempts
+        _delay = delay
+        _retry_on_error_code = retry_on_error_code
+        _retry_on_condition = retry_on_condition
+
+        NodeDisplayType = TypeVar("NodeDisplayType", bound=BaseNodeDisplay)
+
+        def decorator(inner_cls: Type[NodeDisplayType]) -> Type[NodeDisplayType]:
+            node_class = inner_cls.infer_node_class()
+            wrapped_node_class = cast(Type[BaseNode], node_class.__wrapped_node__)
+
+            class RetryNodeDisplay(BaseRetryNodeDisplay[node_class]):  # type: ignore[valid-type]
+                max_attempts = _max_attempts
+                delay = _delay
+                retry_on_error_code = _retry_on_error_code
+                retry_on_condition = _retry_on_condition
+
+            setattr(inner_cls, "__adorned_by__", RetryNodeDisplay)
+
+            # We must edit the node display class to use __wrapped_node__ everywhere it
+            # references the adorned node class, which is three places:
+
+            # 1. The node display class' parameterized type
+            original_base_node_display = get_original_base(inner_cls)
+            original_base_node_display.__args__ = (wrapped_node_class,)
+            inner_cls._node_display_registry[wrapped_node_class] = inner_cls
+
+            # 2. The node display class' output displays
+            old_outputs = list(inner_cls.output_display.keys())
+            for old_output in old_outputs:
+                new_output = getattr(wrapped_node_class.Outputs, old_output.name)
+                inner_cls.output_display[new_output] = inner_cls.output_display.pop(old_output)
+
+            # 3. The node display class' port displays
+            old_ports = list(inner_cls.port_displays.keys())
+            for old_port in old_ports:
+                new_port = getattr(wrapped_node_class.Ports, old_port.name)
+                inner_cls.port_displays[new_port] = inner_cls.port_displays.pop(old_port)
+
+            return inner_cls
+
+        return decorator

vellum_ai-0.14.11/ee/vellum_ee/workflows/display/nodes/vellum/tests/test_retry_node.py

@@ -0,0 +1,47 @@
+from typing import Any, Dict, cast
+
+from vellum.workflows import BaseWorkflow
+from vellum.workflows.errors.types import WorkflowErrorCode
+from vellum.workflows.nodes.bases.base import BaseNode
+from vellum.workflows.nodes.core.retry_node.node import RetryNode
+from vellum_ee.workflows.display.workflows.get_vellum_workflow_display_class import get_workflow_display
+from vellum_ee.workflows.display.workflows.vellum_workflow_display import VellumWorkflowDisplay
+
+
+def test_retry_node_parameters():
+    """Test that RetryNode parameters are correctly serialized."""
+
+    # GIVEN a RetryNode with specific parameters
+    @RetryNode.wrap(max_attempts=5, delay=2.5, retry_on_error_code=WorkflowErrorCode.INVALID_INPUTS)
+    class MyRetryNode(BaseNode):
+        pass
+
+    # AND a workflow using the node
+    class MyWorkflow(BaseWorkflow):
+        graph = MyRetryNode
+
+    # WHEN we serialize the workflow
+    workflow_display = get_workflow_display(base_display_class=VellumWorkflowDisplay, workflow_class=MyWorkflow)
+    serialized_workflow = cast(Dict[str, Any], workflow_display.serialize())
+
+    # THEN the correct inputs should be serialized on the node
+    serialized_node = next(
+        node for node in serialized_workflow["workflow_raw_data"]["nodes"] if node["type"] != "ENTRYPOINT"
+    )
+
+    retry_adornment = next(
+        adornment for adornment in serialized_node["adornments"] if adornment["label"] == "RetryNode"
+    )
+
+    max_attempts_attribute = next(attr for attr in retry_adornment["attributes"] if attr["name"] == "max_attempts")
+    assert max_attempts_attribute["value"]["value"]["value"] == 5
+
+    delay_attribute = next(attr for attr in retry_adornment["attributes"] if attr["name"] == "delay")
+    assert delay_attribute["value"]["value"]["value"] == 2.5
+
+    retry_on_error_code_attribute = next(
+        attr for attr in retry_adornment["attributes"] if attr["name"] == "retry_on_error_code"
+    )
+
+    assert retry_on_error_code_attribute["value"]["value"]["type"] == "STRING"
+    assert retry_on_error_code_attribute["value"]["value"]["value"] == "INVALID_INPUTS"

{vellum_ai-0.14.10 → vellum_ai-0.14.11}/ee/vellum_ee/workflows/display/tests/workflow_serialization/generic_nodes/test_adornments_serialization.py

@@ -29,11 +29,8 @@ class InnerRetryGenericNode(BaseNode):
         output: str
 
 
-class InnerRetryGenericNodeDisplay(BaseNodeDisplay[InnerRetryGenericNode.__wrapped_node__]):  # type: ignore
-    pass
-
-
-class OuterRetryNodeDisplay(BaseRetryNodeDisplay[InnerRetryGenericNode]):  # type: ignore
+@BaseRetryNodeDisplay.wrap(max_attempts=3)
+class InnerRetryGenericNodeDisplay(BaseNodeDisplay[InnerRetryGenericNode]):
     pass
 
 
@@ -44,7 +41,6 @@ def test_serialize_node__retry(serialize_node):
         global_workflow_input_displays={Inputs.input: WorkflowInputsDisplay(id=input_id)},
         global_node_displays={
             InnerRetryGenericNode.__wrapped_node__: InnerRetryGenericNodeDisplay,
-            InnerRetryGenericNode: OuterRetryNodeDisplay,
         },
     )
 

{vellum_ai-0.14.10 → vellum_ai-0.14.11}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "vellum-ai"
-version = "0.14.10"
+version = "0.14.11"
 description = ""
 readme = "README.md"
 authors = []

{vellum_ai-0.14.10 → vellum_ai-0.14.11}/src/vellum/client/core/client_wrapper.py

@@ -18,7 +18,7 @@ class BaseClientWrapper:
         headers: typing.Dict[str, str] = {
             "X-Fern-Language": "Python",
             "X-Fern-SDK-Name": "vellum-ai",
-            "X-Fern-SDK-Version": "0.14.10",
+            "X-Fern-SDK-Version": "0.14.11",
         }
         headers["X_API_KEY"] = self.api_key
         return headers

{vellum_ai-0.14.10 → vellum_ai-0.14.11}/src/vellum/workflows/events/types.py

@@ -75,6 +75,7 @@ class BaseDeploymentParentContext(BaseParentContext):
     release_tag_id: UUID
     release_tag_name: str
     external_id: Optional[str]
+    metadata: Optional[dict]
 
 
 class WorkflowDeploymentParentContext(BaseDeploymentParentContext):
@@ -104,6 +105,10 @@ class WorkflowSandboxParentContext(BaseParentContext):
     scenario_id: UUID
 
 
+class APIRequestParentContext(BaseParentContext):
+    type: Literal["API_REQUEST"] = "API_REQUEST"
+
+
 # Define the discriminated union
 ParentContext = Annotated[
     Union[
@@ -112,6 +117,7 @@ ParentContext = Annotated[
         WorkflowDeploymentParentContext,
         PromptDeploymentParentContext,
         WorkflowSandboxParentContext,
+        APIRequestParentContext,
     ],
     Field(discriminator="type"),
 ]

{vellum_ai-0.14.10 → vellum_ai-0.14.11}/src/vellum/workflows/nodes/displayable/code_execution_node/tests/test_code_execution_node.py

@@ -458,16 +458,7 @@ def main(word: str) -> int:
         node.run()
 
     # THEN the node should have produced the exception we expected
-    assert (
-        exc_info.value.message
-        == """\
-2 validation errors for FunctionCall
-arguments
-  Field required [type=missing, input_value={'n': 'hello', 'a': {}}, input_type=dict]
-name
-  Field required [type=missing, input_value={'n': 'hello', 'a': {}}, input_type=dict]\
-"""
-    )
+    assert exc_info.value.message == "Expected an output of type 'FunctionCall', but received 'dict'"
 
 
 def test_run_node__run_inline__valid_dict_to_pydantic_any_type():
@@ -743,3 +734,46 @@ Node.js v21.7.3
 
     # AND the error should contain the execution error details
     assert exc_info.value.message == message
+
+
+def test_run_node__execute_code__list_extends():
+    # GIVEN a node that will return a list with output type Json
+    class ExampleCodeExecutionNode(CodeExecutionNode[BaseState, Json]):
+        code = """\
+def main(left, right):
+    all = []
+    all.extend(left)
+    all.extend(right)
+    return all
+"""
+        code_inputs = {
+            "left": [1, 2, 3],
+            "right": [4, 5, 6],
+        }
+        runtime = "PYTHON_3_11_6"
+
+    # WHEN we run the node
+    node = ExampleCodeExecutionNode()
+    outputs = node.run()
+
+    # AND the result should be the correct output
+    assert outputs == {"result": [1, 2, 3, 4, 5, 6], "log": ""}
+
+
+def test_run_node__execute_code__non_str_print():
+    # GIVEN a node that will print a non-string value
+    class ExampleCodeExecutionNode(CodeExecutionNode[BaseState, str]):
+        code = """\
+def main():
+    print(type(1))
+    return "hello"
+"""
+        code_inputs = {}
+        runtime = "PYTHON_3_11_6"
+
+    # WHEN we run the node
+    node = ExampleCodeExecutionNode()
+    outputs = node.run()
+
+    # AND the result should be the correct output
+    assert outputs == {"result": "hello", "log": "<class 'int'>\n"}

{vellum_ai-0.14.10 → vellum_ai-0.14.11}/src/vellum/workflows/nodes/displayable/code_execution_node/utils.py

@@ -1,12 +1,12 @@
 import io
 import os
-import re
-from typing import Any, Tuple, Union, get_args, get_origin
+from typing import Any, Tuple, Union
 
-from pydantic import BaseModel, ValidationError
+from pydantic import BaseModel
 
 from vellum.workflows.errors.types import WorkflowErrorCode
 from vellum.workflows.exceptions import NodeException
+from vellum.workflows.nodes.utils import cast_to_output_type
 from vellum.workflows.types.core import EntityInputsInterface
 
 
@@ -67,49 +67,6 @@ def _clean_for_dict_wrapper(obj):
     return obj
 
 
-def _get_type_name(obj: Any) -> str:
-    if isinstance(obj, type):
-        return obj.__name__
-
-    if get_origin(obj) is Union:
-        children = [_get_type_name(child) for child in get_args(obj)]
-        return " | ".join(children)
-
-    return str(obj)
-
-
-def _cast_to_output_type(result: Any, output_type: Any) -> Any:
-    is_valid_output_type = isinstance(output_type, type)
-    if get_origin(output_type) is Union:
-        allowed_types = get_args(output_type)
-        for allowed_type in allowed_types:
-            try:
-                return _cast_to_output_type(result, allowed_type)
-            except NodeException:
-                continue
-    elif get_origin(output_type) is list:
-        allowed_item_type = get_args(output_type)[0]
-        if isinstance(result, list):
-            return [_cast_to_output_type(item, allowed_item_type) for item in result]
-    elif is_valid_output_type and issubclass(output_type, BaseModel) and not isinstance(result, output_type):
-        try:
-            return output_type.model_validate(result)
-        except ValidationError as e:
-            raise NodeException(
-                code=WorkflowErrorCode.INVALID_OUTPUTS,
-                message=re.sub(r"\s+For further information visit [^\s]+", "", str(e)),
-            ) from e
-    elif is_valid_output_type and isinstance(result, output_type):
-        return result
-
-    output_type_name = _get_type_name(output_type)
-    result_type_name = _get_type_name(type(result))
-    raise NodeException(
-        code=WorkflowErrorCode.INVALID_OUTPUTS,
-        message=f"Expected an output of type '{output_type_name}', but received '{result_type_name}'",
-    )
-
-
 def run_code_inline(
     code: str,
     inputs: EntityInputsInterface,
@@ -117,6 +74,11 @@ def run_code_inline(
 ) -> Tuple[str, Any]:
     log_buffer = io.StringIO()
 
+    def _inline_print(*args: Any, **kwargs: Any) -> None:
+        str_args = [str(arg) for arg in args]
+        print_line = f"{' '.join(str_args)}\n"
+        log_buffer.write(print_line)
+
     def wrap_value(value):
         if isinstance(value, list):
             return ListWrapper(
@@ -135,7 +97,7 @@ def run_code_inline(
     exec_globals = {
         "__arg__inputs": {name: wrap_value(value) for name, value in inputs.items()},
         "__arg__out": None,
-        "print": lambda *args, **kwargs: log_buffer.write(f"{' '.join(args)}\n"),
+        "print": _inline_print,
     }
     run_args = [f"{name}=__arg__inputs['{name}']" for name in inputs.keys()]
     execution_code = f"""\
@@ -154,7 +116,6 @@ __arg__out = main({", ".join(run_args)})
     logs = log_buffer.getvalue()
     result = exec_globals["__arg__out"]
 
-    if output_type != Any:
-        result = _cast_to_output_type(result, output_type)
+    result = cast_to_output_type(result, output_type)
 
     return logs, result

{vellum_ai-0.14.10 → vellum_ai-0.14.11}/src/vellum/workflows/nodes/displayable/final_output_node/node.py

@@ -1,7 +1,9 @@
 from typing import Any, Dict, Generic, Tuple, Type, TypeVar, get_args
 
+from vellum.workflows.constants import undefined
 from vellum.workflows.nodes.bases import BaseNode
 from vellum.workflows.nodes.bases.base import BaseNodeMeta
+from vellum.workflows.nodes.utils import cast_to_output_type
 from vellum.workflows.types import MergeBehavior
 from vellum.workflows.types.generics import StateType
 from vellum.workflows.types.utils import get_original_base
@@ -48,4 +50,14 @@ class FinalOutputNode(BaseNode[StateType], Generic[StateType, _OutputType], meta
     class Outputs(BaseNode.Outputs):
         # We use our mypy plugin to override the _OutputType with the actual output type
         # for downstream references to this output.
-        value: _OutputType  # type: ignore[valid-type]
+        value: _OutputType = undefined  # type: ignore[valid-type]
+
+    def run(self) -> Outputs:
+        original_outputs = self.Outputs()
+
+        return self.Outputs(
+            value=cast_to_output_type(
+                original_outputs.value,
+                self.__class__.get_output_type(),
+            )
+        )

vellum_ai-0.14.11/src/vellum/workflows/nodes/displayable/final_output_node/tests/test_node.py

@@ -0,0 +1,20 @@
+import pytest
+
+from vellum.workflows.exceptions import NodeException
+from vellum.workflows.nodes.displayable.final_output_node import FinalOutputNode
+from vellum.workflows.state.base import BaseState
+
+
+def test_final_output_node__mismatched_output_type():
+    # GIVEN a FinalOutputNode with a mismatched output type
+    class StringOutputNode(FinalOutputNode[BaseState, str]):
+        class Outputs(FinalOutputNode.Outputs):
+            value = {"foo": "bar"}
+
+    # WHEN the node is run
+    node = StringOutputNode()
+    with pytest.raises(NodeException) as exc_info:
+        node.run()
+
+    # THEN an error is raised
+    assert str(exc_info.value) == "Expected an output of type 'str', but received 'dict'"

{vellum_ai-0.14.10 → vellum_ai-0.14.11}/src/vellum/workflows/nodes/utils.py

@@ -2,10 +2,12 @@ from functools import cache
 import json
 import sys
 from types import ModuleType
-from typing import Any, Callable, Optional, Type, TypeVar, Union, get_args, get_origin
+from typing import Any, Callable, Dict, ForwardRef, List, Optional, Type, TypeVar, Union, get_args, get_origin
 
-from pydantic import BaseModel
+from pydantic import BaseModel, create_model
 
+from vellum.workflows.errors.types import WorkflowErrorCode
+from vellum.workflows.exceptions import NodeException
 from vellum.workflows.nodes import BaseNode
 from vellum.workflows.nodes.bases.base_adornment_node import BaseAdornmentNode
 from vellum.workflows.ports.port import Port
@@ -141,3 +143,55 @@ def parse_type_from_str(result_as_str: str, output_type: Any) -> Any:
             raise ValueError("Invalid JSON format for result_as_str")
 
     raise ValueError(f"Unsupported output type: {output_type}")
+
+
+def _get_type_name(obj: Any) -> str:
+    if isinstance(obj, type):
+        return obj.__name__
+
+    if get_origin(obj) is Union:
+        children = [_get_type_name(child) for child in get_args(obj)]
+        return " | ".join(children)
+
+    return str(obj)
+
+
+def _clean_output_type(output_type: Any) -> Any:
+    """
+    pydantic currently has a bug where it doesn't support forward references in the create_model function. It will
+    fail due to a max recursion depth error. This negatively impacts our `Json` type, but could also apply to other
+    user defined ForwardRef types.
+    """
+    if get_origin(output_type) is Union:
+        clean_args = [_clean_output_type(child) for child in get_args(output_type)]
+        return Union[tuple(clean_args)]
+
+    if isinstance(output_type, ForwardRef):
+        # Here is where we prevent the max recursion depth error
+        return Any
+
+    if get_origin(output_type) is list:
+        clean_args = [_clean_output_type(child) for child in get_args(output_type)]
+        return List[clean_args[0]]  # type: ignore[valid-type]
+
+    if get_origin(output_type) is dict:
+        clean_args = [_clean_output_type(child) for child in get_args(output_type)]
+        return Dict[clean_args[0], clean_args[1]]  # type: ignore[valid-type]
+
+    return output_type
+
+
+def cast_to_output_type(result: Any, output_type: Any) -> Any:
+    clean_output_type = _clean_output_type(output_type)
+    DynamicModel = create_model("Output", output_type=(clean_output_type, ...))
+
+    try:
+        # mypy doesn't realize that this dynamic model has the output_type field defined above
+        return DynamicModel.model_validate({"output_type": result}).output_type  # type: ignore[attr-defined]
+    except Exception:
+        output_type_name = _get_type_name(output_type)
+        result_type_name = _get_type_name(type(result))
+        raise NodeException(
+            code=WorkflowErrorCode.INVALID_OUTPUTS,
+            message=f"Expected an output of type '{output_type_name}', but received '{result_type_name}'",
+        )

{vellum_ai-0.14.10 → vellum_ai-0.14.11}/src/vellum/workflows/outputs/base.py

@@ -204,13 +204,11 @@ class BaseOutputs(metaclass=_BaseOutputsMeta):
         if not isinstance(other, dict):
             return super().__eq__(other)
 
-        outputs = {
-            name: value for name, value in vars(self).items() if not name.startswith("_") and value is not undefined
-        }
+        outputs = {ref.name: value for ref, value in self if value is not undefined}
         return outputs == other
 
     def __repr__(self) -> str:
-        values = f"{', '.join(f'{k}={v}' for k, v in vars(self).items() if not k.startswith('_'))}"
+        values = f"{', '.join(f'{ref.name}={value}' for ref, value in self if value is not undefined)}"
         return f"{self.__class__.__name__}({values})"
 
     def __iter__(self) -> Iterator[Tuple[OutputReference, Any]]:

{vellum_ai-0.14.10 → vellum_ai-0.14.11}/src/vellum/workflows/workflows/base.py

@@ -24,7 +24,6 @@ from typing import (
     get_args,
 )
 
-from vellum.workflows.context import get_execution_context
 from vellum.workflows.edges import Edge
 from vellum.workflows.emitters.base import BaseWorkflowEmitter
 from vellum.workflows.errors import WorkflowError, WorkflowErrorCode
@@ -481,7 +480,7 @@ class BaseWorkflow(Generic[InputsType, StateType], metaclass=_BaseWorkflowMeta):
         return self.get_inputs_class()()
 
     def get_default_state(self, workflow_inputs: Optional[InputsType] = None) -> StateType:
-        execution_context = get_execution_context()
+        execution_context = self._execution_context
         return self.get_state_class()(
             meta=(
                 StateMeta(