@ancient-wolf-19325 i'm trying to use the new

WebhookTask

feature. i have flytekit 1.16.3 installed and my IDE recognizes the import

from flytekit.extras.webhook import WebhookTask

. however, when i run the workflow with

--remote

i see this error:

Caused by: ModuleNotFoundError: No module named 'flytekit.extras.webhook'

It seems likely that your issue is due to the container image used for remote execution not having the

flytekit

version (1.16.3) that includes the

flytekit.extras.webhook

module. While your local environment recognizes the import, the remote Flyte execution environment may be using an older image or one without the necessary extras installed, leading to the

ModuleNotFoundError

for

flytekit.extras.webhook

(source). To resolve this, ensure that the Docker image specified for your workflow execution includes

flytekit>=1.16.3

and any required plugins. You can do this by building a custom image with the correct dependencies and specifying it via the

--image

flag when running or registering your workflow (example). Would you like more detail on how to build and use a custom image for Flyte workflows? Sources: - https://github.com/flyteorg/flyte/issues/3687 - example 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.

@ancient-wolf-19325 i'm using the latest flytekit image published by Flyte:

@task(container_image="<http://ghcr.io/flyteorg/flytekit:py3.12-latest|ghcr.io/flyteorg/flytekit:py3.12-latest>")

. after a small configuration change, i'm now seeing this related exception:

Trace:

    Traceback (most recent call last):
      File "/root/workflows/ingestionwf.py", line 52, in get_pipeline_config
        from workflows.test import notification_task
      File "/root/workflows/test.py", line 3, in <module>
        from flytekit.extras.webhook import WebhookTask
      File "/usr/local/lib/python3.12/site-packages/flytekit/extras/webhook/__init__.py", line 1, in <module>
        from .connector import WebhookConnector
      File "/usr/local/lib/python3.12/site-packages/flytekit/extras/webhook/connector.py", line 3, in <module>
        import httpx
    ModuleNotFoundError: No module named 'httpx'

    During handling of the above exception, another exception occurred:

    Traceback (most recent call last):
      File "/usr/local/lib/python3.12/site-packages/flytekit/core/base_task.py", line 767, in dispatch_execute
        native_outputs = self.execute(**native_inputs)
                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
      File "/usr/local/lib/python3.12/site-packages/flytekit/core/python_function_task.py", line 214, in execute
        return self._task_function(**kwargs)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
      File "/root/shared/errors.py", line 106, in wrapper
        return func(*args, **kwargs)
               ^^^^^^^^^^^^^^^^^^^^^
      File "/root/workflows/ingestionwf.py", line 56, in get_pipeline_config
        raise ServiceError(
    shared.errors.ServiceError: localerror
 [Caused by: ModuleNotFoundError: No module named 'httpx']

The error occurs because the

WebhookTask

feature requires the

httpx

library, which is not included in the default Flytekit image (

ghcr.io/flyteorg/flytekit:py3.12-latest

). The image contains Flytekit and its core dependencies, but optional dependencies like

httpx

(needed for

flytekit.extras.webhook

) must be installed separately (source). To fix this, you need to build a custom Docker image based on

ghcr.io/flyteorg/flytekit:py3.12-latest

and add

httpx

(e.g., via

pip install httpx

). Then, use this custom image for your task. Would you like detailed steps or a sample Dockerfile for this process? Sources: - https://github.com/flyteorg/flyte/blob/master/docs/user_guide/customizing_dependencies/imagespec.md - https://github.com/flyteorg/flyte/blob/master/docs/user_guide/customizing_dependencies/index.md 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.

@ancient-wolf-19325 what's the syntax to add an annotation to a single task that would include the

httpx

dependency in the default container?

It seems likely that you cannot add a Python dependency like

httpx

to a task via an annotation alone when using the default Flytekit container image. Instead, Flyte recommends specifying dependencies using an

ImageSpec

in your code, which builds a new image with the required packages (source). There is no supported annotation or decorator parameter that injects Python packages into the default image at runtime. Here’s an example using `ImageSpec`:

from flytekit import ImageSpec, task

custom_image = ImageSpec(
    registry="ghcr.io/flyteorg",
    packages=["httpx"]
)

@task(container_image=custom_image)
def my_task():
    import httpx
    # your code here

Would you like more detail on how to use

ImageSpec

or build custom images? Sources: - https://github.com/flyteorg/flytekit/blob/master/flytekit/core/task.py - https://github.com/flyteorg/flyte/blob/master/docs/user_guide/customizing_dependencies/multiple_images_in_a_workflow.md 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.