To use your own custom Docker image in Flyte's ImageSpec, set the

base_image

parameter to your image's name (including registry if needed) when creating the ImageSpec. This will use your image as the starting point for any additional packages or configuration you specify. Example:

from flytekit import ImageSpec

custom_image = ImageSpec(
    base_image="myregistry/my-custom-image:latest",
    packages=["pandas", "numpy"],  # Optional: add extra Python packages
    registry="myregistry",         # Where to push the built image
)

You can then assign this ImageSpec to a task using the

image

(or

container_image

) parameter in the

@task

decorator. For more, see the ImageSpec documentation. Would you like more detail or a full workflow example? Sources: - ImageSpec documentation 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.

If Flyte fails to build the image when running with

--remote

, common causes include Docker not running, platform/architecture mismatches, or missing configuration for the image registry. Ensure Docker is running locally, the

registry

parameter in ImageSpec matches your cluster's registry, and the

platform

parameter matches your cluster's architecture (e.g.,

linux/arm64

for Apple Silicon). Review error logs for details—errors like "no match for platform in manifest" indicate an architecture mismatch, while "Cannot connect to the Docker daemon" means Docker isn't running or accessible. For troubleshooting, see this issue for platform errors and this issue for Docker daemon errors. Also, refer to the ImageSpec troubleshooting guide and docs on configuring local registry and platform. Would you like step-by-step troubleshooting instructions for your specific error? Sources: - ImageSpec documentation - Platform mismatch issue - Docker daemon error - ImageSpec build troubleshooting 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.