square-agency-59624
11/20/2025, 2:56 PMancient-wolf-19325
11/20/2025, 2:56 PMpyflyte run --remote, Flyte does not package your entire local environment. Instead, it "fast-registers" your code by packaging the specified workflow file (and, if needed, its imported modules) into a tarball and injects it into the container image running on the Flyte cluster. The environment (dependencies, Python version, etc.) is determined by the container image specified for the task, not your local environment. If you want to include additional dependencies or files, you must specify them in your Docker image, requirements file, or use options like --copy-all or --include for targeted inclusion during fast registration. The local environment is not automatically replicated in the remote execution environment—only your code and explicitly included files are sent; the base environment is defined by the image used in the cluster execution context.
For more details, see the official documentation on registering workflows and fast registration and the ImageSpec documentation. Would you like a step-by-step breakdown of how the packaging and environment selection works?
Sources:
- ImageSpec documentation
- registering workflows and fast registration 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.square-agency-59624
11/20/2025, 2:59 PMancient-wolf-19325
11/20/2025, 2:59 PM