In short, question is if there is code outside the step functions depending on a particular package, does the flow need to be at the top level of that package to be able to use it? I mean class level imports

Metaflow packages all the Python code from the directory level of the

flow.py

. In the example below, it will package all the Python files under main and therefore you can import them when running it remotely.

.
|-- flow.py
`-- src
    `-- main.py

If you have another type of file added such as an SQL script, by default it will not be packaged, you will need to specify the appropriate extension to package via

--package-suffixes=.sql

.
|-- flow.py
`-- src
    |-- main.py
    `-- query.sql

If you want to package modules that exist outside of the flow directory level, by default it will not work.

.
|-- flows
|   `-- flow.py
`-- src
    |-- main.py
    `-- query.sql

You would need to perform a symlink of the module inside the Flow directory scope via

cd flows && ln -s ../src

The directory will look like this now:

.
|-- flows
|   |-- flow.py
|   `-- src -> ../src
`-- src
    |-- main.py
    `-- query.sql

ok thank you! I will use one of these approaches.