Hi folks, a question about reregistering flows after the flow logic has changed. We have a local agent (running inside a docker container) which executes flows that are stored in S3. Each flow is usually split across multiple Python files, usually with a

main.py

file that imports code from these other files and defines the actual flow logic. If

main.py

is updated, a simple re-registration of the flow seems to be enough to allow the agent to execute the updated flow, because the updated logic is stored on S3 and is then downloaded by the agent during the next execution. However, if one of the other files (which

main.py

imports) is changed, re-registration alone isn't enough to allow the agent to execute the updated flow, seemingly because only the content of

main.py

is stored on S3 at registration. Practically this means that almost every time we make any change any of our flows, we need to rebuild our docker image with the updated logic, redeploy it, and replace the old agent with the new one, before then re-registering the flow. Is there some way for us to register a flow so that all of the flow's code, not just the code in the file that defines the flow, is stored in S3 and we don't need to constantly rebuild and redeploy the agent's image for almost every change? Or is there a cleaner approach to solving this issue which has worked for anybody here? Thanks in advance.

Hey @Greg Roche -- this is a bit tricky because we can't load all of your other files (we don't know where your code is stored and recursing through all your imports and trying to infer where everything is and store it would be unreasonable). The discussion at https://github.com/PrefectHQ/prefect/discussions/4042 may be helpful.

Thanks very much for this info Michael, this will be super helpful for deciding our best course of action 🙂 much appreciated!