Downsides of having a crew per agent?

[flossypurse]

The context surrounding this questions is...
I asked myself, what happens if the process crashes half way through a multi-agent workflow?
So I put each agent into its own crew, and then saved the results of the agent into a database before passing the result to the next agent - so that the workflow could resume from the last agent's successful result.

This appears to work - but I am not sure if CrewAI enables better end results if all the agents are in a single crew?

So, first question - does CrewAI plan on making agent results durable with something like resonatehq.io ? And if not, second question - then what are the downsides of splitting up agents into crews so that I can make results durable in longer running agentic worklflows?
It seems like the only downside is that I have to manually pass a result from 1 crew to another - but want to double check that.

[mouramax]

Regarding "does CrewAI plan on making agent results durable?", I suggest you take a close look at CrewAI's Flow paradigm. Specifically, you can find the documentation on Flow persistence here.

The code presented in the documentation isn't compatible with the latest stable version (0.105.0), but in this response here, I documented a working example that allows you to save and restore the flow's state between executions. I hope this helps you think about it more clearly.

[flossypurse]

This is super helpful - thanks!