Hi there - apologies if this is a slightly vague question, but hoping it might spark an insightful conversation.
Context: I’m a software engineer with 20+ years experience, building end to end systems mostly for e-commerce, and mostly in the .NET space.
I’m starting to play with building my own simple agentic systems and have started with CrewAI. Loving it so far.
What I want to get to, is building systems like those I used to build, but where a component of it is something like CrewAI. Where for just a portion of the system, it calls out to an agentic system and leverages everything it’s got to give, but where the rest of the system comprises of all the standard parts of the tech stack - backends, web apis, frontends, databases, caches, queues, etc.
How would one use CrewAI in this manner? Just import the whole thing in and call the run method in but not from main, just from wherever it fits into the rest of the system?
Is this how CrewAI is meant to be used?
That’s question #1, and it kinda leads onto question #2.
Say I build a simple process where an agent creates a summary of some info it finds on the web, and then I want to save it to a file. The ‘saving to a file’ part - should that also be an agent? I kinda feel like it shouldn’t be - I just want it to write to a given file path so there’s really no need for AI to get involved - or am I thinking about this the wrong way?
The reason I ask is because I tried to build something like the above, but instead I need an agent to read a file and send it via email - and it just wouldn’t honour the file path I was passing it in as input_task. It would often change the path slightly, but enough that the file can’t be found. Maybe I was just doing it wrong - either way, I’m curious 
Any thoughts on the above would be much appreciated!