I'm genuinely curious whether anyone here has actually built and used AI agent automation running entirely on a local LLM.
I'm thinking of automating my current work with CrewAI (my initial goal is office automation like web crawling, data cleaning, etc.).
I'm curious about actual computer specs, how much performance you're getting, and your subjective satisfaction level.
I'd like to build a personal work-automation portfolio with CrewAI and keep researching it on my own after switching jobs,
so I'm wondering whether it's worth investing 5-6 million won in specs for the next Mac Mini, whether I can reasonably build this myself,
how detailed the tasks need to be to get usable performance, and how much abstraction is appropriate, among other things.
In the past, on an M3 MacBook Air 16GB, I couldn't even get OpenClo with QWEN 7B running properly and it kept stuttering badly, so I've been wondering if this is even the right approach.
I'd appreciate any advice from those with more experience!
