MLOps(github.com)

8 pointsby hexer303il y a 3 mois2 comments

Show HN: A collaborative SSH copilot for on-calls/DevOps/MLOps

https://github.com/few-sh/fewshell

3 comments

Going after a problem almost everyone will have at some point, with a safe and secure approach. It seems like a great way to use AI, it boosts your ability to walk away from your laptop and spend time outdoors, as long as you have your phone with you it's possible to keep your long running jobs supervised and moving forward.

How many tokens did it take to develop this project? Do you know what models you used? It feels like a Claude assisted project, but do you have to keep maintaining it with AI or is it a question of architecture skill that keeps it safe and reliable now?

hexer303il y a 3 mois

Thank you. Not many. I just use the Github copilot $40/month plan and it suffices.

I run it in manual approve mode and review/adjust every edit. Mostly auto-select model at 10% discount, or Claude frontier for more complicated tasks.

My dev process follows the same philosophy as the app itself: AI can suggest, but Human must steer and approve.

newsdeskxil y a 3 mois