you might be interested in prompt-guard. It is fully offline and works on individual machine - so sensitive info never leaves user's machine at all - not even to some remote service. https://github.com/chaudharydeepak/prompt-guard happy to explain / collaborate
Very cool project for those that only need protection locally with hand rolled rules. We're aiming to protect data flowing from SaaS solutions to LLMs and for those that don't want to manage the complexity of running a MITM locally.
Thank you for the feedback. There are about 230 rules currently available - with flexibility to add customize more as you see fit. Like you, my vision was, if I want to protect something I would not want to get it on wire in first place, hence local :)
We want to protect custom apps but also between all of the numerous SaaS products that don’t run locally and the guantlet of LLMs. We’re plug and play with virtually every SaaS offering that talks to LLMs.
and as I read more - it seems your solution is more aligned to specific industry vs I am approaching more from developer / general use perspective. cheers!