Hey everyone,
I've been working on a tool called ISNAD for the self-hosted community. It's an open-source (Apache-2.0) engine that creates a locked, cryptographically signed record of an AI's answer—showing who or what handled it at each step.
The goal is to provide "proof of custody" for AI decisions. Instead of just a standard log, you get a tamper-evident record that you can verify offline. This is especially useful for anyone dealing with the EU AI Act's upcoming logging requirements (Art 12 & 19).
It runs entirely on your own infrastructure, so your data never leaves your control.
You can
pip install isnad and find the repo here: https://github.com/alizahidraja/isnadCurious to hear what the self-hosted community thinks about the architecture and approach.
https://redd.it/1x0z15r
@r_SelfHosted
