Been working on a self-hosted AI security agent that runs completely offline — no cloud, no APIs, no external calls.
It ingests system logs, embeds them into vector memory, promotes recurring patterns, and responds based on trust level. The goal is to explore what real autonomous infrastructure might look like: memory-driven, local, and private by default.
Still very much a prototype. Code isn't public yet, but the whitepaper outlines the architecture and direction. Feedback welcome.
1 comment
[ 2.9 ms ] story [ 13.6 ms ] threadIt ingests system logs, embeds them into vector memory, promotes recurring patterns, and responds based on trust level. The goal is to explore what real autonomous infrastructure might look like: memory-driven, local, and private by default.
Still very much a prototype. Code isn't public yet, but the whitepaper outlines the architecture and direction. Feedback welcome.
Whitepaper Paper(View): https://codeberg.org/ptrb25/ptrb25/src/branch/main/whitepape... Whitepaper Paper(Raw): https://codeberg.org/ptrb25/ptrb25/raw/branch/main/whitepape... Repo: https://codeberg.org/ptrb25/ptrb25