This project allows to use a VLM for anomaly identification. It is possible to either use a VLM hosted locally, in a quantized form, or deployed in a distant server.

Refer to sit-aw-api documentation to learn how to get started and use it.

In this documentation, the deployment of a VLM is explained and the way to communicate with one of the two model choices, given CONVINCE use cases. For now only UC1, vacuum cleaner, and UC2, assembly robot. The last of the documentation explains how to custom the communication, to extend the work to other robotic use cases.

The tests have been done only on LINUX.