One control point for every model call and every tool call
Every request below goes through the gateway in this lab: it identifies the caller, masks personal data, applies policy, routes, calls the model or tool, checks the answer and writes the audit record before you see a result. Each step links to the evidence in Kibana.
The model never sees the personal data
Send a prompt as one of two applications. The lab's echo model repeats exactly what reached it, so you can compare what you typed with what the model received.
Send a prompt
What happened
Send a prompt to see the result.
Kill switch
Stop every request from tenant-beta on every gateway node, then release it. Each change is an audited admin action with a reason.