Proxy is ON — every prompt is inspected by Cisco AI Defense before it reaches Ollama.
Fire a test prompt
Conversation
Session log
- No requests sent yet this session.
AI Supply Chain Blocking
Pre-demo checklist
AI Defense — Automated Red Teaming
This demo shows Cisco AI Defense's AI Validation feature running an automated red-team scan against a real model — the Ollama instance running on this same homelab server — rather than inspecting live traffic the way the AI Defense tab does. Validation probes the model directly with adversarial prompts and reports back where it held up and where it didn't.
-
1
Enable the Ollama proxy host
In NGINX Proxy Manager, enable the
ollama.victorialanetempleton.comproxy host. Do this the night before or remotely — it does not need to happen live on stage. -
2
Confirm it's reachable
Visit
https://ollama.victorialanetempleton.comyourself, confirm the basic auth prompt appears, and confirm Ollama responds after authenticating. -
3
Confirm the AI Defense target
In the AI Defense console, confirm the saved validation target (endpoint
https://ollama.victorialanetempleton.com, with the basic auth credential) and profile still exist and are correctly configured. -
4
Run it live
On stage, in the AI Defense console, kick off the validation run against the saved target/profile in front of the room, and wait for the report together.
-
5
Disable the proxy host afterward
Once the demo and event are done, disable the
ollama.victorialanetempleton.comproxy host in NGINX Proxy Manager again — same standing practice as theai.subdomain.
Where things live
- NPM admin: http://nyx:881
- The basic auth credential for the Ollama subdomain is stored in NPM's access list — not printed here.
- If the Ollama subdomain hasn't been used in a while, re-check DNS and cert status before relying on it.