Sentry AI – Resilient Network Guardian

• 

  Sentry AI Dashboard: Real-time network discovery with active 'Protective Mode' alert due to API quota limits.

Inspiration

Most network tools are built on the assumption that cloud services are always available. I wanted to build a security auditor that remains functional and provides visibility even when external APIs or internet connectivity are compromised.

What it does

Sentry AI performs real-time ARP scanning to discover every device on a subnet. It uses Gemini AI to analyze device behavior, but its unique feature is the "Protective Mode"—a fallback engine that maintains a live network feed even if AI quotas are exhausted.

How we built it

Backend: Flask and Python.

Scanning: Custom discovery engine built with Scapy for low-level packet manipulation.

Intelligence: Google Gemini 3 Flash API integration for behavioral auditing.

Challenges we ran into

Handling API Rate Limits (Error 429) was a major hurdle. I had to engineer a specific fallback state in the application to prevent the dashboard from going blank when the quota was hit, ensuring continuous network visibility.

Accomplishments that we're proud of

Successfully creating a resilient hybrid architecture that prioritizes local network visibility while leveraging the power of cloud AI when available.

What we learned

We learned the importance of Graceful Degradation in system design and how to manage namespace conflicts and undefined variables during complex multi-module integration.

Built With

• aur
• flask
• gemini
• python
• scapy