Inspiration

Traditional cybersecurity is reactive, relying on static blacklists that scammers easily bypass. We were inspired to build a "Live Agent" that actually understands the deceptive language of a scam. Aegis-Live was created to give users a proactive partner that not only detects threats but speaks back to ensure they stay safe.

What it does

Aegis-Live is a multimodal browser guardian. When a user visits a site, the extension extracts page intent and sends it to our Gemini-powered backend. If a threat is detected:

  1. A Visual Banner is injected into the page.
  2. An AI Voice Alert triggers to verbally warn the user.
  3. Our Aegis-Scout agent updates a global Firestore database to protect other users from the new threat pattern.

How we built it

  • Backend: Python FastAPI hosted on Google Cloud Run.
  • AI Brain: Gemini 1.5 Flash (via Vertex AI) for high-speed content analysis.
  • Database: Google Cloud Firestore for real-time threat intelligence storage.
  • Frontend: A Manifest V3 Chrome Extension featuring a custom "Aegis Threat Vault" UI.

Challenges we ran into

Navigating the transition to Manifest V3 was a significant hurdle, specifically handling Service Worker imports and asynchronous communication. Additionally, we had to optimize our container deployment on Google Cloud Run to ensure the AI's response time was fast enough for real-time browsing protection.

Accomplishments that we're proud of

We successfully built a "Zero-Error" V3 extension that interacts seamlessly with a cloud-hosted AI. Seeing the agent successfully trigger a voice warning for the first time was a major milestone that proved the power of multimodal safety.

What we learned

We learned the importance of "Agentic Intelligence"—the idea that a tool shouldn't just show data, but should take proactive steps (like speaking) to help a user make better decisions.

What's next for Aegis-Live

We plan to implement regional Indian language support for the voice alerts and add a community-driven reporting system that allows Gemini to learn from user-verified scams in real-time.

Built With

  • chrome-manifest-v3
  • fastapi
  • gemini-2.5-flash
  • google-cloud-firestore
  • google-cloud-run
  • javascript
  • python
  • vertex-ai
Share this project:

Updates