While building ThreatIQ, one challenge was making sure our app could handle real-time threat detection without slowing down. Even though we’re not showing live video to users, we still needed to process video in the background — and that brought its own set of issues.

At first, the alert updates were causing some performance hiccups. Here’s what we did to fix it:

Separated out the alert logic from the rest of the app to reduce re-renders.

Optimized how we fetched and updated threat data.

Used tools to catch unnecessary state updates and clean up the UI performance.

But the biggest challenge was getting a real CCTV feed to test with. Since we didn’t have access to actual CCTV cameras, we had no real input for our detection model.

So we worked around it by:

Setting up a local RTSP stream using a sample video.

Hosting it on localhost to act like a live camera feed.

Feeding that stream into our threat detection model to simulate real use.

Additionally, we faced a limitation with the Groq model, which does not support video analysis. To work around this, we broke the CCTV footage into 5-second chunks, then converted those chunks into a single frame (1 frame per 5 seconds). These frames were then sent to our Groq model for analysis.

This gave us a reliable way to test without needing real CCTV hardware. It also let us focus on what matters — showing users only the threats that are detected, not the raw video — keeping the interface clean, fast, and focused.