Inspiration

Couple of days ago a pack of stray dogs came into my house compound trying to hunt down our cats and it ended up destroying my cover for the car, this is is third time dogs are tearing away the car cover, I wished there was a way to monitor if external dogs come and try to tear of the cover of my car.

What it does

I created AI Vision Monitor at the Bolt.new hackathon because current monitoring tools were either too inflexible or too complex to adapt to my needs. Based on everyday normal experiences—such as keeping an eye on my things while I cook or securing my home office—I needed an easy but effective means to monitor for anything using natural language.

How we built it

I employed Bolt.new's no-code setup to bring all of it together. The UI has an uncluttered live feed viewer, a snug prompt box for commands such as "detect knife" or "spot person," and a responsive alert panel that flashes when the AI picks up something that corresponds to the prompt. In the process, I had to deal with real-time event handling, prompt crafting for robust detection, and prodding the APIs to provide consistent, low-latency results.

Challenges we ran into

The biggest issues were latency and false alarms. Alerts were initially delayed when many frames accumulated, so I noticed adding frame throttling and debouncing could maintain responsiveness. I had issues getting Google cloud vision API keys so I used Open AI key. Fine-tuning those prompts was important to making Vision Monitor accurate and user‑friendly.

Accomplishments that we're proud of

Now we are confident of building any complex apps with Bolt, I have already started to even rebuild some android apps built over many years using Bolt. I understood the value add various partners of Bolt bring to table from Supabase to Netlify to deploy the full stack app.

What we learned

I picked up so much during the project about integrating two powerful AI services—Google Cloud Vision for real-time computer vision and OpenAI for natural‐language understanding. I learned to stream live video from a web cam into the app, pass frames to the vision API for inspection, and then pass that data to OpenAI to figure out what the results indicated based on the typed input of the user. Getting the choreography of these asynchronous calls and showing immediate visual feedback in the browser was a difficult but rewarding task.

What's next for AI Vision Monitor

In the end, I'm glad I've built something that allows anyone to "watch" their camera for anything—safety concerns, package delivery—using nothing but plain language and strong AI. It's exciting to think of how this could evolve—adding voice control, edge-camera support, or smart notifications. For now, AI Vision Monitor is a clean proof‑of‑concept that shows you can push Bolt.new, OpenAI, and Cloud Vision to the limit in one clean, user-friendly app.

Bonus prizes

Creative Use of AI:- This project chains Google Cloud Vision and OpenAI in a novel way: not just detecting objects, but interpreting user‑defined prompts in real time. Users can type anything ("detect package", "spot pets", "flag helmet") and the AI transforms a live video feed into targeted alerts—an unexpected blend of generative and visual AI.It isn’t just “vision” or “language”—it’s a dynamic bridge between the two, enabling a new form of interactive, AI-powered monitoring

Uniquely Useful Tool:- Before this, there wasn’t a simple tool that lets you "watch" a camera for anything, using plain English as the trigger — no complex configurations or rule engines. It solves a real, everyday pain point—from security checks and package deliveries to spotting hazards—by democratizing surveillance through natural language.

We Didn’t Know We Needed This:- It’s a delightfully surprising idea that immediately resonates: “Oh, I really want this.” We often forget we need monitoring until it’s too late. This tool makes it accessible and proactive—you type a prompt and leave it. Its value isn’t obvious until you see it in action. once you do, you’ll wonder how you ever lived without live, AI-driven oversight of your everyday spaces.

Built With

  • bolt
  • openai
Share this project:

Updates