← Back to Projects

MirrAi - A powerful tool for Virtual Try-on

Download Code / Files↓
Mirrai is an AI-powered Chrome extension that enables seamless virtual clothing try-ons directly on e-commerce platforms like Myntra. The project utilizes a serverless architecture with Vercel and Supabase to manage a hybrid credit system.

I just wrapped up the architecture for Mirrai—an AI-powered virtual try-on Chrome Extension that works directly on sites like Myntra.

Here is the exact stack I used to take it from a local ComfyUI workflow to a production-ready micro-SaaS:

The Brain: Headless ComfyUI workflows running Flux models via the RunningHub API.
The Proxy: Vercel Serverless functions securely handling CORS, JWT verification, and API routing.
The Frontend: A Manifest V3 Chrome Extension.
The Infrastructure: Supabase (PostgreSQL) tracking a complex hybrid state—moving users seamlessly from anonymous free credits stored in local storage, straight into authenticated paid sessions.
The Revenue: Razorpay webhooks integrated directly into Vercel to automatically update the Supabase credit ledger upon successful payment.

The biggest technical hurdle? Not the deep learning model itself.

It was translating a chaotic, visual ComfyUI node graph into strict, production-ready JSON payloads that a cloud API wouldn't instantly reject. Moving from visual coding to strict cloud validation breaks everything, and debugging it was a massive learning curve.

If you are building AI products, working with ComfyUI APIs, or dealing with Chrome Extension state management, I’d love to connect and swap notes.

#buildinpublic #AI #softwareengineering #saas #microservices