Analysis Cheque Built-in AI

A Next.js application demonstrating secure, on-device cheque data analysis using Chrome's Built-in AI APIs—no data leaves the browser.

🚀 Live Demo URL: https://demo-analysis-cheque.vercel.app
🏆 Built for the Google Chrome Built-in AI Challenge 2025

This README provides setup and testing instructions for tester and developers. For the full problem statement, architecture, and feature details, see:

• 📄 HACKATHON_SUBMISSION.md
• 🏗️ ARCHITECTURE.md

---

Tech Stack & Tooling

• Framework: Next.js (App Router)
• Styling: Tailwind CSS v4
• UI Components: shadcn/ui
• AI Engine: Chrome Built-in AI APIs (LanguageModel - Gemini Nano)
• Node Requirement: Node.js LTS 22+
• Package Manager: pnpm

---

Prerequisites

To run this project, you must have the following installed:

1. Node.js: Version 22 or higher.

2. pnpm: Install it globally via:

   npm install -g pnpm

3. Google Chrome: Version 140 or newer (or a Canary/Dev build) where the Built-in AI APIs are active.

⚠️ Testing Requirement

The core functionality relies on experimental APIs. If you encounter issues, please ensure the necessary feature flags are enabled in your Chrome browser:

1. Open Chrome and navigate to chrome://flags.
2. Search for flags related to "On-device AI" or "Gemini Nano" and ensure they are set to Enabled.

⚠️ These APIs are experimental.

---

Getting Started

The application runs locally on port 3000.

1. Clone the repository:

   git clone https://github.com/2manoj1/analysis-cheque-built-in-ai.git
   cd analysis-cheque-built-in-ai

2. Install dependencies:

   pnpm i

3. Run the development server:

   pnpm run dev

4. Open the application: The application will be accessible at: http://localhost:3000

---

Testing Scenarios & Verification

1. Visual Analysis & Data Extraction Test

This test verifies the application's core functionality: using on-device AI for visual analysis and structured data extraction.

1. Download a Sample Cheque: Navigate to the sample image repository: https://github.com/2manoj1/ocr-img-cheque/tree/main/images Download any image file (e.g., 1.jpg).

2. Step 1: Upload

   • In the running app, click the "Choose file" button.
   • Select the downloaded cheque image.
   • Click "Extract & Prefill".

3. Expected Result for Step 1: The system should display a status message confirming that extraction is running entirely on-device.

4. Step 2: Review

   • The app should automatically move to the "Review" step (2).
   • The form fields (e.g., Cheque Number, Payee, Amount, IFSC etc) must be accurately populated by the on-device AI.
   • Verify extracted fields (Cheque No., Payee, Amount, IFSC, etc.)
   • Edit if needed

5. Steps 3–4: Add remarks → Run AI analysis → Try translation

All processing happens locally in your browser—no network calls, no cloud.

---

Licensing

This project is released under the MIT License.