Ask Buggu to turn your PDFs and images into structured JSON. Zero server uploads, zero API costs, 100% on-device processing.

Live Demo

Buggu is your personal document extraction assistant that lives entirely in your browser.

Have an invoice? Ask Buggu.
Got a receipt? Ask Buggu.
Need data from a form? Ask Buggu.

The best part? Buggu never sends your documents anywhere. Everything happens on your device using Chrome's Built-in AI.

Every day, millions of people need to extract data from documents:

• Small businesses processing invoices and receipts
• Healthcare workers digitizing patient forms
• Students organizing study materials
• Freelancers tracking expenses

But here's the problem:

Current solutions require uploading sensitive documents to cloud servers:

• ❌ Privacy risks (invoices contain financial data, forms contain PII)
• ❌ API costs (per-document charges add up)
• ❌ Internet dependency (offline = unusable)
• ❌ Vendor lock-in (different APIs for OCR vs. extraction)

What if you could just ask someone you trust?

Buggu leverages Chrome's Built-in AI (Gemini Nano) to extract structured JSON from documents without ever leaving your browser.

✅ 100% Private
Buggu processes everything on your device. Your invoices, receipts, and forms never touch a server.

✅ Natural Language
Just tell Buggu what you want in plain English:
"Hey Buggu, extract the invoice number, date, and line items"

✅ Handles PDFs Like a Pro
Upload multi-page PDFs (up to 3 pages). Buggu converts them to images and "sees" the layout just like you do.

✅ Shows Its Work
Preview your documents before extraction. Watch Buggu process them in real-time.

✅ Zero Cost
No subscriptions, no API keys, no credit cards. Buggu is completely free.

✅ Works Offline
Once Chrome's AI model is downloaded, Buggu works without internet.

Frontend:  React 18 + Vite
Brain:     Chrome Prompt API (Gemini Nano) with Multimodal Input
PDF Eyes:  PDF.js (Canvas rendering at 2x quality)
Style:     CSS3 with modern gradients

┌─────────────────────────────────────────────────────────────────┐
│                     Buggu's Interface                          │
│  User Flow: Upload → Tell Buggu what you want → Get JSON        │
└────────────────────────┬────────────────────────────────────────┘
                         │
                         ▼
┌─────────────────────────────────────────────────────────────────┐
│                  📄 Document Processing                          │
│                                                                  │
│  PDF → PDF.js renders to Canvas → High-res PNG blobs           │
│  Images → Direct blob processing                                │
│                                                                  │
│  Why images? Buggu needs to "see" tables, forms, layouts        │
└──────────────────────────────────┬──────────────────────────────┘
                                   │
                                   ▼
┌─────────────────────────────────────────────────────────────────┐
│                  🧠 Buggu's Brain (Chrome AI)                    │
│                                                                  │
│  System Prompt: "You're Buggu, a data extraction assistant"     │
│  User Message: [What user wants + Document images]              │
│  Gemini Nano: Analyzes images & extracts structured data        │
│  Output: Raw JSON text                                           │
└──────────────────────────────┬───────────────────────────────────┘
                               │
                               ▼
┌─────────────────────────────────────────────────────────────────┐
│                  ✨ JSON Cleanup & Delivery                      │
│                                                                  │
│  -  Remove markdown artifacts                                     │
│  -  Parse & validate JSON                                         │
│  -  Show results with copy/download options                       │
└─────────────────────────────────────────────────────────────────┘

What You Need:

• Chrome Canary 127+ or Chrome Dev 127+

Wake Up Buggu (Enable Chrome AI):

1. Go to chrome://flags/#prompt-api-for-gemini-nano
2. Set to "Enabled"
3. Go to chrome://flags/#optimization-guide-on-device-model
4. Set to "Enabled BypassPerfRequirement"
5. Restart Chrome
6. Visit chrome://components/ and click "Check for update" on "Optimization Guide On Device Model"
7. Wait for download (this is Buggu's brain!)

# Get Buggu
git clone https://github.com/AnchitSingh/buggu.git
cd buggu

# Install dependencies
npm install

# Start Buggu
npm run dev

Visit http://localhost:5173 and meet Buggu!

npm run build
npm run preview

• Click "Convert to JSON"
• Upload images (JPG, PNG, WebP) or PDFs
• Buggu shows you previews (max 3 pages)

Use natural language - just like talking to a friend!

Buggu's Favorite Requests:

For Invoices:

Hey Buggu, extract the invoice number, date, vendor name, 
vendor address, line items with descriptions and prices, 
subtotal, tax, and total amount

For Receipts:

Buggu, get me the store name, date, time, all items with 
quantities and prices, subtotal, tax, and total

For Forms:

Buggu, extract all the form fields and their values

For Study Materials:

Buggu, organize this into sections with headings and 
key points as bullet points

• Click Process Files
• No internet needed (check your network tab!)

• Copy JSON to clipboard
• Download as .json file
• Use in spreadsheets, databases, or apps

✅ Prompt API - Buggu's reasoning engine
✅ Multimodal Input - Buggu's ability to "see" documents

Privacy-first document extraction for sensitive data.

Traditional services force you to upload invoices, medical records, tax forms, and personal documents to strangers' servers.

Buggu changes that.

With Buggu, your documents stay on your device. This means:

• ✅ Healthcare workers can process HIPAA-regulated documents safely
• ✅ Small businesses keep financial data private
• ✅ Students organize notes without data collection
• ✅ Freelancers track expenses without third-party exposure

Everyone with Chrome!

• Regional: Works worldwide (no geo-restrictions)
• Audience: Small businesses, students, healthcare, freelancers, privacy-conscious users
• Languages: English (more coming soon via outputLanguages)

Buggu uses Chrome's multimodal Prompt API to "see" documents:

const bugguMessage = {
  role: 'user',
  content: [
    {
      type: 'text',
      value: 'Extract invoice data: invoice_number, date, items...'
    },
    {
      type: 'image',
      value: invoicePageBlob
    }
  ]
};

const result = await bugguSession.prompt([bugguMessage]);

PDFs have visual structure that plain text loses:

• Table rows and columns
• Form field positions
• Handwritten notes
• Logos and stamps

Buggu renders PDFs to 2x high-res images so Gemini Nano can "see" this structure:

const viewport = page.getViewport({ scale: 2.0 });
await page.render({ canvasContext, viewport }).promise;
const blob = await canvas.toBlob('image/png', 0.95);

❌ Upload your documents
❌ Store your data
❌ Track your extractions
❌ Send telemetry
❌ Call external APIs

✅ Process locally
✅ Use Chrome's on-device AI
✅ Clear memory after extraction
✅ Give YOU full control

Trust, but verify: Open DevTools → Network tab while using Buggu. You'll see zero outbound requests during processing.

MIT License - see LICENSE file

Buggu - Your Privacy-First Document Extractor

Built with ❤️ for the Chrome Built-in AI Challenge 2025

"Ask Buggu. Keep it private."