Privacy-First AI-Powered PDF Form Detection & Filling
100% Client-Side • On-Device Neural Detection • Flawless Complex Unicode & Bengali Support

---

Fillora is a modern, high-performance, browser-native PDF form detection and filling web application inspired by pdf-forms.tools.

It operates 100% on the user's device (client-side). Document files, user entries, uploaded fonts, and extracted metadata never leave the browser sandbox.

---

🌟 Key Features

1. 100% Client-Side Privacy Architecture

• Zero Server Uploads: PDFs, user input, and extracted data are processed purely in-memory within the browser.
• On-Device AI Inference: Uses onnxruntime-web with WebGPU acceleration and multithreaded SIMD WebAssembly (WASM) fallback.
• Offline Capable: All neural models, WASM runtimes, and Bangla fonts are cached locally via the browser Cache API.

2. AI-Powered Form Field Detection

Supports both official CommonForms FFDNet models running locally in a Web Worker:

• FFDNet-S (Small): ~36 MB, ~6M parameters. Optimized for low-latency client-side CPU/GPU inference.
• FFDNet-L (Large): ~97 MB, ~25M parameters. High accuracy model ideal for complex, irregular, or scanned documents.
• Instant Visual & AcroForm Detector: Instant 0ms visual detection for line underlines, boxes, checkboxes, and pre-existing AcroForms.
• Hybrid Mode: Combines neural object detection with structural document contours.
• Detection Scopes: Detect on the Current Page or scan All Pages.

3. Automatic Text Overlays & Direct Form Filling

• Detected bounding boxes immediately receive perfectly aligned text input overlays.
• Locked by Default: Overlays cannot accidentally move or resize while typing, preserving original form alignment.
• Unlock / Lock Controls: Toggle lock to freely reposition and resize fields with 8 draggable handles.
• Auto-Fit Typography: Dynamically scales font size to fit field dimensions, with manual override sliders.
• Single-line & Multiline: Dedicated support for single-line inputs and expansive textarea summary boxes.

4. Flawless Bangla (Bengali) & Complex Unicode Support

• OpenType Complex Script Shaping: Built specifically to avoid broken glyphs and disconnected letters in Indic scripts.
• Full Support for:
  • Bengali vowel signs (কার: া, ি, ী, ু, ূ, ৃ, ে, ৈ, ো, ৌ)
  • Consonant clusters and conjuncts (যুক্তাক্ষর: ক্ত, ক্ষ, জ্ঞ, ষ্ণ, ঙ্ক, ঙ্গ, ত্র, ক্র, ইত্যাদি)
  • Reph (রেফ র্ক) and Phala forms (য-ফলা, র-ফলা, ব-ফলা)
• High-DPI Flattening Pipeline: Renders shaped typography via high-resolution (300+ DPI) transparent canvas layers into pdf-lib, guaranteeing that the exported PDF looks identical to what is displayed on screen without fontkit anchor lookup crashes.
• Font Management:
  • Bundled with Kalpurush, Noto Sans Bengali (Regular & Bold), and standard English typography.
  • Custom font uploader supporting user TTF and OTF files.
  • Apply fonts per-field or globally across the document.

5. Checkbox Detection & Multi-State Interaction

• Automatic detection of checkbox widgets.
• 4-state click cycle:
  1. Check mark (✓)
  2. Cross mark (✗)
  3. Dot (•)
  4. Empty / Unchecked
• Manual checkbox creator to place and resize checkboxes anywhere on any page.

6. Clean Flattened PDF Export

• Generates clean, flattened PDFs preserving original dimensions, vector lines, and page quality.
• Unfilled/empty fields and all editing controls (handles, borders, selections) are completely stripped away.
• Flawless vector and high-DPI output with zero overlay artifacts.

---

🛠️ Technology Stack

• Framework: React 19, TypeScript, Vite
• Styling: Tailwind CSS v4, Lucide Icons
• PDF Engine: pdfjs-dist (rendering, thumbnails, AcroForms) & pdf-lib (document generation and flattening)
• AI Inference Engine: onnxruntime-web (WebGPU / WASM execution providers)
• Concurrency: Dedicated Web Worker (detection.worker.ts) for non-blocking neural inference

---

🚀 Quick Start

Installation

# Clone the repository
git clone https://github.com/rmahbub01/fillora.git
cd fillora

# Install dependencies
npm install

Running the Development Server

npm run dev

Open http://localhost:5173 in your browser.

Building for Production

npm run build
npm run preview

Running Verification Tests

node tests/core-functionality.test.mjs

---

📋 Keyboard Shortcuts

Shortcut	Action
Ctrl + Z / Cmd + Z	Undo
Ctrl + Y / Cmd + Shift + Z	Redo
Ctrl + L / Cmd + L	Lock / Unlock selected field
Tab	Focus next field on current page
Shift + Tab	Focus previous field on current page
Delete / Backspace	Delete selected field
Escape	Deselect active field

---

👨‍💻 Developer

Fillora is designed and developed by Mahbub Rahman.

• GitHub Profile: @rmahbub01
• GitHub Repository: https://github.com/rmahbub01/fillora

---

📄 License

MIT License. Designed and engineered for privacy-conscious users worldwide.