PDFscaler Master Guide
A comprehensive developer and user reference manual on client-side sandboxes, algorithms, and in-browser workflows.
Search the Knowledge Base
💡 Pro-tip: PDFscaler completes all computations locally in your browser memory. We never store or transmit your documents.
Engine Architecture
- Compilation: React 18 / Vite / TypeScript
- Privacy Guard: 100% Volatile Sandboxing
- Execution Layer: Web Workers & HTML5 Canvas
Security & Volatile Memory
How PDFscaler enforces military-grade privacy by isolating document mutations directly within client RAM.
In the modern internet ecosystem, most utility web applications operate as "data funnels." When you click "compress" or "merge," your files are secretly dispatched to remote server clusters, written to persistent disk arrays, and then deleted at a later date. This exposes financial reports, medical charts, and source files to data leaks, system misconfigurations, or network eavesdropping.
PDFscaler breaks this mold entirely. By compiling core computation tools to run directly within your browser’s isolated scripting context, 100% of your raw document content remains in volatile memory (RAM) and vanishes completely the moment you close or refresh the browser tab.
How Client-Side Execution Works:
- Local Array Buffer Access: When you select a file via drag-and-drop, our framework requests a temporary, sandbox-isolated pointer using standard HTML5
FileorBlobAPIs. No upload trigger occurs. - Direct Byte-Stream Mutation: Libraries such as
pdf-libparse the local stream array byte by byte in-memory. Cross-reference indexes, interactive form fields, and object metadata catalogs are manipulated in active JavaScript memory. - Asynchronous Web Workers: Heavy operations like image cropping and file compression run inside secondary background threads (Web Workers). This prevents the browser UI thread from freezing, ensuring seamless, lightning-fast rendering of high-resolution graphic canvases.
Secure Proxy for Advanced AI Features:
PDFscaler includes advanced features like Image to Text (OCR) and Image Translation. Because deep visual transcription models require industrial hardware to execute quickly, these requests are channeled through our secure backend API directly to the Gemini API.
Our server proxies act as a secure pass-through. They attach credential metadata safely and securely, pipe the image data directly to the Gemini models, and yield the text. The proxy never saves, logs, caches, or inspects your images. Once the session concludes, the volatile image buffer is instantly destroyed.