First Release · June 2025
https://perkpdf.com
I built Perk PDF to scratch my own itch: desktop PDF suites are clunky, subscription-locked, or demand installs. Modern browsers + a tiny Flask API can handle heavy PDF ops in milliseconds—no signup, no ads, no data hoarding.
🔧 Tech Stack & Architecture
- Frontend: React + Vite, vanilla CSS-in-JS, intuitive drag-and-drop UX
- Backend: Flask on Render/Vercel (U.S. servers), stateless API, auto-purging temp/ after each request
- PDF Ops:
- PyPDF2 for merge/delete/reorder
- Ghostscript for lossy/lossless compression
- pdf2docx for PDF→Word
- Pillow for images→PDF
- PyMuPDF (fitz) for PDF→PNG
- pdfminer.six + pdfplumber + pandas for text & table extraction
- Security & Privacy: HTTPS/TLS, CORS-restricted, files streamed back and deleted (usually within seconds), GDPR/Iceland-compliant consent mode
🤖 AI-Powered PDF Analysis
- AI Model: Anthropic Claude 3 Haiku - a state-of-the-art AI model optimized for speed and accuracy
- AI Tools:
- Chat with PDF: Interactive Q&A about your document content
- AI Explain PDF: Comprehensive analysis with key topics, main points, and recommendations
- AI Summarize PDF: Intelligent document summarization with structured insights
- How It Works: PDF text is extracted and sent to Anthropic’s Claude API, which processes the content using advanced natural language understanding to provide intelligent responses
- Privacy: AI processing happens on Anthropic’s secure infrastructure with enterprise-grade privacy protection
- Performance: Fast response times with high accuracy for document analysis and summarization
- 100 MB scan → 5 MB compress in ~2 s
- Table extraction from 20-page PDF in 0.5 s
Why It Matters
- No installs: Works on any modern browser (even iPad/Chromebook)
- Zero friction: No accounts, no ads, open-source libs only
- Privacy-first: Files don’t linger—ever
Looking for feedback on:
- Real-world edge-cases (huge page counts, weird PDF flavors)
- AI analysis accuracy and response quality
- Additional AI capabilities (translation, sentiment analysis, etc.)
- Potential optimizations via WebAssembly
- GTM/consent integrations for GDPR
Give it a spin and let me know what breaks!
👉 https://perkpdf.com
Happy to open-source more components or deep-dive into wasm-proof-of-concepts if there’s interest.
