DocuScan

The simple, private, AI-native place to scan, edit, convert, organize, sign, summarize, fill, share and extract data from documents — web-first and mobile-first, no signup needed for basic use.

Most processing runs in your browser: PDF transforms, OCR and text extraction happen on-device, and AI tools send only the extracted text to the API — your files never leave your device.

Stack

Next.js 15 (App Router) · TypeScript · Tailwind v4 · @cantoo/pdf-lib (fork with in-browser encryption) · pdfjs-dist · tesseract.js · OpenCV.js (CDN) · jszip · qrcode · docx · fabric · provider-agnostic AI (@anthropic-ai/sdk / openai, server-only) · @upstash/redis · @aws-sdk/client-s3.

PDF code imports from @cantoo/pdf-lib, not stock pdf-lib. Don't re-add stock pdf-lib.

Getting started

npm install
npm run dev -- -p 3939     # dev (port 3000 is often taken)
npm run build              # production build (type-check + lint)
npm run start -- -p 3939   # production server

Copy .env.example → .env and fill in what you need. Everything is optional; with no keys set the app runs fully (AI features return 503 until a key is added).

Pure-logic tests

# Run one:
node --experimental-strip-types scripts/test-share.mjs

# Run them all (same loop CI uses):
for f in scripts/test-*.mjs; do node --experimental-strip-types "$f"; done

Covers share, smartnotes, templates, invoice, printready, expenses-reminders, editor, docexport, deck.

Continuous integration

.github/workflows/ci.yml runs on every push/PR to main: npm ci, all scripts/test-*.mjs, then npm run build (compile + type-check + lint). It needs no secrets — the app falls back to in-memory backends — so the build must stay green.

Tools

PDF: Scan/Image → PDF, Edit, Merge, Split, Compress, Rotate, Organize (reorder/delete), Watermark, Page numbers, Protect, Unlock, Batch, Convert, Print Ready. Scan: Auto-crop, Enhance scan, Clean scan, Handwriting → text, Enhance image. AI: OCR, Summarize, Chat (Ask PDF), Make editable, Redact, Translate, Compare, Extract to table, Analyze contract, Draft, Rewrite, Study aids, Audio, Slides, Smart assistant (name/tag/extract), Resume Scanner, Receipt Scanner. Forms & signature: Sign, Fillable PDF / Smart form fill, Saved Text Templates. Productivity: Smart Notes (auto-format pasted text → PDF/Word/MD/TXT), Document Reminders, Document QR Labels, Invoice Maker. Sharing: secure temporary share links + QR, WhatsApp/Gmail/native share, Dropbox/Drive save.

See docs/DOCUSCAN_EXPANSION_PLAN.md for the audit + architecture, docs/DOCUSCAN_EXPANSION_TODO.md for the checklist, and docs/QA_CHECKLIST.md for QA.

Architecture notes

• Each tool = a route under app/<tool>/: a "use client" page.tsx + a tiny server layout.tsx that exports toolMetadata(slug) from lib/seo.ts. Register new tools in components/ToolGrid.tsx.
• AI lives in one provider-agnostic module (lib/ai.ts) behind app/api/ai (a task switch, structured JSON, rate-limited 10/5min/IP, 503 without a key).
• No database / no auth yet (by design — no signup for basic use). "Saved" data for the productivity tools is device-local (localStorage), written behind small typed lib/* stores so the backend can later swap to an authenticated API.
• Storage seams: in-memory by default; env-selected Redis (usage/ratelimit) and S3/R2 (shares) activate when their env vars are set.

Privacy

Files are processed in the browser and temporary shared files auto-delete after expiry. AI tools transmit only extracted text. See /privacy and /security.