A chrome/Edge (Manifest V3) extension that reads a Privacy Policy / KVKK notice - from the current web page, or from an uploaded PDF/DOCX file (including scanned/image-only PDFs via on-device OCR) - analyzes it with Claude, flags risky clauses, scores it out of 100, and keeps a local history you can revisit and compare.
- PDF analysis - text-layer extractino via pdf.js
- DOCX analysis - via mammoth.js
- OCR - if a PDF's text layer is nearly empty (scanned document), pages are rasterized and run through Tesseract.js OCR (English + Turkish) entirely on-device before analysis
- History - every analysis (page, PDF or DOCX) is saved locally so u can reopen it later
- Comparison - pick any two saved analyses and get a short AI-Written verdict on which one is safer and why
Carried over from V1: auto page-text extraction, risk score /100, color-coded risky clauses, plain-language explanations and elapsed-time display
Not in this version yet: browser-wide "warn before u accept" mode, data-flow diagrams, a knowledge base of KVKK articles with citations (RAG), local/offline LLM support, and a public API
- Unzip this folder somewhere on ur machine
- Go to 'chrome://extensions'
- Turn on Developer mode (top right)
- Click Load unpacked
- Select this folder ('IReadIt)
- Click the extension icon, then the icon in the popup (or right-click the icon -> Options) to open Settings
- Paste an API key from console.anthropic.com/settings/keys and click Save
The key lives only in this browser's local storage and is sent directly to 'api.anthropic.com' when u run an analysis - no other server is involved
Page tab - open a site's Privacy Policy / Terms / Cookie Policy page, click the extension icon, then Analyze This Page
File tab — drag a .pdf or .docx file into the drop zone (or click to choose one), then
▶ Analyze File. If a PDF turns out to be a scanned image (no real text layer), the extension automatically
falls back to on-device OCR — this is shown as a status message while it runs and flagged on the result screen.
History tab — every analysis you run is listed here with its score, source icon, title, and timestamp. Click an item to reopen its full result. Tick two items and press ⇄ Compare Selected to get a side-by-side score/category comparison plus a short written verdict on which one is safer.
- Very long documents (~20,000 characters) are truncated before being sent for analysis.
- PDF text extraction reads up to the first 40 pages; OCR fallback is capped to the first 12 pages (OCR is slow).
- OCR uses a single bundled WASM build (SIMD); this covers essentially all current desktop Chrome/Edge, but very old browser builds without SIMD support are not handled.
- Language trained-data files for OCR (
eng/tur) are fetched on first use from a public tessdata CDN — this requires an internet connection the first time OCR runs (subsequent runs use the browser's cache). - Password-protected files and pages behind a login wall are not supported.
- Results are informational only and are not legal advice.
IReadIt/
├── manifest.json # Extension definition (Manifest V3)
├── background.js # Anthropic API calls (ANALYZE + COMPARE), prompts, JSON schemas
├── popup.html/.css/.js # Main UI: Page / File / History tabs, shared loading/result/compare views
├── options.html/.css/.js # API key settings + clear-history action
├── lib/
│ ├── pdf-extract.js # pdf.js text extraction + OCR fallback for scanned PDFs
│ ├── docx-extract.js # mammoth.js text extraction
│ ├── ocr.js # Tesseract.js worker wrapper
│ └── history.js # chrome.storage.local-backed history helpers
├── vendor/ # Bundled pdf.js, mammoth.js, tesseract.js (no remote code, MV3-compliant)
└── icons/