1. Added (1). Context glossary: keep professional terms per document scene and let recognition corrections apply automatically). Six built-in scenes (General, Workplace, Legal, Journalist, Medical, Finance) plus your own; term lists for Chinese, English, Japanese and Korean). Entries carry language, scene, pronunciation, note and enabled state, track hit counts, and support CSV template import/export. (2). Visible revisions: a corrected term shows as struck-through original followed by the red replacement, so you can see exactly what changed). Double-tap one occurrence to revert just that one, or revert every replacement on the page from the top of the box. (3). Scene tabs: all scenes laid out in a single row above the result, defaulting to "General" (glossary off); tap a tab to switch). The master switch lives in Settings and is off by default. (4). Full-text history search: saved recognitions and translations are searchable, with the matched term highlighted and a surrounding snippet). Chinese is indexed in two-character grams, so you never have to match a whole sentence. (5). History organisation: tag, favourite and rename saved records, and filter the list down to favourites. (6). Structured info card: amounts, dates, times, phone numbers, emails, links, ID numbers, card numbers, tax IDs, postcodes, IP addresses, percentages and reference numbers are extracted from the recognised text). Tap to copy, or dial, email and open links directly). ID and card numbers are shown masked; copying gives the full value. (7). Privacy self-check: an expandable audit page in Settings listing zero network requests, zero tracking SDKs and zero cloud storage, plus a line-by-line account of which system capabilities are requested (camera, add-to-library) and which never are (read photo library, location, contacts, tracking, Face ID). (8). Fine camera controls: a control strip above the shutter for flash (off/auto/on), zoom stops (0.5x/1x/2x, switching real lenses first so quality is not lost), three output sizes (largest for the current lens by default), and continuous or locked focus. (9). Pinch to zoom: adjust zoom with a two-finger pinch on the preview, with the current factor shown while you pinch). Torch, flash, focus, lens and output size are remembered; zoom always returns to 1x. 2. Changed (1). Results are read-only by default: tapping text no longer drops you into editing). Editing is entered explicitly from the top-right button; while editing, the recognised text is shown as-is and glossary revisions and scene tabs step aside. (2). Proofreading low-confidence lines is now an explicit action inside edit mode instead of jumping in automatically; while proofreading, the bottom bar keeps only the proofreading navigation and hides share/save. (3). The glossary follows the language of the scan rather than the interface language: a Chinese UI scanning a Japanese document uses the Japanese term list. (4). Records saved to History now store the recognised text itself, which is what makes full-text search possible; exports and translations both use the glossary-corrected text. (5). Camera output size changed from "always the largest" to "largest by default, adjustable", so you can trade resolution for storage when you need to. 3. Removed (1). Nothing was removed in this release.

1. Added Minimal single-scan flow: frame right after shooting or picking a photo, confirm and get results immediately; go back to re-adjust, with the previous box pre-filled. QR/barcode recognition: QR, EAN, UPC, Code128 and more. URL codes open automatically in the in-app browser; phone numbers dial, emails compose, locations open Maps; Wi-Fi/vCard/calendar fields are shown with copy; plain text can be copied, searched or translated; a framed URL as text shows a link card with one tap. OCR accuracy: selectable recognition languages (up to 2, default zh/en); upgraded to Vision Rev3 with per-line language detection; photos captured at max resolution; tiny text no longer filtered; new Language Correction toggle for number-heavy documents. Camera torch for faster focus and sharper close-up documents. Translation pack manager: download language packs ahead of time in Settings, with readiness hints for recognition languages. VIP membership & 7+7 trial: 7-day trial extendable once by 7 more days for free; after expiry you can still scan free once per day; a welcome letter for new users with a "don't show today" option. Low-confidence lines are dashed-underlined and can be proofread line by line before translating. Default export titles now use local "YYYY-MM-DD HH:mm:ss", no more same-day duplicates. 2. Changed Translation quality: recognized lines are re-grouped into full paragraphs before translation; a framed single URL becomes a link card instead of being translated. A pinned recognition language is passed directly as the translation source, reducing misdetection from OCR typos. Trial policy: "10 days then locked" replaced by "7 days + one free 7-day extension", with a daily free quota instead of locking features. Restore Purchases moved to the bottom of the membership page; the settings card is streamlined to one entry. Fixed: installed language packs repeatedly asking to download; export titles with wrong dates across time zones, same-day duplicates and illegal filename characters. 3. Removed Multi-page capture and batch recognition: the thumbnail grid, batch progress and "Apply to All" are gone with the single-scan flow; single results still save as PDF/TXT into History. Family/JianBox integration: Send to JianBox, doclens:// import and the App Group shared container are removed — fewer system capabilities requested. Standalone Restore Purchases entry in Settings (moved into the membership page).

DocLens 1.2 DocLens VIP is now available, and your first 10 days are on us. Every feature unlocks free for 10 days the moment you open the app. No card, no sign-up, nothing to cancel. Settings shows a live countdown so you always know where you stand. After that, DocLens VIP Lifetime is a single one-time purchase. Not a subscription. No auto-renewal, no hidden fees. Pay once and keep unlimited OCR, unlimited on-device translation, and unlimited batch processing forever, through every future update. A new Membership section in Settings shows your status and trial days remaining, with one-tap upgrade and Restore Purchase for new devices or reinstalls. Nothing about our privacy stance changed. OCR still runs on Apple's Vision framework and translation on Apple's Translation framework, entirely on your device. Purchases are verified locally by StoreKit, so there are no accounts, no servers, and no payment database. Still no ads, no tracking, and no watermark. Your scans never leave your iPhone.

**DocLens 1.1 — Release Summary: What Changed from 1.0** DocLens 1.1 upgrades the core "select → OCR → translate → export" pipeline into a polished, monetizable product. Everything still runs 100% on-device — no cloud, no tracking, no watermark. **NEW FEATURES** - Confidence visualization: every OCR line now carries its Vision confidence score. Text below 0.7 gets a red dashed underline, 0.7–0.85 yellow, the rest normal — uncertain reads are instantly visible. - In-place editing & proofread mode: tap any low-confidence line to correct it directly; corrections persist in the new ocr_edits SQLite table and flow into exports. - OCR quality: CIFilter preprocessing (contrast stretch + Otsu adaptive threshold) runs before Vision recognition; minimumTextHeight is tunable for dense or tiny text. - Handwriting detection: handwritten blocks trigger a warning banner so users know accuracy may be lower. - New OCR languages: French, German, Spanish added — 7 total (zh-Hans, en-US, ja-JP, ko, fr-FR, de-DE, es-ES). - Translation pre-checks: checkLanguageAvailability reports whether a language pack is installed before translating; requestLanguageDownload triggers the system pack download on iOS 18+. - Translation memory: MD5-hash caching makes repeated phrases translate instantly and consistently. - Layout detection: new LayoutService classifies text layout (7/7 unit tests) to pick the best export format. - Batch processing: background AsyncTask with checkpoint/resume — a large batch survives app suspension; a local completion notification (expo-notifications) fires when it finishes. - Export upgrades: JPG export; PDFKit-based PDF with an invisible, selectable text layer (new doclens-pdf module); ExportService unified for TXT/HTML/PDF/JPG. - Privacy: zero SDKs, zero network calls audited and documented in a new Settings privacy section. **CHANGED** - Share sheet reworked: system share sheet plus JianBox; sharing sends OCR text, image and title through the App Group container via URL scheme. - Family sharing simplified to JianBox only. - App version bumped from 1.0.0 to 1.1.0. **REMOVED** - ClauseCheck / VoxSafe / CalMerge share targets removed from the share menu; JianBox is now the sole family integration. **FIXED** - DocLensShared native module correctly linked (was missing podspec, wrong platform key). - doclens-background podspec fix. - Swift compile fixes: Task type ambiguity, LanguageAvailability.Status enum, missing try await, CGFloat→Float.