iOS WebKit does not provide a reliable native Barcode Detection API, and
ZXing often failed due to strict camera constraints and video startup timing.
- Install @undecaf/barcode-detector-polyfill (ZBar WASM) on Apple devices
- Fall back through progressively looser getUserMedia constraints
- Wait for video metadata/playback before decoding frames
- Throttle native scans on iOS and tune ZXing retry intervals
- Defer scanner startup until the modal video element is mounted
Co-authored-by: Ned Halksworth <hello@nedhalksworth.com>
scanner modal, product preview, verified catalog with 12 red bull barcodes,
appwrite-backed user mappings, lookup pipeline with known/user/partial/unknown result
- Added support for encrypted coach chats with a new `coach_chats` collection in the Appwrite database.
- Updated `.env.example` to include `OLLAMA_API_KEY`, `OLLAMA_MODEL`, and `APPWRITE_API_KEY` for server-side configurations.
- Introduced a setup script in `package.json` for initializing Appwrite database tables.
- Enhanced the Vite configuration to proxy requests to the Ollama API.
- Updated the main application structure to accommodate new chat features and improved theme management.
- Refined CSS styles for better UI consistency and added new components for chat functionality.
- Added a new `barcode_products` collection to the Appwrite setup for managing barcode data.
- Implemented barcode scanning feature with a dedicated modal for scanning and adding products.
- Introduced new components for barcode product preview and management.
- Updated the setup script to seed verified barcode products from a JSON file.
- Enhanced the application state management to handle barcode-related actions and user interactions.
- Added user limits management with daily can and spend limits.
- Integrated onboarding flow to guide users through setting limits.
- Enhanced greeting messages to reflect user limits and violations.
- Updated CSS for new limit-related components and improved UI consistency.
- Refactored coach session to utilize user limits in interactions.
- Replaced Gemini-style glass/blue aesthetic with clean Material You pink theme
- ChatGPT-style message layout: user right-aligned, coach left-aligned with avatars
- User avatar shows initials, coach avatar shows brain icon
- Chat bubble design: rounded cards matching app theme (no glass/blur effects)
- Clean textarea input with Enter to send, Shift+Enter for newlines
- ChatGPT-style empty state with suggestion chips
- Typing indicator with bouncing dots animation
- Sidebar uses Material You surface/container colors
- Status bar shows model name and ready/thinking state
- Error messages styled as Material You error containers
- Bottom hint text: 'coach can make mistakes'
- All colors use CSS custom properties for accent theme consistency
- Added support for encrypted coach chats with a new `coach_chats` collection in the Appwrite database.
- Updated `.env.example` to include `OLLAMA_API_KEY`, `OLLAMA_MODEL`, and `APPWRITE_API_KEY` for server-side configurations.
- Introduced a setup script in `package.json` for initializing Appwrite database tables.
- Enhanced the Vite configuration to proxy requests to the Ollama API.
- Updated the main application structure to accommodate new chat features and improved theme management.
- Refined CSS styles for better UI consistency and added new components for chat functionality.
- Replaced "SF Pro Text" and related fonts with "Google Sans", "Google Sans Text", and "Product Sans" in CSS and Tailwind configuration.
- Adjusted color variables for primary, secondary, and error states to enhance UI consistency.
- Modified background colors and button styles for improved aesthetics and usability.
- Introduced new utility classes for layout and component styling in the CSS.
Cursor Agent
9961