i'm obsessed with re-creating apple products. this is a macos-inspired personal website featuring a full desktop environment with multiple apps.
a macos sierra 10.12 themed desktop with:
- window management: draggable, resizable windows with minimize, maximize, and close
- dock: app launcher with hover tooltips
- menu bar: functional apple menu, file menu, app menu, status menus (wifi, bluetooth, control center), and notification center with live widgets (calendar, messages, weather, photos)
- system states: lock screen, sleep mode, restart, and shutdown overlays
notes - apple notes clone for my personal website content
- public notes viewable by everyone, private notes per browser session
- github flavored markdown with interactive task lists
- image paste/upload support
- swipe gestures on mobile
messages - imessage clone with ai-powered conversations
- chat with ai contacts that have unique personalities (powered by gpt)
- message reactions and sound effects
- typing indicators and read/unread states
- group chats and one-on-one conversations
- pinned conversations and swipe gestures
- @mentions and contact management
- command menu (⌘K) with keyboard shortcuts
- focus mode integration (mutes notifications)
iterm - terminal emulator
- real file system navigation
- github integration (browse your repos)
- basic shell commands (ls, cd, cat, pwd, clear, etc.)
textedit - plain text editor
- opens existing text files in their own windows
- direct opens without a file use the same Finder picker flow
- launching from Finder focuses an open TextEdit document window first, otherwise opens a smaller Finder window at Documents
preview - image and pdf viewer
- opens images and pdfs in their own windows
- direct opens without a file use the same Finder picker flow
- launching from Finder focuses an open Preview document window first, otherwise opens a smaller Finder window at Desktop
- document-app launch behavior is documented in
docs/document-apps.md
finder - file browser
- sidebar navigation (recents, applications, desktop, documents, downloads, projects)
- browse local files and github repositories
- multiple Finder windows on desktop
- launch apps from applications folder
calendar - apple calendar clone
- day, week, month, and year views with smooth navigation
- sample events: exercise, focus time, meetings, dinners
- date night saturdays cycling through sf restaurants
- create, edit, and delete your own events
- drag-to-create events in day/week views
- holidays automatically displayed
- data persisted in localstorage
photos - apple photos clone
- photo library with grid view and full-screen viewer
- collections: flowers, food, friends
- favorites (per-browser, stored in localstorage)
- time filters (today, this week, this month, this year, all)
- keyboard navigation (arrow keys, escape to close)
- upload via ios shortcut with ai auto-categorization
settings - system preferences
- wi-fi and bluetooth panels
- appearance (light/dark/system theme)
- airdrop and focus mode toggles
- about this mac
weather - weather forecast app
- opens from notification center weather widget or
/weather - current conditions with hourly forecast
- non-dock app (appears in dock only while open)
- shared weather scene renderer and palettes documented in
docs/weather-scenes.md
responsive mobile interface with:
- swipe gestures for navigation
- touch-optimized controls
- full app functionality
the app uses next.js app router with a route group for the desktop environment. on desktop screens, all apps render in windows on a shared desktop. on mobile, apps display fullscreen with navigation.
notes use a session-based architecture:
- public notes: managed by the site owner, visible to everyone
- private notes: each browser session gets a unique id (stored in localstorage) linking to notes you create
messages are client-side only:
- conversations stored in localstorage
- ai responses generated via braintrust proxy (openai-compatible)
- no server-side message storage
photos use supabase storage:
- images stored in supabase storage bucket
- metadata (filename, timestamp, collections) in database
- favorites are per-browser (stored in localstorage)
- upload via api with ai auto-categorization (openai gpt-4o-mini)
the app is built with:
- next.js with app router
- typescript for type safety
- supabase for notes database
- braintrust for ai chat responses (openai-compatible proxy)
- react-markdown with github flavored markdown
- tailwind css for styling
the app uses supabase for the postgresql database with row-level security policies to control access to public and private notes.
database schema:
the notes table stores all notes with these fields:
id(uuid): unique identifiertitle(text): note titlecontent(text): markdown contentsession_id(uuid): links notes to browser sessionspublic(boolean): controls visibilityslug(text): url-friendly identifiercategory(text): optional categorizationemoji(text): note iconcreated_at(timestamp): when the note was created
public notes are cached for 24 hours using next.js isr. private notes are always real-time.
to manually revalidate public notes:
set REVALIDATE_TOKEN in environment variables, then:
# revalidate sidebar (when adding/removing public notes)
curl -X POST "https://yourdomain.com/notes/revalidate" \
-H "Content-Type: application/json" \
-H "x-revalidate-token: your-token" \
-d '{"layout": true}'
# revalidate specific note (when updating content)
curl -X POST "https://yourdomain.com/notes/revalidate" \
-H "Content-Type: application/json" \
-H "x-revalidate-token: your-token" \
-d '{"slug": "note-slug"}'or redeploy on vercel to refresh all pages.
REVALIDATE_TOKEN must stay server-side only. do not expose it via NEXT_PUBLIC_* variables.
upload photos directly from your iphone using the share sheet:
- open the shortcuts app on ios
- create a new shortcut with these actions:
- receive images from share sheet
- get details of image → date taken
- resize image to max 2048px (fit)
- convert image to jpeg (quality 0.8)
- encode with base64
- format date → iso 8601
- get contents of url:
- url:
https://yourdomain.com/api/photos/upload - method: POST
- headers:
x-api-key: <your-PHOTOS_UPLOAD_API_KEY> - body: json
{ "image": [base64], "timestamp": [formatted date] }
- url:
- name it "add to website"
- enable "show in share sheet" for images
when you share a photo, the shortcut uploads it to supabase storage and ai automatically categorizes it into collections (flowers, food, friends).
git clone https://github.com/alanagoyal/alanagoyal
this project uses supabase as a backend. to set up the database:
- create a new project and enter your project details
- wait for the database to launch
- navigate to the sql editor in the dashboard
- paste the sql from the migration file into the sql editor and press run
alternatively, use the supabase cli to run migrations locally:
supabase db pushgrab the project url and anon key from the api settings and put them in a new .env.local file in the root directory:
# supabase (required for notes and photos)
NEXT_PUBLIC_SUPABASE_URL="<your-supabase-url>"
NEXT_PUBLIC_SUPABASE_ANON_KEY="<your-anon-key>"
SUPABASE_SERVICE_ROLE_KEY="<your-service-role-key>"
# braintrust (required for messages ai)
BRAINTRUST_API_KEY="<your-braintrust-api-key>"
# photos upload (required for ios shortcut)
PHOTOS_UPLOAD_API_KEY="<generate-random-key>"
OPENAI_API_KEY="<your-openai-api-key>"
# site config (optional)
NEXT_PUBLIC_SITE_URL="https://yourdomain.com"
REVALIDATE_TOKEN="<your-revalidate-token>"
notes:
GITHUB_TOKENis optional but helps avoid rate limits when using iterm/finder github integrationSUPABASE_SERVICE_ROLE_KEYis needed for photo uploads (bypasses RLS)OPENAI_API_KEYis used for ai photo categorization
npm install
run the application in the command line and it will be available at http://localhost:3000.
npm run dev
with BRAINTRUST_API_KEY configured in .env.local, run the Messages group-chat behavior eval:
npm run eval:chat
the eval exercises response routing, waiting for the human, conversation wrap-up, schema validity, repetition, and texting brevity against the same prompt, tools, and model used by the production chat route.
deploy using vercel
notes support github flavored markdown (gfm) with interactive features. here's what you can use:
# heading 1
## heading 2
### heading 3**bold text**
*italic text*
~~strikethrough~~
`inline code`unordered lists:
- item one
- item two
- nested item
- another nested itemordered lists:
1. first item
2. second item
3. third itemtask lists are interactive - click checkboxes to toggle completion:
- [ ] task to do
- [x] completed task
- [ ] another taskthe app automatically updates the markdown when you click checkboxes, so your progress is saved.
create tables using standard markdown table syntax. tables render with a styled dark theme:
| book | author | year read |
|------|--------|-----------|
| the great gatsby | f. scott fitzgerald | 2023 |
| 1984 | george orwell | 2024 |this renders as:
| book | author | year read |
|---|---|---|
| the great gatsby | f. scott fitzgerald | 2023 |
| 1984 | george orwell | 2024 |
table features:
- white borders on dark background
- properly padded cells
- header row styling
- responsive layout
- supports links in cells
[link text](https://example.com)all links automatically open in new tabs for better navigation.
inline code: use backticks for inline code
code blocks: use triple backticks for multi-line code
```javascript
function hello() {
console.log("hello world");
}
```> this is a blockquote
> it can span multiple linespaste images directly into notes by copying any image (screenshot, file, etc.) and pressing ctrl+v (or cmd+v on mac). images are automatically uploaded to supabase storage and inserted as markdown.
you can also manually add images:
supported formats: jpeg, png, gif, webp (including animated gifs) file size limit: 5mb images are automatically resized to fit the note width while maintaining aspect ratio
---licensed under the mit license.