Swipe travel videos. Get a real itinerary.

Soar is a travel planner that turns YouTube Shorts into a personalised day-by-day itinerary. Pick a destination, swipe through Shorts curated for your location and activity vibes, and Gemini 2.5 Flash builds a specific, timed plan inspired by what you liked.

Planning a trip is overwhelming. You spend hours watching travel content, saving videos, reading blogs — and still end up with a generic itinerary that could have been written in 2019.

Soar lets you build a day-by-day travel itinerary by swiping through YouTube Shorts. Like the videos that match your vibe, skip the rest, and get a personalised itinerary powered by Gemini AI.

1. Tell us where: pick a destination, trip length, season, and activity vibes (hiking, dining, nightlife, etc.)
2. Swipe the shorts: a curated feed of location-specific YouTube Shorts plays in a card deck; swipe right to like, left to skip. The feed adapts in real time based on what you engage with.
3. Get the itinerary: liked video URLs are sent to the AI backend, which generates a detailed day-by-day plan grounded in the specific venues and neighbourhoods from those videos.

There is also a YouTube mode that pulls from your own liked videos, favourites, and playlists instead of the live Shorts feed.

The repo contains three independent services that run together:

frontend-main/          React + Vite frontend (TypeScript)
itinerary-backend-main/ Python / Flask AI backend
shorts-backend/         Node.js Shorts feed API

• React 18 with React Router v6 and TypeScript
• Material Design 3 component system (custom, in src/components/M3.tsx)
• Auth via Auth0 (@auth0/auth0-react)
• Three main flows:
  • /create-trip — trip details form → swipe feed → itinerary generation
  • /create-trip-youtube — same flow but sourced from the user's YouTube account
  • /your-trip — renders the generated itinerary with a day-by-day timeline
  • /my-trips — saved itineraries fetched from the backend
• Vite dev server proxies /api/* → localhost:3000 (Shorts backend) and /itinerary-api/* → localhost:8080 (Python backend)

• Plain Node.js HTTP server (no framework)
• Uses yt-dlp to search YouTube for Shorts matching the destination and activity terms
• Rotates through keyword variants and applies a 100k-view floor to filter low-quality results
• Caches feed results for 20 minutes and CDN stream URLs for 90 minutes
• Endpoints:
  • GET /api/feed?page=N&location=...&extra=... — returns up to 8 Short video metadata objects
  • GET /api/stream-url?v=VIDEO_ID — resolves and returns the direct CDN URL as JSON
  • GET /api/proxy?v=VIDEO_ID — resolves the CDN URL and issues a 302 redirect

• Flask with Flask-CORS
• Uses Google Gemini 2.5 Flash (google-generativeai) to generate itineraries as structured JSON
• Auth via Auth0 JWT verification (RS256, JWKS endpoint)
• Itineraries are persisted in Supabase (Postgres)
• Endpoints:
  • POST /generate_itinerary?prompt=...&video_urls=... — generates a new itinerary
  • POST /save_itinerary — saves an itinerary to the authenticated user's account (requires Bearer token)
  • GET /my_itineraries — lists the user's saved itineraries (requires Bearer token)
  • DELETE /delete_itinerary/<id> — deletes a saved itinerary (requires Bearer token)

• Node.js 18+
• Python 3.10+
• yt-dlp installed and on your PATH (pip install yt-dlp or brew install yt-dlp)
• A Google Cloud project with the YouTube Data API v3 enabled
• A Google AI Studio API key (Gemini)
• An Auth0 tenant with a Single Page Application and an API configured
• A Supabase project with an itineraries table (columns: id, user_id, location, data, created_at)

cd shorts-backend
node server.js
# Listens on http://localhost:3000

No dependencies to install — uses only Node built-ins and yt-dlp.

cd itinerary-backend-main
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Create a .env file:

GOOGLE_API_KEY=your_gemini_api_key

AUTH0_DOMAIN=your-tenant.auth0.com
AUTH0_AUDIENCE=https://your-api-identifier

SUPABASE_URL=https://your-project.supabase.co
SUPABASE_SERVICE_KEY=your_supabase_service_role_key

Start the server:

python server.py
# Listens on http://localhost:8080

cd frontend-main
npm install

Create a .env file:

VITE_AUTH0_DOMAIN=your-tenant.auth0.com
VITE_AUTH0_CLIENT_ID=your_auth0_spa_client_id
VITE_AUTH0_AUDIENCE=https://your-api-identifier
VITE_GOOGLE_CLIENT_ID=your_google_oauth_client_id

Start the dev server:

npm run dev
# Listens on http://localhost:5173

All three services need to be running at the same time for the full flow to work.

Made with ❤️ by the Soar Team