This is a submission for the Hacktoberfest Open-Source AI Challenge: Week 2 - Side Hustle.)
πΊ Watch the 30-Second Demo Video
π Watch Spot-Undoo in Action (Interactive Demo)
π What I Built
Spot-Undoo is an open-source, local-first micro-travel planner and hidden gem discovery engine tailored for travelers and transit commuters across Kerala.
Most travel applications (Google Maps, TripAdvisor) funnel visitors to the exact same overcrowded commercial tourist traps, completely missing quiet backwaters, heritage artisan alleys, and authentic street food stalls that define local culture. Furthermore, transit commuters stuck on railway layovers (like Shoranur Junction or Ernakulam South) suffer from extreme train delay anxiety and cannot find curated routes within strict time and rupee limits.
Spot-Undoo solves this by pairing locally hosted open-weight AI models (fine-tuned Qwen 2.5:3B on Ollama) with MongoDB Atlas, Render cloud deployment, and a Native Android APK to generate constraint-guaranteed itineraries on the fly.
π Key Features
-
πΊοΈ 70 Authentically Verified Kerala Spots:
- Curated across 7 hubs: Shoranur Jn Hub, Fort Kochi / Ernakulam, Calicut / Kozhikode, Trivandrum Central, Palakkad Town, Kollam Jn, and Wayanad.
- Includes real photography, coordinate bounds, expense ratings, and local Kerala community reviews.
-
β‘ Zero-Click Dual-Strategy Fast GPS:
- Automatically acquires live GPS with high accuracy and low-latency cached fallback.
- Snaps to the nearest Kerala railway hub automatically upon opening.
-
π¨ Neo-Brutalist Vector Adventure Canvas:
- Edge-to-edge mobile-first UI with high-contrast neo-brutalist styling.
- Alternating spot cards with large image previews, entry fees, and rating badges.
- Seamless switching between Canvas Mode, Live OpenStreetMap (Leaflet), and High-Density Grid.
-
π§ On-Device Qwen 2.5 Edge AI Inference:
- Runs locally using Ollama (
spot-undo-qwenorqwen2.5:3b). - ~71.4 tokens/second on an RTX 3050 mobile GPU (warm latency: 2.39 seconds).
- Generates 3 custom plan archetypes: Best Mood Match, Ultra-Budget Saver, and Hidden Gem Route.
- Safe Return Guarantee: The trail visually loops back and terminates at a mandatory Safe Return card before train departure.
- Runs locally using Ollama (
-
π MongoDB Atlas Cloud Integration & User Authentication:
- Multi-user authentication system (User Registration, Login, and JWT-authenticated sessions).
- User-specific itinerary history: every generated plan can be saved to the traveler's personal MongoDB collection.
- Cloud persistence for community-uploaded spots and traveler reviews.
- Built-in Geospatial Duplicate Guard (rejects identical spots within a Β±0.00001 coordinate tolerance).
-
βοΈ Multi-Cloud Hosting Ready: Render & DigitalOcean:
-
Render Blueprint (
render.yaml): One-click deployment for the Node/Express backend API and static frontend with automatic SSL and zero-config builds. -
DigitalOcean Droplet Setup (
scripts/deploy-digitalocean-ollama.sh): Automated script leveraging Hacktoberfest DigitalOcean cloud credits to host a remote Ollama LLM endpoint with GPU/CPU support, CORS headers, and systemd automation. -
Docker Support (
Dockerfile): Lightweight multi-stage container build for seamless portable hosting.
-
Render Blueprint (
-
π± Native Android APK:
- Packaged with Capacitor 6 for native Android deployment (5.2 MB footprint).
- Custom app launcher icons, native splash screen, and full offline fallback routing engine.
πΌ The Side Hustle: Monetization & Micro-Tourism SaaS
Spot-Undoo is designed as a Micro-Tourism Concierge Engine targeting three distinct revenue streams:
-
Backpacker Hostels & Boutique Homestays (B2B SaaS):
- Hostels in Fort Kochi, Varkala, and Wayanad constantly field layover questions ("I have 3 hours and βΉ400 before my train, what should I do?").
- Spot-Undoo provides branded digital concierge kiosks / QR codes for guests. Hostels pay a βΉ1,499/month subscription to offer hyper-local, white-labeled AI itineraries to their guests.
-
Local Auto Drivers & Freelance Guides (B2B2C Marketplace):
- Kerala autorickshaw drivers and local guides curate custom trails ("Old Calicut Halwa & Biryani Trail" or "Shoranur River & Pottery Tour").
- Spot-Undoo lets them turn their local knowledge into monetized digital guide cards that tourists purchase for βΉ49 to βΉ99 per trail.
-
Hyper-Local Food & Craft Vendors:
- Small local eateries (like Shaji's Tapioca Stall or century-old tea shops) get verified discovery without paying exorbitant 30% aggregator commissions.
π οΈ Tech Stack & Architecture
-
AI Model: Ollama running fine-tuned
spot-undo-qwen(Qwen 2.5 3B Q4_K_M GGUF) or cloud-hosted on DigitalOcean. -
Database: MongoDB Atlas (M0 free-tier cloud cluster) managing
users,places,reviews, andplans. - Backend API: Node.js & Express 5 REST API on port 4000 with CORS and JWT auth.
- Frontend: React 18, Vite, Tailwind CSS, Lucide Icons, Leaflet OpenStreetMap.
-
Cloud Infrastructure: Render (
render.yaml) + DigitalOcean Droplet + Docker multi-stage container. - Mobile Engine: Capacitor 6 native Android bridge with Gradle build.
π Code Repository & Links
- Interactive Demo Video: Watch on ngram
- GitHub Repository: https://github.com/FM9447/Spot-Undoo
- Slide Deck (.pptx): Spot-Undoo Presentation Deck
- MLH Submission: MLH Project #01a14abe
Built with β€οΈ for Hacktoberfest 2026.
Top comments (0)