Lanka Heritage AI

• 

  Core Hub – An intuitive dual-portal landing view letting users effortlessly switch between Heritage Explorer and Agri-Care AI diagnostics.

• 

  Context Command Center – Where users trigger live GPS telemetry tracking, access the Voice Input Pod, or deploy instant Vision Diagnostics.

• 

  A responsive Leaflet-driven geographical matrix pinpointing historical heritage sites and agricultural zones across Sri Lanka.

• 

  Structured markdown content layer breaking down complex historical timelines and seasonal event details seamlessly with low latency.

💡 Inspiration

Sri Lanka possesses a deeply rich history spanning thousands of years, but modern travelers and tech-savvy locals often lack an interactive, smart context-aware guide. Traditional cultural tracking tools are rigid and non-responsive. We built Lanka Heritage AI to breathe digital life into history—deploying an intelligent assistant that doesn't just chat, but visually analyzes heritage vectors, speaks to the user dynamically, and uses telemetry arrays to reveal hidden historical metrics.

⚙️ What it does

• Auto-Prompt Injected URL Links: Instantly spins up targeted contexts (such as Sigiriya) upon custom direct URL hits without requiring initial text entry.
• Scan Historical Path (GPS Telemetry): Performs one-tap tracking that safely reverse-geocodes your active location and serves an immediate structural timeline.
• Voice Input Pod & Audio Guide: Provides hands-free English speech dictation synced with fluid background speech synthesis models.
• Vision Diagnostics: Allows users to snap or upload ruins, ancient artifacts, or local agricultural assets for rapid structural triage and historical summary.

🛠️ How we built it

We architected the web layer entirely around React and TypeScript, leveraging TailwindCSS to construct a lightweight, high-performance viewport optimized for mobile screens. The platform utilizes standard browser-level Geolocation APIs for immediate telemetry polling. The intelligence core orchestrates custom OpenAI prompts tuned with dual simple/expert reading parameters.

🛑 Challenges we overcame

• Overcoming asynchronous memory leaks inside traditional browser speech synthesis engines during continuous view re-renders.
• Ensuring live device camera media streams safely release on layout shifts to prevent persistent hardware battery drain.
• Synchronizing query state variables instantly upon custom deep-link activation.

🏅 What we're proud of

Successfully compiling a deeply technical context layer into an incredibly intuitive, responsive user interface that runs gracefully without any initial cloud-database latency.

🧠 What we learned

We mastered browser telemetry life-cycles, advanced state tracking across deep-link URL paths, and edge-case handling for native mobile media permissions.

🚀 What's next for Lanka Heritage AI

Integrating true offline neural translation layers for native Sinhala/Tamil support, and expanding the vision model to read historical stone inscriptions (Inscriptions/Sela Lipi) directly from the live feed.

Built With

• geolocation-api
• openai-api
• react
• tailwindcss
• typescript
• vite