mAccess Map

An inclusive, community-powered accessibility review platform — with AI-driven feature suggestions and NFT badges to reward contributors.

Inspiration

We believe everyone deserves equal access to the built world. Too often, information about ramps, Braille signage, talking elevators or automatic doors lives in scattered spreadsheets or hard-to-find forums. At the 3PercentClub hackathon, we saw an opportunity to empower under-represented voices by letting anyone—with or without deep Web3 experience—crowdsource and share real-world accessibility data. By marrying AI, blockchain badges, and a familiar map interface, we wanted to lower barriers for both contributors and consumers of accessibility information.

What it does

Search & discover public spaces (restaurants, transit, parks, etc.) on a map and see an accessibility summary (rating, features, comments).
Add reviews using Google Places Autocomplete + map-based geocoding to capture location, rating, comments.
AI suggestions: write a quick comment (“good lighting, clear signage”) and our HuggingFace model recommends relevant features to tag (e.g. visual-aids, audio-cues).
Supabase backend with Row-Level Security ensures only authenticated users can write, while the world can read reviews in real time.
NFT badges: every verified review earns you on-chain badges (First Steps, Community Builder, etc.). Metadata lives on IPFS and OpenZeppelin’s ERC-721 contract mints directly to your wallet.

How we built it

Frontend: React + Vite + Tailwind.
- Google Maps JavaScript API (Places Autocomplete → PlaceAutocompleteElement migration ahead)
- Contexts for Auth (Supabase) and Wallet (Ethers.js)
- PhotoUpload component storing images in Supabase Storage
Backend: Supabase
- Tables for locations, reviews, user_profiles, nft_badges
- RLS policies: authenticated users can INSERT/UPDATE their own data; public role can SELECT
AI: HuggingFace Zero-shot classification
- Candidate labels matched to our feature keys
- Explanation generation via template on top suggestions
Blockchain:
- ERC-721 contract (OpenZeppelin’s ERC721URIStorage + Ownable) deployed on Sepolia
- BadgeMinted event for easy log parsing
- Metadata uploaded to IPFS via Pinata/Infura
- Ethers.js in mintBadgeNFT() to send transactions, parse BadgeMinted or fallback to Transfer.
Badges System: badgeSystem.ts orchestrates eligibility → metadata → on-chain mint → Supabase record.

Challenges we ran into

Row-Level Security: initially blocked INSERTs until we fine-tuned policies (INSERT for authenticated role, public SELECT).
Token ID extraction: our first contract emitted only Transfer; parsing logs to get the right tokenId took some trial and error. Now we emit a custom BadgeMinted event for reliability.
Google Places Autocomplete deprecation: migrating to PlaceAutocompleteElement is next on our list.
Supabase timeouts: dealing with “raw fetch POST 401” until we unlocked RLS and refreshed anon keys correctly.
UI state: keeping auth, maps, AI suggestions, photo uploads, and submit spinner in sync required careful effect and error handling.

Accomplishments that we're proud of

End-to-end flow from typing an address → AI-powered feature tags → on-chain badge minting.
Robust error handling with toasts, console logs and user-friendly messages for almost every edge case.
Badge automation: users can click “Claim Badge” immediately after their first review, or watch it mint automatically on submission if they’ve connected a wallet.
Polished UX: mobile-first, PWA install prompt, “Locate me” button, “Submitting…” spinners, and smooth map integration.

What we learned

Security first: even in a hackathon, RLS + AccessControl patterns matter for protecting data integrity.
Contract design: custom events (BadgeMinted) pay dividends when you parse logs off-chain.
AI integration: zero-shot classification can drive real-time UX features, but you need thresholds to avoid spamming suggestions.
Coordination: juggling Supabase auth state, map loading, photo uploads, AI calls, and wallet connections pushes React hooks to their limits—centralized state and disciplined logging saved the day.

What’s next for mAccess Map

🔄 PlaceAutocompleteElement migration for future-proof address input.
🌐 Multi-chain support (Polygon, BSC) so badges live in mainnet worlds.
📊 Enhanced Analytics dashboard to surface trending features and underserved neighborhoods using AI.
🌍 Offline reviews via Service Worker sync for when connectivity is spotty.
🛠️ AI Review Flag for reviewing & verifying user-submitted reviews to reduce manual approval timeframe.
🏆 Leaderboard Social Share: Show top contributors, share badges on Twitter / Discord.