EcoVerse AI

Shape your future. One choice at a time.

EcoVerse AI is a climate companion built for the PromptWars Challenge 3. It turns everyday habits into readable consequences, a visual future city, and personalized guidance users can act on right away.

What it does

• Onboarding flow collects location, diet, commute, and energy habits.
• Future city simulation generates a localized 50-year narrative for the user's city.
• AI Carbon Coach now streams its guidance in real time and converts daily actions into carbon consequences and lower-impact alternatives.
• Daily quests suggest small, achievable actions based on the user profile.
• Accessibility tools include ELI10 mode, high contrast mode, voice input, text-to-speech support, a skip link, reduced-motion support, and live announcements for AI responses.
• Animated 3D background adds a more immersive presentation without breaking reduced-motion preferences.
• Firestore-backed caching and rate limiting keep repeated AI requests fast and deployment-friendly on Cloud Run.

Tech stack

• Next.js 16
• React 19
• TypeScript
• Zustand
• Gemini API via @google/genai
• Zod
• Firebase config and Firestore rules
• Security headers
• Firestore-backed rate limiting with in-memory fallback
• Firestore-backed AI response caching with in-memory fallback

Gemini setup

The app uses GEMINI_MODEL when provided, and defaults to gemini-2.5-flash.

Required environment variables:

GEMINI_API_KEY=your_key_here

Optional:

GEMINI_MODEL=gemini-2.5-flash

Firebase variables are also included in the repo for local and future integration:

NEXT_PUBLIC_FIREBASE_API_KEY=
NEXT_PUBLIC_FIREBASE_AUTH_DOMAIN=
NEXT_PUBLIC_FIREBASE_PROJECT_ID=
NEXT_PUBLIC_FIREBASE_STORAGE_BUCKET=
NEXT_PUBLIC_FIREBASE_MESSAGING_SENDER_ID=
NEXT_PUBLIC_FIREBASE_APP_ID=
FIREBASE_CLIENT_EMAIL=
FIREBASE_PRIVATE_KEY=

Local development

npm install
npm run build
npm run start

For a full local check:

npm ci
npm run build
npm run test

Test coverage

• src/app/api/__tests__/routes.test.ts covers AI success, validation, rate limit, malformed JSON, and streaming paths.
• src/components/__tests__/OnboardingFlow.test.tsx smoke-tests the onboarding flow.
• src/app/dashboard/__tests__/page.test.tsx smoke-tests the dashboard render path.

Deployment

The app is deployed on Google Cloud Run.

Live URL:

https://ecoverse-ai-655808244864.asia-south1.run.app

LinkedIn post:

https://www.linkedin.com/posts/pavan-karthikeya-atchyuta-3a5040354_buildwithai-promptwarsvirtual-googleai-ugcPost-7470847484966207488-CwrI/?utm_source=share&utm_medium=member_desktop&rcm=ACoAAFhIRi0Bgd3lw3t73PozqDSMLW6jDLjPfL8

Notes

• Docker uses a standalone Next.js output.
• The production image now runs on Node 22 Alpine.
• The deployment issue was fixed by syncing the lockfile and switching Gemini routes away from gemini-2.5-pro.
• next.config.ts now sets security headers for production.
• Gemini usage is centralized through a shared singleton client in src/lib/gemini.ts.
• Streaming coach replies and cached AI payloads are handled through shared helpers in src/lib/ai-cache.ts.