This is a submission for Weekend Challenge: Earth Day Edition
What I Built
EcoSense AI is an AI-powered carbon footprint analyzer that helps people understand their environmental impact through a simple 4-step questionnaire — with persistent memory and carbon offset donations.
Users answer questions about their:
- 🚗 Transportation — commute method & distance
- 🍽️ Diet — from heavy meat to vegan
- ⚡ Home Energy — fossil fuels to 100% renewable
- 🛍️ Shopping — fast fashion to minimal/second-hand
Then the app:
- Google Gemini 2.0 Flash analyzes their habits and returns an eco score, CO2 estimate, personalized tips, and Earth Day pledge
- Backboard saves each assessment to persistent memory threads, enabling progress tracking over time
- Solana enables carbon offset tree-planting donations via SOL transfers
Demo
🌍 Live: https://ecosense-ai.pages.dev
Try it now — takes about 60 seconds!
Features:
- Eco Score (0-100) with SVG donut chart visualization
- Letter grade (A+ to F) with estimated annual CO₂
- Impact breakdown by category with color-coded status bars
- 5 personalized tips from Gemini AI
- Earth Day pledge with copy-to-clipboard sharing
- Carbon offset via Solana donation
- Memory persistence via Backboard threads
Code
x-tahosin
/
ecosense-ai
AI Carbon Footprint Analyzer - Google Gemini + Backboard + Solana | Earth Day 2026
🌍 EcoSense AI
AI-Powered Carbon Footprint Analyzer | Earth Day 2026
What It Does
Answer 4 quick questions about your daily habits and Google Gemini analyzes your carbon footprint in real-time:
- Eco Score (0-100) with letter grade and SVG donut chart
- Estimated annual CO2 compared to global average
- Impact breakdown by category (transport, diet, energy, shopping)
- 5 personalized tips to reduce your footprint
- Earth Day pledge — copy and share on social media
- Carbon offset — donate SOL via Solana to plant trees
- Progress tracking — Backboard memory saves your assessments
Architecture
Browser → Static HTML/JS (Cloudflare Pages)
↓
/api/generate → Cloudflare Function → Google Gemini API
/api/memory → Cloudflare Function → Backboard API
API keys NEVER reach the browser.
Tech Stack
| Technology | Purpose |
|---|---|
| Next.js 16 | Static export frontend |
| Google Gemini 2.0 Flash | AI analysis engine |
| Backboard | Persistent memory threads |
| Solana | Carbon offset donations |
| Cloudflare Pages | Hosting + serverless functions |
Key files:
-
app/page.tsx— Full React UI with step wizard, score visualization, Solana integration, and Backboard memory status -
functions/api/generate.js— Cloudflare Function proxying Gemini (API key never reaches browser) -
functions/api/memory.js— Cloudflare Function proxying Backboard memory API
How I Built It
Tech Stack
- Next.js 16 with static export
- Tailwind CSS — custom green/earth color palette
- Google Gemini 2.0 Flash — AI analysis engine
- Backboard — persistent memory for tracking assessments over time
- Solana — carbon offset tree-planting donations
- Cloudflare Pages — hosting + serverless functions
- Lucide React — iconography
- GitHub Copilot — used throughout development for rapid iteration
Architecture: Everything Server-Side
Browser → Static HTML/JS (Cloudflare Pages)
↓
/api/generate → Cloudflare Function → Google Gemini (key server-only)
/api/memory → Cloudflare Function → Backboard API (key server-only)
No API keys ever touch the browser. Both Gemini and Backboard calls go through Cloudflare Functions.
How Each Technology Is Used
Google Gemini — The core analysis engine. I send a structured prompt with the user's 4 selections and ask for specific JSON output: score, grade, CO₂ estimate, impact breakdown, tips, and pledge. Temperature 0.7 gives the best variety/accuracy balance.
Backboard — After each analysis, results are saved to a Backboard memory thread keyed by session ID. This enables returning users to see progress over time — "Your score improved from 45 to 62 since last month!" The memory API gracefully falls back if unavailable.
Solana — The results page calculates how many trees would offset the user's footprint (~1 tree absorbs ~22kg CO₂/year) and offers a Solana donation option. SOL is ideal for micro-donations: sub-second finality, fraction-of-a-cent fees, and Solana's Proof of Stake uses 99.9% less energy than Proof of Work chains.
GitHub Copilot — Used throughout development for the step wizard logic, SVG chart math, and Tailwind styling. Copilot's inline suggestions accelerated the build significantly.
UI Highlights
- Step wizard with animated progress bar
- SVG donut chart for the eco score
- Color-coded impact badges (red → green)
- Category breakdown bars in the results
- Purple Solana donation panel with explorer link
- Blue Backboard memory status indicator
- Copy-to-clipboard Earth Day pledge for social sharing
Prize Categories
- Best Use of Google Gemini — Gemini is the core engine producing calibrated environmental analysis from 4 simple inputs
- Best Use of Solana — Carbon offset donations via SOL, leveraging Solana's eco-friendly PoS consensus
- Best Use of GitHub Copilot — Copilot accelerated the entire build from UI components to API integration
- Best Use of Backboard — Persistent memory threads track assessment history for progress monitoring
Top comments (3)
Let's goooo! How was your experience with backboard? would you recommend?
Thanks Jonathan! Backboard was great to integrate — I used it for the memory/pledge system in EcoSense AI so users' carbon footprint commitments persist across sessions. The API is straightforward and the free tier covers everything I needed. Definitely recommend it if you need a simple key-value store or memory layer for AI apps. What are you building with it?
I'm building Backboard 😉