🌍 EcoScan AI — Climate + Data = Action

Turning everyday decisions into climate action — powered by open-source AI

🧩 The Problem

Every day, millions of individuals and small businesses make decisions that silently damage the climate — but they never find out. Generic sustainability advice ("drive less, eat less meat") doesn't work because it isn't personal, isn't quantified, and doesn't tell you what exactly to do differently.

Meanwhile:

The average person emits ~10 tonnes of CO₂ per year through consumption alone
UK small businesses waste an estimated £1.6 billion annually on avoidable energy costs
80% of people want to act on climate change but don't know where to start

The gap isn't awareness. It's actionable, personalised intelligence.

💡 The Solution

EcoScan AI is a two-in-one Streamlit application powered by Llama-3.1-8B-Instruct (open-source LLM via Hugging Face Inference Providers) that bridges the gap between climate data and real-world action.

🛒 Mode 1 — Purchase Carbon Scanner

Paste your shopping list (groceries, travel, electronics — anything). EcoScan:

Estimates the kg CO₂e footprint of each item based on lifecycle data
Flags high / medium / low impact items
Suggests a specific, named lower-carbon alternative with exact % savings
Shows your total footprint vs. potential footprint if you make the switches

Example: 500g beef mince (6.8 kg CO₂e) → Quorn mince (0.4 kg CO₂e) = 94% reduction

🏢 Mode 2 — Business Energy Advisor

Input your business type, monthly energy bill, heating system, and lighting. EcoScan generates a personalised 5-step energy reduction plan with:

Specific action titles and explanations
Estimated monthly £ savings per action
Payback period in months
Difficulty rating (easy / medium / hard)
CO₂ impact level

Example: A restaurant spending £800/month could save £96/month just by switching to LED lighting — payback in 8 months.

⚙️ How It Works

User Input (Streamlit UI)
        │
        ▼
Input Sanitization + Prompt Engineering
        │
        ▼
Hugging Face Inference Providers Router
(meta-llama/Llama-3.1-8B-Instruct)
        │
        ▼
Structured JSON Response Parsing
        │
        ▼
Interactive Results Dashboard (Streamlit)

The app uses structured JSON prompting — the LLM is given a strict system message instructing it to return only a valid JSON array with defined keys. A robust extract_json() parser handles edge cases like markdown fences, partial output, and nested objects.

🛠️ Tech Stack

Layer	Technology
Frontend & app framework	Streamlit
LLM	Llama-3.1-8B-Instruct (Meta, open-source)
LLM API	Hugging Face Inference Providers (`router.huggingface.co`)
Language	Python 3.10+
Dependencies	`streamlit`, `requests` (minimal footprint)

No paid APIs. No proprietary models. Fully open-source stack.

🌱 Real-World Impact

Target	Impact
Individual consumers	Personalised carbon footprint breakdown per purchase with actionable swaps
Small businesses (5.5M in UK alone)	AI energy audit replacing £500+ consultant visits
NGOs & educators	Free tool to run carbon literacy workshops
Local governments	Scalable citizen engagement tool for net-zero programmes

🏆 Why EcoScan AI Wins

Innovation

Not a dashboard — a decision engine. Most climate apps show you data. EcoScan tells you exactly what to buy instead and exactly how much you'll save. Structured JSON prompting of an open-source LLM for real-time sustainability intelligence is a novel approach at this scale.

Functionality

Two fully working AI-powered modes. Live LLM inference. Parsed, structured output rendered as interactive metrics and cards. Handles edge cases gracefully with robust error handling.

Presentation

Clean, branded UI with a clear narrative arc: here's your footprint → here's what to change → here's the impact. Demo-ready in 30 seconds — judges can use it live.

Problem Solving

Attacks climate change at two of its highest-leverage points simultaneously: individual consumption and small business energy waste — both quantified, both actionable, both personalised.

🚀 How to Run

# 1. Clone the repo
git clone https://github.com/khannakiran2001-beep/EcoScan-AI

# 2. Install dependencies (only 2 packages)
pip install -r requirements.txt

# 3. Launch
streamlit run app.py

Get a free Hugging Face token at huggingface.co/settings/tokens and paste it into the sidebar.