As climate awareness grows, individuals are looking for actionable ways to reduce their personal carbon footprints. However, most carbon calculators are either too complex or offer generic, unhelpful advice.
To solve this, I built CarbonWise—a production-ready Carbon Footprint Awareness Platform. It combines deterministic scientific carbon calculations with real-time, personalized AI reduction strategies using the Groq LLM API.
Here is a technical deep-dive into how I built, secured, and optimized this application for the PromptWars: Virtual challenge.
The application is designed to be lightweight, secure, and highly performant, avoiding heavy framework overhead.
┌──────────────────────────────────────────────────────────┐
│ User Browser │
└─────────────┬──────────────────────────────▲─────────────┘
│ HTTPS (POST / GET) │ Rendered HTML/CSS
┌─────────────▼──────────────────────────────┴─────────────┐
│ Flask App (app.py) │
└─────────────┬──────────────┬───────────────▲─────────────┘
│ │ │
┌─────────────▼──────────┐ ┌─▼─────────────┐ │
│ SQLite DB (carbon.db) │ │Secure Session │ │
│ - Users & Logs │ │ Cookies │ │
│ - WAL Mode Enabled │ └───────────────┘ │
└────────────────────────┘ │ Structured JSON Insights
┌────────────────────────────────────────────┴─────────────┐
│ Groq API (Llama 3.1) │
│ - Model: llama-3.1-8b-instant │
└──────────────────────────────────────────────────────────┘
llama-3.1-8b-instant
).carbon_engine.py
) To prevent calculation drift, the engine uses fixed conversion factors for logging three main categories:
0.21
kg/km), Motorbike (0.05
kg/km), Bus (0.08
kg/km), and Flight (0.255
kg/km).0.5
kg/meal) and Non-vegetarian meal (2.5
kg/meal).0.82
kg/kWh).
def calculate_co2(activity_type: str, quantity: float) -> float:
if activity_type not in EMISSION_FACTORS:
raise ValueError(f"Unknown activity type '{activity_type}'")
if quantity <= 0:
raise ValueError("Quantity must be positive")
return round(EMISSION_FACTORS[activity_type] * quantity, 4)
ai_engine.py
) To keep the dashboard fast and prevent redundant external HTTP calls, I implemented a session caching mechanism:
summary
, top_emission_sources
, suggestions
(3–5 tips), eco_score
(0–100), and motivation
./log
, the cache is invalidated so the dashboard gets fresh insights on the next load.
insights = session.get("ai_insights")
if not insights:
all_acts = _get_all_activities_7days(user_id)
insights = get_ai_insights(all_acts, daily_total, category_breakdown)
session["ai_insights"] = insights
To make sure data queries are incredibly fast, I optimized the database schema by introducing a covering index:
CREATE INDEX IF NOT EXISTS idx_activities_covering
ON activities(user_id, timestamp, activity_type, co2_kg);
With this index in place, queries for the weekly analytics dashboard and category breakdowns are satisfied directly from index pages, bypassing expensive full-table scans. We also optimized category queries to aggregate in a single SQL operation:
SELECT
CASE
WHEN activity_type IN ('car', 'bike', 'bus', 'flight') THEN 'travel'
WHEN activity_type IN ('veg', 'non_veg') THEN 'food'
WHEN activity_type IN ('electricity') THEN 'electricity'
END AS category,
COALESCE(SUM(co2_kg), 0) AS total
FROM activities
WHERE user_id = ? AND timestamp >= DATE('now', '-6 days')
GROUP BY category
X-Content-Type-Options: nosniff
X-Frame-Options: DENY
X-XSS-Protection: 1; mode=block
Referrer-Policy: strict-origin-when-cross-origin
Content-Security-Policy
(CSP) restricting assets to 'self'
and specific secure fonts.Secure
, HttpOnly
, and SameSite='Lax'
flags.<span role="img" aria-label="...">
tags for screen-reader compatibility..sr-only
class to visually hide labels while exposing them to accessibility scanners.<h1>
header.4.5:1
on dark backgrounds.I built a comprehensive test suite in tests/test_app.py
containing 47 total assertions across five key test domains:
test_carbon_calculation
test_api_response
test_ai_json_structure
test_login
test_input_validation
pytest tests/test_app.py -v
Building this platform with Google Antigravity shifted my focus from syntax debugging to high-level intent design.
The resulting codebase is performant, accessible, secure, and ready to scale. Let's make the web a greener place! 🌿
This build is part of my journey with RohithBuilds, where I document building real AI applications with zero fluff.
If you want to learn how to build production-grade AI systems, check out my courses at rohith-builds.onrender.com.