DEV Community

autozbudoucnosti
autozbudoucnosti

Posted on

Stop Letting AI Hallucinate Your Sustainability Data: Why I Built a Deterministic API

#python #fastapi #api #webdev

The Problem: AI Guesses, Math Doesn't

We live in the age of LLMs. You can ask ChatGPT, "What is the carbon footprint of a cotton t-shirt?" and it will give you a convincing answer. But ask it three times, and you might get three different numbers.

For e-commerce brands and supply chain tools, "hallucinated" data isn't good enough. You need deterministic, consistent, and auditable numbers.

That’s why I built the Product Impact API. It’s a dedicated engine that uses hard-coded environmental factors to calculate a consistent "Impact Score" for products based on their materials and shipping weight.

The Tech Stack

I chose this stack for speed and type safety:

Language: Python 3.12 (because modern Python is fast).

Framework: FastAPI (for async performance and auto-generated docs).

Validation: Pydantic (to ensure no bad data gets in).

Deployment: Railway (for effortless CI/CD).

How It Works

The core logic revolves around a weighted scoring system. A product is defined by its material (e.g., Cotton, Polyester) and its weight.

Here is the Pydantic model that enforces the data structure:

Python
from pydantic import BaseModel

class Product(BaseModel):
    name: str
    material: str
    weight_kg: float
    shipping_distance_km: float

    # Example payload for documentation
    model_config = {
        "json_schema_extra": {
            "examples": [
                {
                    "name": "Eco T-Shirt",
                    "material": "cotton",
                    "weight_kg": 0.2,
                    "shipping_distance_km": 1500
                }
            ]
        }
    }
Enter fullscreen mode Exit fullscreen mode

The calculation engine uses a dictionary lookup for materials. This is O(1) complexity—super fast compared to querying an LLM or a heavy database.

Python

MATERIAL_SCORES = {
    "cotton": 4.5,    # High water usage
    "polyester": 5.5, # Microplastics/Oil based
    "bamboo": 2.0,    # Sustainable
    "recycled": 1.5   # Best option
}

def calculate_impact(product: Product):
    # Base score from material
    base_score = MATERIAL_SCORES.get(product.material.lower(), 5.0)

    # Logistics penalty (0.1 points per 100km)
    transport_penalty = (product.shipping_distance_km / 100) * 0.1

    total_score = base_score * product.weight_kg + transport_penalty
    return round(total_score, 2)
Enter fullscreen mode Exit fullscreen mode

The Challenge: CORS & Security

One hurdle I hit during development was connecting my frontend (a simple HTML/JS dashboard) to the hosted API. I kept getting 401 Unauthorized errors.

I learned that even for public APIs, you need to be careful with CORS (Cross-Origin Resource Sharing). I had to explicitly allow my frontend's origin in the FastAPI middleware:

Python

app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"], # In production, lock this down!
    allow_methods=["*"],
    allow_headers=["*"],
)
Enter fullscreen mode Exit fullscreen mode

I also implemented API Key authentication using fastapi.security.api_key to prevent abuse.

The Result

The API is now live. It accepts a JSON payload and returns a detailed breakdown of the environmental cost, CO2 estimates, and a "Green Score."

You can check out the auto-generated Swagger documentation here: https://web-production-a1f91.up.railway.app/docs#/

What's Next?

I plan to expand the MATERIAL_SCORES database and add a "Suggestion Engine" that recommends greener alternatives (e.g., "Switching to Bamboo would save 1.2kg of CO2").

If you're looking for a developer who understands both backend logic and sustainability metrics, feel free to connect!

Top comments (0)