PurePath

Inspiration

Honestly, we were just tired of the marketing fog. It feels like every brand is suddenly "eco friendly" just because they put a leaf on their packaging. With nearly 80% of investors saying greenwashing is getting worse, we realized that if the pros are confused, the rest of us don't stand a chance. We wanted to build something that actually cuts through the noise and checks if those Net Zero promises are real or just a way to sell more stuff.

What it does

PurePath is basically a BS detector for sustainability. Scrape Promises: It reads corporate websites to pull out specific commitments like "Net Zero by 2030" or "Plastic-Free". Checks the Receipts: It scans news, lawsuits, and reports for anything that contradicts those claims like hidden emissions or shady lobbying. Gives a Truth Score: It wraps everything up into a simple 0–100 score so you can see if a brand is actually a leader or just another greenwasher.

How we built it

We used Yellowcake API to handle structured web scraping of the corporate websites. Gemini 2.5 Flash does the heavy lifting, comparing what the company said to what the news says happened. A Node.js server manages API keys and routes data between the frontend and the AI models. A Streamlit dashboard because it was the fastest way to show the data streaming in real time.

Challenges we ran into

API Limits: We hit the "Resource Exhausted" wall pretty early with experimental models, so we had to pivot to Gemini 2.5 Flash and write some logic to handle retries. Messy Data: Corporate ESG reports were a nightmare to parse. We had to use Pydantic schemas to force the AI to give us clean JSON instead of a wall of text.

Accomplishments that we're proud of

1. We got the Server-Sent Events (SSE) working so you can see the claims pop up as the AI finds them. 2.We're proud that we turned thousands of data points into one simple "Truth Score" that actually makes sense.

Built With

• geminiapi
• python
• streamlit
• yellowcakeapi