Prompt Footprint

Inspiration

We wanted to make a tool that helped quantify and reduce the environmental impacts of AI because we often saw news about the rise of AI data centers, but lacked the proper knowledge on the specifics. We decided that a chrome extension, which lives directly in the browser that most people use AI on would be the best way to implement this.

What it does

PromptFootprint tracks token usage in your conversations with ChatGPT, Claude, and Gemini, identifies the model being used and it's associated energy and water usage, as well as carbon dioxide emissions, to calculate the user's environmental impact. For more detailed statistics, the user can navigate to the full stats website from the extension pop-up to view visualized data, including weekly graphs, past sessions, and data animations, to help the user fully understand their impact. Following that, to help people actually reduce their footprint, we decided to add a prompt optimizing feature, which allows people to see a popup directly in the browser that helps cut down on their token usage per prompt, therefore reducing their environmental impact.

How we built it

We built it using standard languages for web development (CSS, HTML, JavaScript), as well as deployment tools such as Vercel for hosting and Railway for our backend. We use modals and popups for the in-the-browser features, tokenizers such as OpenAI's tokenizer to calculate token usage, and a database with filler words and common phrases that were unneeded in prompts to check against for the prompt optimizing feature.

Challenges we ran into

One struggle we had was deciding how to go about the prompt optimizing feature and to check which parts were unneeded, as using an AI to analyze those prompts would be the opposite of what we were trying to solve, so we eventually landed on a database to check against.