Inspiration

  • Dealing with not getting the outputs you want from certain AI engines, and having that slow down the coding process.

What it does

  • MetaPrompt is a browser extension that transforms highlighted text into optimized AI prompts—saving time, boosting response quality, and integrating seamlessly with ChatGPT, Gemini, and Claude.

How we built it

  • Tech Stack
    • Frontend: Browser Extension (Manifest V3), HTML, CSS, JavaScript
    • Backend: Server Written with FastAPI, Pydantic, and Dotenv. Prompt Optimization Powered by Gemini API
    • Platform Integrations: ChatGPT, Gemini, Claude, FetchAPI

Challenges we ran into

  • Countless JavaScript errors, mostly related to permissions and clipboard access. The permission to copy to clipboard requires a button press to give web-admin privilege but it takes 3s on average to query our database in which that privilege can then expires. Backend errors, sleep deprivation :)

Accomplishments that we're proud of

  • Actually creating a project that we will be useful in each of our lives that we'll actually be able to use outside of this and also be proud to show off.

What we learned

  • We learned how to effectively work with JavaScript, CSS, FastAPI, and Gemini's API. We have a new appreciation for whiteboarding, and how it can break really complicated concepts down.

What's next for MetaPrompt

  • If we were to continue with MetaPrompt, we would first host our 'local server' and API on AWS along with upgrade our free Gemini token to a paid Gemini token to boost performance and lower overall runtime. We would then upload our current production build to the Chrome Web Store and Firefox Web Store. We would also add Analytics to our current project that way we can look to improve the model, and improve the UI / UX.

Built With

Share this project:

Updates