Inspiration
The idea for PromptDiff came from the frustration of managing and testing LLM prompts. Even small changes in prompts can drastically affect outputs, yet there were no tools to track versions, compare differences, and test efficiently. So i create a tool that brings developer-friendly version control and clarity to prompt engineering.
What it does
It compare your prompt so you can check and testing which version of your code performing best.
How we built it
Frontend: React (for prompt editor, diff viewer, and output panel) Backend/API: Gemini LLM API (default) Database: Local storage for prompt history and versioning Features include:
- Selecting any two versions of a prompt to compare
- Side-by-side output comparison with token usage
- Session-based BYO API key support
Challenges we ran into
- Designing a clear, Git-like diff view for text prompts
- Calculating token usage across multiple LLM outputs
- Balancing feature scope with hackathon time constraints
Accomplishments that we're proud of
- An clean and minimal UI/UX design
- fast response
What we learned
- Time management
- How to handle api key and different work enviroment.
- LLM prompt behavior and token usage
- Versioning concepts similar to Git applied to text -Integrating AI APIs in a web environment
What's next for PromptDiff
PromptDiff: Git for LLM prompts — version, compare, test, and optimize your prompts with ease for smarter AI outputs
Built With
- api
- browser-local-storage
- gemini
- gemini-api-key
- javascript
- react
- tailwindcss
Log in or sign up for Devpost to join the conversation.