JudgementDay

An AI prompt version control system with an LLM-as-judge experiment framework. Built with React/Next.js, JudgementDay supports datasets (CSV uploads), experiments tied to prompt versions, live output viewing during experiment runs, and persistent saving to localStorage.

Features

• Prompt Version Control: Git-like branching and merging for AI prompts
• Dataset Management: Upload and manage CSV datasets for experiments
• Experiment Framework: Create experiments with custom judge prompts
• Live Output: Real-time viewing of experiment results as they run
• API Integration: Support for OpenAI and Anthropic APIs
• Persistent Storage: All data saved to localStorage

Getting Started

First, run the development server:

npm run dev

Open http://localhost:3000 with your browser to see the result.

Usage

1. Create a Prompt: Start by creating a new prompt with initial parameters
2. Upload Datasets: Upload CSV files to use in experiments
3. Create Experiments: Set up experiments with custom judge prompts
4. Run Experiments: Execute experiments with live output viewing
5. Version Control: Use branching and merging for prompt evolution

API Configuration

Configure your API keys in the settings panel:

• OpenAI API key for GPT models
• Anthropic API key for Claude models

Tech Stack

• Frontend: React/Next.js with TypeScript
• UI: Radix UI components with Tailwind CSS
• State Management: React hooks with localStorage persistence
• APIs: OpenAI and Anthropic integration