demo/README.md

CorrSteer Frontend

Overview

CorrSteer demonstrates how text classification datasets can be used to steer large language models (LLMs), correlating with SAE (Sparse Autoencoder) features. This demo incorporates a modern tech stack for a seamless and efficient experience.

---

How to Run the Demo

1. Set Environment Variables: Create a .env file in the demo directory and include the following:

   VITE_API_BASE_URL=<your-api-url>
   

2. Install Dependencies:

   pnpm i
   

3. Start the Development Server:

   pnpm dev
   

   The application will be available at http://localhost:5173 by default.

4. Build for Production (Optional):

   pnpm build
   pnpm preview
   

---

Key Features

• Dataset & Model Selection: Select datasets and models using dropdown menus.
• Streaming Outputs: Generate outputs from multiple models with live updates as data streams.
• Interactive Tabs: Switch between different categories for customized prompts.

---

Technology Stack

1. Vite:

   • Development server and build tool.

2. React:

   • UI library for building components and managing state.

3. Tailwind CSS:

   • CSS framework for styling.

4. ShadCN/UI:

   • Pre-built component library for UI elements.

---

License

This project is licensed under the MIT License. ```