CursorWeb

[📄arXiv] | [🤗HF Paper] | [🤖Models] | [🛠️Code] | [Web] | [Discord]

---

• CursorWeb
  • Introduction
  • Features
  • Installation
    • Prerequisites
    • Frontend Setup
    • Backend Setup
  • Usage
  • Citation
  • Contribution

---

Introduction

CursorWeb is the frontend display interface of CursorCore, designed to implement popular features of Cursor in the browser without relying on specific editor or IDE. The application is built using Vue.js for the frontend and FastAPI for the backend. Please read our paper to learn more.

Features

• Chat: A conversational interface similar to ChatGPT.
• Inline Chat: An inline chat feature similar to Github Copilot Chat or Cursor Command K.
• Tab: An automated editing feature similar to Cursor Copilot++

Installation

Prerequisites

• Node.js (version 14 or higher)
• Python (version 3.7 or higher)

Frontend Setup

1. Enter the frontend directory:

   cd frontend

2. Install the dependencies:

   yarn install

3. Start the development server:

   yarn serve

Backend Setup

1. Enter the backend directory:

   cd backend

2. Install the required Python packages:

   pip install -r requirements.txt

3. Start the model inference service CursorCore uses vllm's OpenAI-compatible server to run its services. Below is an example command to start the service:

   python -m vllm.entrypoints.openai.api_server --port 10086 --model TechxGenus/CursorCore-Yi-1.5B --dtype bfloat16 --enable-prefix-caching

   For more details on model inference configuration and optimization options, refer to the documentation of vllm.

   Note: The backend leverages extra parameters specific to vllm's OpenAI-compatible server for handling custom chat templates and special tokens. Other OpenAI-compatible inference services may not be directly applicable.

4. Configure model service settings: Define the model inference service parameters in model_map.json. An example configuration is as follows:

   {
       "TechxGenus/CursorCore-Yi-1.5B": {
           "base": "http://127.0.0.1:10086/v1",
           "api": "sk-xxx"
       }
   }

5. Run the FastAPI server:

   python main.py --model_map model_map.json

Usage

Open your browser and go to http://localhost:8080 to access the interface.

The left window provides the chat interface, while the code editor on the right displays the inline chat and tab features. To use the inline chat, press Ctrl+I (or Command+I on Mac). To accept automated editing suggestions, press Tab.

Citation

@article{jiang2024cursorcore,
  title   = {CursorCore: Assist Programming through Aligning Anything},
  author  = {Hao Jiang and Qi Liu and Rui Li and Shengyu Ye and Shijin Wang},
  year    = {2024},
  journal = {arXiv preprint arXiv: 2410.07002}
}

Contribution

Contributions are welcome! If you find any bugs or have suggestions for improvements, please open an issue or submit a pull request.