This project provides a proxy worker for integrating Ollama with the Cursor IDE, enabling local LLM inference for Cursor's AI features. It addresses the limitation where Cursor's default server configuration prevents direct communication with local Ollama instances, offering a solution for users who want to leverage their own models.

How It Works

Curxy acts as an intermediary server, forwarding requests from the Cursor IDE to a running Ollama instance. It then relays the Ollama response back to Cursor. This approach bypasses Cursor's default server routing, allowing seamless integration with locally hosted LLMs. The project utilizes Deno for its runtime environment.

Quick Start & Requirements

• Install/Run: deno run -A jsr:@ryoppippi/curxy
• Prerequisites: Deno runtime, Ollama server running locally.
• Configuration: Set "Override OpenAI Base URL" in Cursor to the URL provided by curxy (e.g., https://your-curxy-url.trycloudflare.com/v1). Add desired model names to Cursor's configuration.
• Optional: Set OPENAI_API_KEY environment variable for access restriction.

Highlighted Details

• Provides a public URL via Cloudflare Tunnel for external access.
• Supports access restriction via an OPENAI_API_KEY environment variable.

Maintenance & Community

No specific information on contributors, sponsorships, or community channels is provided in the README.

Licensing & Compatibility

• License: MIT
• Compatibility: Permissive MIT license allows for commercial use and integration with closed-source applications like Cursor.

Limitations & Caveats

The project is described as a "simple proxy worker," implying potential limitations in advanced features or robustness compared to more comprehensive solutions. The reliance on Cloudflare Tunnel for public URLs might introduce external dependencies or potential points of failure.