strudel-llm-mirror

Strudel LLM Fork

This is a fork of Strudel for experimenting with Large Language Model (LLM) integration.

Original Strudel

Strudel is a JavaScript port of TidalCycles, bringing live coding patterns to the web browser. For the original project, see:

• Original Repository: https://github.com/tidalcycles/strudel
• Website: https://strudel.cc/
• Documentation: https://strudel.cc/learn/

LLM Integration Features

This fork adds:

• MCP (Model Context Protocol) Server: FastAPI server for LLM communication
• WebSocket Support: Real-time communication between LLMs and Strudel
• Session Management: Multi-session support for concurrent LLM connections
• Code Execution: LLMs can play, stop, and retrieve code from Strudel
• Docker Production Build: Complete containerization for deployment

MCP Server - The Core Innovation

This fork's main feature is the Python FastAPI MCP (Model Context Protocol) Server that enables LLMs to interact with Strudel in real-time.

What the MCP Server Does

The MCP server (strudel_mcp_server.py) provides:

• WebSocket Communication: Real-time bidirectional communication with LLMs
• Session Management: Multiple concurrent LLM sessions with unique session IDs
• Code Execution: LLMs can play/stop patterns and retrieve current code
• Static Site Serving: Serves the built Strudel website
• Documentation Routes: Comprehensive routing for all Strudel docs

LLM Integration Features

• Play Code: LLMs send code to execute in Strudel
• Stop Playback: LLMs can stop current patterns
• Get Current Code: LLMs can retrieve what's currently in the editor
• Session Tracking: Each LLM connection gets a unique session ID
• Real-time Status: Connection status monitoring

Architecture

LLM (Claude Code) ←→ WebSocket ←→ MCP Server ←→ Strudel Web Interface

1. LLM connects via WebSocket with session ID
2. MCP server manages connections and message routing
3. Strudel interface displays session info and executes code
4. Real-time sync between LLM commands and audio output

Deployment

Production (MCP Server)

# Build production container with MCP server
docker build -t strudel-prod .

# Run MCP server on port 8080
docker run -p 8080:8080 strudel-prod

Development (Optional)

# Traditional Strudel development
pnpm install && pnpm dev  # Port 4321

Docker Build Script

./docker-build.sh prod    # MCP production build
./docker-build.sh dev     # Development build
./docker-build.sh         # Build both

Using the MCP Server

1. Deploy the container to your server (Coolify, etc.)
2. Access the web interface: http://your-server:8080/strudel/
3. Connect LLMs via WebSocket: ws://your-server:8080/ws?session_id=your_id
4. LLMs can now play code, stop playback, and get current code
5. Session info displays in the Strudel header

Deployment

This fork is designed for deployment with:

• Coolify on Digital Ocean
• Docker containers
• FastAPI backend
• Static site serving

GitHub Mirror

A deployment mirror is maintained at: https://github.com/calvinw/strudel-llm-mirror

Use ./sync-to-github.sh to sync changes to the mirror for deployment.

License

Same as original Strudel: AGPL-3.0-or-later

Contributing

This is an experimental fork. For the main Strudel project, contribute to: https://github.com/tidalcycles/strudel