Co-Pilot / 辅助式
更新于 24 days ago

plugin-freedom-system

Gglittercowboy
0.1k
glittercowboy/plugin-freedom-system
86
Agent 评分

💡 摘要

一个通过对话设计创建VST3和AU音频插件的AI辅助系统。

🎯 适合人群

希望创建自定义音频效果的音乐家希望在不编码的情况下原型化插件的声音设计师需要为其工作流程量身定制音频工具的制作人对DSP感兴趣但不想编程的音频工程师教授音频插件开发的教育工作者

🤖 AI 吐槽:看起来很能打,但别让配置把人劝退。

安全分析中风险

风险:Medium。建议检查:是否执行 shell/命令行指令;是否发起外网请求(SSRF/数据外发);API Key/Token 的获取、存储与泄露风险。以最小权限运行,并在生产环境启用前审计代码与依赖。

Plugin Freedom System

An AI-assisted JUCE plugin development system that enables conversational creation of professional VST3 and AU audio plugins for macOS. Design and build custom audio processors through natural dialogue with Claude Code—no programming experience required.

Created by TÂCHES

Watch the Demo

Why This Exists

Traditional plugin development demands deep expertise in C++, DSP algorithms, and the JUCE framework. This barrier keeps plugin creation restricted to experienced programmers, leaving musicians, producers, and sound designers dependent on commercial tools.

The Plugin Freedom System removes that barrier entirely.

By enabling conversational plugin development, this system:

  • Democratizes creation: Anyone with an idea can build it, regardless of technical background
  • Prioritizes creativity: Focus on sonic goals and UX, not implementation details
  • Accelerates iteration: Ideas become working plugins in hours, not weeks
  • Removes gatekeeping: Opens audio software development to the people who actually use these tools

What You Can Build

  • Effects: Reverbs, delays, distortion, modulation, filters, dynamics processors
  • Synthesizers: Subtractive, FM, wavetable, granular, additive
  • Utilities: Analyzers, meters, routing tools, MIDI processors
  • Experimental: Custom DSP algorithms, hybrid processors, generative tools

All plugins compile to native VST3/AU formats compatible with any DAW (Ableton, Logic, Reaper, etc.).

How It Works

1. Dream (/dream)

Brainstorm your plugin concept through conversation:

  • Creative brief - Vision, sonic goals, UX principles
  • Parameter specification - All controls, ranges, and mappings
  • UI mockups - Visual design and layout

2. Plan (/plan)

Research and design the technical architecture:

  • DSP architecture - Signal flow and processing strategy
  • Implementation plan - Technical approach and complexity analysis

3. Implement (/implement)

Transform your specifications into a fully functional plugin through an automated workflow:

  • Build System Ready (Stage 1): Project structure, CMake configuration, and all parameters implemented - validated automatically
  • Audio Engine Working (Stage 2): DSP algorithms and audio processing complete - validated automatically
  • UI Integrated (Stage 3): WebView interface connected to audio engine (or skip for headless plugins) - validated automatically with runtime tests
  • After Stage 3 validation passes: Plugin complete, ready to install

4. Deploy & Iterate

  • /install-plugin - Install to system folders for DAW use
  • /test - Run automated validation suite
  • /improve - Add features or fix bugs (with regression testing)
  • /reconcile - Reconcile state between planning and implementation

Modern Interface Design

Plugins use web-based interfaces (HTML/CSS/JS) rendered via JUCE's WebView instead of traditional GUI frameworks. This enables:

  • Rapid prototyping: See design changes instantly without rebuilding
  • Modern aesthetics: Leverage contemporary web design patterns and animations
  • Interactive mockups: Test and refine interfaces before implementation
  • Familiar tools: Use web technologies many creators already understand
  • Responsive layouts: Easily adapt UIs to different sizes and contexts

GUI-Optional Workflow

Plugins can skip custom UI and ship as "headless" plugins using DAW-provided controls:

  • Faster iteration: Test DSP immediately without waiting for UI implementation
  • Progressive enhancement: Add custom UI later via /improve
  • Flexibility: Decide when/if to build visual interface
  • Zero overhead: Smaller binary, faster compile, all parameters exposed to DAW

Key Features

Automated Build Pipeline

7-phase build system (scripts/build-and-install.sh) handles validation, compilation, installation, and verification. No manual CMake commands or Xcode configuration required.

Quality Assurance

  • Automatic validation after each stage (compile-time + runtime tests)
  • validation-agent runs pluginval automatically (VST3/AU validation)
  • Validation is blocking - errors must be fixed before progressing
  • Regression testing on modifications
  • Backup verification before destructive operations
  • Build failure detection and troubleshooting

Knowledge Base

Dual-indexed troubleshooting database (troubleshooting/) captures solutions to build failures, runtime issues, GUI problems, and API misuse. The system learns from every problem encountered.

Required Reading (juce8-critical-patterns.md) automatically prevents repeat mistakes by injecting proven patterns into all subagent contexts.

Graduated Research Protocol

3-level investigation system (/research) for complex problems:

  • Quick: Single-agent targeted investigation (1-2 min)
  • Moderate: Multi-agent parallel search (3-5 min)
  • Deep: Comprehensive multi-level analysis (5-10 min)

Version Management

  • Semantic versioning on improvements
  • Git-based state tracking
  • Safe rollback capabilities
  • Backup verification before destructive operations

Workflow Modes

  • Manual mode (default): Present decision menus at each checkpoint for full control
  • Express mode: Auto-progress through implementation stages without intermediate menus
  • Configurable: Set preferences in .claude/preferences.json or use --express/--manual flags
  • Safe: Express mode drops to manual on any error, ensuring oversight when needed

Lifecycle Management

  • /install-plugin - Deploy to system folders
  • /uninstall - Remove binaries (keep source)
  • /reset-to-ideation - Roll back to concept stage
  • /destroy - Complete removal with verified backup
  • /clean - Interactive cleanup menu

System Architecture

Contracts (Single Source of Truth)

Every plugin has immutable contracts in plugins/[Name]/.ideas/:

  • creative-brief.md - Vision, sonic goals, UX principles
  • parameter-spec.md - Complete parameter definitions
  • architecture.md - DSP design and signal flow
  • plan.md - Implementation strategy
  • ui-mockups/ - Visual design references

Zero drift: All stages reference the same specs. No "telephone game" across workflows.

Dispatcher Pattern

Each implementation stage runs in a fresh subagent context:

  • foundation-shell-agent (Stage 1) - Project structure and parameter management
  • dsp-agent (Stage 2) - Audio processing
  • gui-agent (Stage 3) - WebView UI
  • validation-agent (after each stage) - Automatic validation with runtime tests

No context accumulation: Clean separation prevents cross-contamination and keeps token usage minimal.

Discovery Through Play

All features discoverable via:

  • Slash command autocomplete (type / in Claude Code)
  • Numbered decision menus at checkpoints
  • Interactive skill prompts

No manual required: Learn by exploring, not reading docs.

Checkpoint Protocol

At every completion point:

  1. Auto-commit changes
  2. Update state files (.continue-here.md, PLUGINS.md)
  3. Present numbered decision menu
  4. Wait for user response
  5. Execute chosen action

Never auto-proceeds: You stay in control.

Quick Start

Prerequisites

  • macOS (Sonoma or later recommended)
  • Claude Code CLI

All other dependencies (Xcode Command Line Tools, JUCE, CMake, Python, pluginval) can be validated and installed via /setup.

First-Time Setup

# Validate and configure your system dependencies /setup # The setup wizard will: # - Detect your platform and installed tools # - Offer to install missing dependencies automatically or guide manual installation # - Save configuration for build scripts # - Generate a system report

Create Your First Plugin

# 1. Dream the concept /dream # Brainstorm your plugin idea through conversation # Creates: creative brief, parameter spec, UI mockups # 2. Plan the architecture /plan # Research and design the technical implementation # Creates: DSP architecture, implementation plan # 3. Build it /implement # Automated workflow builds the plugin # 4. Install and test /install-plugin YourPluginName # Plugin now available in your DAW

Improve an Existing Plugin

# Fix bugs or add features /improve MyPlugin # Describe the change # System handles versioning, testing, and rollback safety

Resume Interrupted Work

# Pick up where you left off /continue MyPlugin # System loads checkpoint and presents next steps

Complete Command Reference

Setup

  • /setup - Validate and configure system dependencies (first-time setup)
    • Detects platform, checks for required tools
    • Offers automated installation or guided manual setup
    • Saves configuration to .claude/system-config.json

Development Workflow

  • /dream - Brainstorm concept, create creative brief, parameter spec, and UI mockups
  • /plan - Research and design DSP architecture and implementation strategy
  • /implement [Name] - Build plugin through automated 3-stage workflow with continuous validation
  • /continue [Name] - Resume paused workflow
  • /improve [Name] - Modify completed plugin (with regression testing)

Quality Assurance

  • /test [Name] - Run automated validation suite
  • /research [topic] - Deep investigation (3-level protocol)
  • /doc-fix - Document solved problems for knowledge base
  • /add-critical-pattern - Add current problem to Required Reading

Deployment

  • /install-plugin [Name] - Install to system folders
  • /uninstall [Name] - Remove binaries (keep source)
  • /show-standalone [Name] - Preview UI in standalone mode

Lifecycle

  • /clean - Interactive cleanup menu (uninstall/reset/destroy)
  • /reconcile [Name] - Reconcile state between
五维分析
清晰度9/10
创新性9/10
实用性9/10
完整性8/10
可维护性8/10
优缺点分析

优点

  • 使非程序员能够民主化插件创建
  • 通过自动化工作流程加速开发
  • 支持现代网页技术进行UI设计
  • 全面的验证和测试过程

缺点

  • 需要macOS,限制了可访问性
  • 对于完全初学者来说学习曲线陡峭
  • 对Claude Code CLI的依赖可能会让一些用户却步
  • 管理插件版本的复杂性

相关技能

pytorch

S
toolCode Lib / 代码库
92/ 100

“它是深度学习的瑞士军刀,但祝你好运能从47种安装方法里找到那个不会搞崩你系统的那一个。”

agno

S
toolCode Lib / 代码库
90/ 100

“它承诺成为智能体领域的Kubernetes,但得看开发者有没有耐心学习又一个编排层。”

nuxt-skills

S
toolCo-Pilot / 辅助式
90/ 100

“这本质上是一份组织良好的小抄,能把你的 AI 助手变成一只 Nuxt 框架的复读机。”

免责声明:本内容来源于 GitHub 开源项目,仅供展示和评分分析使用。

版权归原作者所有 glittercowboy.