K.G.Studio 是 AI Skill Hub 本期精选Agent工作流之一。综合评分 7.5 分，整体质量较高。我们推荐使用将其纳入你的 AI 工具库，帮助提升工作效率。
📚 深度解析
K.G.Studio 工作流的设计遵循"最小配置，最大复用"原则：核心逻辑已经封装好，用户只需配置自己的 API Key 和业务参数即可快速上手。工作流内置错误处理和重试机制，在网络波动或 API 限速等情况下仍能稳定运行，适合作为生产环境的自动化基础设施。
在实际部署时，建议先在测试环境中运行 3-5 次，验证各个环节的输出结果符合预期，再部署到生产环境。AI Skill Hub 评分 7.5 分，是同类 Agent 工作流中的精选推荐。
📋 工具概览
K.G.Studio 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排，将复杂的多步骤任务拆解为清晰的自动化流程，实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成，适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
📖 中文文档
K.G.Studio 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排，将复杂的多步骤任务拆解为清晰的自动化流程，实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成，适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
- 可视化 Agent 工作流编排，无需编写复杂代码
- 支持多步骤自动化任务链，实现全流程无人值守
- 与外部 API、数据库和第三方服务无缝集成
- 内置错误处理与自动重试机制，保障稳定运行
- 提供可复用的自动化模板，快速在同类场景部署
- 自动化日常重复性工作，将精力集中于创造性任务
- 构建数据采集 → 处理 → 输出的完整自动化管线
- 实现跨平台、跨系统的数据流转和业务协同
# 方式一：npm 全局安装 npm install -g k.g.studio # 方式二：npx 直接运行（无需安装） npx k.g.studio --help # 方式三：项目依赖安装 npm install k.g.studio # 方式四：从源码运行 git clone https://github.com/KGAudioLab/K.G.Studio cd K.G.Studio npm install npm start
- 访问 GitHub 仓库获取工作流文件
- 在对应平台（Dify / Flowise / Make 等）中找到「导入工作流」功能
- 上传工作流文件
- 按照提示配置必要的环境变量和 API Key
- 运行测试确认流程正常后投入使用
# 命令行使用
k.g.studio --help
# 基本用法
k.g.studio [options] <input>
# Node.js 代码中使用
const k.g.studio = require('k.g.studio');
const result = await k.g.studio.run(options);
console.log(result);
# k.g.studio 配置说明 # 查看配置选项 k.g.studio --config-example > config.yml # 常见配置项 # output_dir: ./output # log_level: info # workers: 4 # 环境变量（覆盖配置文件） export K.G.STUDIO_CONFIG="/path/to/config.yml"
简介
About the agent and LLM providers
For security, when using K.G.Studio from a non‑local host, API keys are not persisted in IndexedDB by default; you will need to input your API key each time you start K.G.Studio. To opt‑in to persistence on non‑local hosts, enable "Persist API Keys on Non-Localhost" in Settings (not recommended for shared/production environments).
K.G.Studio does not provide or host any of the models listed above, nor is it affiliated with any model provider. All data is stored locally on your device; K.G.Studio does not collect or transmit your data. You are solely responsible for any data you provide to third‑party model providers.
Highlights
- K.G.Studio Musician Assistant: Chat with the LLM‑powered AI agent; it automatically executes tools to make music edits scoped to your selected region.
- Browser-embedded LLM — no API key required: Run Gemma 4 E4B entirely in your browser via WebGPU (LiteRT-LM). No API calls, no cost, no data leaving your device. Model is downloaded once and cached locally.
- Browser-embedded stem separation — no server required: Split any audio region into stems using UVR-MDX-NET-Inst_HQ_3 (2-stem: Vocals / Instrumental) or Demucs htdemucs_4s (4-stem: Vocals / Drums / Bass / Others) — both run entirely in-browser via ONNX Runtime WebGPU, no K.G.One server needed.
- Audio chord detection: Open the piano roll on an audio region and run Detect Chords to automatically analyse the recording and populate the global Chord Track using a zero-dependency FFT pipeline, with configurable sensitivity, stability, and seventh-chord detection.
- Tempo detection with auto-align beats: Run Detect Tempo in the piano roll toolbar to analyse an audio region for BPM and optionally realign the project's Tempo Track to match.
- Global Track System: Four persistent global tracks — Marker, Tempo, Key Signature, and Chord — provide project-wide structure that all features (playback timing, chord detection, sheet notation) reference.
- K.G.One Music Studio integration: When connected to a local K.G.One server, unlock GPU-accelerated Full Song Generation (ACE-Step 1.5), Clip & MIDI Loop Generation (Foundation-1), and additional Stem Separation models.
- Multiple LLM providers: OpenAI, Claude / Gemini (via OpenRouter), OpenAI-compatible (Ollama, vLLM, etc.), or the built-in Local Browser LLM — no key required.
- Track & Region editing: Add/reorder tracks, create/move/resize regions, multi-select with lasso, bulk move/resize, merge & split regions, and full undo/redo.
- Piano roll: Notes, pitch bend, and MIDI CC automation lanes; staff notation (sheet music) view via VexFlow; spectrogram overlay for audio-to-MIDI reference.
- Real instruments: Tone.js‑based sampler with high-quality FluidR3 soundfonts. Record audio directly from your microphone into an audio track.
- Intelligent Chord Assistant: Real-time chord suggestions based on functional harmony (T/S/D) with visual preview and one-click chord creation.
- Persistence with privacy: Projects and configuration saved entirely in your browser (OPFS / IndexedDB). No first-party servers.
For a deeper technical overview, see overview.md.
Upcoming Features
Feature priorities might change.
Install dependencies
npm install
Getting Started
Make sure you have Node.js >= 20.19.3 installed
Quick Start
Demo Videos
K.G.One Music Studio |
Short Demo (DAW Only) |
Full Demo (DAW Only) |
Setting up K.G.Studio Musician Assistant
There are two ways to use K.G.Studio Musician Assistant: Local LLM (Browser) — runs entirely in your browser with no API key, no cost, and no data leaving your device — or an external LLM provider for higher-quality responses.
Option A: Local LLM (Browser) — No API Key Required ✦
K.G.Studio can run Gemma 4 E4B entirely inside your browser using WebGPU acceleration. No API calls are made, no cost is incurred, and your data never leaves your machine.
- Click here to start using the app online: K.G.Studio (kgaudiolab.github.io/kgstudio)
- In Settings ⚙️ → General → LLM Provider, select Local LLM (Browser) (this is the default).
- The model (~2.8 GB) downloads automatically the first time you open the chat and is cached in your browser's OPFS (Origin Private File System) for instant subsequent launches.
- Optionally, configure the Context Length (32k / 64k / 128k tokens) — larger values require more VRAM.
- Start chatting! No key, no account, no network traffic after the initial model download.
Requirements for Local LLM: A WebGPU-capable browser (Chrome 113+ or Edge 113+) running in a secure context (HTTPS or localhost). Recommended hardware: a GPU with at least 8 GB VRAM, or a system with at least 16 GB unified RAM.
Note: The local model's quality cannot be compared to commercial models like GPT or Claude series. For complex music editing tasks, an external provider will generally produce better results.
Option B: External LLM Provider (Higher Quality)
- Click here to start using the app online: K.G.Studio (kgaudiolab.github.io/kgstudio)
- Click here to get a free OpenRouter API Key (you may need an OpenRouter account).
- In Settings ⚙️ → General → LLM Provider, select OpenAI Compatible.
- In OpenAI Compatible Server → Key, paste your key. (Note: on non‑localhost, your key isn't persisted by default for security; you can enable "Persist API Keys on Non-Localhost" in Settings to persist them, though this may increase XSS risk.)
- In OpenAI Compatible Server → Model, enter
openai/gpt-oss-120b:free. (Note: this is a free model; non‑free models may require billing; free model providers may collect your data, check the model page for details; this project is not affiliated with OpenRouter or any model provider.) - In OpenAI Compatible Server → Base URL, enter
https://openrouter.ai/api/v1.
Tips: - You can also use the official OpenAI API, other OpenAI-compatible services, or a self-hosted LLM server (e.g., Ollama, vLLM). Note that model quality varies — not all models perform equally well for music editing tasks. For local hosting, we recommend qwen3.5-35b-a3b as a good balance between generation quality and hardware requirements. - If you have an active subscription with OpenAI or another LLM provider, you can use CLIProxyAPI to run a local proxy server that routes requests through your existing subscription, without needing a separate API key.
Configuration
K.G.Studio loads defaults from ./public/config.json (with an internal fallback) and persists user edits in the browser via ConfigManager + IndexedDB. IndexedDB is your browser’s own local database stored on your device; it does not leave your machine and is cleared if you clear this site’s data. Modify settings via the in‑app Settings panel.
- General
- LLM provider: OpenAI, or OpenAI‑compatible
- API keys and models for the selected provider
- Persist API Keys on Non-Localhost: Enable to persist API keys on non-localhost environments (security opt-in, not recommended for shared/production environments)
- OpenAI‑compatible base URL (for self‑hosted gateways)
- Soundfont base URL (CDN for instrument samples)
- Behavior
- Chatbox default open on startup
- Templates
- Custom instructions used by the AI assistant
Configuring Your LLM Provider
Navigate to Settings ⚙️ → General → LLM Provider. Depending on your chosen provider, you will need to supply the appropriate API Key and, if applicable, a custom base URL (for non-official OpenAI-compatible services such as Ollama, OpenRouter, etc.).
Note: due to CORS limitations with some providers, Google Gemini and Anthropic Claude are currently supported via OpenRouter only.
创新性的AI音乐工作流
- 需要 K-G-Studio 解决具体问题的开发者与运营人员
- 先在测试环境跑通最小用例，再接入生产数据
- API key 直接提交到 git 仓库（请用 .env 并加入 .gitignore）
- 云端托管：可放在 Vercel / Railway / Fly.io 等 PaaS 平台
⚡ 核心功能
- 可视化 Agent 工作流编排，无需编写复杂代码
- 支持多步骤自动化任务链，实现全流程无人值守
- 与外部 API、数据库和第三方服务无缝集成
- 内置错误处理与自动重试机制，保障稳定运行
- 提供可复用的自动化模板，快速在同类场景部署
- 需要 K-G-Studio 解决具体问题的开发者与运营人员
- 先在测试环境跑通最小用例，再接入生产数据
- API key 直接提交到 git 仓库（请用 .env 并加入 .gitignore）
👥 适合人群
🎯 使用场景
- 自动化日常重复性工作，将精力集中于创造性任务
- 构建数据采集 → 处理 → 输出的完整自动化管线
- 实现跨平台、跨系统的数据流转和业务协同
⚖️ 优点与不足
- +大幅减少重复性人工操作
- +可视化流程，清晰直观
- +可扩展性强，支持复杂场景
- −初始配置和调试需投入一定时间
- −强依赖外部服务的稳定性
- −复杂场景需具备一定技术基础
该工具使用 NOASSERTION 协议，商用场景请仔细阅读协议条款，必要时咨询法律意见。
AI Skill Hub 为第三方内容聚合平台，本页面信息基于公开数据整理，不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后，再部署至生产环境，并做好必要的安全评估。
📄 NOASSERTION — 请查阅原始协议条款了解具体使用限制。
🔗 相关工具推荐
❓ 常见问题 FAQ
经综合评估，K.G.Studio 在Agent工作流赛道中表现稳健，质量良好。如果你已有明确的使用需求，可以直接上手体验；如果还在评估阶段，建议对比同类工具后再做决策。
| 原始名称 | K-G-Studio |
| 原始描述 | 开源AI工作流：K.G.Studio is a lightweight, browser‑based DAW with an LLM‑powered AI Agent "K.G。⭐18 · TypeScript |
| Topics | AIDAWLLM |
| GitHub | https://github.com/KGAudioLab/K.G.Studio |
| License | NOASSERTION |
| 语言 | TypeScript |
收录时间：2026-05-29 · 更新时间：2026-05-31 · License：NOASSERTION · AI Skill Hub 不对第三方内容的准确性作法律背书。