Your Private, Local-First AI Desktop Assistant
Chat · Roleplay · Study Tools · Knowledge Base · MCP Tools — All in One App
Quick Start · Features · Download · 国内下载
Zero cloud dependency. Your data never leaves your machine.
Most AI chat apps force you to pick a lane — casual chat, roleplay, or productivity. EchoMuse doesn't. It's a single lightweight desktop app that handles daily conversations, deep character roleplay, academic study tools, and document-powered knowledge retrieval — all running locally on your machine with optional cloud API support.
零云端依赖，数据始终留在本地。
大多数 AI 聊天工具只能做一件事。EchoMuse 不一样 — 日常聊天、深度角色扮演、学习工具、知识库检索，一个桌面应用全部搞定，本地运行，也可接入云端 API。
Run Ollama models locally (Qwen3, Mistral, Llama, etc.) or connect to 7+ cloud APIs (OpenAI, DeepSeek, Kimi, Claude, SiliconFlow, Zhipu, Baichuan) — or mix both. Every model supports Flash (fast response) and Thinking (deep reasoning with chain-of-thought) dual modes.
本地 Ollama 模型 + 7 种以上云端 API，可混合使用。每个模型支持 Flash（快速）和 Thinking（深度推理）双模式切换。
- Multi-session management with search, favorites, and auto-titling
- Message branching — regenerate or edit any message to create tree-structured conversation paths
- Web search injection — DuckDuckGo results fed directly into context
- Real-time translation overlay — translate AI replies into 8 languages while keeping the original
多会话管理 · 消息分支（树状对话） · 联网搜索注入 · 实时翻译覆盖层（8 语言）
- Custom characters with avatars, expressions, memory text, and relationship settings
- Tavern card import — drag-and-drop PNG/JSON character cards
- Lorebook — keyword-triggered world-building lore auto-injected into context
- Multi-character group chat with spectator mode
- Message branching for exploring different story paths
自定义角色卡 · Tavern 卡导入 · Lorebook 世界观书 · 多角色群聊 · 旁观模式 · 剧情分支探索
Upload documents (txt / md / pdf / docx) → automatic chunking → BM25 + vector hybrid retrieval. Toggle it on in any chat and the AI answers with your documents as context.
上传文档自动分块索引，BM25 + 向量混合检索，聊天中一键启用。
Connect external MCP servers (stdio / SSE / HTTP) to extend the AI's capabilities — web scraping, sequential thinking, note-taking, and more. Supports multi-turn automatic tool calling.
接入外部 MCP 服务器，扩展 AI 能力 — 网页抓取、深度推理、笔记等，支持多轮自动工具调用。
- Exam Review Pack Generator — outlines, key points, question banks (MCQ / fill-in / short answer), Anki flashcard export
- Paper & Report Generator — lab reports, course papers, structured academic writing
期末复习包生成器（大纲 + 题库 + Anki 卡片） · 论文/实验报告生成器
- TTS — AI reads replies aloud with configurable voice packs
- STT — browser speech recognition + local Whisper offline transcription
TTS 语音朗读 + 本地 Whisper 语音输入
- Export to JSON / Anki CSV / PDF / Word
- Team sharing — OpenAI-compatible API relay with member tokens and usage tracking
- LAN access — QR code for mobile/tablet access over local network
多格式导出 · 团队共享（OpenAI 兼容中继） · 局域网 QR 码访问
- 8 languages — Chinese, English, Japanese, Korean, French, German, Spanish, Russian
- 3 themes — Light, Dark, System auto
- First-run setup wizard — guided onboarding for beginners
- PWA support — also works as a web app in your browser
8 语言 UI · 3 种主题 · 新手引导向导 · PWA 支持
Go to GitHub Releases or Gitee Releases (国内) and grab the latest
.exe— NSIS installer or portable version.
git clone https://github.com/YPDWHM/EchoMuse.git
cd EchoMuse
npm install
# Web version (local only)
npm run local
# Desktop version (Electron)
npm run desktop:devGitee 用户：
git clone https://gitee.com/tutuHM/echo-muse.git
LAN sharing:
npm run share
npm run bootstrap:winEchoMuse works with both local and cloud models — use one or both.
Install Ollama, then pull a model:
ollama pull qwen3:8b # Recommended — 8B general purpose
ollama pull mistral:7b # Alternative — fast and lightweightAdd a Provider in Settings. Supported:
| Provider | Notes |
|---|---|
| OpenAI | GPT series |
| DeepSeek | Cost-effective |
| Kimi (Moonshot) | Long context |
| SiliconFlow | China-optimized CDN |
| Zhipu AI | GLM series |
| Baichuan | Chinese-optimized |
| Anthropic | Claude series |
| Custom | Any OpenAI-compatible endpoint |
graph LR
subgraph Frontend
A[Vanilla JS SPA] --> B[Chat UI]
A --> C[Settings]
A --> D[Artifact Drawer]
end
subgraph Backend
E[Express Server]
E --> F[Ollama Local]
E --> G[Cloud APIs]
E --> H[MCP Manager]
E --> I[RAG Engine]
E --> J[Translation]
E --> K[Team Relay]
end
A <-->|HTTP + SSE| E
subgraph Desktop
L[Electron Shell] --> E
end
| Layer | Tech |
|---|---|
| Frontend | Vanilla JS, KaTeX, Marked |
| Backend | Node.js, Express |
| Desktop | Electron |
| LLM | Ollama, OpenAI API, Anthropic API |
| Retrieval | BM25 + Vector Embedding (Cosine Similarity) |
| Tools | MCP (Model Context Protocol) |
| Export | jsPDF, docx, html2canvas |
EchoMuse/
├── server.js # Express backend (API, LLM, RAG, MCP, translation, team sharing)
├── mcp-manager.js # MCP client connection manager
├── public/
│ ├── index.html # Main page
│ ├── app.js # Frontend logic
│ ├── styles.css # Styles (light/dark themes)
│ └── js/ # Modular JS (utils, domain, chat render, voice, setup wizard)
├── desktop/
│ ├── main.js # Electron main process
│ └── preload.js # Electron preload
├── prompts/ # Prompt templates
└── scripts/ # Startup & build scripts
Q: I don't know which model to pick. / 不懂选什么模型？ Add a cloud API Provider in Settings (e.g. DeepSeek) — no local model needed. 在设置里添加云端 API Provider（如 DeepSeek），不装本地模型也能用。
Q: Can my PC run local models? / 电脑能跑本地模型吗？ 16GB RAM can handle 8B models (qwen3:8b, mistral:7b). 13B+ needs better hardware. 16GB 内存可跑 8B 级别模型，更大模型需要更好配置。
Q: How to import character cards? / 怎么导入角色卡？ Sidebar → Contacts → Import Card. Supports PNG (embedded data) and JSON. 侧边栏「联系人」→「导入卡片」，支持 PNG 和 JSON 格式。
Q: How to share with others on LAN? / 怎么局域网共享？ Enable Team Sharing in Settings, generate member tokens, others connect via your LAN IP. 设置里开启「团队共享」，生成成员 Token，对方通过局域网 IP 访问。
Issues and feature requests are welcome on GitHub or Gitee.
欢迎在 GitHub 或 Gitee 提交 Issue 和功能建议。
EchoMuse Source Available License — Free for personal, educational, and non-commercial use. Commercial use requires author authorization.
个人使用、学习、教育用途免费。商业用途需获得作者授权。
If EchoMuse helps you, consider giving it a ⭐ — it means a lot!
如果 EchoMuse 对你有帮助，点个 ⭐ Star 支持一下吧！