neuron-explorer
Qwen3.5‑4B 전용 조향형 뉴런 탐색기, 모든 MLP 뉴런을 밀어 결과를 보관·탐색
QevosAgent: 무료 오픈소스 로컬 AI 에이전트
Free & open-source local AI agent. Supports Qwen3, Qwen3.6 27B, DeepSeek V4, Gemma 4, Claude and more. Persistent memory, tool calling, web dashboard. Zero API cost, data never leaves your machine.
개발 도구Python · HTMLGitHub @HongyunQiu온라인운영자 미확인
써 보셨어요? 한 번만 눌러 주세요. 로그인은 필요 없어요.
QevosAgent is a minimalist agent runtime for real tool-calling workflows with persistent artifacts, recoverable memory, and observable execution.
저장소 README에서 가져옴
README 전체를 불러오는 중입니다. 아래는 저장된 발췌입니다.
🇺🇸 English (EN) • 🇨🇳 中文 (ZH) 🦊 Your Local AI Agent, Ready Out of the Box A local AI agent truly designed for everyone — native Windows/macOS/Linux installation, no WSL required, works out of the box. The core design philosophy of QevosAgent is to keep the execution process as simple as possible, move complexity out of the runtime logic, and encapsulate capabilities into an extensible tool system.README 전문 보기 ↗
🇺🇸 English (EN) • 🇨🇳 中文 (ZH)
🦊 Your Local AI Agent, Ready Out of the Box
A local AI agent truly designed for everyone — native Windows/macOS/Linux installation, no WSL required, works out of the box.
The core design philosophy of QevosAgent is to keep the execution process as simple as possible, move complexity out of the runtime logic, and encapsulate capabilities into an extensible tool system.
Instead of relying on complex and opaque internal workflows, QevosAgent defines capabilities as tools. These tools can continuously evolve and improve, which means the capability boundary of QevosAgent is not fixed. It can expand over time as the tool ecosystem grows, giving it a broad and open-ended potential.
At the same time, QevosAgent emphasizes observability and transparency. Through a visual dashboard, users can monitor the Agent’s execution process, task status, and key information in real time, making the progress of complex work clear, visible, and easy to understand.
Unlike other AI Agents, QevosAgent was designed for desktop users from day one:
QevosAgent is not just a tool for developers:
| Platform | Download |
|---|---|
| 🪟 Windows | Download Windows Installer |
| 🍎 macOS Apple Silicon | Download macOS ARM |
| 🍎 macOS Intel | Download macOS Intel |
| 🐧 Linux | Download Linux AppImage |
git clone https://github.com/HongyunQiu/QevosAgent.git
cd QevosAgent
pip install -r requirements.txt
copy .env.example .env
# Edit .env and fill in your API Key
# Launch the web dashboard
cd dashboard
npm install
cd ..
node dashboard/server.js
Any capability can be registered as a tool. Through tool evolution, you can massively expand the Agent's abilities — the possibilities are limitless.
Every task is automatically saved to disk — logs, scratchpads, summaries and final answers, auditable at any time.
Memory persists across sessions. The Agent picks up right where it left off — the more you use it, the smarter it gets.
When a tool fails, the Agent automatically diagnoses and fixes it — no infinite loops, no deadlocks.
Launch tasks, inject commands, and browse history from your browser — fully visual.
Deep support for Qwen3, Gemma4, and any OpenAI-compatible endpoint. Zero API cost, data stays on your machine.
File read/write, Python execution, shell commands, web search, memory management — all ready out of the box.
A dedicated high-level advisor LLM provides strategic guidance at critical moments, keeping the Agent on track.
| Feature | QevosAgent | OpenClaw | Hermes | OpenCode |
|---|---|---|---|---|
| Desktop App Download | ✅ One-click | ❌ | ❌ | ❌ |
| Windows Native | ✅ Full support | ❌ Requires WSL | ⚠️ Partial | ❌ Linux-first |
| Setup Difficulty | ⭐ Easy | ⭐⭐⭐ Complex | ⭐⭐ Medium | ⭐⭐⭐ Complex |
| Target Users | Everyone, office-focused | Requires some expertise | Requires some expertise | Developers |
QevosAgent's unique positioning:
Auto-analyze codebases, generate documentation, run shell scripts — let the Agent handle the tedious work.
Auto-collect references, run Python analyses, generate fully documented research reports.
Run background shell tasks, batch file processing, workflow automation — no manual supervision needed.
Let AI run Python scripts, process datasets, generate charts — fully auditable and reproducible.
Q&A, paper assistance, resource organization — AI accompanies your learning and remembers your knowledge base.
Connect local models, data never leaves the network — low-cost private AI Agent for your team.
Make sure Python 3.10 or higher is installed:
python --version
git clone https://github.com/HongyunQiu/QevosAgent.git
cd QevosAgent
pip install -r requirements.txt
copy .env.example .env
Edit the .env file and set your API key (or use a local model for zero cost).
cd dashboard
npm install
cd ..
node dashboard/server.js
Open http://localhost:8765 in your browser to monitor task execution in real time.
The dashboard is included in the repository — launch tasks, stop tasks, inject commands, inspect history, and browse run artifacts.
This project is licensed under the MIT License — free and open source forever.
Thanks to all users and contributors!
If QevosAgent has been helpful to you, please give it a ⭐ Star!
QevosAgent — Local First · Private · Ready Out of the Box
🦊 你的本地 AI 智能体,开箱即用
一个真正为所有人设计的本地 AI 智能体——支持 Windows/macOS/Linux 原生安装,无需 WSL,开箱即用。
QevosAgent 的核心设计哲学是:尽可能简化运行流程,将复杂性从系统逻辑中剥离出来,并沉淀到可扩展的工具体系中。
它不依赖复杂、封闭的运行机制,而是将各种能力定义为工具。工具可以持续发展、不断进化,因此 QevosAgent 的能力边界并不是固定的,而是可以随着工具生态的扩展而不断延伸,具备广阔的成长空间。
同时,QevosAgent 倡导运行过程的可观测化与透明化。通过可视化看板,用户可以实时观察 Agent 的执行过程、任务状态和关键信息,让复杂任务的推进过程清晰可见、一目了然。
与其他 AI Agent 不同,QevosAgent 从第一天起就为桌面用户设计:
QevosAgent 不是只给程序员用的工具:
| 平台 | 下载 |
|---|---|
| 🪟 Windows | 下载 Windows 安装器 |
| 🍎 macOS Apple Silicon | 下载 macOS ARM 版 |
| 🍎 macOS Intel | 下载 macOS Intel 版 |
| 🐧 Linux | 下载 Linux AppImage |
git clone https://github.com/HongyunQiu/QevosAgent.git
cd QevosAgent
pip install -r requirements.txt
copy .env.example .env
# 编辑 .env 填入 API Key
# 使用 Dashboard 看板
cd dashboard
npm install
cd ..
node dashboard/server.js
任何能力都可以注册为工具,通过工具进化可以大幅度扩展 Agent 能力,具有无限可能性。
每个任务自动保存到磁盘——日志、草稿本、摘要和最终答案,随时可审计。
记忆跨会话持久化。Agent 从上次中断处继续——用得越多,越聪明。
工具失败时,Agent 自动诊断并修复——无无限循环,无卡死。
从浏览器启动任务、注入命令、浏览历史记录——完全可视化。
深度支持 Qwen3、Gemma4 及任何 OpenAI 兼容端点。零 API 成本,数据留在本机。
文件读写、Python 执行、Shell 命令、网络搜索、记忆管理——开箱即用。
独立的高级指导员 LLM 在关键时刻提供战略指导,确保 Agent 不偏航。
| 特性 | QevosAgent | OpenClaw | Hermes | OpenCode |
|---|---|---|---|---|
| 桌面程序下载 | ✅ 一键安装 | ❌ | ❌ | ❌ |
| Windows 原生 | ✅ 完美支持 | ❌ 需要 WSL | ⚠️ 部分支持 | ❌ Linux 优先 |
| 安装难度 | ⭐ 简单 | ⭐⭐⭐ 复杂 | ⭐⭐ 中等 | ⭐⭐⭐ 复杂 |
| 目标用户 | 所有人,面向办公场景 | 需要一定基础 | 需要一定基础 | 程序员 |
QevosAgent 的独特定位:
自动分析代码库、生成文档、运行 Shell 脚本——让 Agent 处理繁琐工作。
自动收集参考文献、运行 Python 分析、生成完整记录的研究报告。
运行后台 Shell 任务、批量文件处理、工作流自动化——无需人工监督。
让 AI 运行 Python 脚本、处理数据集、生成图表——完全可审计、可复现。
问答、论文辅助、资源整理——AI 陪伴学习并记住你的知识库。
连接本地模型,数据永不离开网络——为团队提供低成本私有 AI Agent。
确保已安装 Python 3.10 或更高版本:
python --version
git clone https://github.com/HongyunQiu/QevosAgent.git
cd QevosAgent
pip install -r requirements.txt
copy .env.example .env
编辑 .env 文件,设置你的 API 密钥(或使用本地模型实现零成本)。
cd dashboard
npm install
cd ..
node dashboard/server.js
在浏览器打开 http://localhost:8765,即可实时监控任务执行。
Dashboard 已包含在仓库中,可以启动任务、停止任务、注入命令、检查历史、浏览运行产物。
本项目采用 MIT 许可证 — 永久免费开源。
感谢所有使用者和贡献者!
如果 QevosAgent 对你有帮助,请给一个 ⭐ Star!
QevosAgent — 本地优先 · 数据私有 · 开箱即用
让 AI 成为你的智能助手,而不仅仅是开发工具。