strix/strix_hub
2026-08-27 00:00:53 +08:00
..
__init__.py feat(hub): add Strix Hub multi-tenant task orchestration & dual-channel model routing console 2026-08-26 14:59:35 +08:00
db.py fix(hub): resolve command injection, run isolation, credentials logging, and CLI argv 2026-08-26 15:07:20 +08:00
main.py fix(hub): resolve command injection, run isolation, credentials logging, and CLI argv 2026-08-26 15:07:20 +08:00
model_router.py revert(router): preserve 100% untouched strix message payloads 2026-08-26 23:32:30 +08:00
README.md feat(hub): add Strix Hub multi-tenant task orchestration & dual-channel model routing console 2026-08-26 14:59:35 +08:00
server.py feat(server): auto-load .env and resolve channels with robust fallback 2026-08-27 00:00:53 +08:00
task_manager.py feat(server): auto-load .env and resolve channels with robust fallback 2026-08-27 00:00:53 +08:00

🦅 Strix Hub (Multi-Tenant & Dual-Channel Orchestration Console)

English | 简体中文


English

Strix Hub is an open-source, multi-tenant task orchestration and dual-channel model routing platform designed for Strix autonomous AI penetration testing.

✨ Key Features

  1. Zero-Intrusion Facade Architecture:
    • Operates completely outside the core strix/ codebase as a management facade.
    • 100% compatible with current and future upstream Strix versions without code merge conflicts.
  2. Independent Dual-Channel Model Routing:
    • Root Agent (Brain): High-reasoning cloud models (e.g. Gemini 3.1 Pro, Claude 3.7 Sonnet, GPT-4o) for strategic planning and vulnerability discovery.
    • Sub-agents (Muscles): Private/local models (e.g. Qwen 3.8 / DeepSeek / Ollama / vLLM / SGLang) for high-concurrency port scanning, fuzzing, and payload execution with zero token costs and zero rate limits.
    • Built-in Auto Tool-Call Recovery to seamlessly convert text-based XML tool outputs into standard OpenAI function calls.
  3. Live Hot-Reloading:
    • Switch model providers, base URLs, and API keys on-the-fly without interrupting running penetration tasks.
  4. OS-Level Process Control:
    • True process group lifecycle management supporting Start, Pause (SIGSTOP with 0 CPU & 0 Token consumption), Resume (SIGCONT), and Terminate.
  5. Multi-Tenancy & Role-Based Access Control (RBAC):
    • Built-in SQLite database with user isolation (Users manage their own scans; Admins monitor the entire fleet).
  6. Zero External Dependencies:
    • Powered purely by Python standard libraries (http.server, sqlite3, subprocess, threading) and a single-file modern Dark Mode React SPA.

🚀 Quick Start

# 1. Install & build Strix
git clone https://github.com/usestrix/strix.git
cd strix
make dev-install

# 2. Launch Strix Hub (default port 8888)
python -m strix_hub.main --port 8888

Open http://localhost:8888 in your browser.

  • Default Admin Account: admin / admin123

⚙️ Environment Variables (Optional)

export LOCAL_LLM_MODEL="openai/Qwen3.8-27B-abliterated"
export LOCAL_LLM_URL="http://127.0.0.1:8000/v1"
export LOCAL_LLM_KEY="your-api-key"

简体中文

Strix Hub 是专为 Strix 打造的现代化、零侵入、多租户渗透测试任务编排与双渠道模型网关管理平台。

🌟 核心特性

  1. 零侵入门面架构 (Zero-Intrusion Facade):
    • 100% 独立于官方 strix/ 核心代码运行,完全解耦。
    • 官方上游仓库执行 git pull 升级时零代码合并冲突,天然适配当前与未来所有 Strix 版本。
  2. 独立双渠道模型路由网关 (Dual-Channel Routing):
    • 主控大脑 (Root Agent):可自由配置云端顶尖推理模型(如 Gemini 3.1 Pro、Claude 3.7 Sonnet、GPT-4o),负责全局渗透决策与漏洞挖掘。
    • 探测打手 (Sub-agents):可无缝对接局域网私有化模型(如本地部署的 Qwen 3.8-27B 无审查特化版 / DeepSeek / SGLang / vLLM),负责高并发端口扫描与 Web 模糊测试,零 Token 成本、零外网限流!
    • 内置 Tool-Call 自动解析自愈器,自动将开源模型输出的 XML 格式工具调用转为标准 OpenAI 协议,确保子智能体触手稳定并发派生。
  3. 运行时配置热更新 (Live Hot-Reload):
    • 任务在运行中或暂停中,均可在 Web 控制台一键热更模型渠道、Base URL 与 API Key,底层网关实时生效。
  4. 操作系统进程级精准调度:
    • 采用进程组信号控制,真正实现任务 一键暂停 (SIGSTOP 瞬间零 CPU、零 Token 消耗挂起)、继续 (SIGCONT 唤醒) 与 强力终止。
  5. 多租户与 RBAC 权限隔离:
    • 内置轻量 SQLite 持久化,普通用户仅能查看与操作自己创建的任务,管理员全局统一纳管。
  6. 零额外依赖 & 极速部署:
    • 后端基于 Python 标准库实现,前端内置现代化 Single-File React SPA 暗黑大屏,一键即可启动。

🚀 快速启动

# 1. 克隆并安装 Strix
git clone https://github.com/usestrix/strix.git
cd strix
make dev-install

# 2. 启动 Strix Hub 服务 (默认端口 8888)
python -m strix_hub.main --port 8888

浏览器访问 http://localhost:8888 即可开始使用。

  • 默认管理员账户:admin / admin123(登录后可在用户管理中修改或新增成员)

📄 License

Apache License 2.0