AI Skill Hub 推荐使用:MingJian Agent工作流 是一款优质的Agent工作流。AI 综合评分 7.5 分,在同类工具中表现稳健。如果你正在寻找可靠的Agent工作流解决方案,这是一个值得深入了解的选择。
MingJian Agent工作流 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。
MingJian Agent工作流 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。
# 方式一:pip 安装(推荐)
pip install mingjian
# 方式二:虚拟环境安装(推荐生产环境)
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install mingjian
# 方式三:从源码安装(获取最新功能)
git clone https://github.com/dashitongzhi/MingJian
cd MingJian
pip install -e .
# 验证安装
python -c "import mingjian; print('安装成功')"
# 命令行使用
mingjian --help
# 基本用法
mingjian input_file -o output_file
# Python 代码中调用
import mingjian
# 示例
result = mingjian.process("input")
print(result)
# mingjian 配置文件示例(config.yml) app: name: "mingjian" debug: false log_level: "INFO" # 运行时指定配置文件 mingjian --config config.yml # 或通过环境变量配置 export MINGJIAN_API_KEY="your-key" export MINGJIAN_OUTPUT_DIR="./output"
<br />
git checkout -b feature/amazing-feature
pip install -e ".[dev]"
cd frontend-v2 npm install cd ..
| Package | Version | Purpose |
|---|---|---|
| **FastAPI** | 0.110+ | High-performance async API framework |
| **SQLAlchemy** | 2.0+ | Database ORM |
| **Alembic** | 1.16+ | Database migrations |
| **Pydantic** | 2.11+ | Data validation |
| **OpenAI** | 2.28+ | OpenAI API client |
| **Anthropic** | 0.52+ | Anthropic API client |
| **Redis** | 6.2+ | Event bus and caching |
| **pgvector** | 0.3+ | Vector similarity search |
| **MinIO** | 7.2+ | Object storage |
| **HTTPX** | 0.28+ | Async HTTP client |
| **Uvicorn** | 0.35+ | ASGI server |
| Package | Version | Purpose |
|---|---|---|
| **Vite** | 8+ | Frontend build tool |
| **React** | 19+ | UI library |
| **TypeScript** | 6.0+ | Type safety |
| **Tailwind CSS** | 4.2+ | Utility-first CSS |
| **React Router** | 7+ | Client-side routing |
| **Recharts** | 3.8+ | Charting library |
| Package | Version | Purpose |
|---|---|---|
| **pytest** | 8.4+ | Testing framework |
| **pytest-asyncio** | 1.1+ | Async test support |
| **Ruff** | 0.12+ | Python linter |
| **ESLint** | 9+ | JavaScript linter |
| **Prettier** | 3+ | Code formatter |
| **vitest** | ^4.1.5 | Unit testing framework |
| **@testing-library/react** | ^16.x | React component testing |
| **@testing-library/jest-dom** | ^6.x | Custom Jest matchers |
---
python tests/stress_test.py ```
Latest Results: - ✅ Backend: 92 unit tests passing (0.26s) - ✅ Frontend: 16 component tests passing (0.55s) - ✅ Stress Test: 112 pass, 0 fail, 2 warnings - 🔥 Concurrent: 20 users, 844 RPS, P50=1ms, zero 500 errors
---
The fastest way to run 明鉴 locally is the Docker setup script. It checks Docker, creates .env from .env.example, asks for your OpenAI API key, and starts the full stack.
Install Docker Desktop first, then run:
chmod +x setup.sh
./setup.sh
When the script finishes, open:
| Service | URL |
|---|---|
| Frontend | http://localhost:3001 |
| API | http://localhost:8000 |
| MinIO Console | http://localhost:9001 |
MinIO login: use PLANAGENT_MINIO_ACCESS_KEY / PLANAGENT_MINIO_SECRET_KEY from your .env.
To stop the Docker stack:
docker compose -f docker-compose.yml down
Use this path if you want to run the backend and frontend directly on your machine for development.
frontend-v2/ is the active Vite frontend. The older frontend/ and frontend-new/ experiments have been retired, so new product UI, builds, and Docker development should target frontend-v2/.
Before you begin, ensure you have the following installed:
| Requirement | Version | Installation |
|---|---|---|
| **Python** | 3.12+ | [python.org](https://www.python.org/downloads/) |
| **Node.js** | 18+ | [nodejs.org](https://nodejs.org/) |
| **npm** | 9+ | Comes with Node.js |
| **Git** | 2.30+ | [git-scm.com](https://git-scm.com/) |
| **PostgreSQL** | 14+ (optional) | [postgresql.org](https://www.postgresql.org/download/) |
| **Redis** | 7+ (optional) | [redis.io](https://redis.io/download) |
| Component | Minimum | Recommended |
|---|---|---|
| **CPU** | 2 cores | 4+ cores |
| **RAM** | 4 GB | 8+ GB |
| **Storage** | 10 GB | 50+ GB |
| **OS** | macOS, Linux, Windows | macOS or Linux |
Create a .env file in the project root with the following variables:
```bash
| Use Case | Description | Benefit |
|---|---|---|
| **📊 Investment Research** | Analyze market trends, debate investment theses | Faster research, better decisions |
| **🏭 Corporate Strategy** | Competitive intelligence, scenario planning | Data-driven decisions, reduced risk |
| **⚔️ Military Planning** | Operational analysis, logistics optimization | Strategic advantage, better outcomes |
| **🛡️ Risk Management** | Multi-perspective risk assessment | Reduced uncertainty |
| **📈 Market Analysis** | Real-time market intelligence | Faster insights, better positioning |
| **🎯 Policy Analysis** | Multi-stakeholder impact assessment | Informed policy, better outcomes |
---
python -m venv .venv source .venv/bin/activate # On Windows: .venv\Scripts\activate
cp .env.example .env
Question
-> source discovery
-> evidence extraction
-> analysis and simulation
-> role-based debate
-> recommendation
-> version timeline
-> scheduled / source-change refresh
-> decision record and outcome feedback
| Stage | Purpose | User-Visible Proof |
|---|---|---|
| Gather | Collect multi-source evidence from news, code, social, and public data feeds. | Source list, cursor health, last checked time. |
| Reason | Convert evidence into arguments, risks, simulations, and structured claims. | Analysis artifacts, report sections, debate rounds. |
| Debate | Let role-based agents critique, revise, and arbitrate the decision. | Support/challenge/arbiter traces. |
| Remember | Persist sessions, recommendation versions, refresh triggers, and outcomes. | Timeline, recommendation history, feedback records. |
---
python -m pytest tests/ -v ```
| Feature | 明鉴 | Manus | Traditional AI | Single-Agent | LangChain |
|---|---|---|---|---|---|
| **Data Sources** | ✅ 10+ real-time | ⚠️ General search | ❌ Manual input | ⚠️ Limited | ⚠️ Limited |
| **Evidence Chain** | ✅ Full traceability | ❌ No tracking | ❌ No tracking | ❌ No tracking | ❌ No tracking |
| **Multi-Agent Debate** | ✅ 9-agent adversarial | ⚠️ Orchestrator + sub-agents | ❌ Single model | ❌ Single model | ⚠️ Basic |
| **Decision Traces** | ✅ Deterministic | ❌ Black box | ❌ Black box | ❌ Black box | ❌ Black box |
| **Self-Repair** | ✅ Jarvis engine | ⚠️ Dynamic re-planning | ❌ None | ❌ None | ❌ None |
| **Streaming Analysis** | ✅ Real-time | ✅ Real-time | ❌ Batch only | ❌ Batch only | ⚠️ Limited |
| **Continuous Monitoring** | ✅ WatchRule + auto-update | ❌ One-shot tasks | ❌ None | ❌ None | ❌ None |
| **Corporate Domain** | ✅ Full support | ❌ Generic | ⚠️ Generic | ❌ Generic | ❌ Generic |
| **Military Domain** | ✅ Full support | ❌ Generic | ⚠️ Generic | ❌ Generic | ❌ Generic |
| **Scenario Branching** | ✅ Beam-search | ❌ None | ❌ Manual | ❌ None | ❌ None |
| **Knowledge Graph** | ✅ Embedding-backed | ❌ None | ❌ None | ❌ None | ❌ None |
| **Code Execution** | ⚠️ Planned | ✅ Full sandbox VM | ❌ None | ⚠️ Limited | ❌ None |
| **Data Sovereignty** | ✅ Self-hosted | ❌ Cloud only | ⚠️ Varies | ⚠️ Varies | ✅ Self-hosted |
| **Open Source** | ✅ Apache 2.0 | ❌ Closed source | ⚠️ Varies | ⚠️ Varies | ✅ Various |
---
MingJian(明鉴)是一款专为深度分析设计的智能决策辅助工具。本项目旨在通过先进的 AI 技术,为用户提供高效的信息处理与逻辑推理能力,帮助用户在复杂的信息环境中快速洞察本质,实现从数据到决策的智能化跨越。
MingJian 提供强大的核心功能,支持多维度的信息分析与逻辑推演。无论是进行投资研究、企业战略规划,还是复杂的军事行动分析,��统都能通过集成 AI 能力,协助用户进行市场趋势分析、竞争情报收集及场景模拟,从而提升决策的科学性与战略优势。
在开始开发或部署之前,请确保您的系统已安装必要的开发环境。本项目依赖 Python 环境进行后端逻辑处理,并需要 Node.js 环境来驱动前端工程。在进行开发模式安装时,请使用 pip 安装包含 dev 依赖的 Python 包,并进入 frontend 目录执行 npm install 完成前端依赖的构建。
本项目提供两种安装方式:推荐使用 Docker 一键部署,通过运行 setup.sh 脚本,系统会自动检查 Docker 环境、配置 .env 文件并引导您输入 OpenAI API key,实现全栈快速启动。若需进行本地开发,请手动配置后端与前端环境,并注意 frontend-v2/ 目录为当前活跃的 Vite 前端开发目录。
MingJian 适用于多种高阶应用场景。在投资研究领域,它可以辅助分析市场趋势并进行投资论证;在企业战略层面,可用于竞争情报分析与风险规避;在军事规划中,则能通过作战分析与物流优化提供战略优势。通过将复杂数据转化为结构化洞察,助力用户做出更精准的决策。
项目配置主要通过环境变量进行管理。您可以根据需求配置 AI Model 相关的参数,并支持可选的 Database(本地开发默认为 SQLite)与 Redis(用于生产环境的 event bus)。请务必在 .env 文件中正确配置相关的 API key,以确保 AI 服务能够正常调用。
MingJian 的接入非常便捷,您仅需准备一个 API key 即可开启使用。在启动服务前,请先在配置文件中设置好 PLANAGENT_OPENAI_PRIMARY_API_KEY,并运行 provider 连接测试,以确保 API 通道畅通无阻,从而实现与 AI 模型的稳定交互。
项目的开发与测试流程遵循标准化的工程实践。在完成功能开发后,建议运行集成测试命令 `python -m pytest tests/ -v` 来确保各模块逻辑的正确性。通过严谨的测试工作流,可以保障从后端逻辑到前端交互的整体稳定性。
该项目提供了一个开源的AI工作流平台,支持多智能体和证据驱动的场景模拟,具有较高的开发性和扩展性,但仍需要进一步的优化和完善。
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。
✅ Apache 2.0 — 宽松开源协议,可商用,需保留版权声明和 NOTICE 文件,含专利授权条款。
总体来看,MingJian Agent工作流 是一款质量良好的Agent工作流,在同类工具中具备一定竞争力。AI Skill Hub 将持续追踪其更新动态,建议收藏备用,结合自身场景选择合适时机引入使用。
| 原始名称 | MingJian |
| 原始描述 | 开源AI工作流:AI-powered multi-agent platform for evidence-driven scenario simulation and stra。⭐19 · Python |
| Topics | workflowagentaicorporatedecision-makingevidence-basedpython |
| GitHub | https://github.com/dashitongzhi/MingJian |
| License | Apache-2.0 |
| 语言 | Python |
收录时间:2026-05-17 · 更新时间:2026-05-19 · License:Apache-2.0 · AI Skill Hub 不对第三方内容的准确性作法律背书。
选择 Agent 类型,复制安装指令后粘贴到对应客户端