🚀 ProjectAI: Immersive Project & Projection Sandbox

ProjectAI is a project projection system based on LangGraph and Large Language Models (LLMs). By simulating the collaboration between a "Project Manager" and "Domain Experts," it helps users fully project the entire implementation process of an Idea from a conceptual level before hands-on practice.

Note on the name: "ProjectAI" plays on the double meaning of "Project"—as a noun (the project itself) and as a verb (to project/predict future outcomes).

✨ Core Features

• Background Awareness: The system dynamically adjusts projection logic based on the team background you input (e.g., high school students, startup company, hobbyists).
• Append-only Progress: Each expert performs projections and appends contributions based on existing progress, ensuring everyone sees each other's progress without overwriting.
• Dynamic Expert Expansion: The Project Manager "summons" the most suitable domain experts in real-time based on task requirements.
• Prompt Decoupling: All core system prompts have been extracted into external Markdown files for easy personalization and fine-tuning.
• Final Achievement Report: Upon project completion, the system automatically summarizes the required talent portraits (real expert names and their prompts) and outputs a final project archive.

🛠️ Technical Architecture

• Core Engine: LangGraph (State-driven workflow)
• Large Model: Any compatible Chat model (Configurable via secrets.py)
• Data Structures: Pydantic & TypedDict
• Prompt Management: Dynamic loading of Markdown templates

📂 Project Structure

ProjectAI/
├── main.py            # Entry point and workflow orchestration
├── agents/            # Agent logic
│   ├── ProjectManager.py  # Project Manager Agent
│   ├── ExpertAgent.py     # Domain Expert Agent
│   └── ...
├── prompts/           # Prompt library (Markdown)
│   ├── manager_system_prompt.md
│   └── expert_system_prompt.md
├── secrets.py         # Sensitive information & configuration (not in Git)
└── .gitignore         # Git ignore rules

🚀 Quick Start

1. Environment Preparation: Ensure langchain-openai, langgraph, and pydantic are installed.

2. Configure Provider & Keys: Create or edit the secrets.py file in the root directory and fill in your details:

   # API Keys
   SILICONFLOW_API_KEY = "YOUR_API_KEY"
   
   # Model Provider Configuration
   MODEL_BASE_URL = "https://api.siliconflow.cn/v1"
   MODEL_NAME = "deepseek-ai/DeepSeek-R1" # Or any other model

3. Run Projection (CLI):

   python main.py

4. Launch Professional Web Dashboard:

   • Backend:

     python server.py

   • Frontend (React):

     cd frontend
     npm install
     npm run dev

   • Open http://localhost:5173 in your browser.

🖥️ Web Dashboard Features

• Modern UI: Built with React, Tailwind CSS, and Framer Motion for smooth animations.
• Command Center Design: A professional, dark-themed dashboard inspired by control rooms.
• Real-time Timeline: Watch the Project Manager and Experts collaborate in a vertical feed.
• Internal Thought View: See the action_log for each expert in a monospace terminal block.
• Dynamic Talent Tracker: Monitor the expert pool as it expands in the sidebar.
• Closure Achievement: A grand "Projection Closure Report" with a trophy theme when the goal is reached.

4. Input Guidance:
   • Idea: e.g., "Build a water quality monitor for a national competition"
   • Background: e.g., "Three high school students with basic physics and a 3D printer"

📝 Customizing Prompts

You can change the AI's personality, projection depth, or behavioral norms by editing the .md files in the prompts/ directory. The system will automatically load these updates every time it runs.

---

Let every Idea experience a brilliant baptism in ProjectAI before it lands in reality.

⚖️ License

This project is licensed under the MIT License - see the LICENSE file for details.