CodeEdu Crew is a multi-agent Coding Education system built with CrewAI, exposed via a Flask backend with a streaming-capable API. It allows AI agents to collaborate on complex tasks such as Python code execution, report writing, research, and education.
🧠 What It Can Do 🧑🏫 Explain complex tech topics with code examples
👨💻 Analyze, execute, and improve uploaded Python files
📊 Generate summaries and structured reports
🤝 Autonomously assign agents and tasks via planner
📂 Stream logs, thoughts, and outputs to your frontend
src/
└── codeedu/
├── main.py # Entry Flask App
├── config/
│ ├── agents.yaml
│ ├── tasks.yaml
│ └── paths.py
├── core/
│ ├── executor.py
│ ├── stream.py
│ ├── registry.py
│ └── output.py
├── routes/
│ ├── chat.py
│ ├── upload.py
│ └── conversation.py
├── tools/
│ ├── custom_tool.py
├── utils/
│ ├── session.py
│ ├── intention.py
│ └── log_queues.py
├── agent_pool.py
├── task.py
├── conversations/ # Will be created at runtime
├── output/ # Will be created at runtime
Python >= 3.10 and < 3.13
Docker (for running .py file or container deployment)
git clone https://github.com/your-org/codeedu.git
cd codeedu/src/codeedu
pip install -r requirements.txtpython main.pyBy default, it will start on http://localhost:5000.
./build_and_run.shThen visit: http://localhost:5000
docker build -t codeedu-backend .
docker run -d -p 5000:5000 \
-e PROJECT_ROOT="/app/src/codeedu" \
--name codeedu-app \
codeedu-backendCreate a .env file or export these before running:
MODEL=your-llm-model e.g. openai/gpt-4o-mini
OPENAI_API_KEY=your-api-key
OPENROUTER_API_KEY=your-api-key (optional)
BASE_URL=https://api.openai.com/v1 (optional)
SERPER_API_KEY=optional
PROJECT_ROOT=/home/user/.../codeedu/src/codeedu
STORAGE_PATH=${PROJECT_ROOT}/conversations (conversation history)
OUTPUT_PATH=${PROJECT_ROOT}/output (output file)
OTHER_MODEL=other-tool-model (optional)| Endpoint | Method | Description |
|---|---|---|
/chat |
POST | Multi-agent reasoning for custom input |
/execute_code_snippet |
POST | Upload a Python file and get agent analysis |
/upload_code |
POST | Inject a code snippet into session context |
/output/<filename> |
GET | Download files produced by agents |
/new_conversation |
POST | Start a new session |
/conversations |
GET | List all saved conversations |
/conversation/<cid> |
GET | Get full history of a specific conversation |
/delete_conversation/<cid> |
DELETE | Delete a conversation and its history |
educator:
role: "AI Educator"
goal: "Explain technical concepts clearly with code examples"
backstory: "You are a brilliant teacher known for simplifying complex ideas."| Feature | Status |
|---|---|
| ✅ MAS coordination | Completed ✅ |
| ✅ Flask API with streaming | Completed ✅ |
| ✅ File upload + code execution | Completed ✅ |
| ✅ Agent auto-assignment | Completed ✅ |
| ✅ LLM summarization of thought | Completed ✅ |
| ✅Frontend log visualization | Completed ✅ |
| 🔄 LangGraph integration | In Progress |
All generated reports, fixed code files, and result documents are saved in the /output directory and can be downloaded via:
GET /output/<filename>
| Name | Role | Description |
|---|---|---|
| Jianing ZHAO | Backend Architect & AI Agent Designer | Led the entire backend system architecture and implementation. Designed and built all AI agents, task flows, session management, streaming, and API layers. |
| ACCFOOL | Frontend Developer | Responsible for designing and implementing the frontend interface, enabling seamless interaction between users and the multi-agent system. |
| YIN Jianing | Frontend Developer | Responsible for designing and implementing the frontend interface, enabling seamless interaction between users and the multi-agent system. |
| Try1234567 | Co-Designer | Contributed in the early stage of the project, helping design agent roles and task planning. Provided conceptual input to the overall architecture direction. |
Jianing ZHAO 💻 🎨 📆 |
ACCFOOL 💻 🎨 |
YIN Jianing 💻 🎨 |
Try1234567 🎨 💻 |