README.md

metadata

title: OpenManus2
emoji: 🔥
colorFrom: yellow
colorTo: red
sdk: docker
pinned: false

README.md

🧠 OpenManus2 – Agentic AI FastAPI Interface

This private Hugging Face Space deploys the OpenManus AI agent using a FastAPI backend and template-based UI. Designed for long-running, streaming LLM tasks via external endpoints.

---

🚀 Features

• ✅ FastAPI backend with SSE streaming
• 🧩 Jinja2 templates + custom UI themes
• 🔁 Asynchronous task queue with status updates
• 🌐 Integrates with external LLM tools via OPENMANUS_ENDPOINT_URL

---

🛠️ Run Locally

# Clone the repo
git clone https://huggingface.co/spaces/kstang88/OpenManus2
cd OpenManus2

# Install dependencies
pip install -r requirements.txt

# Set the backend endpoint
export OPENMANUS_ENDPOINT_URL=https://your-llm-backend.com/api

# Run the app
uvicorn app:app --host 0.0.0.0 --port 7860

---

🧱 Project Structure

OpenManus2/
├── app.py                # FastAPI app backend
├── config/
│   └── config.toml       # Optional: LLM endpoint config
├── requirements.txt      # Python dependencies
├── Dockerfile            # Hugging Face Docker Space
├── static/               # CSS/JS + theme assets
└── templates/            # HTML pages (Jinja2)

---

⚙️ Environment Variables

Name	Description
OPENMANUS_ENDPOINT_URL	External LLM API endpoint (required)

Set via .env file locally or Space Secrets (UI → Settings → Secrets).

---

📡 API Endpoints

Method	Path	Description
GET	/	Theme selector page
GET	/chat?theme=openmanus	Chat UI with custom theme
POST	/tasks	Create a new task
GET	/tasks	List all tasks
GET	/tasks/{task_id}	View task status
GET	/tasks/{task_id}/events	SSE task stream updates
GET	/download?file_path=	Download generated files

---

🧪 Test Prompt

Send a prompt to your running agent:

curl -X POST http://localhost:7860/tasks \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Summarize GPT-4 architecture."}'

Then stream results via:

curl http://localhost:7860/tasks/<task_id>/events

---

🛡 License

MIT License © 2025 kstang88

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference