AgentWeaver – Vertical AI Agent Framework

AgentWeaver is a modular multi-agent framework tailored for domain-specific AI automation. It supports orchestrating autonomous LLM-driven agents across industry verticals such as Healthcare, Legal, and Finance.

🚀 Features

• 🤖 Multi-agent system with roles: Planner, Retriever, Executor, Evaluator
• 🧠 Integration with LLMs (OpenAI, Claude, HuggingFace)
• 🔌 Plug-and-play support for domain tools and APIs
• 📊 Streamlit-based observability dashboard
• 🗃️ Vector database integration (Chroma, Weaviate)

🏗️ Project Structure

agentweaver-vertical-ai/
├── agents/                  # Core autonomous agent modules
├── verticals/               # Domain-specific logic (healthcare, legal, finance)
├── tools/                   # Generic tools for agents (search, APIs)
├── vector_store/            # Embedding + vector DB integration
├── orchestration/           # Messaging and agent communication layer
├── dashboard/               # Streamlit interface for real-time agent monitoring
├── tests/                   # Unit/integration tests
├── requirements.txt         # Python dependencies
└── .env.example             # Environment variables template

📦 Installation

git clone https://github.com/yourusername/agentweaver-vertical-ai.git
cd agentweaver-vertical-ai
python -m venv venv
source venv/bin/activate  # or venv\Scripts\activate on Windows
pip install -r requirements.txt

⚙️ Configuration

Create a .env file based on the .env.example:

OPENAI_API_KEY=your_key
VECTOR_DB_PATH=./vector_store/index

🧪 Running the Dashboard

streamlit run dashboard/streamlit_app.py

📚 Example Use Case: Healthcare EHR Assistant

• Planner Agent receives a goal: "Summarize and flag unusual patient risks"
• Retriever queries vector DB for relevant context
• Executor processes EHR file with LLM
• Evaluator ensures findings meet quality thresholds

📌 Roadmap

• Agent memory and state persistence
• Fine-tuned vertical models
• Agent routing and skill tagging
• Docker & deployment templates

🤝 Contributing

PRs are welcome! Please include tests and follow PEP8.

📄 License

MIT License © 2025 AgentWeaver Contributors