Agent Architectures

### Giai đoạn: Nền Tảng & Khái Niệm Cốt Lõi **Agent Architectures** Agent architectures define how an AI agent is structured: how it reasons, selects tools, manages memory, and decides when to stop. Common patterns include ReAct (reason + act), plan-and-execute, and tree-of-thought. Choosing the right architecture depends on how much control the application needs over model behavior and how tolerant the customer is of unpredictable agent outputs. ### Khái niệm cốt lõi: 1. **Common patterns include ReAct (reason + act), plan-and-execute, and tree-of-thought** ### Tài liệu tham khảo chính thống: - [What is agentic architecture?](https://www.ibm.com/think/topics/agentic-architecture) (article) - [The Complete Guide to AI Agent Architecture](https://medium.com/@amarg3891/the-complete-guide-to-ai-agent-architecture-25dc2cbe7016) (article) - [AI Agents, Clearly Explained](https://www.youtube.com/watch?v=FwOTs4UxQS4) (video)