HauUhang/Agents

# Technical Summary Provides a comprehensive Chinese translation of Google's 2025 AI Agent technical whitepaper, covering agent architecture decomposed into three core components: reasoning model, actionable tools, and orchestration layer managing the agentic loop. The framework describes agents as context window management systems that iteratively execute a five-step cycle—acquire task, scan environment, reason deeply, execute actions via tools/APIs, and observe results—enabling autonomous problem-solving beyond static language model capabilities. Hierarchical classification spans from basic reasoning systems to self-evolving multi-agent organizations, with production considerations for model selection (balancing quality, latency, and cost), tool integration patterns including RAG and external APIs, and governance strategies for enterprise-scale deployment.