《a-practical-guide-to-building-agents》导读

Open AI 发布的《a-practical-guide-to-building-agents.》旨在帮助开发者理解和构建基于人工智能的智能体（Agents）。智能体是一个能够感知其环境并根据感知到的信息采取行动以实现特定目标的实体。在 AI 领域，智能体通常指的是能够自主执行任务、做出决策并与环境交互的软件或系统。

LLM Agent

• Situation
  • Complex decision-making
  • Difficult to maintain rules
  • Heavy reliance on unstructured data
• Foundations
  • Model
    • Set up evals to establish a performance baseline
    • Focus on meeting your accuracy target with the best models available
    • Optimize for cost and latency by replacing larger models with smaller ones where possible
  • Tools
    • Data
    • Action
    • Orchestration
  • Instructions
    • Use existing documents
    • Prompt agents to break down tasks
    • Define clear actions
• Orchestration
  • Single-agent systems
  • Multi-agent systems
    • Scenario
      • Complex logic
      • Tool overload
    • Types
      • Manager (agents as tools)
      • Decentralized (agents handing off to agents)
• Guardrails
  • Types
    • Relevance classifier
    • Safety classifier
    • PII filter
    • Moderation
    • Tool safeguards
    • Rules-based protections
    • Output validation
  • Build Guardrails
    • Focus on data privacy and content safety
    • Add new guardrails based on real-world edge cases and failures you encounter
    • Optimize for both security and user experience, tweaking your guardrails as youragent evolves.