文档：https://docs.langchain.com/oss/python/deepagents/overview
示例：https://github.com/langchain-ai/deepagents/tree/main/examples
项目示例分析
Based on the GitHub page, here are the examples ranked by beginner-friendliness:
| Example | Beginner Level | Why |
|---|
text-to-sql-agent
⭐ Best for beginners
Clean, focused use case (NL → SQL); demonstrates planning + skills; uses well-known Chinook database
content-builder-agent
⭐⭐ Good second step
Shows memory (AGENTS.md), skills, subagents — more concepts but well-structured
deep_research
⭐⭐⭐ Intermediate
Multi-step web research with parallel sub-agents — powerful but complex
ralph_mode
⭐⭐⭐ Advanced pattern
Autonomous looping with filesystem persistence — architectural pattern, not starter material
downloading_agents
Utility
Just shows how to download/run agents, not for learning concepts
My recommendation for you
Start with text-to-sql-agent:
- Familiar domain — SQL is something you already know (Java backend background)
- Clear input/output — natural language question → SQL query
- Demonstrates core Deep Agents concepts:
- Planning （TODO列表）
- Skill-based workflows
- Tool use (database)
- Self-contained — uses Chinook demo database, no external API keys needed beyond LLM
After you understand that, move to content-builder-agent to see how memory and subagents work.
