本目录包含 5 个自包含 Python 示例，不依赖 experiments/shared/。用于快速理解 Agent 循环、工具校验、流式拼接、简易记忆与多 Agent 协作。
| 文件 | 内容 |
|---|---|
01_mini_agent.py |
最小 while 循环 + 计算器工具；无 API key 时使用 mock。 |
02_tool_use.py |
用 dataclasses 做输入校验、注册表与调度。 |
03_streaming.py |
模拟异步 SSE 事件流，拼接文本与 tool_use JSON。 |
04_memory.py |
文件型记忆 + 简易 TF-IDF 召回。 |
05_multi_agent.py |
asyncio 队列：Leader 分派任务，两名 Worker 并行处理。 |
cd learn_claude_code/examples
python3 01_mini_agent.py
python3 02_tool_use.py
python3 03_streaming.py
python3 04_memory.py
python3 05_multi_agent.py可选：若已 pip install anthropic 并配置 ANTHROPIC_API_KEY，01_mini_agent.py 会调用真实 API（否则自动 mock）。
Five standalone Python scripts with no imports from experiments/shared/. They demonstrate the agent loop, tool validation, streaming reassembly, file-backed memory with TF‑IDF recall, and a tiny leader/worker multi-agent demo.
| File | What it shows |
|---|---|
01_mini_agent.py |
Minimal loop + calculator tool; mock LLM without an API key. |
02_tool_use.py |
Dataclass validation, registry, and dispatch. |
03_streaming.py |
Mock async SSE stream; merge text and tool JSON fragments. |
04_memory.py |
Topic files plus a simple TF‑IDF index for recall. |
05_multi_agent.py |
asyncio queues: leader assigns work to two workers. |
cd learn_claude_code/examples
python3 01_mini_agent.py
python3 02_tool_use.py
python3 03_streaming.py
python3 04_memory.py
python3 05_multi_agent.pyOptional: with pip install anthropic and ANTHROPIC_API_KEY, 01_mini_agent.py calls the real API; otherwise it stays on the mock path.