nanoagent agent framework clean + zero magic + less dependcy
more use builtin function less dependcy keep it code readable and clean more use asyncio async
mouble design as for framework when create a new feature , add use age to example folder
core: Maintain simplicity in your agent's design. Prioritize transparency by explicitly showing the agent’s planning steps. Carefully craft your agent-computer interface (ACI) through thorough tool documentation and testing. tool :
- Choosing the right tools to implement (and not to implement)
- Namespacing tools to define clear boundaries in functionality
- Returning meaningful context from tools back to agents
- Optimizing tool responses for token efficiency
- Prompt-engineering tool descriptions and specs
evaluation : Generate real-world evaluation tasks: Avoid overly simplistic "sandbox" environments. Complex task design: Robust evaluation tasks may require multiple (even dozens of) tool calls. Verifiable outcomes: Each evaluation prompt should correspond to a verifiable answer or result. Specify expected tool calls: You can define expected tool sequences to measure whether the agent truly understands tool utility. Programmatic execution: It is recommended to run evaluations programmatically by directly calling LLM APIs. Comprehensive metrics collection: Collect various metrics including execution time, number of tool calls, token consumption, and tool errors. Result analysis: Analyze the results to identify specific areas where the agent encounters difficulties or confusion.
optinal: Building an evaluation measure the performance of your tools,You can automatically optimize your tools against this evaluation. rule : TDD release dependcy every change git commit | git push must run ruff check --fix && ruff format
测试框架：tests/agent/
# unit (mock): uv run pytest tests/agent/ -m unit -v
# integration (real): uv run pytest tests/agent/ -m integration -v新工具测试：
tools/<tool>.py→ 注册 →core/agent.py示例 →tests/agent/test_<tool>.py@mark.unit+@mark.integration双模式测试AgentTestHarness.assert_tool_called("tool_name")验证工具被调用