data_juicer_agents.core#

Core modules for Data Juicer Agent.

data_juicer_agents.core.create_agent(name: str, sys_prompt: str, toolkit: Toolkit, description: str | None = None, model: ChatModelBase | None = None, formatter: FormatterBase | None = None, memory: MemoryBase | None = None, max_iters: int = 10, parallel_tool_calls: bool = False, **kwargs) → ReActAgent[源代码]#

Create a ReActAgent with standardized configuration.

参数:

  • name -- Agent identifier

  • sys_prompt -- System prompt template (supports {name} placeholder)

  • toolkit -- Toolkit instance

  • model -- Language model (defaults to GPT-4o)

  • formatter -- Message formatter (defaults to OpenAIChatFormatter)

  • memory -- Memory instance (defaults to InMemoryMemory)

  • max_iters -- Maximum reasoning iterations

  • parallel_tool_calls -- Enable parallel tool execution

  • **kwargs -- Additional ReActAgent arguments

返回:

Configured ReActAgent instance

示例

>>> agent = create_agent(
...     name="sql_expert",
...     sys_prompt="You are {name}, a SQL database expert",
...     tools=sql_tools
... )
data_juicer_agents.core.register_dj_agent_hooks(agent)[源代码]#

Register cleaning hooks for DataJuicer agent.

参数:

agent -- ReActAgent instance to register hooks for