Skip to content

agent_tools

Managed agent tools

Attributes

AGENT_SYSTEM_PROMPT module-attribute

AGENT_SYSTEM_PROMPT = ''

Classes

Functions

derive_persona

derive_persona(
    *,
    agent: Agent,
    include_description: bool = True,
    include_special_instructions: bool = True,
    wrap: bool = True
) -> str

Applies consistent logic when deriving an Agent’s persona.

Including special instructions for this function is typically used for completion-style inferencing. For chat-style inferencing it is usually better to add special instructions as a system-level chat message at the end if the input context so it has the greatest effect.

Parameters:

  • agent (Agent) –

    ORM reference of the agent to derive the persona for.

  • include_description (bool, default: True ) –

    Include the agent’s (physical) description in the persona, this is useful for roleplaying however is just noise for utility / reasoning agents. Defaults to True.

  • include_special_instructions (bool, default: True ) –

    Include the agent’s special instructions in the persona, see description above. Defaults to True.

  • wrap (bool, default: True ) –

    Wrap the persona in a PEM-style header/footer for the LLM. Defaults to True.

Returns:

  • str ( str ) –

    The derived persona for the agent.

derive_special_instructions

derive_special_instructions(*, agent: Agent) -> str

Derives special instructions for an agent.

Parameters:

  • agent (Agent) –

    ORM reference of the agent to derive the special instructions for.

Returns:

  • str ( str ) –

    Special instructions for the agent.