Ctrl+K

agent_studio.agent.direct_agent

agent_studio.agent.direct_agent#

Module Contents#

Classes#

DirectAgent

Zero-shot LLM agents.

Attributes#

config

logger

agent_studio.agent.direct_agent.config[source]#
agent_studio.agent.direct_agent.logger[source]#
class agent_studio.agent.direct_agent.DirectAgent[source]#

Bases: agent_studio.agent.base_agent.BaseAgent

Zero-shot LLM agents.

name: str = 'direct'[source]#
reset(instruction: str) → None[source]#
trajectory2intermediate_msg() → list[dict[str, Any]][source]#

Converts the trajectory to intermediate messages.

Returns:

The intermediate messages.
  • role:

    • system

    • user

    • assistant

  • content: The content of the message. content can either be a string or a np.array. If it is a np.array, it should be in RGB format.

Return type:

list[dict[str, Any]]

eval(final_obs: numpy.ndarray | None = None) → dict[str, Any][source]#
Contents