Agent

dopus.core.Agent

Bases: ABC

Abstract base class of an agent capable of running tools and managing conversations. The agent interfaces with a language model (LLM) and has access to various tools through the tool registry. It is designed to handle conversations where tools may be called upon dynamically.

Attributes:

Name	Type	Description
provider	Provider	The language model provider used by the agent.
name	str	The name of the agent.
convo	Convo	Conversation handler that manages the dialogue context.
registry	dict	A registry of tools that can be utilized by the agent.
tool_manager	ToolRunner	Manager that handles tool execution and lifecycle events.

__init__

__init__(provider: Provider, name: str = 'Agent', convo: Convo = None, registry: dict = None, tool_manager: ToolRunner = None)

Parameters:

Name	Type	Description	Default
name	str	Name of the agent.	'Agent'
provider	Provider	language model provider instance.	required

get_actions

get_actions()

Get the list of actions the agent has taken

reset

reset()

Resets the conversation context for the agent. Clears the current conversation history.

run

run(message: str = None)

Runs the agent, initiating a conversation or tool execution loop. If a message is provided, it appends it to the conversation.

Parameters:

Name	Type	Description	Default
message	str	Message from the user to be processed by the agent.	None

Returns:

Type	Description
	Result of the tool execution loop managed by the tool manager.

stop

stop(result=None)

Stops the current tool execution or conversation flow and returns a result.

Parameters:

Name	Type	Description	Default
result		Optional result to return when stopping.	None

on_tool_use

on_tool_use(tool: str, func)

Registers a callback function to be triggered when a tool is used.

Parameters:

Name	Type	Description	Default
tool	str	Name of the tool.	required
func	function	Function to be called when the tool is triggered.	required

add_tool

add_tool(tool: str)

Adds a tool by name to the tool manager, making it available for the agent to use.

Parameters:

Name	Type	Description	Default
tool	str	Name of the tool to add.	required

add_tools

add_tools(tools: list)

Adds a list of tools to the tool manager

Parameters:

Name	Type	Description	Default
tools	list	List of tool names to add.	required

prompt abstractmethod

prompt() -> str

Abstract method to provide the agent's system prompt or message. Must be implemented by subclasses.

Returns:

Name	Type	Description
str	str	The prompt or message that represents the agent's behavior.