AgentRuntime

The AgentRuntime class is a high-level wrapper around ChatObject that provides a reusable agent operation interface.

This class encapsulates the complexity of ChatObject and provides a simplified API for agent interactions. It maintains session state, configuration, and strategy settings, making it a reusable object for multiple agent operations within the same context.

Properties

• strategy (type[AgentStrategy]): Agent strategy class used for execution
• session_id (str): Session ID for the agent
• slot (BackendSlots): Backend slots providing memory and ability backends
• preset (ModelPreset): Model preset configuration
• config (AmritaConfig): Amrita configuration object
• train (Message[str]): Training data (system prompts)
• template (Template): Jinja2 template used to render system role message

Constructor Parameters

• config (AmritaConfig): Amrita configuration object containing global configuration settings
• preset (ModelPreset): Model preset configuration defining basic model parameters and settings
• train (dict[str, str] | Message[str]): System prompt for the agent (dict or Message object)
• strategy (type[AgentStrategy], optional): Agent strategy class, defaults to ReActAgentStrategy
• template (Template | str, optional): Jinja2 template used to render the system prompt, defaults to DEFAULT_TEMPLATE
• session_id (str | None, optional): Session identifier string. If None, a new UUID-based ID is generated. The session_id is passed to every ChatObject created by this runtime, allowing the Backend to isolate memory and abilities per session
• backend (BackendSlots | None, optional): Backend slots providing memory and ability backends. If None, a LegacyBackend is used for both slots, which stores data in global in-process containers

Methods

set_strategy(strategy)

Set the agent strategy to be used for execution.

Parameters:

• strategy (type[AgentStrategy]): The agent strategy to be used for execution

get_chatobject(input, **kwargs)

Get a chat object for a specific interaction.

Parameters:

• input (USER_INPUT): Input from the user
• **kwargs: Additional keyword arguments passed to ChatObject constructor

Returns: ChatObject - A configured ChatObject instance ready for execution

Usage Example

from amrita_core import create_agent

# Create an agent using the factory function
agent = create_agent(
    "https://api.example.com",
    "your-api-key",
    model="gpt-4",
    model_config={"temperature": 0.7},
)

# Get a chat object for interaction
chat = agent.get_chatobject("Hello, what can you do?")

# Execute the interaction
async with chat.begin():
    response = await chat.full_response()
    await chat  # Wait for the task to finish before exiting
    print(response)