Skip to main content

​
AgentRuntime Class Documentation

Description: The AgentRuntime class provides the runtime environment for executing agent workflows. It manages the lifecycle of agent execution, including input processing, payment validation, response generation, and output delivery. Class Definition:

​
Parameters:

  • inputs (List[AgentInput]):
    • Description: A list of AgentInput objects. These define the sources from which the agent receives messages.
    • Requirement: Each AgentInput object must implement the start method, which pushes messages to a queue for processing.
  • outputs (List[AgentOutput]):
    • Description: A list of AgentOutput objects. These define the destinations to which the agent sends responses.
    • Requirement: Each AgentOutput object must implement the send method, which delivers a response message to its destination.
  • agent (Agent):
    • Description: An Agent object that represents the core intelligence of the system.
    • Requirement: This object must implement the execute method, which processes a single request and generates a response.
  • pricing (Optional[Pricing], defaults to None):
    • Description: An optional Pricing object that defines the cost of using the agent and the wallet address to which payments should be sent.
    • Default: If None, no payment validation is performed.
  • debug (bool, defaults to False):
    • Description: A boolean indicating whether debug mode is enabled. In debug mode, additional logging information is printed.
  • enable_logs (bool, defaults to False):
    • Description: A boolean indicating whether logging is enabled. When enabled, logs are exported to the Galadriel platform.
Variables:
  • inputs (List[AgentInput]): Stores the input sources for the agent.
  • outputs (List[AgentOutput]): Stores the output destinations for responses.
  • agent (Agent): Stores the agent implementation being used.
  • pricing (Optional[Pricing]): Stores the payment configuration, if required.
  • spent_payments (Set[str]): A set to keep track of already processed transaction signatures to prevent duplicate processing.
  • debug (bool): Flag to enable debug mode for more verbose logging.
  • enable_logs (bool): Flag to enable log exporting.
Methods:

​
__init__(self, inputs: List[AgentInput], outputs: List[AgentOutput], agent: Agent, pricing: Optional[Pricing] = None, debug: bool = False, enable_logs: bool = False)

Initializes the AgentRuntime with input sources, output destinations, an agent implementation, pricing configuration, debug mode flag, and enable logs flag.

​
run(self)

Starts the agent runtime loop, continuously processing incoming requests.

​
_run_request(self, request: Message)

Processes a single request through the agent pipeline, handling payment validation, agent execution, and response delivery.

​
_get_memory(self) -> List[Dict[str, str]]

Retrieves the current state of the agent’s memory.

​
_generate_proof(self, request: Message, response: Message) -> str

Generates a cryptographic proof of the request and response.

​
_publish_proof(self, request: Message, response: Message, proof: str)

Publishes the cryptographic proof. Data Types:
  • List[AgentInput]: A list of objects conforming to the AgentInput interface.
  • List[AgentOutput]: A list of objects conforming to the AgentOutput interface.
  • Agent: An object conforming to the Agent interface.
  • Optional[Pricing]: An optional Pricing object with the following attributes:
    • cost (float): The cost of the task in SOL.
    • wallet_address (str): The Solana wallet address where payment should be sent.
  • bool: A boolean value (True or False).
  • Message: An object with the following attributes:
    • content (str): The content of the message.
    • conversation_id (Optional[str], defaults to None): An optional conversation ID.
    • type (Optional[str], defaults to None): An optional message type.
    • additional_kwargs (Optional[Dict], defaults to None): An optional dictionary of additional keyword arguments.