BaseActor subclassing#

To create a subclass of BaseActor, the following methods should be overridden:

_observe(self, observation: str | Any, time: Any = None, data: Any = None)#

This method handles the observation process. It should take an observation and handle it as necessary for the actor's functionality. time and data may contain optional additional attributes to be stored. Note that some agents may ignore either of these, and not all memory systems support them.

from interlab import BaseActor

class MyActor(BaseActor):
    def _observe(self, observation):
        # Process the observation
        pass

_query(self, prompt: Any = None, **kwargs) -> Any#

This method processes a query and returns the result. The prompt and any keyword arguments can be used to tailor the query/response logic of the actor.

class MyActor(BaseActor):
    # ... (other methods)

    def _query(self, prompt: Any = None, **kwargs) -> Any:
        # Process the query and return the result
        return some_result

copy(self)#

Copy of an actor. By default it makes a deep copy but you can override it.

ActorWithMemory Subclassing#

When subclassing ActorWithMemory, only the _query method is compulsory to override as it already provides implementations for _observe.

_query(self, prompt: Any = None, **kwargs) -> Any#

Just like with the BaseActor, you need to implement query processing logic that suits your actor's role.

from interlab import ActorWithMemory

class MyMemoryActor(ActorWithMemory):
    # ... (other methods, if any)

    def _query(self, prompt: Any = None, **kwargs) -> Any:
        # Access the memory and process the query
        return some_result