MemoryManagers

Memory Managers

In Pantella, memory managers are used to manage the memory of characters in the game. The memory manager is responsible for managing the memory of a character after they're added to a conversation, saving memories, managing the character in the game, and managing the character's memory manager.

Memory Manager Class

The memory manager class is a class that manages the memory of characters in the game. The memory manager class has the following methods and properties:

• __init__(self,character_manager): Initializes the memory manager. The character_manager parameter is the character manager that the memory manager is associated with.
• name: Returns the name of the character.
• race: Returns the race of the character.
• gender: Returns the gender of the character.
• base_id: Returns the base ID of the character.
• ref_id: Returns the reference ID of the character.
• messages: Returns the messages of the conversation manager.
• after_step(self): Performs a step in the memory manager after the conversation step.
• before_step(self): Performs a step in the memory manager before the conversation step.
• reached_conversation_limit(self): Ran when the conversation limit is reached. Some memory managers may need to perform some action when the conversation limit is reached.
• add_message(self, message): Adds a message to the memory manager.
• load_messages(self): Loads messages from the memory manager. Some memory managers may need to load messages from a file or database, and can also use this method to load old messages into the conversation_manager's messages.
• memories: Returns the memories of the character's memory manager.
• memory_offset: Returns the memory depth of the character.
• memory_offset_direction: Returns the memory offset direction of the character.

Method Details

__init__(self, character_manager) Method

Initializes the memory manager. The character_manager parameter is the character manager that the memory manager is associated with. When creating a new memory manager, you should call the parent class's __init__ method to ensure that the default memory manager is loaded.

after_step(self) Method

Perform after dialogue generation in the memory manager - Some memory managers may need to perform some action every step.

before_step(self) Method

Performed before dialogue generation in the memory manager - Some memory managers may need to perform some action every step.

reached_conversation_limit(self) Method

Ran when the conversation limit is reached, or the conversation is ended - Some memory managers may need to perform some action when the conversation limit is reached.

add_message(self, message) Method

Add a message to the memory manager.

load_messages(self) Method

Load messages from the memory manager - Some memory managers may need to load messages from a file or database, and can also use this method to load old messages into the conversation_manager's messages.

memories Property

Return a string representation of the memories stored in the memory manager - Some memory managers have updating memories strings.

memory_offset Property

Return the memory depth of the character.

memory_offset_direction Property

Return the memory offset direction of the character.

Example Memory Manager

from src.logging import logging

manager_slug = "example_memory_manager"

class MemoryManager():
    def __init__(self,character_manager):
        self.conversation_manager = character_manager.conversation_manager
        self.character_manager = character_manager
        self.config = self.conversation_manager.config
        # Add your memory manager initialization code here

    def after_step(self):
        logging.info("Performing a step in the memory manager after the conversation step.")

    def before_step(self):
        logging.info("Performing a step in the memory manager before the conversation step.")

    def reached_conversation_limit(self):
        logging.info("Conversation limit reached, nothing to do in the memory manager.")

    def add_message(self, message):
        logging.info(f"Adding message to the memory manager: {message}")

    def load_messages(self):
        logging.info("Loading messages from the memory manager.")

    @property
    def memories(self):
        logging.info("Returning the memories of the character's memory manager.")
        return [{
            "role": "example",
            "content": "This is an example memory.",
            "type": "memory"
        }]