Memory

Memory

Bases: MemoryMeta

Concrete implementation of MemoryMeta for storing and managing conversational memory. Memory is persisted in a JSON Lines file, with support for user-specific data and graph-based representations.

Methods:

Name	Description
save_memory	Save a list of memory entries for a specific user to the memory file.
load_all_memory	Load all memory data from the memory file.
load_memory_by_user	Load memory data for a specific user from the memory file.
save_short_term_memory	Convert a message to a graph using a language model and update the user's memory.
update_memory	Update the user's memory by adding or updating graph entries, avoiding duplicates.

save_memory

save_memory(obj: list, memory_path: Path, user_id: str = 'unknown_user')

Save a list of memory entries for a specific user to the memory file.

Parameters:

Name	Type	Description	Default
obj	list	List of memory entries to save.	required
memory_path	Path	Path to the memory file.	required
user_id	str	The user identifier. Defaults to "unknown_user".	'unknown_user'

load_all_memory

load_all_memory()

Load all memory data from the memory file.

Returns:

Name	Type	Description
dict		The entire memory data as a dictionary, with user IDs as keys and lists of memory entries as values.

load_memory_by_user

load_memory_by_user(load_type: Literal['list', 'string'] = 'list', user_id: str = 'unknown_user')

Load memory data for a specific user from the memory file.

Parameters:

Name	Type	Description	Default
load_type	Literal['list', 'string']	Format of the returned data ("list" or "string"). Defaults to "list".	'list'
user_id	str	The user identifier. Defaults to "unknown_user".	'unknown_user'

Returns:

Type	Description
	Union[List[dict], str]: List of memory entries if load_type is "list", or a string representation if "string".

save_short_term_memory

save_short_term_memory(llm: Union[ChatTogether, BaseLanguageModel, BaseChatOpenAI], message: str, user_id: str = 'unknown_user', *args, **kwargs)

Convert a message to a graph using a language model and update the user's memory.

Parameters:

Name	Type	Description	Default
llm	Union[ChatTogether, BaseLanguageModel, BaseChatOpenAI]	Language model for graph generation.	required
message	str	The message to convert and store.	required
user_id	str	The user identifier. Defaults to "unknown_user".	'unknown_user'
*args,	**kwargs	Additional arguments for flexibility.	required

Returns:

Name	Type	Description
list		The generated graph representation of the message.

update_memory

update_memory(graph: list, user_id: str = 'unknown_user')

Update the user's memory by adding or updating graph entries, avoiding duplicates.

Parameters:

Name	Type	Description	Default
graph	list	List of graph entries, each with head, head_type, relation, relation_properties, tail, and tail_type.	required
user_id	str	The user identifier. Defaults to "unknown_user".	'unknown_user'

Returns:

Name	Type	Description
list		The updated list of memory entries for the user.