memory¶

Simple converation memory component for user-assistant conversations.

current_conversation¶

Stores the current active conversation.

turn_db¶

Database for storing all conversation turns.

conver_db¶

Database for storing complete conversations.

Classes

class ConversationMemory(turn_db: LocalDB = None, user_id: str = None)[source]¶

Bases: Component

call(metadata_filter: List[str] | None = None) → str[source]¶

Returns the current conversation history as a formatted string.

Parameters:

metadata_filter (Optional[List[str]]) – List of metadata keys to include. If None, all metadata is included.

Returns:

Formatted conversation history with alternating user and assistant messages.: Returns empty string if no conversation history exists.

Return type:

str

add_dialog_turn(user_query: str | UserQuery, assistant_response: str | AssistantResponse)[source]¶

Add a new dialog turn to the current conversation.

Parameters:

add_user_query(user_query: str | UserQuery) → str[source]¶

Add a user query to start a new dialog turn.

Parameters:: user_query (Union[str, UserQuery]) – The user’s input message. If UserQuery object, can include metadata.
Returns:: The ID of the pending dialog turn.
Return type:: str
Raises:: ValueError – If there’s already a pending user query without an assistant response.

add_assistant_response(assistant_response: str | AssistantResponse) → str[source]¶

Add an assistant response to complete the current dialog turn.

Parameters:: assistant_response (Union[str, AssistantResponse]) – The assistant’s response message.
Returns:: The ID of the completed dialog turn.
Return type:: str
Raises:: ValueError – If there’s no pending user query to respond to.

reset_pending_query()[source]¶: Reset any pending query state. Useful for cleanup after cancellations or before starting new queries.