Parameters
| Parameter | Type | Default | Description |
|---|---|---|---|
storage | Storage | Required | Storage backend to use for persistence |
session_id | Optional[str] | None | Session ID for session-based memory features |
user_id | Optional[str] | None | User ID for user-specific memory features |
full_session_memory | bool | False | Enable full session memory storage |
summary_memory | bool | False | Enable session summary generation and storage |
user_analysis_memory | bool | False | Enable user trait analysis and profile building |
user_profile_schema | Optional[Type[BaseModel]] | None | Custom schema for user profile data |
dynamic_user_profile | bool | False | Enable dynamic user profile schema generation |
num_last_messages | Optional[int] | None | Limit conversation history to last N messages |
model | Optional[Union[Model, str]] | None | Model provider for memory analysis tasks |
debug | bool | False | Enable debug logging |
feed_tool_call_results | bool | False | Include tool call results in memory analysis |
user_memory_mode | Literal['update', 'replace'] | 'update' | Mode for updating user memory (update existing or replace) |
Functions
prepare_inputs_for_task
Gathers all relevant memory data before a task execution, correctly parsing and limiting the chat history.
Returns:
Dict[str, Any]: Dictionary containing:message_history: List of validated and limited message historycontext_injection: Context string for session summarysystem_prompt_injection: System prompt injection for user profile
update_memories_after_task
Updates all relevant memories after a task has been completed, saving the chat history in the correct format.
Parameters:
model_response: The model response containing new messages to store
_update_interaction_session
Helper method to handle updating the InteractionSession object.
Parameters:
model_response: The model response containing new messages
_update_user_profile
Helper method to handle updating the UserProfile object.
Parameters:
model_response: The model response containing new messages
_limit_message_history
Limits conversation history to the last N runs, creating a new synthetic first request that combines the original system prompt with the user prompt from the beginning of the limited window.
Parameters:
message_history(List): The full, flat list of ModelRequest and ModelResponse objects
List: A new, limited message history list
_generate_new_summary
Generates a new session summary based on the previous summary and new conversation turn.
Parameters:
previous_summary(str): The previous session summarymodel_response: The model response containing new messages
str: The updated session summary
_extract_user_prompt_content
Extracts the content string from all UserPromptParts in a list of messages.
Parameters:
messages(list): List of messages to extract user prompts from
list[str]: List of user prompt content strings
_analyze_interaction_for_traits
Analyzes user interaction to extract traits using the configured model provider.
Parameters:
current_profile(dict): Current user profile datamodel_response: The model response containing new messages
Tuple[dict, Literal['update', 'replace']]: Updated traits and update mode