Struct SessionManager 

pub struct SessionManager { /* private fields */ }

Orchestrates session operations using an Arc<dyn SessionStore>.

Analogous to Python’s SessionManager. Loads conversation history as Vec<Message> for LLM multi-turn conversations, and saves Q&A entries after each search completion.

Implementations

impl SessionManager

pub fn new(store: Arc<dyn SessionStore>) -> SessionManager

pub fn with_llm(self, llm: Arc<dyn Llm>) -> SessionManager

pub fn with_default_session_id(self, id: impl Into<String>) -> SessionManager

pub fn with_history_limit(self, limit: usize) -> SessionManager

pub async fn load_history_messages( &self, session_id: Option<&str>, user_id: Option<&str>, ) -> Result<Vec<Message>, SessionError>

Load conversation history as alternating User/Assistant messages.

Returns the last history_limit Q&A pairs as: [User(q1), Assistant(a1), User(q2), Assistant(a2), ...]

pub async fn load_history_both( &self, session_id: Option<&str>, user_id: Option<&str>, ) -> Result<(Vec<Message>, String), SessionError>

Load history as structured messages AND a formatted string, with a single store round-trip.

pub async fn save_qa( &self, session_id: Option<&str>, user_id: Option<&str>, question: &str, answer: &str, context: Option<&str>, used_graph_element_ids: Option<UsedGraphElementIds>, ) -> Result<String, SessionError>

Save a Q&A exchange to the session. Returns the generated qa_id.

used_graph_element_ids carries the node/edge IDs that were consulted during retrieval so the memify pipeline can trace which graph elements produced the answer (mirrors Python session_manager.py:492-525, add_qa(..., used_graph_element_ids=used_graph_element_ids)).

pub async fn delete_session( &self, session_id: Option<&str>, user_id: Option<&str>, ) -> Result<bool, SessionError>

Delete all Q&A entries for a session.

pub fn format_entries(entries: &[SessionQAEntry]) -> String

Format Q&A entries as a human-readable string (for debugging / compatibility with Python’s SessionManager.format_entries).

When include_context is true, the context field is included between QUESTION and ANSWER (matching the Python include_context parameter).

pub fn format_entries_with_context( entries: &[SessionQAEntry], include_context: bool, ) -> String

Format Q&A entries, optionally including context.

pub async fn update_qa( &self, session_id: Option<&str>, user_id: Option<&str>, qa_id: &str, updates: SessionQAUpdate, ) -> Result<bool, SessionError>

Update arbitrary fields on a QA entry.

pub async fn add_feedback( &self, session_id: Option<&str>, user_id: Option<&str>, qa_id: &str, feedback_text: Option<&str>, feedback_score: Option<i32>, ) -> Result<bool, SessionError>

Add or update feedback on a QA entry (convenience over update_qa).

Resets memify_metadata.feedback_weights_applied to false so that the memify pipeline will re-apply weights on the next run.

pub async fn delete_feedback( &self, session_id: Option<&str>, user_id: Option<&str>, qa_id: &str, ) -> Result<bool, SessionError>

Clear feedback from a QA entry.

pub async fn latest_qa_id( &self, session_id: Option<&str>, user_id: Option<&str>, ) -> Result<Option<String>, SessionError>

Return the qa_id of the most-recent Q&A entry in the session.

Returns None when the session has no entries yet. Used to route conversationally-detected feedback to the prior QA entry before saving the new turn (mirrors Python session_manager.py:462-469).

pub async fn get_graph_context( &self, session_id: Option<&str>, user_id: Option<&str>, ) -> Result<Option<String>, SessionError>

Retrieve graph knowledge snapshot for a session.

pub async fn set_graph_context( &self, session_id: Option<&str>, user_id: Option<&str>, context: &str, ) -> Result<(), SessionError>

Store graph knowledge snapshot for a session.

pub async fn add_agent_trace_step( &self, user_id: &str, session_id: Option<&str>, origin_function: &str, status: &str, memory_query: &str, memory_context: &str, method_params: Value, method_return_value: Option<Value>, error_message: &str, generate_feedback: bool, ) -> Result<String, SessionError>

Append one agent-trace step to the session and return the generated trace_id (UUID4).

Mirrors Python’s SessionManager.add_agent_trace_step.

When generate_feedback is true, this method attempts to use the configured LLM (with_llm) to summarize method_return_value; if no LLM is wired or generation fails, it falls back to deterministic feedback (<origin> succeeded/failed).

pub async fn get_agent_trace_session( &self, user_id: &str, session_id: Option<&str>, last_n: Option<usize>, ) -> Result<Vec<SessionTraceStep>, SessionError>

Retrieve agent-trace steps for a session, oldest-first.

If last_n is Some(n), the trailing n entries are returned (mirrors Python’s entries[-last_n:]).

Trait Implementations

impl Clone for SessionManager

fn clone(&self) -> SessionManager

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.