Struct ProspectiveStore 

pub struct ProspectiveStore { /* private fields */ }

Storage and query engine for prospective memory (reminders)

Implementations

impl ProspectiveStore

pub fn column_family_descriptors() -> Vec<ColumnFamilyDescriptor>

Return the column family descriptors needed by ProspectiveStore.

Call this when opening the shared DB so the CFs are registered.

pub fn new(db: Arc<DB>, storage_path: &Path) -> Result<Self>

Create a new prospective store backed by the given shared DB.

The DB must have been opened with the column families returned by [column_family_descriptors()]. On first run after the migration, data from the old separate tasks/ and index/ sub-DBs is copied into the corresponding CFs and the old directories are renamed.

pub fn flush(&self) -> Result<()>

Flush all column families to disk (critical for graceful shutdown)

pub fn databases(&self) -> Vec<(&str, &Arc<DB>)>

Get references to all RocksDB databases for backup

pub fn store(&self, task: &ProspectiveTask) -> Result<()>

Store a new prospective task

pub fn get( &self, user_id: &str, task_id: &ProspectiveTaskId, ) -> Result<Option<ProspectiveTask>>

Get a task by ID

pub fn update(&self, task: &ProspectiveTask) -> Result<()>

Update a task (e.g., mark as triggered/dismissed)

Atomic: reads old task, builds a single WriteBatch containing old-index deletes + new task write + new index writes, then commits once.

pub fn delete(&self, user_id: &str, task_id: &ProspectiveTaskId) -> Result<bool>

Delete a task (atomic: removes task + all indices in single WriteBatch)

pub fn list_for_user( &self, user_id: &str, status_filter: Option<ProspectiveTaskStatus>, ) -> Result<Vec<ProspectiveTask>>

List all tasks for a user, optionally filtered by status

pub fn get_due_tasks(&self, user_id: &str) -> Result<Vec<ProspectiveTask>>

Get all due time-based tasks for a user

Returns tasks where:

• Trigger is time-based (AtTime or AfterDuration)
• Trigger time <= now
• Status is Pending

pub fn get_all_due_tasks(&self) -> Result<Vec<(String, ProspectiveTask)>>

Scan ALL users for due reminders (used by active reminder scheduler).

Returns (user_id, task) pairs for all pending tasks whose due time has passed. Leverages the zero-padded due:{timestamp}:{task_id} index for efficient scanning: lexicographic order = chronological, so we stop at the first future timestamp.

pub fn check_context_triggers( &self, user_id: &str, context: &str, ) -> Result<Vec<ProspectiveTask>>

Check for context-triggered reminders based on text content (keyword match only)

Returns tasks where:

• Trigger is OnContext
• Any keyword matches the context text
• Status is Pending

For semantic matching, use check_context_triggers_semantic instead.

pub fn check_context_triggers_semantic<F>( &self, user_id: &str, context: &str, context_embedding: &[f32], embed_fn: F, ) -> Result<Vec<(ProspectiveTask, f32)>>

where F: Fn(&str) -> Option<Vec<f32>>,

Check for context-triggered reminders using both keyword AND semantic matching

Returns tasks that either:

• Have keyword matches in the context (score = 1.0), OR
• Have semantic similarity above their threshold

Arguments

• user_id - User to check reminders for
• context - Current context text (for keyword matching)
• context_embedding - Precomputed embedding of the context
• embed_fn - Closure to compute embedding for task content

Returns

Vector of (task, score) tuples sorted by score (highest first)

pub fn mark_triggered( &self, user_id: &str, task_id: &ProspectiveTaskId, ) -> Result<bool>

Mark a task as triggered

pub fn mark_dismissed( &self, user_id: &str, task_id: &ProspectiveTaskId, ) -> Result<bool>

Mark a task as dismissed

pub fn pending_count(&self, user_id: &str) -> Result<usize>

Get count of pending tasks for a user

pub fn find_by_prefix( &self, user_id: &str, id_prefix: &str, ) -> Result<Option<ProspectiveTask>>

Find a task by ID prefix (for short ID lookups)

Allows users to dismiss reminders using short IDs like “d8cdc580” instead of full UUIDs like “d8cdc580-bf96-403a-85c5-57098c7b1786”

pub fn cleanup_old_tasks( &self, user_id: &str, older_than_days: i64, ) -> Result<usize>

Cleanup expired/dismissed tasks older than given days