Struct CheckpointsRepo 

pub struct CheckpointsRepo<'a> { /* private fields */ }

Implementations

impl CheckpointsRepo<'_>

pub fn create(&self, new: NewCheckpoint) -> Result<CheckpointRecord, Error>

pub fn get(&self, id: &str) -> Result<Option<CheckpointRecord>, Error>

pub fn set_approval(&self, id: &str, approval_id: &str) -> Result<(), Error>

pub fn delete(&self, id: &str) -> Result<bool, Error>

Delete a checkpoint row outright. Returns whether a row was removed.

F23 (RC-F): coordinates the on-disk checkpoint-dir GC (crate::checkpoint::gc_old_checkpoint_dirs) with the DB. The dir GC prunes by mtime regardless of archive state, while storage Self / gc() only removes ARCHIVED checkpoint rows — so a never-archived old checkpoint would lose its directory while its row survived, and a later restore_checkpoint would fail on the missing manifest. The dir GC now calls this so list() and the on-disk dirs stay in agreement.

pub fn list(&self, limit: usize) -> Result<Vec<CheckpointRecord>, Error>

pub fn list_all(&self, limit: usize) -> Result<Vec<CheckpointRecord>, Error>

pub fn list_for_session( &self, session_id: &str, after_message_index: i64, ) -> Result<Vec<CheckpointRecord>, Error>

Unarchived checkpoints of session_id anchored STRICTLY past after_message_index, oldest first. Strict > is the fork-boundary invariant: a fork at user-message index k keeps messages[..k], and a checkpoint stamped message_index == k snapshotted state from BEFORE that user message existed — it belongs to the kept prefix, not the discarded timeline. Oldest-first because each checkpoint is a PRE-mutation snapshot: the oldest one past the cut holds the file state closest to the fork point.

pub fn archive(&self, ids: &[String], reason: &str) -> Result<usize, Error>

pub fn count_archived(&self) -> Result<usize, Error>