Struct RuntimeStore 

pub struct RuntimeStore { /* private fields */ }

SQLite-backed durable runtime state.

Implementations

impl RuntimeStore

pub fn open_default() -> Result<Self>

Open the store at its default location under the app data dir, creating and (best-effort) locking down that dir first.

Errors

Resolving the data dir, creating it, and everything Self::open reports. Tightening permissions is best-effort on both platforms and never fails the open, so an Ok store is not proof the data dir is owner-only.

pub fn open(path: impl AsRef<Path>) -> Result<Self>

Open (creating if needed) the SQLite store at path, then apply the connection pragmas and run the schema migration.

Errors

Creating the parent directory, opening the database — a corrupt file or a directory the process cannot write — setting the connection pragmas, and the schema migration. A concurrent opener is not among them: WAL plus the busy timeout is exactly what keeps a second process from failing here with SQLITE_BUSY.

pub fn path(&self) -> &Path

pub fn sessions(&self) -> SessionsRepo<'_>

pub fn messages(&self) -> MessagesRepo<'_>

pub fn tasks(&self) -> TasksRepo<'_>

pub fn tool_runs(&self) -> ToolRunsRepo<'_>

pub fn approvals(&self) -> ApprovalsRepo<'_>

pub fn processes(&self) -> ProcessesRepo<'_>

pub fn checkpoints(&self) -> CheckpointsRepo<'_>

pub fn compactions(&self) -> CompactionsRepo<'_>

pub fn plugins(&self) -> PluginsRepo<'_>

pub fn provider_probes(&self) -> ProviderProbesRepo<'_>

pub fn pairing_tokens(&self) -> PairingTokensRepo<'_>

pub fn outcomes(&self) -> OutcomesRepo<'_>

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

Recover state stranded by a previous daemon’s crash/stop (#120, #118). A Running task’s worker died with the daemon, so it can never finish — mark it failed with an event. An approval left in the transient approving claim state (a replay that crashed mid-effect, #118) is reset to undecided so it reappears as pending and stays re-runnable. Call once on daemon startup, before serving. Returns (tasks_reset, claims_released).

F18 (RC-E): only daemon-owned running tasks are reset. The store is shared with interactive mermaid CLI runs; their tasks are created with a NULL owner_kind and are LEFT RUNNING here, so a live CLI session isn’t wrongly flipped to failed (with a spurious “interrupted” event) just because the daemon restarted. The daemon tags the tasks it runs in-process via NewTask::daemon_owned.

Errors

Taking the BEGIN IMMEDIATE write lock — which waits out a concurrent writer for busy_timeout before giving up — and any statement in the pass. Every failure rolls the transaction back, so recovery is all-or-nothing: no task is left flipped to failed without its interrupted event, and no claim is released without the tasks beside it. The caller may simply run it again.

pub fn gc( &self, retention_days: i64, outcomes_retention_days: i64, ) -> Result<u64>

Best-effort retention GC (#130, F22/RC-F): prune archived approvals/checkpoints, the events of long-finished tasks, terminal tasks, and the high-churn / old rows of the remaining tables, all older than retention_days. The append-only outcomes reward table — the self-improving-loop training corpus — is pruned on its own, longer outcomes_retention_days window so a large training history survives the shorter task/session window. Deletes only archived, finished, or terminal-and-old rows — active data is never touched (a running task, a still-open tool run, a live process, or a recently-updated session all survive). Returns the number of rows removed.

Errors

Opening the transaction and any DELETE in it. The whole pass is one transaction, so a failure prunes nothing and the returned count is never partial. Having nothing to prune is Ok(0).