Struct AdvisorRuntime 

pub struct AdvisorRuntime { /* private fields */ }

Drives the advisor agent. Construct via AdvisorRuntime::new, wrap in Arc, then call AdvisorRuntime::install_self so it can self-spawn its drain task.

Implementations

impl AdvisorRuntime

pub fn new( agent: Arc<dyn AdvisorAgent>, host: Arc<dyn AdvisorRuntimeHost>, retry_delay: Duration, ) -> Self

Construct. retry_delay is the backoff between failed advisor turns (omp default 1000ms).

pub fn install_self(&self, weak: Weak<AdvisorRuntime>)

Install the weak self-reference required for self-spawning the drain task. Call once after wrapping in Arc.

pub fn backlog(&self) -> u64

Current backlog (primary turns not yet digested by the advisor).

pub fn is_disposed(&self) -> bool

Whether the runtime has been disposed.

pub fn on_turn_end(&self, messages: Vec<Message>)

Feed one primary turn’s transcript to the advisor. Renders the delta (new messages since the last drain), queues it, and spawns a drain task if none is running. omp onTurnEnd.

pub async fn wait_for_catchup(&self, max: Duration, threshold: u64)

Block until the backlog drops below threshold, or max elapses. omp waitForCatchup. Registration + backlog-check happen under the waiters lock so a concurrent notify_waiters cannot miss the waiter (the catchup lost-wakeup fix).

pub fn reset(&self)

Re-prime the advisor after a history rewrite (compaction, session switch/resume, branch). Clears the advisor’s context and rewinds the cursor so the next turn replays the full current transcript. omp reset.

pub fn seed_to(&self, count: u64)

Seed the cursor to the current transcript length when the advisor is enabled mid-session, so the next turn doesn’t replay the entire history. omp seedTo.

pub fn dispose(&self)

Stop the runtime permanently. Aborts the advisor agent and drops all pending state. omp dispose.