Skip to main content

RunEventKind

Enum RunEventKind 

Source
pub enum RunEventKind {
Show 32 variants SuperStepStart, SuperStepEnd, NodeInput, NodeOutput, ModelCall, ToolCall, RemoteCall, WasmCall, Interrupt, Resume, RoutingDecision, CheckpointWritten, EffectReceipt, AgentSpawn, AgentExit, MailboxSend, MailboxReceive, SupervisionEvent, CoordinationStart, CoordinationEnd, MemoryRead, MemoryWrite, MemoryForget, CandidateCreated, CandidateEvaluated, CandidatePromoted, CandidateRolledBack, PolicyDecision, CapsuleResolved, CapsuleCall, CapsuleDenied, SigningKeyRotated,
}
Expand description

What a RunEvent records. Closed set; replay and analysis code matches exhaustively on it.

Variants§

§

SuperStepStart

A super-step began; input lists the activated node set.

§

SuperStepEnd

A super-step merged at the barrier; output carries the post-reducer channel values (the reducer result for the step).

§

NodeInput

A node invocation was scheduled; input is its (scoped) state snapshot.

§

NodeOutput

A node invocation finished; output is its partial updates plus any routing command, with the measured latency.

§

ModelCall

A chat-model call; input is the request (messages + tool schemas), output the response, with model identity, token usage, and cost where reported.

§

ToolCall

A tool invocation; input is the arguments, output the result.

§

RemoteCall

A remote-node call to a worker over the wire protocol; input is the NodeTask, output the NodeTaskResponse payload.

§

WasmCall

A WASM guest-module invocation; input is the guest input, output the guest output.

§

Interrupt

A node suspended the run; input is the interrupt payload.

§

Resume

The run resumed from a checkpoint; input carries the checkpoint id and, when present, the resume value.

§

RoutingDecision

The routing phase selected the next active set; output describes the planned invocations (including Send fan-outs).

§

CheckpointWritten

A checkpoint was persisted; output carries the checkpoint id, step, and journal head reference stamped into it.

§

EffectReceipt

An Idempotent effect’s own confirmation, journaled by the effect’s recipient (R0.6 Durable Work): output carries the EffectReceipt — the provider’s confirmation id plus the idempotency key the effect ran under. Exact replay serves the receipt instead of re-sending the effect: the same journaled-output rule model and tool calls follow, extended across the crash boundary between a run and its durable tasks.

§

AgentSpawn

An agent was spawned (R0.7 Agent Fabric): input carries the agent id, its pinned crate::agents::CapabilityManifest, and the declared crate::agents::StateScopes — the manifest’s scopes are journaled with the spawn so later access checks replay against the same declaration. Wave 1 lands the variant as an inert contract; the agent host emits it.

§

AgentExit

An agent terminated (R0.7): output carries the terminal disposition (completed, failed, cancelled) and the final checkpoint reference. Wired in wave 2 for the cancellation tree: the server journals it into the agent’s supervision journal when an agent/team cancel actually touches work. Host-emitted terminal exits land with the agent host.

§

MailboxSend

A mailbox message was submitted to an agent’s mailbox (R0.7): output carries the recipient (agent:{agent_id}), the message kind, and the task id the queue assigned. The sender-side half of the mailbox journal pair; the envelope’s parent links it into the team’s causal tree. Wired in wave 3: the coordination runtime journals it into the pattern’s journal for every member task it submits. Host-emitted sends land with the agent host.

§

MailboxReceive

An agent’s activation began a turn on a mailbox message (R0.7): input carries the task id and idempotency key the turn is processing. The recipient-side half of the mailbox pair. Wired in wave 3 as the settlement observation: the coordination runtime journals it when a member task settles, with the terminal status and result — the pattern’s evidence half of the pair. Host-emitted turn begins land with the agent host.

§

SupervisionEvent

A supervision decision was made (R0.7 wave 2): output carries the policy (permanent / transient / temporary), the triggering failure’s EventStatus / error class, and the restart ordinal — the journaled decision record that makes “no restart without a journaled decision” auditable. Wired in wave 2: the server journals it into the agent’s supervision journal on restart / escalate / manual-restart decisions.

§

CoordinationStart

A coordination pattern (delegate / fan-out / race / quorum) began (R0.7 wave 3): output carries the pattern’s typed contract — members, thresholds, effect declarations. The team’s causal root for every event the pattern spawns. Wired in wave 3: the coordination runtime journals it on the pattern’s first drive, before any member task exists.

§

CoordinationEnd

A coordination pattern settled (R0.7 wave 3): output carries the result reference and the per-member dispositions (completed, failed, cancelled) — the fan-in evidence record. Wired in wave 3: the coordination runtime journals it exactly once, as the pattern’s terminal fact.

§

MemoryRead

A governed memory retrieval was performed (R0.8 Rusty Learn, wave 1): input is the canonical request — the resolved crate::memory::MemoryQuery (with as_of stamped through the run’s clock) plus the crate::memory::ContextBudget — and output is the journaled crate::memory::MemoryAssembly: the record ids and their order, the packed records, and the token accounting. Declared Effect::ReadOnly: exact replay serves the journaled assembly instead of re-querying the store — the same journaled-output rule model and tool calls follow, applied to the memory seam so candidate evaluation is reproducible.

§

MemoryWrite

A governed memory write landed (R0.8 wave 1): an Effect::Idempotent effect under the derived key memory:{scope}:{memory_id} (see crate::memory::memory_effect_key), so retried submissions converge. Input carries the effect key and the memory id; output carries the stored crate::memory::MemoryRecord with its provenance — the write’s attribution is journaled, not implied.

§

MemoryForget

A memory record was forgotten (R0.8 wave 2): real deletion of derived state, with the journaled tombstone as the erasure receipt. An Effect::Idempotent effect under the derived key memory_forget:{scope}:{memory_id} (see crate::memory::memory_forget_effect_key), so retried erasures converge. Input carries the effect key and the memory id; output carries the crate::memory::MemoryForgetTombstone — the id, scope, reason, and dependent invalidations, metadata by construction: the tombstone has no content field, so the forgotten bytes cannot leak through the receipt of their erasure. Corrections are not a separate variant: they journal through MemoryWrite with the correction’s attribution in the derived record’s provenance.

§

CandidateCreated

A learning candidate was created (R0.8 wave 3): a distiller’s proposed change landed as an immutable, content-addressed crate::learn::Candidate — identity is integrity, so two distillations of the same change converge on one id and a tampered candidate fails its own address. An Effect::Idempotent effect under the derived key candidate:{candidate_id} (see crate::learn::candidate_effect_key), so retried submissions converge. Input carries the effect key and the candidate id; output carries the candidate itself — the distiller’s identity and the evidence span it read are journaled with it, not implied.

§

CandidateEvaluated

A candidate was evaluated against recorded evidence (R0.8 wave 3): replay re-drove recorded runs with the candidate applied, the experiment runner graded it over the versioned dataset, and the comparison diffed it against the baseline. An Effect::Idempotent effect under the derived key evaluation:{candidate_id}:{dataset_version} (see crate::learn::evaluation_effect_key): re-evaluation against the same dataset version converges; a new dataset version is a new evaluation. Output carries the journaled crate::learn::CandidateEvaluation — the replay divergence summary, the report pair, the dataset version, and the verdict. The evaluation is evidence, not a log line.

§

CandidatePromoted

A candidate was promoted (R0.8 wave 3): the active-version pointer for its production surface moved. An Effect::Idempotent effect under the derived key promotion:{candidate_id} (see crate::learn::promotion_effect_key) — retried promotions converge, and recovery re-derives the same key. Output carries the crate::learn::PromotionReceipt: the pointer’s previous value and the promotion’s authority — the envelope version (the standing approval) or the approval token’s approved_by for an out-of-envelope promotion.

§

CandidateRolledBack

A promoted candidate was rolled back (R0.8 wave 3): the surface’s active-version pointer re-pointed to the previous candidate. Byte-exact because candidates are content-addressed and immutable — the restored version is the one that previously served, not a reconstruction. An Effect::Idempotent effect under the derived key rollback:{surface}:{candidate_id} (see crate::learn::rollback_effect_key). Output carries the crate::learn::RollbackReceipt: from, to, and the causing evidence. New runs bind the re-pointed version at admission; in-flight runs keep the version their checkpoint header pins.

§

PolicyDecision

An executor policy decision was made and recorded (R0.8 wave 4): the policy plane emitted a DecisionEvent at a decision point — v1 wires the retry classifier (crate::durable::classify_retry). An Effect::Pure record: the decision was already applied by the scheduler that emitted it; this event is the evidence of why, not a command to apply it again. Output carries the journaled DecisionEvent — features, the closed legal-action set, the selected action, the propensity, and the policy version that decided.

§

CapsuleResolved

A run manifest’s capsule version pin was resolved to a content address (R0.9 Rusty Capsules, wave 1): the server’s capsule registry mapped (identity, version) → crate::capsule::CapsuleId at admission and journaled the resolution, so the full chain — header pin → journaled resolution → receipt — reaches the manifest digest. An Effect::ReadOnly record (a registry lookup): output carries the journaled crate::capsule::CapsuleResolution — the pin name, the version string, the resolved capsule id, and the build digest the registry holds for it. A pin that resolves to a manifest failing its own content address fails admission instead — tampering is an admission error, never a journaled resolution.

§

CapsuleCall

A granted capsule capability was exercised (R0.9 wave 1): the guest called a linked import and the host performed (or attempted) the operation. The capsule rule’s “every use is journaled” half: output carries the journaled crate::capsule::CapsuleUse — the capsule id, the capability kind, the operation, and the request/response summaries. The effect class is the operation’s own (a GET fetch records Effect::ReadOnly; a writing method records Effect::NonIdempotent), and a failed operation is journaled with EventStatus::Error — a granted call that fails is still a use.

§

CapsuleDenied

A capsule capability attempt was refused (R0.9 wave 1): either the guest probed an import its manifest’s grants never linked (the structural denial — the import does not exist), or it called a granted import outside the grant’s scope (a network grant naming host A, a fetch attempted against host B). An Effect::Pure record: nothing executed, so there is no external effect to classify — the event is the evidence that nothing happened. Output carries the journaled crate::capsule::CapsuleDenial — the capsule id, the requested capability, and the manifest grant that was absent (the grant that would have permitted the attempt), so the denial is attributable to a declaration, not to a stack trace.

§

SigningKeyRotated

The deployment’s receipt signing key changed (R0.9 wave 3): a rotation was performed (or a host with no local secret joined a shared store) and the server journaled the new key id into the deployment’s receipts journal — the run id receipt-keys, not any tenant’s run. An Effect::Pure record: the key operation is local key material, so there is no external effect to classify — the event is the lineage evidence. Output carries the journaled crate::receipt::SigningKeyRotation — the previous key id (None on genesis), the new key id, its public half, and the rotation instant — so “which key signed what, from when” is a chained fact, and old receipts keep verifying against the key history the journal attests.

Trait Implementations§

Source§

impl Clone for RunEventKind

Source§

fn clone(&self) -> RunEventKind

Returns a duplicate of the value. Read more
1.0.0 (const: unstable) · Source§

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

Performs copy-assignment from source. Read more
Source§

impl Copy for RunEventKind

Source§

impl Debug for RunEventKind

Source§

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter. Read more
Source§

impl<'de> Deserialize<'de> for RunEventKind

Source§

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>
where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer. Read more
Source§

impl Eq for RunEventKind

Source§

impl Hash for RunEventKind

Source§

fn hash<__H: Hasher>(&self, state: &mut __H)

Feeds this value into the given Hasher. Read more
1.3.0 · Source§

fn hash_slice<H>(data: &[Self], state: &mut H)
where H: Hasher, Self: Sized,

Feeds a slice of this type into the given Hasher. Read more
Source§

impl PartialEq for RunEventKind

Source§

fn eq(&self, other: &RunEventKind) -> bool

Equality operator ==. Read more
1.0.0 (const: unstable) · Source§

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=. Read more
Source§

impl Serialize for RunEventKind

Source§

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>
where __S: Serializer,

Serialize this value into the given Serde serializer. Read more
Source§

impl StructuralPartialEq for RunEventKind

Auto Trait Implementations§

Blanket Implementations§

Source§

impl<T> Any for T
where T: 'static + ?Sized,

Source§

fn type_id(&self) -> TypeId

Gets the TypeId of self. Read more
Source§

impl<T> Borrow<T> for T
where T: ?Sized,

Source§

fn borrow(&self) -> &T

Immutably borrows from an owned value. Read more
Source§

impl<T> BorrowMut<T> for T
where T: ?Sized,

Source§

fn borrow_mut(&mut self) -> &mut T

Mutably borrows from an owned value. Read more
Source§

impl<ST, DT> CastableFrom<ST, Initialized, Initialized> for DT
where ST: ?Sized, DT: ?Sized,

Source§

impl<ST, DT> CastableFrom<ST, Uninit, Uninit> for DT
where ST: ?Sized, DT: ?Sized,

Source§

impl<T> CloneToUninit for T
where T: Clone,

Source§

unsafe fn clone_to_uninit(&self, dest: *mut u8)

🔬This is a nightly-only experimental API. (clone_to_uninit)
Performs copy-assignment from self to dest. Read more
Source§

impl<T> DeserializeOwned for T
where T: for<'de> Deserialize<'de>,

Source§

impl<T> From<T> for T

Source§

fn from(t: T) -> T

Returns the argument unchanged.

Source§

impl<T> Instrument for T

Source§

fn instrument(self, span: Span) -> Instrumented<Self> ⓘ

Instruments this type with the provided Span, returning an Instrumented wrapper. Read more
Source§

fn in_current_span(self) -> Instrumented<Self> ⓘ

Instruments this type with the current Span, returning an Instrumented wrapper. Read more
Source§

impl<T, U> Into<U> for T
where U: From<T>,

Source§

fn into(self) -> U

Calls U::from(self).

That is, this conversion is whatever the implementation of From<T> for U chooses to do.

Source§

impl<T> PolicyExt for T
where T: ?Sized,

Source§

fn and<P, B, E>(self, other: P) -> And<T, P>
where T: Sized + Policy<B, E>, P: Policy<B, E>,

Create a new Policy that returns Action::Follow only if self and other return Action::Follow. Read more
Source§

fn or<P, B, E>(self, other: P) -> Or<T, P>
where T: Sized + Policy<B, E>, P: Policy<B, E>,

Create a new Policy that returns Action::Follow if either self or other returns Action::Follow. Read more
Source§

impl<T> Read<Exclusive, BecauseExclusive> for T
where T: ?Sized,

Source§

impl<T> Same for T

Source§

type Output = T

Should always be Self
Source§

impl<T> ToOwned for T
where T: Clone,

Source§

type Owned = T

The resulting type after obtaining ownership.
Source§

fn to_owned(&self) -> T

Creates owned data from borrowed data, usually by cloning. Read more
Source§

fn clone_into(&self, target: &mut T)

Uses borrowed data to replace owned data, usually by cloning. Read more
Source§

impl<T, U> TryFrom<U> for T
where U: Into<T>,

Source§

type Error = Infallible

The type returned in the event of a conversion error.
Source§

fn try_from(value: U) -> Result<T, <T as TryFrom<U>>::Error>

Performs the conversion.
Source§

impl<T, U> TryInto<U> for T
where U: TryFrom<T>,

Source§

type Error = <U as TryFrom<T>>::Error

The type returned in the event of a conversion error.
Source§

fn try_into(self) -> Result<U, <U as TryFrom<T>>::Error>

Performs the conversion.
Source§

impl<V, T> VZip<V> for T
where V: MultiLane<T>,

Source§

fn vzip(self) -> V

Source§

impl<T> WithSubscriber for T

Source§

fn with_subscriber<S>(self, subscriber: S) -> WithDispatch<Self> ⓘ
where S: Into<Dispatch>,

Attaches the provided Subscriber to this type, returning a WithDispatch wrapper. Read more
Source§

fn with_current_subscriber(self) -> WithDispatch<Self> ⓘ

Attaches the current default Subscriber to this type, returning a WithDispatch wrapper. Read more