Enum Event 

pub enum Event {
    RunStarted {
        feature: String,
        tasks: Vec<PlannedTask>,
    },
    TaskStarted {
        feature: String,
        task: String,
        title: String,
    },
    Stage {
        task: String,
        stage: Stage,
        state: StageState,
        detail: Option<String>,
        elapsed_ms: Option<u64>,
        tail: Vec<String>,
    },
    Resumed {
        task: String,
        stage: Option<Stage>,
        last_call: Option<String>,
    },
    ApprovalRequested {
        task: String,
        call: String,
        tool: String,
        what: String,
        class: String,
        object: String,
    },
    ApprovalDecided {
        task: String,
        call: String,
        decision: String,
        scope: Option<String>,
        by: String,
        how: String,
        class: String,
        object: String,
    },
    ToolCall {
        task: String,
        turn: u32,
        tool: String,
        target: String,
        decision: String,
        ok: Option<bool>,
        tail: Vec<String>,
    },
    Usage {
        task: String,
        input_tokens: u64,
        output_tokens: u64,
        cache_read_tokens: u64,
        cache_write_tokens: u64,
        usd: Option<f64>,
    },
    TaskFinished {
        task: String,
        outcome: String,
        message: String,
        run: Option<String>,
        input_tokens: u64,
        output_tokens: u64,
        cache_read_tokens: u64,
        cache_write_tokens: u64,
        usd: Option<f64>,
        regression: Option<Regression>,
    },
    CollectionChecked {
        task: String,
        files: Vec<CollectionFile>,
    },
    RegressionBisected {
        task: String,
        runs: u32,
        result: Option<String>,
    },
    AgentStarted {
        task: String,
        agent: String,
        mode: String,
        decides_alone: bool,
        known: bool,
    },
    AgentUnasked {
        task: String,
        calls: u32,
    },
    AgentPlan {
        task: String,
        entries: Vec<PlanEntry>,
    },
    Said {
        task: String,
        turn: u32,
        text: String,
    },
    Reasoning {
        task: String,
        turn: u32,
        text: String,
    },
    Notice {
        text: String,
    },
    Bug {
        task: String,
        step: BugEvent,
    },
    Oracle {
        step: OracleEvent,
    },
    PlanningStopped {
        unit: Option<String>,
        kind: String,
        next_skill: Option<String>,
    },
    LadderOffered {
        unit: String,
        task: String,
        signals: Vec<String>,
    },
    UnitPromoted {
        unit: String,
        from_unit: String,
        requirement: String,
        task: String,
        requirement_sha256: String,
        conversation_sha256: Vec<String>,
        signals: Vec<String>,
        by: String,
        via: String,
    },
    Project {
        step: ProjectEvent,
    },
    AskStep {
        conversation: String,
        turn: u32,
        tool: String,
        target: String,
        read: Option<String>,
        refused: bool,
    },
    AskText {
        conversation: String,
        turn: u32,
        delta: String,
    },
    AskTurnEnded {
        tag: String,
        output: Box<Value>,
    },
    AskDone {
        report: Box<AskReport>,
    },
}

Something a run did.

Variants

RunStarted

A feature run begins: its tasks, as the graph has them (editors show them as a plan).

Fields

feature: String

Feature directory name.

tasks: Vec<PlannedTask>

Every task, in document order.

TaskStarted

A task attempt run begins.

Fields

feature: String

Feature directory name.

task: String

Task ID.

title: String

Task title.

Stage

A stage of the task’s pipeline began, ended or does not apply.

Fields

task: String

Task ID.

stage: Stage

Which stage.

state: StageState

Where it is.

detail: Option<String>

1/2 · sample 2/3 for an attempt, a short commit id, a reason.

elapsed_ms: Option<u64>

How long it took, once it ended.

tail: Vec<String>

The last lines of a failed check (at most five).

Resumed

A run picks up an attempt a killed process left behind.

Fields

task: String

Task ID.

stage: Option<Stage>

The stage it was in, when known.

last_call: Option<String>

The last tool call, when known.

ApprovalRequested

A call waits for a person: notify them.

Fields

task: String

Task ID.

call: String

The model’s call id.

tool: String

The tool.

what: String

What waits, e.g. cargo add x (no control characters, ≤ 60 chars).

class: String

The class of the object a decision covers: write, command, network or mcp (empty: none, the call is decided once).

object: String

The object’s short form: the path, argv#<12 hex>, <tool>@<12 hex> — no argv beyond what what shows.

ApprovalDecided

A person decided a call inline.

Fields

task: String

Task ID.

call: String

The model’s call id.

decision: String

approve or reject.

scope: Option<String>

once, task or run, for an approval.

by: String

Who decided.

how: String

How that is known: asserted, connected (or expired).

class: String

The object’s class, as for Event::ApprovalRequested.

object: String

The object’s short form.

ToolCall

A tool call the model asked for, with the policy’s decision.

Fields

task: String

Task ID.

turn: u32

Model turn (1-based) within the attempt.

tool: String

The tool.

target: String

Its target: a path, a command, a pattern.

decision: String

allow, deny, approval, unknown, stop.

ok: Option<bool>

Whether an allowed call succeeded.

tail: Vec<String>

The last lines of a failed command (at most five).

Usage

Tokens so far in the current attempt.

Fields

task: String

Task ID.

input_tokens: u64

Input tokens.

output_tokens: u64

Output tokens.

cache_read_tokens: u64

Input tokens read from the cache.

cache_write_tokens: u64

Input tokens written to the cache.

usd: Option<f64>

USD, when priced.

TaskFinished

The task run ended.

Fields

task: String

Task ID.

outcome: String

verified, regression_only, failed, blocked, …

message: String

The run’s message.

run: Option<String>

The attempt’s agentplane run, when one ran.

input_tokens: u64

Input tokens, all attempts and the acceptance call.

output_tokens: u64

Output tokens.

cache_read_tokens: u64

Input tokens read from the cache.

cache_write_tokens: u64

Input tokens written to the cache.

usd: Option<f64>

USD, when priced.

regression: Option<Regression>

A regression-only tick: why, worded.

CollectionChecked

The attempt’s changed test files were checked against what the run collected (collection.checked).

Fields

task: String

Task ID.

files: Vec<CollectionFile>

Each judged file: path, state, what decided it.

RegressionBisected

The bisection for an already-passing tick ended (regression.bisected).

Fields

task: String

Task ID.

runs: u32

Acceptance runs it took.

result: Option<String>

The task that made the tests pass, before_feature, or null.

AgentStarted

An external agent makes the attempt: its name, and what its permission mode is (decides_alone: calls it makes without asking never reach speccoder, so an empty Needs You proves nothing).

Fields

task: String

Task ID.

agent: String

The agent’s name.

mode: String

Its permission mode, in a sentence.

decides_alone: bool

It is in a mode that decides some calls without asking.

known: bool

A recording of its version says whether it asks (false: not known, and said so).

AgentUnasked

An external agent changed something without asking (its own settings let it): the count so far.

Fields

task: String

Task ID.

calls: u32

Calls it reported making without a request of speccoder’s.

AgentPlan

An external agent’s plan for its attempt, as it last sent it: its claim of what it will do, never evidence.

Fields

task: String

Task ID.

entries: Vec<PlanEntry>

The entries, in its order.

Said

What the model said in a turn, beside its tool calls: its account of what it is doing — a claim, never evidence. Control characters escaped; at most 2000 characters.

Fields

task: String

Task ID.

turn: u32

Model turn (1-based) within the attempt.

text: String

The words.

Reasoning

The model’s summary of its reasoning in a turn, as the provider gave it ([model] reasoning_summary): its account of why — a claim, never evidence. Control characters escaped; at most 4000 characters.

Fields

task: String

Task ID.

turn: u32

Model turn (1-based) within the attempt.

text: String

The summary.

Notice

Anything else worth a line (convergence rounds, notices).

Fields

text: String

The line.

Bug

A bug’s step (bug.* on the event stream).

Fields

task: String

The task (T003, bug:<slug>).

step: BugEvent

What happened.

Oracle

The oracle at review (oracle.* on the event stream): stage names, counts and digests — never a test’s content, never held-out material.

Fields

step: OracleEvent

What happened.

PlanningStopped

Plan through stopped (planning.stopped).

Fields

unit: Option<String>

The unit, once it exists.

kind: String

questions, review, failed, done.

next_skill: Option<String>

The skill a continuation runs next.

LadderOffered

A quick change or bug fix that grew is offered as a spec of its own (ladder.offered): the signals that hold, by id.

Fields

unit: String

Its unit.

task: String

The task.

signals: Vec<String>

S1…S4.

UnitPromoted

A unit was promoted from a quick change or bug fix (unit.promoted): ids and digests, never the words.

Fields

unit: String

The new unit.

from_unit: String

Where it came from.

requirement: String

The requirement promoted.

task: String

The task it came from.

requirement_sha256: String

The requirement’s words’ digest.

conversation_sha256: Vec<String>

The digests of the conversations carried.

signals: Vec<String>

The signals that held.

by: String

Who.

via: String

Through what.

Project

A new project’s step, its creation or its undo (scaffold.*, project.*): step ids, argv, states, digests — never the idea, never the folder’s path.

Fields

step: ProjectEvent

What happened.

AskStep

An Ask turn read something (ask.step).

Fields

conversation: String

The conversation.

turn: u32

The turn.

tool: String

The tool.

target: String

What it read, in one line.

read: Option<String>

The read id the answer may cite, when it returned something.

refused: bool

It failed or was refused.

AskText

An Ask turn’s answer as it is written (ask.text).

Fields

conversation: String

The conversation.

turn: u32

The turn.

delta: String

The next words.

AskTurnEnded

Inside the Ask plane: a turn’s raw end, which ask::turn turns into Event::AskDone; never forwarded.

Fields

tag: String

ask:<conversation>#<turn>.

output: Box<Value>

The loop’s output.

AskDone

An Ask turn ended (ask.done).

Fields

report: Box<AskReport>

The turn as shown.

Trait Implementations

impl Clone for Event

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for Event

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

Formats the value using the given formatter.

impl PartialEq for Event

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

Equality operator ==.

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

Inequality operator !=.

impl Serialize for Event

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

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for Event