Enum RunStatus 

pub enum RunStatus {
    Prep,
    Implementing,
    Judging,
    Deliberating,
    Voting,
    Reviewing,
    Gating,
    Landing,
    Merged,
    Ready,
    Stalled,
    Blocked,
    Failed,
    VerifiedNoop,
}

Where a run got to.

Variants

Prep

Worktrees being prepared.

Implementing

Candidates being implemented.

Judging

Judges ranking blind.

Deliberating

Judges deliberating after a split.

Voting

Final votes being collected privately.

Reviewing

Winner in the review + verification loop.

Gating

Gate commands running. Transient: graph::Runner::gate and graph::Runner::merge always move a run on from here, whether or not any gate commands are configured — see RunState::gate_status and SCHEMA’s doc for schema 7, which fixed a repo with an empty verify.gate stranding a review-only run in Gating forever.

Landing

Inside crate::land’s post-merge loop: watching CI, running a fix round, rebasing onto a moved base, or waiting on the owner’s merge approval. A run parked here while an approval is outstanding has handed its daemon slot back — see crate::daemon — and resumes through exactly this status, not a fresh competition.

Merged

Winner merged.

Ready

Winner passed the gate; merge was not requested.

Stalled

The judgement did not gather enough judges (e.g. rate limiting took out seats), so the verdict is not trustworthy. The run stopped and kept its work so it can be resumed or folded — it must never be confused with a healthy Ready.

Blocked

Review rounds exhausted with findings still open, or the gate failed.

Failed

The graph could not complete.

VerifiedNoop

Every candidate wrote nothing, and every one of them said why in a way that survived crate::graph::Runner’s adoption guard: a clean CLI exit, an actually-empty tree, no command left with an unconfirmed exit status, and non-empty evidence. Distinct from Failed on purpose — see SCHEMA’s doc for schema 10 — because the two read identically to an operator glancing at a card (“nothing happened”) while meaning opposite things: one is an agent that could not do the work, the other is an agent that checked and the work was already done. Settles the task through crate::queue::Task::handed_off, not crate::queue::Task::fail: a human still has to look — the claim is unverified by magi itself — and Held (not Failed-and-requeued) means nothing retries the task unattended on the same unconfirmed claim while that look is pending.

Implementations

impl RunStatus

pub fn done(self) -> bool

Is this a terminal state?

pub fn as_str(self) -> &'static str

The name this status is written and shown under, matching the snake_case serde spelling so a log line, an error message and the JSON a phone reads all say the same word.

pub fn display_label(self) -> &'static str

Label for a human-facing listing or report — the same word as Self::as_str except where the machine spelling would read harsher than the state actually is. VerifiedNoop is the one case: its own as_str exists for logs, JSON and event messages, none of which should quietly grow a second vocabulary, but a bare “verified_noop” in a report reads like an error code, not the qualified, evidence-backed claim it actually is.

pub fn resumable(self) -> bool

Can this run be carried on from where it stopped?

Everything except a finished run and a failed one. execute skips nodes already recorded, so re-entering is cheap wherever the run stopped, and the alternative is always a fresh competition against work that already exists.

• Stalled re-asks only the seats whose absence collapsed the panel, keeping the candidates that were already paid for.
• Blocked re-enters the review loop against a branch that is built.
• A non-terminal status means the run was interrupted: a parked run waiting for its upgrade, or one whose daemon was killed. This used to be excluded, which left run 4043 stuck at reviewing with the deck telling the operator it could not be resumed - the one state where resuming is the only sensible answer.

Failed does not qualify: the graph could not complete and there is no established point to continue from. Nor does a finished run, whose answer is a new competition. Nor does VerifiedNoop: every candidate already agreed nothing belongs in this worktree, and resuming would only re-ask the same question — the answer is for a human to check the evidence, not for the graph to run again.

Whether anything is already driving the run is a separate question, answered by daemon::is_working_on at the callers that need it.

Trait Implementations

impl Clone for RunStatus

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for RunStatus

impl Debug for RunStatus

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for RunStatus

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Eq for RunStatus

impl PartialEq for RunStatus

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

Equality operator ==.

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

Inequality operator !=.

impl Serialize for RunStatus

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 RunStatus