Struct ProcStatus 

pub struct ProcStatus {
    pub state: ProcState,
    pub ready: Option<bool>,
    pub pid: Option<u32>,
    pub uptime_secs: Option<u64>,
    pub version: Option<String>,
    pub detail: Option<String>,
    pub endpoints: BTreeMap<String, String>,
    pub metrics: BTreeMap<String, f64>,
}

A workload’s self-description. Only Self::state is required.

@yah:relay(R866, “Deployed-credential drift: apply-time value fingerprints reported back over the ProcStatus rail”) @yah:at(2026-09-05T08:47:33Z) @yah:status(open) @yah:assignee(agent:bundle-anthropic-ashguard) @arch:see(.yah/docs/working/W337-credential-health-and-rotation.md) @yah:depends_on(R556-F6) @yah:next(“THE PROBLEM. A credential used by a cloud workload is injected at yah cloud apply time from the local vault via resolve_serve_env (grep fn resolve_serve_env in app/yah/cli/src/cloud.rs — :7979 on 2026-09-05, but it has moved three times). That splits rot in two: (a) the vault copy rotted, which a local yah keys doctor probe already catches, and (b) THE DEPLOYED COPY DRIFTED FROM THE VAULT COPY — someone rotated the vault and never re-applied, or applied and the workload never restarted. In case (b) the vault probe is GREEN and the service is DOWN. This relay is (b) and only (b).”) @yah:next(“SCOPE THE DELTA. The apply-time PRESENCE half is already built and must NOT be rebuilt: resolve_serve_env is FATAL on an empty resolution, with an error naming the slot and the yah keys set fix, so a serve process is never forked with a blank credential. What it cannot catch is a value that resolves fine and is DEAD, or one that DRIFTED after apply.”) @yah:next(“SHAPE. At apply time store a SALTED hash of each injected value; have the workload report the hash of what it is actually running with; compare. That detects (b) without moving a secret anywhere. HARD CONSTRAINT: no secret value, and no secret LENGTH, may appear in the fingerprint path, in the reported status document, or in any log.”) @yah:next(“USE THE EXISTING ProcStatus RAIL — do not build a bespoke per-workload health endpoint. Right here in this file: workloads publish a ProcStatus self-description document (:155), fetched by fetch_status (:234) over a ControlEndpoint that is Socket(PathBuf) OR Http(String) (:211, doc: \"for a process that already serves HTTP, including every cloud-tier workload\"), at conventional path DEFAULT_HTTP_PATH = \"/_yah/status\" (:111, overridable per-workload via [process.control] http_path). Producer side is a published two-line helper crate, oss/kamaji/crates/procctl (serve_env / serve_at / ControlServer, lib.rs:78). Riding this rail makes the feature work for EVERY procctl-conforming workload rather than mesofact alone.”) @yah:next(“THE ONE GAP, and it is the first edit: ProcStatus has NO field a hex fingerprint fits. metrics is BTreeMap<String, f64> (numeric only), endpoints is addresses, detail is documented as ONE HUMAN LINE. Add a string-valued field — env_fingerprint, or a general free-form labels: BTreeMap<String, String>; THAT CHOICE IS A NAMING CALL, make it deliberately — carrying #[serde(default)] so workloads built before this change keep deserializing.”) @yah:next(“SURFACE IT IN THE EXISTING TABLE, not a new command. yah cloud mirror-status already does declared-vs-observed comparison for replicas and already has a –drift filter: handle_mirror_status at app/yah/cli/src/cloud.rs:11862, row type MirrorStatusRow at :6774, –drift applied at :11929. Add a row type; do not add a command. (Older prose cites :9684 for this — that was never a mirror-status line.)”) @yah:gotcha(“THERE IS NO LIVE CONSUMER YET, AND THAT GATES THIS RELAY — it is why depends_on(R556-F6) is set. Measured 2026-09-05: ZERO uncommented vault: declarations in any tracked TOML. All four hits are commented out — .yah/services/yah-analytics/mirrors/cloud.toml:616-618 (the #! cut-over block) and .yah/qed/gha-actions.toml:25 — so today there is nothing for an apply-time hash to hash and the reporting half would be dead code on both ends. Step (4) of R556-F6’s cut-over uncomments that block and creates the first live declaration. Confirm the field shape against what R556-T12 actually SHIPPED, not against the comment: the comment predates it and names cloudflare-r2-endpoint, a slot the vault does not have.”) @yah:gotcha(“RIPGREP TRAP that has already cost two sessions a false reading: rg skips hidden directories by default, and every vault: declaration in this tree lives under .yah/. So rg '=\\s*\"vault:\"' --glob '*.toml' WITHOUT –hidden returns clean over a tree that is not clean. Always pass –hidden when re-measuring the trigger.”) @yah:gotcha(“BLAST RADIUS IS WIDER THAN IT LOOKS — weigh it before starting. This touches oss/yubaba AND oss/kamaji, which are INDEPENDENT Cargo workspaces excluded from the yah root workspace (so no workspace = true inheritance from the root inside them), and procctl is a crates.io PUBLISH surface, meaning a ProcStatus field change is a wire-format change for external consumers (noisetable, in the entambi repo, is named as one in this file’s own R-notes). #[serde(default)] on the new field is not optional politeness; it is what keeps already-deployed workloads deserializing.”) @yah:gotcha(“HISTORY: this was R856-F8, deferred unbuilt across three sessions (2026-09-03/04/05) because the trigger never fired. Operator decision 2026-09-05 re-filed it here as its own relay rather than holding R856 open — R856’s remaining work is vault-local and finished, while this spans two oss workspaces and a publish surface. R856-F8 is archived; its design record is W337 §5.”) @yah:verify(“Rotating a vault slot without re-applying shows as drift in yah cloud mirror-status --drift, while the local yah keys doctor probe still reports Valid on the same slot”) @yah:verify(“No secret value and no secret LENGTH appears in the fingerprint path, in the reported ProcStatus document, or in any log”) @yah:verify(“A workload built before the new ProcStatus field still deserializes (pin it with a test that feeds the pre-change JSON through serde), so a partial fleet roll cannot break status reporting”) @yah:verify(“cargo test –manifest-path oss/yubaba/Cargo.toml -p yah-cloud proc_control # the rail’s existing serde round-trip tests still pass (see proc_control.rs:493)”)

Fields

state: ProcState

Lifecycle state, in kamaji’s vocabulary.

ready: Option<bool>

Redundant convenience mirror of state == running, accepted from producers that emit it. Never trusted over state — a document claiming {"state":"starting","ready":true} is a producer bug, and believing the optimistic half of it is how a supervisor reports a half-booted process as up.

pid: Option<u32>

Process id, when the process knows and cares to say.

uptime_secs: Option<u64>

Seconds since the process considered itself started.

version: Option<String>

Build/version string, for an operator staring at two of these.

detail: Option<String>

One human line elaborating on state — “replaying WAL 3/7”, “waiting for GPU”. This is the field that replaces log-grepping.

endpoints: BTreeMap<String, String>

Named addresses the process serves — {"http":"http://127.0.0.1:4325"}. A portless process may legitimately name a non-URL surface here.

metrics: BTreeMap<String, f64>

Numeric gauges the process wants surfaced. Free-form on purpose: this is a status channel, not a metrics pipeline.

Implementations

impl ProcStatus

pub fn is_ready(&self) -> bool

Ready iff the state says so. See Self::ready for why the producer-supplied boolean does not get a vote.

pub fn summary(&self) -> String

One-line rendering for an operator-facing note or log line.

Trait Implementations

impl Clone for ProcStatus

fn clone(&self) -> ProcStatus

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ProcStatus

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for ProcStatus

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl PartialEq for ProcStatus

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

Equality operator ==.

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

Inequality operator !=.

impl Serialize for ProcStatus

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 ProcStatus