@obsunified/collector 2.2.0 → 2.2.1

package/dist/index.d.ts

@@ -154,6 +154,8 @@ interface HousekeepingSummary {
     ontology_nodes?: number;
     change_events?: number;
     incident_investigations?: number;
+    /** `leased` incidents whose lease expired and were returned to `open` (RFC 0013 § 11). */
+    incident_leases_reaped?: number;
     investigation_findings?: number;
     remediation_proposals?: number;
 }
@@ -1116,7 +1118,9 @@ declare function computeChangeBlastRadius(db: SqlDb, projectId: string, changeId
     minComparableSample?: number;
 }): Promise<ChangeBlastRadiusResult | null>;
 
-declare function recordInvestigationFinding(db: SqlDb, projectId: string, input: InvestigationFindingInput): Promise<InvestigationFinding>;
+declare function recordInvestigationFinding(db: SqlDb, projectId: string, input: InvestigationFindingInput, options?: {
+    logger?: Logger;
+}): Promise<InvestigationFinding>;
 declare function getInvestigationFinding(db: SqlDb, projectId: string, findingId: string, environment?: string): Promise<InvestigationFinding | null>;
 declare function searchInvestigationFindings(db: SqlDb, options: {
     projectId: string;
@@ -1274,13 +1278,30 @@ declare function acquireIncidentLease(db: SqlDb, projectId: string, incidentId:
 }>;
 /**
  * SSRF-safe, HMAC-SHA256 signed AutoSRE incident webhook outbox delivery
- * (RFC 0013 § 6.A).
+ * (RFC 0013 § 6.A). Mirrors the alert webhook delivery: literal host check,
+ * DNS classification (fail closed on private/unresolvable), no redirects and
+ * a 5s timeout.
+ *
+ * Grant handoff (RFC 0013 § 6.B): when `webhook.grant` is supplied the payload
+ * is an `autosre.remediation.grant_issued` event carrying `proposalId`,
+ * `status` and the plaintext `grantToken`. The plaintext token exists only at
+ * approve/propose time (the DB keeps its hash), so the caller that mints it
+ * must hand it to this function in the same request. Grant deliveries do not
+ * count against the incident's 5-attempt dispatch budget.
  */
 declare function dispatchIncidentWebhook(db: SqlDb, projectId: string, incidentId: string, webhook: {
     url: string;
     secret?: string;
+    environment?: string;
     allowPrivateNetworks?: boolean;
+    hostLookup?: HostLookup;
     fetchImpl?: typeof fetch;
+    logger?: Logger;
+    grant?: {
+        proposalId: string;
+        status: string;
+        grantToken: string;
+    };
 }): Promise<{
     delivered: boolean;
     status?: number;
@@ -1293,7 +1314,9 @@ declare class RemediationGateError extends Error {
 /**
  * Pre-execution remediation proposal gate (RFC 0013 § 6.B & P1 #4).
  */
-declare function proposeRemediationAction(db: SqlDb, projectId: string, input: RemediationProposalInput): Promise<RemediationProposal>;
+declare function proposeRemediationAction(db: SqlDb, projectId: string, input: RemediationProposalInput, options?: {
+    logger?: Logger;
+}): Promise<RemediationProposal>;
 declare function getRemediationProposal(db: SqlDb, projectId: string, proposalId: string, environment?: string): Promise<RemediationProposal | null>;
 declare function listRemediationProposals(db: SqlDb, options: {
     projectId: string;
@@ -1306,6 +1329,10 @@ declare function listRemediationProposals(db: SqlDb, options: {
 /**
  * Human dashboard session approval for `pending_human_approval` proposals
  * (RFC 0013 § 6.B & P1 #4). Mints a single-use `grantToken`.
+ *
+ * The status precondition lives in the UPDATE's WHERE clause (D1 and Postgres
+ * have no multi-statement transactions beyond `batch`), so two concurrent
+ * approvals cannot both mint a token: only one sees `meta.changes === 1`.
  */
 declare function approveRemediationProposal(db: SqlDb, projectId: string, proposalId: string, options: {
     environment?: string;
@@ -1319,19 +1346,31 @@ declare function rejectRemediationProposal(db: SqlDb, projectId: string, proposa
 /**
  * Single-use grant redemption with optimistic `expectedCurrentState` verification
  * against `ontology_nodes` (RFC 0013 § 6.B & P1 #4).
+ *
+ * The status, token-hash and expiry preconditions are part of the claiming
+ * UPDATE, and the outcome is read from `meta.changes`: of N concurrent claims
+ * for one token exactly one sees `changes === 1`; the rest get a 409. The
+ * incident transition rides in the same `batch` and is guarded by an EXISTS on
+ * the winning claim, so a losing claim cannot touch the incident.
  */
 declare function claimRemediationGrant(db: SqlDb, projectId: string, proposalId: string, options: {
     environment?: string;
     grantToken: string;
     runnerId?: string;
     nowIso?: string;
+    logger?: Logger;
 }): Promise<RemediationProposal>;
 /**
  * Records remediation completion, updates `ontology_nodes` to desired state,
  * and arms the +2m/+5m/+15m post-remediation watchdog on the finding.
+ *
+ * Only the runner that holds the claim may complete, and only while the claim
+ * window is open (`now <= claim_expires_at`); otherwise 409. All follow-up
+ * writes are guarded on the completion UPDATE having applied.
  */
 declare function completeRemediationProposal(db: SqlDb, projectId: string, proposalId: string, options: {
     environment?: string;
+    runnerId?: string;
     resultingChangeId?: string | null;
     remediationAgentRunId?: string | null;
     baselineRequestRatePerMin?: number | null;
@@ -1339,6 +1378,7 @@ declare function completeRemediationProposal(db: SqlDb, projectId: string, propo
 }): Promise<RemediationProposal>;
 declare function failRemediationProposal(db: SqlDb, projectId: string, proposalId: string, options: {
     environment?: string;
+    runnerId?: string;
     failureReason: string;
     nowIso?: string;
 }): Promise<RemediationProposal>;
@@ -1349,6 +1389,8 @@ interface WatchdogEvaluationSummary {
     stillDegraded: number;
     timedOutClaims: number;
     expiredGrants: number;
+    /** Findings whose evaluation threw; logged and retried on the next tick. */
+    failed: number;
 }
 /**
  * Post-Remediation Watchdog & Lease/Claim Timeout Evaluator (RFC 0013 § 5.B, § 6.B & P1 #3).
@@ -1357,10 +1399,18 @@ interface WatchdogEvaluationSummary {
  *  - `recovery_status`: whether the service recovered under comparable traffic (`N >= minComparableRequests`)
  *  - `causal_verification_status`: `"operationally_corroborated"` (never promoted to `"counterfactual_verified"`
  *    by telemetry recovery alone!).
+ *
+ * Checkpoints are scheduled at +2m/+5m/+15m measured from the remediation's
+ * completion time (not from whenever a tick happened to run), and a finding
+ * only becomes `recovered` after two consecutive healthy checkpoints that each
+ * had sufficient traffic. Pass `projectId` to scope the sweep to one tenant
+ * (HTTP path); the cron path omits it and sweeps every project.
  */
 declare function evaluateRemediationWatchdogs(db: SqlDb, options?: {
     nowIso?: string;
     minComparableRequests?: number;
+    projectId?: string;
+    logger?: Logger;
 }): Promise<WatchdogEvaluationSummary>;
 
 /**