@obsunified/collector 2.1.1 → 2.2.0

package/src/lib/eval-cases-store.ts

@@ -76,6 +76,10 @@ export interface EvalRunCandidate {
 	modelProvider: string | null;
 	model: string | null;
 	modelVersion: string | null;
+	environment?: string | null;
+	serviceVersion?: string | null;
+	commitSha?: string | null;
+	configDigest?: string | null;
 }
 
 export interface EvalRunSourceCase {
@@ -99,6 +103,12 @@ export interface EvalRun {
 	evalCaseId: string | null;
 	status: EvalRunStatus;
 	candidate: EvalRunCandidate;
+	environment?: string;
+	serviceVersion?: string | null;
+	commitSha?: string | null;
+	configDigest?: string | null;
+	inputPayloadHash?: string | null;
+	failureFingerprint?: string | null;
 	startedAt: string | null;
 	endedAt: string | null;
 	totalCount: number;
@@ -117,6 +127,12 @@ export interface EvalRunInput {
 	evalCaseId?: string | null;
 	status?: EvalRunStatus;
 	candidate?: Partial<EvalRunCandidate>;
+	environment?: string | null;
+	serviceVersion?: string | null;
+	commitSha?: string | null;
+	configDigest?: string | null;
+	inputPayloadHash?: string | null;
+	failureFingerprint?: string | null;
 	startedAt?: string | null;
 	endedAt?: string | null;
 	totalCount?: number;
@@ -144,6 +160,12 @@ interface EvalRunRow {
 	candidate_model_provider: string | null;
 	candidate_model: string | null;
 	candidate_model_version: string | null;
+	environment?: string | null;
+	service_version?: string | null;
+	commit_sha?: string | null;
+	config_digest?: string | null;
+	input_payload_hash?: string | null;
+	failure_fingerprint?: string | null;
 	status: string;
 	started_at: string | null;
 	ended_at: string | null;
@@ -412,7 +434,17 @@ const rowToEvalRun = (row: EvalRunRow): EvalRun => {
 			modelProvider: row.candidate_model_provider ?? null,
 			model: row.candidate_model ?? null,
 			modelVersion: row.candidate_model_version ?? null,
+			environment: row.environment ?? "default",
+			serviceVersion: row.service_version ?? null,
+			commitSha: row.commit_sha ?? null,
+			configDigest: row.config_digest ?? null,
 		},
+		environment: row.environment ?? "default",
+		serviceVersion: row.service_version ?? null,
+		commitSha: row.commit_sha ?? null,
+		configDigest: row.config_digest ?? null,
+		inputPayloadHash: row.input_payload_hash ?? null,
+		failureFingerprint: row.failure_fingerprint ?? null,
 		startedAt: row.started_at ?? null,
 		endedAt: row.ended_at ?? null,
 		totalCount: row.total_count ?? 0,
@@ -560,6 +592,16 @@ export class EvalCasesStore {
 		}
 
 		const candidate = input.candidate ?? {};
+		const environment =
+			input.environment?.trim() || candidate.environment?.trim() || "default";
+		const serviceVersion =
+			input.serviceVersion?.trim() || candidate.serviceVersion?.trim() || null;
+		const commitSha =
+			input.commitSha?.trim() || candidate.commitSha?.trim() || null;
+		const configDigest =
+			input.configDigest?.trim() || candidate.configDigest?.trim() || null;
+		const inputPayloadHash = input.inputPayloadHash?.trim() || null;
+		const failureFingerprint = input.failureFingerprint?.trim() || null;
 		const totalCount = nonNegativeInteger(input.totalCount);
 		const passCount = nonNegativeInteger(input.passCount);
 		const failCount = nonNegativeInteger(input.failCount);
@@ -580,10 +622,12 @@ export class EvalCasesStore {
 					candidate_agent_id, candidate_agent_version,
 					candidate_prompt_id, candidate_prompt_version,
 					candidate_model_provider, candidate_model, candidate_model_version,
+					environment, service_version, commit_sha, config_digest,
+					input_payload_hash, failure_fingerprint,
 					status, started_at, ended_at,
 					total_count, pass_count, fail_count, average_score,
 					metadata_json, created_at
-				) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)`,
+				) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)`,
 			)
 			.bind(
 				id,
@@ -596,6 +640,12 @@ export class EvalCasesStore {
 				candidate.modelProvider ?? null,
 				candidate.model ?? null,
 				candidate.modelVersion ?? null,
+				environment,
+				serviceVersion,
+				commitSha,
+				configDigest,
+				inputPayloadHash,
+				failureFingerprint,
 				status,
 				startedAt,
 				endedAt,
@@ -679,6 +729,14 @@ export class EvalCasesStore {
 					average_score = (
 						SELECT AVG(score) FROM eval_case_results
 						WHERE project_id = ? AND run_id = ? AND score IS NOT NULL
+					),
+					input_payload_hash = COALESCE(
+						input_payload_hash,
+						(SELECT input_payload_hash FROM eval_case_results WHERE project_id = ? AND run_id = ? AND input_payload_hash IS NOT NULL LIMIT 1)
+					),
+					failure_fingerprint = COALESCE(
+						failure_fingerprint,
+						(SELECT failure_fingerprint FROM eval_case_results WHERE project_id = ? AND run_id = ? AND failure_fingerprint IS NOT NULL LIMIT 1)
 					)
 				WHERE project_id = ? AND id = ?`,
 			)
@@ -693,6 +751,10 @@ export class EvalCasesStore {
 				id,
 				projectId,
 				id,
+				projectId,
+				id,
+				projectId,
+				id,
 			)
 			.run();
 	}

package/src/lib/evidence-references.ts

@@ -146,6 +146,12 @@ export const evidenceRouteFor = (
 			return traceRoute(entityId);
 		case "docs":
 			return entityId.startsWith("#/") ? entityId : `#/docs/${entityId}`;
+		case "change":
+			return `#/changes/${encodeURIComponent(entityId)}`;
+		case "finding":
+			return `#/findings/${encodeURIComponent(entityId)}`;
+		case "incident":
+			return `#/incidents/${encodeURIComponent(entityId)}`;
 	}
 	const exhaustive: never = entityKind;
 	return exhaustive;

package/src/lib/housekeeping.test.ts

@@ -58,6 +58,12 @@ describe("purgeSecondaryRows", () => {
 			metric_series: 1,
 			evidence_refs: 1,
 			evidence_ref_expansions: 1,
+			ontology_edges: 0,
+			ontology_nodes: 0,
+			change_events: 0,
+			incident_investigations: 0,
+			remediation_proposals: 0,
+			investigation_findings: 0,
 		});
 		const ids = (sql: string) =>
 			(raw.prepare(sql).all() as Array<{ id: string }>).map((r) => r.id).sort();
@@ -70,4 +76,78 @@ describe("purgeSecondaryRows", () => {
 		expect(ids("SELECT id FROM evidence_ref_expansions")).toEqual(["e2"]);
 		db.close();
 	});
+
+	it("enforces RFC 0013 AutoSRE retention windows while preserving counterfactual_verified and operationally_corroborated findings", async () => {
+		const db = createMigratedSqliteDb();
+		const raw = db.raw;
+		raw.exec(`
+			INSERT INTO projects (id, name, slug, created_at) VALUES ('alpha', 'A', 'a', '${iso(3000)}');
+		`);
+		// Ontology edges/nodes (30d = 720h)
+		raw
+			.prepare(
+				`INSERT INTO ontology_nodes (id, project_id, environment, kind, name, first_seen_at, last_seen_at) VALUES (?, 'alpha', 'prod', 'service', ?, ?, ?)`,
+			)
+			.run("n-old", "old-svc", iso(800), iso(750));
+		raw
+			.prepare(
+				`INSERT INTO ontology_nodes (id, project_id, environment, kind, name, first_seen_at, last_seen_at) VALUES (?, 'alpha', 'prod', 'service', ?, ?, ?)`,
+			)
+			.run("n-fresh", "fresh-svc", iso(100), iso(10));
+		raw
+			.prepare(
+				`INSERT INTO ontology_edges (id, project_id, environment, source_node_id, predicate, target_node_id, relationship_certainty, first_seen_at, last_seen_at) VALUES (?, 'alpha', 'prod', 'n-old', 'calls', 'n-fresh', 'exact', ?, ?)`,
+			)
+			.run("edge-old", iso(800), iso(750));
+
+		// Change events (90d = 2160h)
+		raw
+			.prepare(
+				`INSERT INTO change_events (id, project_id, environment, kind, source, target_node_id, summary, occurred_at, received_at) VALUES (?, 'alpha', 'prod', 'deploy', 'ci_webhook', 'n-fresh', 'old deploy', ?, ?)`,
+			)
+			.run("chg-old", iso(2200), iso(2200));
+		raw
+			.prepare(
+				`INSERT INTO change_events (id, project_id, environment, kind, source, target_node_id, summary, occurred_at, received_at) VALUES (?, 'alpha', 'prod', 'deploy', 'ci_webhook', 'n-fresh', 'recent deploy', ?, ?)`,
+			)
+			.run("chg-recent", iso(100), iso(100));
+
+		// Findings (90d = 2160h, unless counterfactual_verified or operationally_corroborated)
+		const insFinding = raw.prepare(
+			`INSERT INTO investigation_findings (id, project_id, environment, fingerprint, fingerprint_input_json, title, summary, causal_proof_tier, causal_verification_status, confidence, confidence_breakdown_json, symptom_entity_kind, symptom_entity_id, fault_locus_entity_kind, fault_locus_entity_id, causal_chain_json, cited_evidence_ids_json, actor_type, actor_id, created_at, updated_at) VALUES (?, 'alpha', 'prod', 'fp_1', '{}', 't', 's', 'corroborated_candidate', ?, 0.8, '{}', 'trace', 't1', 'span', 's1', '[]', '["t1"]', 'agent', 'autosre', ?, ?)`,
+		);
+		insFinding.run("fnd-old-unverified", "unverified", iso(2200), iso(2200));
+		insFinding.run(
+			"fnd-old-verified",
+			"counterfactual_verified",
+			iso(2200),
+			iso(2200),
+		);
+		insFinding.run(
+			"fnd-old-corroborated",
+			"operationally_corroborated",
+			iso(2200),
+			iso(2200),
+		);
+
+		const summary = await purgeSecondaryRows(db, {
+			retentionHours: 72,
+			now: () => NOW,
+		});
+		expect(summary.ontology_edges).toBe(1);
+		expect(summary.ontology_nodes).toBe(1);
+		expect(summary.change_events).toBe(1);
+		expect(summary.investigation_findings).toBe(1);
+
+		const remainingFindings = (
+			raw
+				.prepare(`SELECT id FROM investigation_findings ORDER BY id ASC`)
+				.all() as Array<{ id: string }>
+		).map((r) => r.id);
+		expect(remainingFindings).toEqual([
+			"fnd-old-corroborated",
+			"fnd-old-verified",
+		]);
+		db.close();
+	});
 });

package/src/lib/housekeeping.ts

@@ -13,6 +13,7 @@
  *    the signal they reference); expansions follow their ref.
  */
 
+import { evaluateRemediationWatchdogs } from "./autosre-remediation-gate";
 import type { SqlDb } from "./sql-db";
 
 export const ALERT_EVALUATIONS_MAX_AGE_HOURS = 720;
@@ -24,6 +25,12 @@ export interface HousekeepingSummary {
 	metric_series: number;
 	evidence_refs: number;
 	evidence_ref_expansions: number;
+	ontology_edges?: number;
+	ontology_nodes?: number;
+	change_events?: number;
+	incident_investigations?: number;
+	investigation_findings?: number;
+	remediation_proposals?: number;
 }
 
 export interface HousekeepingOptions {
@@ -42,6 +49,7 @@ export async function purgeSecondaryRows(
 	const nowMs = options.now?.() ?? Date.now();
 	const isoBefore = (hours: number) =>
 		new Date(nowMs - hours * 3_600_000).toISOString();
+	const nowIso = isoBefore(0);
 
 	const alertEvaluations = await db
 		.prepare(`DELETE FROM alert_evaluations WHERE evaluated_at < ?`)
@@ -71,7 +79,7 @@ export async function purgeSecondaryRows(
 			`DELETE FROM evidence_retrieval_refs
 			 WHERE expires_at IS NOT NULL AND expires_at < ?`,
 		)
-		.bind(isoBefore(0))
+		.bind(nowIso)
 		.run();
 	const expansions = await db
 		.prepare(
@@ -83,11 +91,103 @@ export async function purgeSecondaryRows(
 		)
 		.run();
 
+	// RFC 0013 — AutoSRE ontology, change events, incident investigations,
+	// investigation findings, and remediation proposals retention + watchdog.
+	const thirtyDaysAgoIso = isoBefore(30 * 24);
+	const ninetyDaysAgoIso = isoBefore(90 * 24);
+	let ontologyEdgesPurged = 0;
+	let ontologyNodesPurged = 0;
+	let changeEventsPurged = 0;
+	let incidentInvestigationsPurged = 0;
+	let investigationFindingsPurged = 0;
+	let remediationProposalsPurged = 0;
+	try {
+		const edgesRes = await db
+			.prepare(`DELETE FROM ontology_edges WHERE last_seen_at < ?`)
+			.bind(thirtyDaysAgoIso)
+			.run();
+		ontologyEdgesPurged = changes(edgesRes);
+
+		const nodesRes = await db
+			.prepare(
+				`DELETE FROM ontology_nodes
+				 WHERE last_seen_at < ?
+				   AND NOT EXISTS (
+				     SELECT 1 FROM ontology_edges e
+				     WHERE e.project_id = ontology_nodes.project_id
+				       AND e.environment = ontology_nodes.environment
+				       AND (e.source_node_id = ontology_nodes.id OR e.target_node_id = ontology_nodes.id)
+				   )`,
+			)
+			.bind(thirtyDaysAgoIso)
+			.run();
+		ontologyNodesPurged = changes(nodesRes);
+
+		const proposalsRes = await db
+			.prepare(
+				`DELETE FROM remediation_proposals
+				 WHERE (expires_at IS NOT NULL AND expires_at < ?)
+				    OR (expires_at IS NULL AND created_at < ?)`,
+			)
+			.bind(nowIso, ninetyDaysAgoIso)
+			.run();
+		remediationProposalsPurged = changes(proposalsRes);
+
+		const findingsRes = await db
+			.prepare(
+				`DELETE FROM investigation_findings
+				 WHERE causal_verification_status NOT IN ('counterfactual_verified', 'operationally_corroborated')
+				   AND ((expires_at IS NOT NULL AND expires_at < ?)
+				     OR (expires_at IS NULL AND created_at < ?))`,
+			)
+			.bind(nowIso, ninetyDaysAgoIso)
+			.run();
+		investigationFindingsPurged = changes(findingsRes);
+
+		// Preserve change_events referenced by active investigation_findings
+		const changesRes = await db
+			.prepare(
+				`DELETE FROM change_events
+				 WHERE ((expires_at IS NOT NULL AND expires_at < ?)
+				     OR (expires_at IS NULL AND occurred_at < ?))
+				   AND NOT EXISTS (
+				     SELECT 1 FROM investigation_findings f
+				     WHERE f.project_id = change_events.project_id
+				       AND f.environment = change_events.environment
+				       AND f.contributing_change_id = change_events.id
+				   )`,
+			)
+			.bind(nowIso, ninetyDaysAgoIso)
+			.run();
+		changeEventsPurged = changes(changesRes);
+
+		const incidentsRes = await db
+			.prepare(
+				`DELETE FROM incident_investigations
+				 WHERE status IN ('resolved', 'escalated')
+				   AND ((expires_at IS NOT NULL AND expires_at < ?)
+				     OR (expires_at IS NULL AND updated_at < ?))`,
+			)
+			.bind(nowIso, ninetyDaysAgoIso)
+			.run();
+		incidentInvestigationsPurged = changes(incidentsRes);
+
+		await evaluateRemediationWatchdogs(db, { nowIso });
+	} catch {
+		// Gracefully ignore when running against partial test schemas without 050
+	}
+
 	return {
 		alert_evaluations: changes(alertEvaluations),
 		ask_evidence_events: changes(askEvents),
 		metric_series: changes(series),
 		evidence_refs: changes(refs),
 		evidence_ref_expansions: changes(expansions),
+		ontology_edges: ontologyEdgesPurged,
+		ontology_nodes: ontologyNodesPurged,
+		change_events: changeEventsPurged,
+		incident_investigations: incidentInvestigationsPurged,
+		investigation_findings: investigationFindingsPurged,
+		remediation_proposals: remediationProposalsPurged,
 	};
 }

package/src/lib/store/ingest.ts

@@ -1,4 +1,8 @@
 import type { StoredSpan } from "@obsunified/types";
+import {
+	extractSpanLineageAndTopology,
+	upsertOntologyFromSpans,
+} from "../autosre-ontology";
 import type { SqlDb } from "../sql-db";
 
 export async function ingestTelemetrySpans(
@@ -7,10 +11,29 @@ export async function ingestTelemetrySpans(
 ): Promise<{ inserted: number; traceCount: number }> {
 	if (spans.length === 0) return { inserted: 0, traceCount: 0 };
 
-	const statements = spans.map((span) => {
+	const enrichedSpans = spans.map((span) => {
 		if (!span.projectId)
 			throw new Error("TelemetryStore.ingest: span.projectId is required");
-		return db
+		if (
+			span.environment !== undefined &&
+			span.serviceVersion !== undefined &&
+			span.targetNodeId !== undefined
+		) {
+			return span;
+		}
+		const lineage = extractSpanLineageAndTopology(span);
+		return {
+			...span,
+			environment: span.environment ?? lineage.environment,
+			serviceVersion: span.serviceVersion ?? lineage.serviceVersion,
+			commitSha: span.commitSha ?? lineage.commitSha,
+			targetNodeId: span.targetNodeId ?? lineage.targetNode?.id ?? null,
+			errorSignature: span.errorSignature ?? lineage.errorSignature,
+		};
+	});
+
+	const statements = enrichedSpans.map((span) =>
+		db
 			.prepare(`
         INSERT OR IGNORE INTO telemetry_spans (
           project_id, trace_id, span_id, parent_span_id, trace_state,
@@ -19,8 +42,9 @@ export async function ingestTelemetrySpans(
           attributes_json, dropped_attributes_count,
           resource_attributes_json, events_json, dropped_events_count,
           links_json, dropped_links_count, received_at, expires_at,
-          session_id, interaction_id, telemetry_sdk_name
-        ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+          session_id, interaction_id, telemetry_sdk_name,
+          environment, service_version, commit_sha, target_node_id, error_signature
+        ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
       `)
 			.bind(
 				span.projectId,
@@ -50,10 +74,74 @@ export async function ingestTelemetrySpans(
 				span.sessionId ?? null,
 				span.interactionId ?? null,
 				span.telemetrySdkName ?? null,
+				span.environment ?? "default",
+				span.serviceVersion ?? null,
+				span.commitSha ?? null,
+				span.targetNodeId ?? null,
+				span.errorSignature ?? null,
+			),
+	);
+
+	try {
+		await db.batch(statements);
+	} catch (err) {
+		const msg = err instanceof Error ? err.message : String(err);
+		if (msg.includes("no such column: environment")) {
+			// Fallback for unit tests running against pre-050 partial schemas
+			const legacyStatements = enrichedSpans.map((span) =>
+				db
+					.prepare(`
+            INSERT OR IGNORE INTO telemetry_spans (
+              project_id, trace_id, span_id, parent_span_id, trace_state,
+              service_name, scope_name, scope_version, span_name, span_kind,
+              status_code, status_message, start_time, end_time, duration_ms,
+              attributes_json, dropped_attributes_count,
+              resource_attributes_json, events_json, dropped_events_count,
+              links_json, dropped_links_count, received_at, expires_at,
+              session_id, interaction_id, telemetry_sdk_name
+            ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+          `)
+					.bind(
+						span.projectId,
+						span.traceId,
+						span.spanId,
+						span.parentSpanId,
+						span.traceState,
+						span.serviceName,
+						span.scopeName,
+						span.scopeVersion,
+						span.spanName,
+						span.spanKind,
+						span.statusCode,
+						span.statusMessage,
+						span.startTime,
+						span.endTime,
+						span.durationMs,
+						span.attributesJson,
+						span.droppedAttributesCount,
+						span.resourceAttributesJson,
+						span.eventsJson,
+						span.droppedEventsCount,
+						span.linksJson,
+						span.droppedLinksCount,
+						span.receivedAt,
+						span.expiresAt,
+						span.sessionId ?? null,
+						span.interactionId ?? null,
+						span.telemetrySdkName ?? null,
+					),
 			);
-	});
+			await db.batch(legacyStatements);
+		} else {
+			throw err;
+		}
+	}
 
-	await db.batch(statements);
+	try {
+		await upsertOntologyFromSpans(db, enrichedSpans);
+	} catch {
+		// Non-fatal on partial-schema unit test DBs
+	}
 
 	// Trace instrumentation gaps are computed lazily at read time (see
 	// `getTelemetryTraceGaps`), not materialized on ingest. This keeps the