feat(miner-discovery-plane): add anonymized telemetry event schema for the hosted plane (#4301)

packages/gittensory-engine/src/index.ts

@@ -137,6 +137,15 @@ export {
   type GovernorLedgerEventType,
   type NormalizedGovernorLedgerEvent,
 } from "./governor-ledger.js";
+export {
+  MINER_TELEMETRY_EVENT_TYPES,
+  MINER_TELEMETRY_OUTCOME_BUCKETS,
+  normalizeMinerTelemetryEvent,
+  type MinerTelemetryEvent,
+  type MinerTelemetryEventType,
+  type MinerTelemetryOutcomeBucket,
+  type NormalizedMinerTelemetryEvent,
+} from "./miner-telemetry.js";
 export {
   ATTEMPT_LOG_EVENT_TYPES,
   createAttemptLogBuffer,

packages/gittensory-engine/src/miner-telemetry.ts

@@ -0,0 +1,106 @@
+// Anonymized discovery-plane telemetry event schema (pure) — #4301, Wave 2 tracker #2353 Phase 6.
+//
+// Typed event shapes for the OPTIONAL hosted discovery-index service (#4250) — which candidates a miner
+// fetched/ranked and whether a soft-claim succeeded or collided — so that shared service can be operated and
+// debugged WITHOUT ever holding source, diffs, or credentials. This mirrors governor-ledger.ts's pure
+// validate/normalize shape (fixed fail-closed vocabulary, JSON-round-trip-verified payload) and copies the
+// anonymization POSTURE of src/selfhost/orb-collector.ts (the one shipped precedent for "anonymized telemetry
+// leaving an instance"): repo/issue identifiers are HMAC hashes keyed by a per-instance secret the collector
+// never holds, and free-text-adjacent fields are collapsed to a fixed low-cardinality bucket rather than raw text.
+//
+// NEVER INCLUDED in a telemetry event (the discovery-plane analogue of orb-collector.ts:1-18's inventory): no
+// source contents, no diffs, no GitHub tokens or credentials, no full issue bodies or titles, no commit SHAs, and
+// no RAW repo/issue identifiers — only the exporter's per-instance HMAC hashes reach this shape. This module is
+// SCHEMA/TYPES ONLY: it does not export events, hash anything itself (the exporter does that at #4250's boundary),
+// or wire into an endpoint. It only defines and validates the on-the-wire contract.
+
+/** Immutable discovery-plane telemetry event vocabulary — an unknown value fails closed before it is recorded. */
+export const MINER_TELEMETRY_EVENT_TYPES = Object.freeze([
+  "query_issued",
+  "candidates_returned",
+  "soft_claim_attempted",
+  "soft_claim_succeeded",
+  "soft_claim_collided",
+] as const);
+
+export type MinerTelemetryEventType = (typeof MINER_TELEMETRY_EVENT_TYPES)[number];
+
+/** Fixed low-cardinality outcome buckets — the discovery-plane analogue of orb-collector's `bucketReasonCode`, so a
+ *  free-text reason can never leak through the telemetry surface. */
+export const MINER_TELEMETRY_OUTCOME_BUCKETS = Object.freeze([
+  "ok",
+  "empty",
+  "collision",
+  "rate_limited",
+  "error",
+  "other",
+] as const);
+
+export type MinerTelemetryOutcomeBucket = (typeof MINER_TELEMETRY_OUTCOME_BUCKETS)[number];
+
+/** A single discovery-plane telemetry event, pre-anonymization-checked. `repoHash`/`issueHash` are the exporter's
+ *  per-instance HMAC hashes (orb-collector's `getOrCreateAnonSecret`/`hmacField` posture) — NEVER a raw
+ *  `owner/repo` or issue number. `metrics` is count-only quantitative data (e.g. `candidatesReturned`), never text. */
+export type MinerTelemetryEvent = {
+  eventType: MinerTelemetryEventType;
+  repoHash?: string | null | undefined;
+  issueHash?: string | null | undefined;
+  outcome: MinerTelemetryOutcomeBucket;
+  metrics?: Record<string, number> | undefined;
+};
+
+/** The normalized, storage/transport-ready form: hashes coerced to `string | null`, metrics serialized to JSON. */
+export type NormalizedMinerTelemetryEvent = {
+  eventType: MinerTelemetryEventType;
+  repoHash: string | null;
+  issueHash: string | null;
+  outcome: MinerTelemetryOutcomeBucket;
+  metricsJson: string;
+};
+
+const telemetryEventTypeSet = new Set<string>(MINER_TELEMETRY_EVENT_TYPES);
+const telemetryOutcomeSet = new Set<string>(MINER_TELEMETRY_OUTCOME_BUCKETS);
+
+/** Coerce an optional anonymized identifier. Present values must be a non-empty opaque hash — an anti-leak guard
+ *  rejects anything that looks like a RAW identifier (contains `/`, i.e. `owner/repo`, or any whitespace), so a
+ *  caller cannot accidentally ship an un-hashed `repoFullName` through the anonymized surface. */
+function normalizeOptionalHash(value: unknown, code: string): string | null {
+  if (value === undefined || value === null) return null;
+  if (typeof value !== "string") throw new Error(code);
+  const trimmed = value.trim();
+  if (!trimmed || trimmed.includes("/") || /\s/.test(trimmed)) throw new Error(code);
+  return trimmed;
+}
+
+/** Serialize the count-only metrics map, rejecting any non-finite-number value (so no free text or NaN/Infinity can
+ *  slip into the telemetry payload). Absent metrics normalize to an empty object. */
+function normalizeMetrics(metrics: unknown): string {
+  if (metrics === undefined) return "{}";
+  if (metrics === null || typeof metrics !== "object" || Array.isArray(metrics)) throw new Error("invalid_metrics");
+  for (const value of Object.values(metrics as Record<string, unknown>)) {
+    if (typeof value !== "number" || !Number.isFinite(value)) throw new Error("invalid_metrics");
+  }
+  return JSON.stringify(metrics);
+}
+
+/**
+ * Validate and normalize a discovery-plane telemetry event before it is recorded/exported. Fail-closed, mirroring
+ * {@link normalizeGovernorLedgerEvent}: an unknown `eventType`/`outcome`, a non-hash identifier, or a non-numeric
+ * metric throws rather than silently shipping malformed or de-anonymizing data. Defines the contract only — it does
+ * NOT perform the HMAC hashing (that is the exporter's job at #4250's boundary) or send anything.
+ */
+export function normalizeMinerTelemetryEvent(input: unknown): NormalizedMinerTelemetryEvent {
+  if (!input || typeof input !== "object") throw new Error("invalid_event");
+  const event = input as Partial<MinerTelemetryEvent>;
+  const eventType = typeof event.eventType === "string" ? event.eventType.trim() : "";
+  if (!telemetryEventTypeSet.has(eventType)) throw new Error("invalid_event_type");
+  const outcome = typeof event.outcome === "string" ? event.outcome.trim() : "";
+  if (!telemetryOutcomeSet.has(outcome)) throw new Error("invalid_outcome");
+  return {
+    eventType: eventType as MinerTelemetryEventType,
+    repoHash: normalizeOptionalHash(event.repoHash, "invalid_repo_hash"),
+    issueHash: normalizeOptionalHash(event.issueHash, "invalid_issue_hash"),
+    outcome: outcome as MinerTelemetryOutcomeBucket,
+    metricsJson: normalizeMetrics(event.metrics),
+  };
+}

test/unit/miner-telemetry.test.ts

@@ -0,0 +1,78 @@
+import { describe, expect, it } from "vitest";
+import {
+  MINER_TELEMETRY_EVENT_TYPES,
+  MINER_TELEMETRY_OUTCOME_BUCKETS,
+  normalizeMinerTelemetryEvent,
+  type MinerTelemetryEvent,
+} from "../../packages/gittensory-engine/src/miner-telemetry";
+
+const base: MinerTelemetryEvent = { eventType: "candidates_returned", outcome: "ok" };
+
+describe("miner-telemetry schema (#4301)", () => {
+  it("freezes fixed, low-cardinality vocabularies", () => {
+    expect(Object.isFrozen(MINER_TELEMETRY_EVENT_TYPES)).toBe(true);
+    expect(Object.isFrozen(MINER_TELEMETRY_OUTCOME_BUCKETS)).toBe(true);
+    expect([...MINER_TELEMETRY_EVENT_TYPES]).toContain("soft_claim_collided");
+    expect([...MINER_TELEMETRY_OUTCOME_BUCKETS]).toContain("collision");
+  });
+
+  it("normalizes a full event: hashed identifiers pass through, metrics serialize to JSON", () => {
+    expect(
+      normalizeMinerTelemetryEvent({
+        eventType: "candidates_returned",
+        repoHash: "  a1b2c3  ",
+        issueHash: "deadbeef",
+        outcome: "ok",
+        metrics: { candidatesReturned: 12, rankMs: 3 },
+      }),
+    ).toEqual({
+      eventType: "candidates_returned",
+      repoHash: "a1b2c3",
+      issueHash: "deadbeef",
+      outcome: "ok",
+      metricsJson: '{"candidatesReturned":12,"rankMs":3}',
+    });
+  });
+
+  it("defaults absent identifiers to null and absent metrics to an empty object", () => {
+    expect(normalizeMinerTelemetryEvent({ eventType: "query_issued", outcome: "empty" })).toEqual({
+      eventType: "query_issued",
+      repoHash: null,
+      issueHash: null,
+      outcome: "empty",
+      metricsJson: "{}",
+    });
+    // explicit null identifiers are also accepted
+    expect(normalizeMinerTelemetryEvent({ ...base, repoHash: null, issueHash: null }).repoHash).toBeNull();
+  });
+
+  it("fails closed on a non-object, an unknown event type, or an unknown outcome bucket", () => {
+    expect(() => normalizeMinerTelemetryEvent(null)).toThrow("invalid_event");
+    expect(() => normalizeMinerTelemetryEvent("nope")).toThrow("invalid_event");
+    expect(() => normalizeMinerTelemetryEvent({ ...base, eventType: "mystery" })).toThrow("invalid_event_type");
+    expect(() => normalizeMinerTelemetryEvent({ eventType: 7, outcome: "ok" })).toThrow("invalid_event_type");
+    expect(() => normalizeMinerTelemetryEvent({ ...base, outcome: "great" })).toThrow("invalid_outcome");
+    expect(() => normalizeMinerTelemetryEvent({ eventType: "query_issued", outcome: 5 })).toThrow("invalid_outcome");
+  });
+
+  it("anti-leak guard: rejects a non-hash identifier (raw owner/repo, whitespace, non-string, or empty)", () => {
+    expect(() => normalizeMinerTelemetryEvent({ ...base, repoHash: "acme/widgets" })).toThrow("invalid_repo_hash");
+    expect(() => normalizeMinerTelemetryEvent({ ...base, repoHash: "has space" })).toThrow("invalid_repo_hash");
+    expect(() => normalizeMinerTelemetryEvent({ ...base, repoHash: "   " })).toThrow("invalid_repo_hash");
+    expect(() => normalizeMinerTelemetryEvent({ ...base, repoHash: 123 })).toThrow("invalid_repo_hash");
+    expect(() => normalizeMinerTelemetryEvent({ ...base, issueHash: "owner/12" })).toThrow("invalid_issue_hash");
+  });
+
+  it("rejects non-numeric or malformed metrics (no free text or NaN can leak through)", () => {
+    expect(() => normalizeMinerTelemetryEvent({ ...base, metrics: { n: "twelve" } })).toThrow("invalid_metrics");
+    expect(() => normalizeMinerTelemetryEvent({ ...base, metrics: { n: Number.NaN } })).toThrow("invalid_metrics");
+    expect(() => normalizeMinerTelemetryEvent({ ...base, metrics: [1, 2] })).toThrow("invalid_metrics");
+    expect(() => normalizeMinerTelemetryEvent({ ...base, metrics: null })).toThrow("invalid_metrics");
+  });
+
+  it("is re-exported from the package barrel", async () => {
+    const barrel = await import("../../packages/gittensory-engine/src/index");
+    expect(typeof barrel.normalizeMinerTelemetryEvent).toBe("function");
+    expect(barrel.MINER_TELEMETRY_EVENT_TYPES).toBe(MINER_TELEMETRY_EVENT_TYPES);
+  });
+});