fix(miner): drop discovery-index candidates whose repo bans AI contributions (#9680)

packages/loopover-miner/lib/discover-cli.ts

@@ -186,11 +186,19 @@ async function supplementWithDiscoveryIndex(
   if (!isDiscoveryPlaneEnabled(env)) return fanOut;
   const queryIndex = options.queryDiscoveryIndex ?? queryDiscoveryIndex;
   const response = await queryIndex(queryScope, { env });
-  recordDiscoveryTelemetry("discover_query", response.candidates.length > 0 ? "supplemented" : "empty", { env });
-  if (response.candidates.length === 0) return fanOut;
+  // #9680: the hosted index preserves a repo's AI-contribution ban (normalizeDiscoveryIndexCandidate keeps
+  // aiPolicyAllowed:false), but nothing downstream re-checks it -- the `as RawCandidateIssue` cast below would
+  // launder a `false` through a type declared as the literal `true`, so an AI-banned repo's issue would be ranked
+  // and enqueued. Enforce the ban here exactly as the local fetchTargetIssues does (`if (!verdict.allowed) return
+  // []`): drop it, not down-rank or warn. A candidate that omits the field is kept (`!== false`, matching
+  // normalizeDiscoveryIndexCandidate's own default).
+  const aiAllowed = response.candidates.filter((candidate) => candidate.aiPolicyAllowed !== false);
+  const droppedAiBanned = response.candidates.length - aiAllowed.length;
+  recordDiscoveryTelemetry("discover_query", response.candidates.length > 0 ? "supplemented" : "empty", { env, droppedAiBanned });
+  if (aiAllowed.length === 0) return fanOut;
 
   const seen = new Set(fanOut.issues.map((issue) => dedupeKey(issue.repoFullName, issue.issueNumber)));
-  const supplemented = response.candidates
+  const supplemented = aiAllowed
     .filter((candidate) => !seen.has(dedupeKey(candidate.repoFullName, candidate.issueNumber)))
     // DiscoveryIndexCandidate is a near-superset of RawCandidateIssue; copy the real assignees through when the
     // hosted contract carried them (#7442), falling back to [] only when the served response genuinely omitted the

packages/loopover-miner/lib/discovery-index-client.ts

@@ -171,9 +171,16 @@ export async function submitSoftClaim(
 export function recordDiscoveryTelemetry(
   event: string,
   outcome: string,
-  options: { env?: Record<string, string | undefined> } = {},
+  options: { env?: Record<string, string | undefined>; droppedAiBanned?: number } = {},
 ): void {
   const env = options.env ?? process.env;
   if (!isDiscoveryPlaneEnabled(env) || !isDiscoveryTelemetryEnabled(env)) return;
-  getLogger().info("discovery_plane_telemetry", { event, outcome });
+  // #9680: low-cardinality count of index candidates dropped for an AI-contribution ban this query, when the
+  // caller supplies it. Spread conditionally so callers that don't pass it keep the exact { event, outcome }
+  // shape (and payload) they emitted before.
+  getLogger().info("discovery_plane_telemetry", {
+    event,
+    outcome,
+    ...(options.droppedAiBanned !== undefined ? { droppedAiBanned: options.droppedAiBanned } : {}),
+  });
 }

test/unit/miner-discover-cli.test.ts

@@ -2035,6 +2035,89 @@ describe("discovery-index supplementation (#7168)", () => {
     ]);
   });
 
+  it("REGRESSION (#9680): drops an index candidate whose repo bans AI contributions, keeps the allowed one, and reports the dropped count", async () => {
+    const clientModule = await import("../../packages/loopover-miner/lib/discovery-index-client");
+    const telemetrySpy = vi.spyOn(clientModule, "recordDiscoveryTelemetry");
+    const portfolioQueue = tempQueueStore();
+    const fetchCandidateIssuesWithSummary = vi.fn(async () => ({
+      issues: [fanOutIssue({ issueNumber: 1, title: "Local candidate" })],
+      warnings: [],
+      rateLimitRemaining: 100,
+      rateLimitResetAt: null,
+    }));
+    const queryDiscoveryIndex = vi.fn(async () => ({
+      contractVersion: 1,
+      candidates: [
+        indexCandidate({ issueNumber: 2, title: "AI-banned repo issue", aiPolicyAllowed: false }),
+        indexCandidate({ issueNumber: 3, title: "Allowed index issue", aiPolicyAllowed: true }),
+      ],
+      nextCursor: null,
+    }));
+    const log = vi.spyOn(console, "log").mockImplementation(() => undefined);
+
+    const exitCode = await runDiscover(["acme/widgets", "--json"], {
+      nowMs: NOW,
+      env: { LOOPOVER_MINER_DISCOVERY_PLANE: "true" },
+      initPortfolioQueue: () => portfolioQueue,
+      initPolicyDocCache: () => tempPolicyDocCacheStore(),
+      initPolicyVerdictCache: () => tempPolicyVerdictCacheStore(),
+      initRankedCandidatesStore: () => tempRankedCandidatesStore(),
+      fetchCandidateIssuesWithSummary,
+      queryDiscoveryIndex: queryDiscoveryIndex as never,
+    });
+
+    expect(exitCode).toBe(0);
+    const payload = JSON.parse(String(log.mock.calls[0]?.[0]));
+    const issueNumbers = payload.ranked
+      .map((entry: { issueNumber: number }) => entry.issueNumber)
+      .sort((a: number, b: number) => a - b);
+    // #2 (aiPolicyAllowed:false) is dropped before ranking; the local #1 and the allowed index #3 remain.
+    // Before the fix #2 would also be ranked and enqueued into the miner's backlog.
+    expect(issueNumbers).toEqual([1, 3]);
+    expect(telemetrySpy).toHaveBeenCalledWith(
+      "discover_query",
+      "supplemented",
+      expect.objectContaining({ droppedAiBanned: 1 }),
+    );
+  });
+
+  it("REGRESSION (#9680): keeps an index candidate that OMITS aiPolicyAllowed (field-absent is not a ban)", async () => {
+    const portfolioQueue = tempQueueStore();
+    const fetchCandidateIssuesWithSummary = vi.fn(async () => ({
+      issues: [fanOutIssue({ issueNumber: 1, title: "Local candidate" })],
+      warnings: [],
+      rateLimitRemaining: 100,
+      rateLimitResetAt: null,
+    }));
+    const candidateNoField = indexCandidate({ issueNumber: 4, title: "Older-build candidate" });
+    delete (candidateNoField as { aiPolicyAllowed?: unknown }).aiPolicyAllowed;
+    const queryDiscoveryIndex = vi.fn(async () => ({
+      contractVersion: 1,
+      candidates: [candidateNoField],
+      nextCursor: null,
+    }));
+    const log = vi.spyOn(console, "log").mockImplementation(() => undefined);
+
+    const exitCode = await runDiscover(["acme/widgets", "--json"], {
+      nowMs: NOW,
+      env: { LOOPOVER_MINER_DISCOVERY_PLANE: "true" },
+      initPortfolioQueue: () => portfolioQueue,
+      initPolicyDocCache: () => tempPolicyDocCacheStore(),
+      initPolicyVerdictCache: () => tempPolicyVerdictCacheStore(),
+      initRankedCandidatesStore: () => tempRankedCandidatesStore(),
+      fetchCandidateIssuesWithSummary,
+      queryDiscoveryIndex: queryDiscoveryIndex as never,
+    });
+
+    expect(exitCode).toBe(0);
+    const payload = JSON.parse(String(log.mock.calls[0]?.[0]));
+    const issueNumbers = payload.ranked
+      .map((entry: { issueNumber: number }) => entry.issueNumber)
+      .sort((a: number, b: number) => a - b);
+    // A candidate that never carried aiPolicyAllowed must NOT become a fail-closed drop.
+    expect(issueNumbers).toEqual([1, 4]);
+  });
+
   it("uses the --search term as the discovery-index scope in search mode", async () => {
     const portfolioQueue = tempQueueStore();
     const searchCandidateIssuesWithSummary = vi.fn(async () => ({

test/unit/miner-discovery-index-client.test.ts

@@ -225,6 +225,18 @@ describe("recordDiscoveryTelemetry (#7168)", () => {
     });
     expect(logSpy.info).toHaveBeenCalledWith("discovery_plane_telemetry", { event: "discover_query", outcome: "supplemented" });
   });
+
+  it("#9680: carries the droppedAiBanned count in the payload when the caller supplies it", () => {
+    recordDiscoveryTelemetry("discover_query", "supplemented", {
+      env: { [DISCOVERY_PLANE_FLAG]: "true", [DISCOVERY_TELEMETRY_FLAG]: "true" },
+      droppedAiBanned: 2,
+    });
+    expect(logSpy.info).toHaveBeenCalledWith("discovery_plane_telemetry", {
+      event: "discover_query",
+      outcome: "supplemented",
+      droppedAiBanned: 2,
+    });
+  });
 });
 
 describe("defaulted options (real process.env / real global fetch) (#7168)", () => {