fix(ci): base-reds round 3b — stale sibling tests + mode-pack weight contract

- check-docs-counts-sync.test.ts: drop the imports/subtests of the four helpers #10196 removed from the gate script (readMcpFactsFromSource, listLocalizedDocs, makeRequiredCountsValidator, checkFreeTierInventory) — the new-API tests that #10196 added stay; the file now loads again under the node runner - quota-connection-recovery.test.ts: convert from vitest APIs to node:test — the file lives in tests/unit/*.test.ts (node-runner glob) and the vitest runtime crashes when imported outside vitest, killing the whole shard entry - modePacks.ts: re-normalize all six mode packs to sum 1.0 — #8940 added sessionAvailability: 0.05 to every pack without rebalancing (1.05 total); ratios preserved exactly (÷1.05), so post-normalizeScoringWeights behavior is unchanged; restores the declared sum-to-1.0 contract the 4235 test pins Refs #9985
2026-08-13 18:52:18 +03:00 · 2026-08-12 18:17:17 -03:00
parent 24c95aac78
commit 76d825ce76
3 changed files with 76 additions and 123 deletions
--- a/open-sse/services/autoCombo/modePacks.ts
+++ b/open-sse/services/autoCombo/modePacks.ts
@@ -14,84 +14,85 @@ export const MODE_PACKS: Record<string, ScoringWeights> = {
  // Prioritize latency → health. tierPriority replaces 0.05 from stability.
  // tierAffinity/specificityMatch stay at 0 (manifest-routing-only weights).
  "ship-fast": {
-    quota: 0.14,
-    health: 0.28,
-    costInv: 0.05,
-    latencyInv: 0.32,
-    taskFit: 0.1,
-    stability: 0.0,
-    tierPriority: 0.05,
+    quota: 0.1333,
+    health: 0.2667,
+    costInv: 0.0476,
+    latencyInv: 0.3048,
+    taskFit: 0.0952,
+    stability: 0,
+    tierPriority: 0.0476,
    tierAffinity: 0,
    specificityMatch: 0,
-    contextAffinity: 0.01,
-    sessionAvailability: 0.05,
+    contextAffinity: 0.0095,
+    sessionAvailability: 0.0476,
    resetWindowAffinity: 0,
-    connectionDensity: 0.05,
+    connectionDensity: 0.0476,
  },
  // Prioritize cost. tierPriority replaces 0.05 from stability.
  "cost-saver": {
-    quota: 0.14,
-    health: 0.19,
-    costInv: 0.37,
-    latencyInv: 0.05,
-    taskFit: 0.1,
-    stability: 0.05,
-    tierPriority: 0.05,
+    quota: 0.1333,
+    health: 0.181,
+    costInv: 0.3524,
+    latencyInv: 0.0476,
+    taskFit: 0.0952,
+    stability: 0.0476,
+    tierPriority: 0.0476,
    tierAffinity: 0,
    specificityMatch: 0,
-    contextAffinity: 0.0,
-    sessionAvailability: 0.05,
+    contextAffinity: 0,
+    sessionAvailability: 0.0476,
    resetWindowAffinity: 0,
-    connectionDensity: 0.05,
+    connectionDensity: 0.0476,
  },
  // Prioritize task fitness. tierPriority replaces 0.05 from latencyInv.
  "quality-first": {
-    quota: 0.1,
-    health: 0.18,
-    costInv: 0.05,
-    latencyInv: 0.05,
-    taskFit: 0.37,
-    stability: 0.15,
-    tierPriority: 0.05,
+    quota: 0.0952,
+    health: 0.1714,
+    costInv: 0.0476,
+    latencyInv: 0.0476,
+    taskFit: 0.3524,
+    stability: 0.1429,
+    tierPriority: 0.0476,
    tierAffinity: 0,
    specificityMatch: 0,
-    contextAffinity: 0.0,
-    sessionAvailability: 0.05,
+    contextAffinity: 0,
+    sessionAvailability: 0.0476,
    resetWindowAffinity: 0,
-    connectionDensity: 0.05,
+    connectionDensity: 0.0476,
  },
  // Prioritize quota availability. tierPriority replaces 0.05 from taskFit.
  "offline-friendly": {
-    quota: 0.37,
-    health: 0.28,
-    costInv: 0.1,
-    latencyInv: 0.05,
-    taskFit: 0.0,
-    stability: 0.1,
-    tierPriority: 0.05,
+    quota: 0.3524,
+    health: 0.2667,
+    costInv: 0.0952,
+    latencyInv: 0.0476,
+    taskFit: 0,
+    stability: 0.0952,
+    tierPriority: 0.0476,
    tierAffinity: 0,
    specificityMatch: 0,
-    contextAffinity: 0.0,
-    sessionAvailability: 0.05,
+    contextAffinity: 0,
+    sessionAvailability: 0.0476,
    resetWindowAffinity: 0,
-    connectionDensity: 0.05,
+    connectionDensity: 0.0476,
  },
  // #4235 `:reliable` — prioritize healthy, low-variance providers (high availability).
-  // health (circuit-breaker) + stability (latency std-dev) dominate; weights sum to 1.0.
+  // health (circuit-breaker) + stability (latency std-dev) dominate; weights sum to ~1.0
+  // (re-normalized after #8940 added sessionAvailability without rebalancing — #9985).
  "reliability-first": {
-    quota: 0.14,
-    health: 0.37,
-    costInv: 0.04,
-    latencyInv: 0.05,
-    taskFit: 0.1,
-    stability: 0.2,
-    tierPriority: 0.05,
+    quota: 0.1333,
+    health: 0.3524,
+    costInv: 0.0381,
+    latencyInv: 0.0476,
+    taskFit: 0.0952,
+    stability: 0.1905,
+    tierPriority: 0.0476,
    tierAffinity: 0,
    specificityMatch: 0,
-    contextAffinity: 0.0,
-    sessionAvailability: 0.05,
+    contextAffinity: 0,
+    sessionAvailability: 0.0476,
    resetWindowAffinity: 0,
-    connectionDensity: 0.05,
+    connectionDensity: 0.0476,
  },
  // Chaos mode — priority: health > stability > taskFit > latency.
  // Selects top-N healthy providers for parallel dispatch. Favors providers with
@@ -100,19 +101,19 @@ export const MODE_PACKS: Record<string, ScoringWeights> = {
  // to picking the most stable providers); connectionDensity boosted slightly to
  // prefer providers with multiple accounts (more resilient to per-account rate limits).
  "chaos-mode": {
-    quota: 0.05,
-    health: 0.42,
-    costInv: 0.02,
-    latencyInv: 0.03,
-    taskFit: 0.2,
-    stability: 0.18,
-    tierPriority: 0.02,
+    quota: 0.0476,
+    health: 0.4,
+    costInv: 0.019,
+    latencyInv: 0.0286,
+    taskFit: 0.1905,
+    stability: 0.1714,
+    tierPriority: 0.019,
    tierAffinity: 0,
    specificityMatch: 0,
-    contextAffinity: 0.03,
-    sessionAvailability: 0.05,
+    contextAffinity: 0.0286,
+    sessionAvailability: 0.0476,
    resetWindowAffinity: 0,
-    connectionDensity: 0.05,
+    connectionDensity: 0.0476,
  },
 };

--- a/tests/unit/check-docs-counts-sync.test.ts
+++ b/tests/unit/check-docs-counts-sync.test.ts
@@ -8,9 +8,6 @@ import {
  tallyDrift,
  readProviderTotal,
  countLocales,
-  readMcpFactsFromSource,
-  listLocalizedDocs,
-  makeRequiredCountsValidator,
 } from "../../scripts/check/check-docs-counts-sync.mjs";

 // Explicit types for the .mjs exports — keep the test at 0 no-explicit-any warnings.
@@ -27,11 +24,6 @@ const tally = tallyDrift as (
 ) => { strict: number; soft: number; lines: string[] };
 const readTotal = readProviderTotal as () => number;
 const locales = countLocales as () => number;
-const mcpFacts = readMcpFactsFromSource as () => { tools: number; scopes: number } | null;
-const localizedDocs = listLocalizedDocs as (relativePath: string) => string[];
-const requireCounts = makeRequiredCountsValidator as (
-  requirements: { label: string; value: number }[]
-) => (content: string) => { ok: boolean; detail: string };

 const here = path.dirname(fileURLToPath(import.meta.url));
 const GATE = path.resolve(here, "../../scripts/check/check-docs-counts-sync.mjs");
@@ -100,31 +92,6 @@ test("countLocales reads a real, positive locale count from config/i18n.json", (
  assert.ok(locales() >= 40, "i18n config should define at least 40 locales");
 });

-test("source-only MCP fallback matches the canonical inventory and scope union", () => {
-  assert.deepEqual(mcpFacts(), { tools: 107, scopes: 32 });
-});
-
-test("localized-doc discovery returns every locale root document", () => {
-  const readmes = localizedDocs("README.md");
-  assert.equal(readmes.length, 42);
-  assert.ok(readmes.includes("docs/i18n/pt-BR/README.md"));
-  assert.ok(readmes.includes("docs/i18n/zh-CN/README.md"));
-});
-
-test("required-count validator reports precisely which live markers are missing", () => {
-  const validate = requireCounts([
-    { label: "providers", value: 327 },
-    { label: "MCP tools", value: 107 },
-    { label: "MCP scopes", value: 32 },
-  ]);
-  assert.equal(validate("327 providers; 107 tools; 32 scopes").ok, true);
-  const stale = validate("290 providers; 104 tools; 31 scopes");
-  assert.equal(stale.ok, false);
-  assert.match(stale.detail, /providers=327/);
-  assert.match(stale.detail, /MCP tools=107/);
-  assert.match(stale.detail, /MCP scopes=32/);
-});
-
 // --- live gate smoke -----------------------------------------------------------------

 test("the gate exits 0 against the current (synced) repo state", () => {
@@ -138,7 +105,6 @@ test("the gate exits 0 against the current (synced) repo state", () => {
 // down to 1.37B, because no gate watched that number.
 import {
  checkFreeTierHeadline,
-  checkFreeTierInventory,
  extractHeadlineClaims,
 } from "../../scripts/check/check-docs-counts-sync.mjs";

@@ -180,23 +146,6 @@ test("free-tier gate passes when a file carries no headline at all", () => {
  assert.equal(checkHeadline("no figures here", TOTALS).ok, true);
 });

-const checkInventory = checkFreeTierInventory as (
-  content: string,
-  totals: { pools: number; models: number }
-) => { ok: boolean; detail: string };
-
-test("free-tier inventory gate accepts the live pool/model counts", () => {
-  assert.equal(
-    checkInventory("43 provider pools / 522 model budget entries", { pools: 43, models: 522 }).ok,
-    true
-  );
-});
-
-test("free-tier inventory gate rejects stale model counts", () => {
-  const result = checkInventory("43 provider pools / 516 models", { pools: 43, models: 522 });
-  assert.equal(result.ok, false);
-  assert.match(result.detail, /live catalog has 43 pools \/ 522 model budget entries/);
-});

 // --- Generic numeric-claim gate (engines / MCP tools / scopes / CLI) --------
 // Extends the same drift guard to the counts that silently drifted in v3.8.49:
--- a/tests/unit/quota-connection-recovery.test.ts
+++ b/tests/unit/quota-connection-recovery.test.ts
@@ -1,4 +1,5 @@
-import { describe, expect, it, vi } from "vitest";
+import { describe, it, mock } from "node:test";
+import assert from "node:assert/strict";
 import {
  CREDITS_EXHAUSTED_STATUS,
  isCreditsExhaustedReprobeCandidate,
@@ -17,7 +18,7 @@ describe("connectionRecovery — credits_exhausted reprobe", () => {
      testStatus: CREDITS_EXHAUSTED_STATUS,
      rateLimitedUntil: new Date(nowMs - 5000).toISOString(),
    };
-    expect(isRecoverableCooldownConnection(conn, nowMs)).toBe(false);
+    assert.equal(isRecoverableCooldownConnection(conn, nowMs), false);
  });

  it("should reprobe credits_exhausted when >30m has elapsed since lastErrorAt", () => {
@@ -27,7 +28,7 @@ describe("connectionRecovery — credits_exhausted reprobe", () => {
      testStatus: CREDITS_EXHAUSTED_STATUS,
      lastErrorAt: thirtyOneMinAgo,
    };
-    expect(isCreditsExhaustedReprobeCandidate(conn, nowMs)).toBe(true);
+    assert.equal(isCreditsExhaustedReprobeCandidate(conn, nowMs), true);
  });

  it("should NOT reprobe credits_exhausted when <30m has elapsed since lastErrorAt", () => {
@@ -37,7 +38,7 @@ describe("connectionRecovery — credits_exhausted reprobe", () => {
      testStatus: CREDITS_EXHAUSTED_STATUS,
      lastErrorAt: tenMinAgo,
    };
-    expect(isCreditsExhaustedReprobeCandidate(conn, nowMs)).toBe(false);
+    assert.equal(isCreditsExhaustedReprobeCandidate(conn, nowMs), false);
  });

  it("should reprobe credits_exhausted if no timestamp is present (first tick after startup)", () => {
@@ -45,7 +46,7 @@ describe("connectionRecovery — credits_exhausted reprobe", () => {
      id: "conn-1",
      testStatus: CREDITS_EXHAUSTED_STATUS,
    };
-    expect(isCreditsExhaustedReprobeCandidate(conn, nowMs)).toBe(true);
+    assert.equal(isCreditsExhaustedReprobeCandidate(conn, nowMs), true);
  });

  it("selectRecoverableConnections includes both transient cooldowns and expired credits_exhausted", () => {
@@ -69,18 +70,18 @@ describe("connectionRecovery — credits_exhausted reprobe", () => {
      [activeTransient, expiredCredits, freshCredits],
      nowMs
    );
-    expect(selected.map((c) => c.id)).toEqual(["t-1", "c-1"]);
+    assert.deepEqual(selected.map((c) => c.id), ["t-1", "c-1"]);
  });

  it("runConnectionRecoveryTick calls clearConnectionError for reprobe candidates", async () => {
-    const loadConnections = vi.fn().mockResolvedValue([
+    const loadConnections = mock.fn(async () => [
      {
        id: "c-1",
        testStatus: CREDITS_EXHAUSTED_STATUS,
        lastErrorAt: new Date(nowMs - thirtyMinMs - 1000).toISOString(),
      },
    ]);
-    const clearConnectionError = vi.fn().mockResolvedValue(undefined);
+    const clearConnectionError = mock.fn(async () => undefined);

    const res = await runConnectionRecoveryTick({
      nowMs,
@@ -88,8 +89,10 @@ describe("connectionRecovery — credits_exhausted reprobe", () => {
      clearConnectionError,
    });

-    expect(res.recovered).toBe(1);
-    expect(res.recoveredIds).toEqual(["c-1"]);
-    expect(clearConnectionError).toHaveBeenCalledWith("c-1", expect.anything());
+    assert.equal(res.recovered, 1);
+    assert.deepEqual(res.recoveredIds, ["c-1"]);
+    assert.equal(clearConnectionError.mock.callCount(), 1);
+    assert.equal(clearConnectionError.mock.calls[0].arguments[0], "c-1");
+    assert.notEqual(clearConnectionError.mock.calls[0].arguments[1], undefined);
  });
 });