@claude-flow/cli 3.10.46 → 3.11.0

package/package.json

@@ -1,134 +1,136 @@
-{
-  "name": "@claude-flow/cli",
-  "version": "3.10.46",
-  "type": "module",
-  "description": "Ruflo CLI - Enterprise AI agent orchestration with 60+ specialized agents, swarm coordination, MCP server, self-learning hooks, and vector memory for Claude Code",
-  "main": "dist/src/index.js",
-  "types": "dist/src/index.d.ts",
-  "sideEffects": false,
-  "bin": {
-    "cli": "./bin/cli.js",
-    "claude-flow": "./bin/cli.js",
-    "claude-flow-mcp": "./bin/mcp-server.js"
-  },
-  "homepage": "https://github.com/ruvnet/claude-flow#readme",
-  "bugs": {
-    "url": "https://github.com/ruvnet/claude-flow/issues"
-  },
-  "repository": {
-    "type": "git",
-    "url": "https://github.com/ruvnet/claude-flow.git",
-    "directory": "v3/@claude-flow/cli"
-  },
-  "keywords": [
-    "claude",
-    "claude-code",
-    "anthropic",
-    "ai-agents",
-    "multi-agent",
-    "swarm",
-    "mcp",
-    "model-context-protocol",
-    "llm",
-    "cli",
-    "orchestration",
-    "automation",
-    "developer-tools",
-    "coding-assistant",
-    "vector-database",
-    "embeddings",
-    "self-learning",
-    "enterprise"
-  ],
-  "author": {
-    "name": "RuvNet",
-    "email": "ruv@ruv.io",
-    "url": "https://ruv.io"
-  },
-  "license": "MIT",
-  "exports": {
-    ".": {
-      "types": "./dist/src/index.d.ts",
-      "import": "./dist/src/index.js"
-    },
-    "./dist/*": "./dist/*",
-    "./ruvector": {
-      "types": "./dist/src/ruvector/index.d.ts",
-      "import": "./dist/src/ruvector/index.js"
-    },
-    "./ruvector/*": {
-      "types": "./dist/src/ruvector/*.d.ts",
-      "import": "./dist/src/ruvector/*.js"
-    },
-    "./mcp-tools": {
-      "types": "./dist/src/mcp-tools/index.d.ts",
-      "import": "./dist/src/mcp-tools/index.js"
-    },
-    "./memory": {
-      "types": "./dist/src/memory/memory-initializer.d.ts",
-      "import": "./dist/src/memory/memory-initializer.js"
-    },
-    "./intelligence": {
-      "types": "./dist/src/memory/intelligence.d.ts",
-      "import": "./dist/src/memory/intelligence.js"
-    }
-  },
-  "files": [
-    "dist",
-    "bin",
-    "scripts",
-    ".claude",
-    "README.md"
-  ],
-  "scripts": {
-    "build": "tsc",
-    "test": "vitest run",
-    "test:plugin-store": "npx tsx src/plugins/tests/standalone-test.ts",
-    "test:pattern-store": "npx tsx src/transfer/store/tests/standalone-test.ts",
-    "postinstall": "node ./scripts/postinstall.cjs",
-    "prepublishOnly": "cp ../../../README.md ./README.md",
-    "release": "npm version prerelease --preid=alpha && npm run publish:all",
-    "publish:all": "./scripts/publish.sh"
-  },
-  "devDependencies": {
-    "typescript": "^5.3.0",
-    "vitest": "^4.0.16"
-  },
-  "dependencies": {
-    "@claude-flow/cli-core": "^3.7.0-alpha.5",
-    "@claude-flow/mcp": "^3.0.0-alpha.8",
-    "@claude-flow/neural": "^3.0.0-alpha.9",
-    "@claude-flow/shared": "^3.0.0-alpha.7",
-    "@noble/ed25519": "^2.1.0",
-    "@ruvector/rabitq-wasm": "^0.1.0",
-    "semver": "^7.6.0",
-    "yaml": "^2.8.0"
-  },
-  "optionalDependencies": {
-    "@claude-flow/aidefence": "^3.0.2",
-    "@claude-flow/codex": "^3.0.0-alpha.8",
-    "@claude-flow/embeddings": "^3.0.0-alpha.18",
-    "@claude-flow/guidance": "^3.0.0-alpha.1",
-    "@claude-flow/memory": "^3.0.0-alpha.20",
-    "@claude-flow/plugin-gastown-bridge": "^0.1.3",
-    "@claude-flow/security": "^3.0.0-alpha.10",
-    "@ruvector/attention": "^0.1.32",
-    "@ruvector/attention-darwin-arm64": "0.1.32",
-    "@ruvector/diskann": "^0.1.0",
-    "@ruvector/learning-wasm": "^0.1.29",
-    "@ruvector/router": "^0.1.30",
-    "@ruvector/ruvllm-wasm": "^2.0.2",
-    "@ruvector/rvagent-wasm": "^0.1.0",
-    "@ruvector/sona": "^0.1.6",
-    "agentdb": "^3.0.0-alpha.16",
-    "agentic-flow": "^3.0.0-alpha.1",
-    "ruvector": "^0.2.27"
-  },
-  "publishConfig": {
-    "access": "public",
-    "tag": "latest"
-  },
-  "overrides": {
-    "protobufjs": ">=7.5.6"
-  }
-}
+{
+  "name": "@claude-flow/cli",
+  "version": "3.11.0",
+  "type": "module",
+  "description": "Ruflo CLI - Enterprise AI agent orchestration with 60+ specialized agents, swarm coordination, MCP server, self-learning hooks, and vector memory for Claude Code",
+  "main": "dist/src/index.js",
+  "types": "dist/src/index.d.ts",
+  "sideEffects": false,
+  "bin": {
+    "cli": "./bin/cli.js",
+    "claude-flow": "./bin/cli.js",
+    "claude-flow-mcp": "./bin/mcp-server.js"
+  },
+  "homepage": "https://github.com/ruvnet/claude-flow#readme",
+  "bugs": {
+    "url": "https://github.com/ruvnet/claude-flow/issues"
+  },
+  "repository": {
+    "type": "git",
+    "url": "https://github.com/ruvnet/claude-flow.git",
+    "directory": "v3/@claude-flow/cli"
+  },
+  "keywords": [
+    "claude",
+    "claude-code",
+    "anthropic",
+    "ai-agents",
+    "multi-agent",
+    "swarm",
+    "mcp",
+    "model-context-protocol",
+    "llm",
+    "cli",
+    "orchestration",
+    "automation",
+    "developer-tools",
+    "coding-assistant",
+    "vector-database",
+    "embeddings",
+    "self-learning",
+    "enterprise"
+  ],
+  "author": {
+    "name": "RuvNet",
+    "email": "ruv@ruv.io",
+    "url": "https://ruv.io"
+  },
+  "license": "MIT",
+  "exports": {
+    ".": {
+      "types": "./dist/src/index.d.ts",
+      "import": "./dist/src/index.js"
+    },
+    "./dist/*": "./dist/*",
+    "./ruvector": {
+      "types": "./dist/src/ruvector/index.d.ts",
+      "import": "./dist/src/ruvector/index.js"
+    },
+    "./ruvector/*": {
+      "types": "./dist/src/ruvector/*.d.ts",
+      "import": "./dist/src/ruvector/*.js"
+    },
+    "./mcp-tools": {
+      "types": "./dist/src/mcp-tools/index.d.ts",
+      "import": "./dist/src/mcp-tools/index.js"
+    },
+    "./memory": {
+      "types": "./dist/src/memory/memory-initializer.d.ts",
+      "import": "./dist/src/memory/memory-initializer.js"
+    },
+    "./intelligence": {
+      "types": "./dist/src/memory/intelligence.d.ts",
+      "import": "./dist/src/memory/intelligence.js"
+    }
+  },
+  "files": [
+    "dist",
+    "bin",
+    "scripts",
+    ".claude",
+    "README.md"
+  ],
+  "scripts": {
+    "build": "tsc",
+    "test": "vitest run",
+    "test:plugin-store": "npx tsx src/plugins/tests/standalone-test.ts",
+    "test:pattern-store": "npx tsx src/transfer/store/tests/standalone-test.ts",
+    "postinstall": "node ./scripts/postinstall.cjs",
+    "prepublishOnly": "cp ../../../README.md ./README.md",
+    "release": "npm version prerelease --preid=alpha && npm run publish:all",
+    "publish:all": "./scripts/publish.sh"
+  },
+  "devDependencies": {
+    "typescript": "^5.3.0",
+    "vitest": "^4.0.16"
+  },
+  "dependencies": {
+    "@claude-flow/cli-core": "^3.7.0-alpha.5",
+    "@claude-flow/mcp": "^3.0.0-alpha.8",
+    "@claude-flow/neural": "^3.0.0-alpha.9",
+    "@claude-flow/shared": "^3.0.0-alpha.7",
+    "@noble/ed25519": "^2.1.0",
+    "@ruvector/rabitq-wasm": "^0.1.0",
+    "semver": "^7.6.0",
+    "yaml": "^2.8.0"
+  },
+  "optionalDependencies": {
+    "@claude-flow/aidefence": "^3.0.2",
+    "@claude-flow/codex": "^3.0.0-alpha.8",
+    "@claude-flow/embeddings": "^3.0.0-alpha.18",
+    "@claude-flow/guidance": "^3.0.0-alpha.1",
+    "@claude-flow/memory": "^3.0.0-alpha.20",
+    "@claude-flow/plugin-gastown-bridge": "^0.1.3",
+    "@claude-flow/security": "^3.0.0-alpha.10",
+    "@metaharness/router": "^0.3.2",
+    "@ruvector/attention": "^0.1.32",
+    "@ruvector/attention-darwin-arm64": "0.1.32",
+    "@ruvector/diskann": "^0.1.0",
+    "@ruvector/learning-wasm": "^0.1.29",
+    "@ruvector/router": "^0.1.30",
+    "@ruvector/ruvllm-wasm": "^2.0.2",
+    "@ruvector/rvagent-wasm": "^0.1.0",
+    "@ruvector/sona": "^0.1.6",
+    "@ruvector/tiny-dancer": "^0.1.22",
+    "agentdb": "^3.0.0-alpha.16",
+    "agentic-flow": "^3.0.0-alpha.1",
+    "ruvector": "^0.2.27"
+  },
+  "publishConfig": {
+    "access": "public",
+    "tag": "latest"
+  },
+  "overrides": {
+    "protobufjs": ">=7.5.6"
+  }
+}

package/scripts/beir-bootstrap-significance.mjs

@@ -1,162 +1,162 @@
-#!/usr/bin/env node
-// beir-bootstrap-significance.mjs — paired bootstrap significance test for
-// BEIR retrieval results (ADR-086).
-//
-// Given a run JSON with per-query nDCG@10 and a baseline mean nDCG@10,
-// estimates whether our result is significantly above the baseline.
-//
-// Method: paired bootstrap — resample N=10000 with replacement from our
-// per-query nDCG scores, compute mean each time, take 2.5%-97.5% percentile
-// for 95% CI. Compare baseline to our CI: if baseline < CI lower bound,
-// the difference is significant at p < 0.05.
-//
-// For a paired test against ANOTHER per-query run (e.g. our pure-BM25 vs
-// our BGE-base), resample paired DIFFERENCES instead. Pass --paired
-// path/to/baseline-run.json.
-//
-// Usage:
-//   node scripts/beir-bootstrap-significance.mjs <run-json-path>
-//   node scripts/beir-bootstrap-significance.mjs <ours> --paired <baseline>
-//
-// Example:
-//   node scripts/beir-bootstrap-significance.mjs \
-//     docs/benchmarks/runs/beir-nfcorpus-bge-latest.json
-
-import { readFileSync } from 'node:fs';
-import { performance } from 'node:perf_hooks';
-
-const args = process.argv.slice(2);
-if (args.length === 0) {
-  console.error('Usage: node beir-bootstrap-significance.mjs <run-json> [--paired <baseline-run-json>]');
-  process.exit(1);
-}
-const RUN_PATH = args[0];
-const PAIRED_IDX = args.indexOf('--paired');
-const PAIRED_PATH = PAIRED_IDX > 0 ? args[PAIRED_IDX + 1] : null;
-const ITERATIONS = Number(process.env.BOOTSTRAP_ITERATIONS) || 10000;
-
-// Deterministic PRNG (Mulberry32) — so re-runs are reproducible.
-const SEED = Number(process.env.BOOTSTRAP_SEED) || 42;
-function mulberry32(a) {
-  return function () {
-    a |= 0; a = (a + 0x6D2B79F5) | 0;
-    let t = a;
-    t = Math.imul(t ^ (t >>> 15), t | 1);
-    t ^= t + Math.imul(t ^ (t >>> 7), t | 61);
-    return ((t ^ (t >>> 14)) >>> 0) / 4294967296;
-  };
-}
-
-function mean(arr) { let s = 0; for (const v of arr) s += v; return s / arr.length; }
-
-function percentile(sortedArr, p) {
-  const idx = Math.max(0, Math.min(sortedArr.length - 1, Math.floor(p * sortedArr.length)));
-  return sortedArr[idx];
-}
-
-function pointEstimate(scores) {
-  return {
-    mean: mean(scores),
-    n: scores.length,
-  };
-}
-
-function bootstrapCI(scores, iterations, rng, percentiles = [0.025, 0.975]) {
-  const means = new Float64Array(iterations);
-  const n = scores.length;
-  for (let i = 0; i < iterations; i++) {
-    let s = 0;
-    for (let j = 0; j < n; j++) s += scores[Math.floor(rng() * n)];
-    means[i] = s / n;
-  }
-  const sorted = Array.from(means).sort((a, b) => a - b);
-  return {
-    point: mean(scores),
-    lower: percentile(sorted, percentiles[0]),
-    upper: percentile(sorted, percentiles[1]),
-    p_value_vs_baseline_zero: null,
-  };
-}
-
-// Paired bootstrap: compare our per-query scores to a baseline's per-query
-// scores. Resample (idx) and average paired differences. The 95% CI on the
-// mean difference tells us if it's significantly non-zero.
-function pairedBootstrap(oursByQid, baselineByQid, iterations, rng) {
-  const sharedIds = [...oursByQid.keys()].filter((id) => baselineByQid.has(id));
-  const diffs = sharedIds.map((id) => oursByQid.get(id) - baselineByQid.get(id));
-  const n = diffs.length;
-  const means = new Float64Array(iterations);
-  for (let i = 0; i < iterations; i++) {
-    let s = 0;
-    for (let j = 0; j < n; j++) s += diffs[Math.floor(rng() * n)];
-    means[i] = s / n;
-  }
-  const sorted = Array.from(means).sort((a, b) => a - b);
-  // One-sided p ≈ fraction of resamples where mean diff ≤ 0
-  let negCount = 0;
-  for (const m of means) if (m <= 0) negCount++;
-  return {
-    n_paired: n,
-    point_diff: mean(diffs),
-    lower_95: percentile(sorted, 0.025),
-    upper_95: percentile(sorted, 0.975),
-    p_one_sided: negCount / iterations,
-  };
-}
-
-function loadRun(path) {
-  const data = JSON.parse(readFileSync(path, 'utf-8'));
-  if (!data.perQuery || !Array.isArray(data.perQuery)) {
-    console.error(`Run JSON at ${path} has no perQuery array — re-run the bench with the updated harness.`);
-    process.exit(2);
-  }
-  return data;
-}
-
-const ours = loadRun(RUN_PATH);
-console.log(`# BEIR significance test (ADR-086)`);
-console.log(`Run: ${RUN_PATH}`);
-console.log(`Dataset: ${ours.dataset ?? '?'}  Model: ${ours.model ?? '?'}  Queries: ${ours.perQuery.length}`);
-console.log(`Bootstrap: ${ITERATIONS} iterations, seed=${SEED}`);
-
-const rng = mulberry32(SEED);
-const ourScores = ours.perQuery.map((q) => q.ndcg10);
-
-const t0 = performance.now();
-const oursCI = bootstrapCI(ourScores, ITERATIONS, rng);
-console.log(`\n=== Our nDCG@10 (1-sample bootstrap CI) ===`);
-console.log(`  point:    ${oursCI.point.toFixed(4)}`);
-console.log(`  95% CI:   [${oursCI.lower.toFixed(4)}, ${oursCI.upper.toFixed(4)}]`);
-
-if (ours.baselines) {
-  console.log(`\n=== vs each published baseline (CI overlap) ===`);
-  for (const [name, score] of Object.entries(ours.baselines)) {
-    const ciDiff = oursCI.point - score;
-    const lowerDiff = oursCI.lower - score;
-    const upperDiff = oursCI.upper - score;
-    const direction = ciDiff > 0 ? '↑ above' : '↓ below';
-    const sig = (lowerDiff > 0 && upperDiff > 0) ? 'p<0.05 (95% CI excludes baseline)'
-              : (lowerDiff < 0 && upperDiff < 0) ? 'p<0.05 (95% CI below baseline)'
-              : 'n.s. (95% CI overlaps baseline)';
-    console.log(`  ${score.toFixed(3)}  ${name.padEnd(28)}  Δ=${ciDiff >= 0 ? '+' : ''}${ciDiff.toFixed(4)}  ${direction}  [${sig}]`);
-  }
-}
-
-if (PAIRED_PATH) {
-  const baseline = loadRun(PAIRED_PATH);
-  console.log(`\n=== Paired bootstrap vs ${PAIRED_PATH} ===`);
-  const oursByQid = new Map(ours.perQuery.map((q) => [q.qid, q.ndcg10]));
-  const baseByQid = new Map(baseline.perQuery.map((q) => [q.qid, q.ndcg10]));
-  const rng2 = mulberry32(SEED + 1);
-  const paired = pairedBootstrap(oursByQid, baseByQid, ITERATIONS, rng2);
-  console.log(`  paired queries: ${paired.n_paired}`);
-  console.log(`  Δ (ours - baseline): ${paired.point_diff >= 0 ? '+' : ''}${paired.point_diff.toFixed(4)}`);
-  console.log(`  95% CI: [${paired.lower_95.toFixed(4)}, ${paired.upper_95.toFixed(4)}]`);
-  console.log(`  one-sided p (Δ ≤ 0): ${paired.p_one_sided.toFixed(4)}`);
-  const sig = paired.lower_95 > 0 ? 'SIGNIFICANT improvement (p<0.025 one-sided)' :
-              paired.upper_95 < 0 ? 'SIGNIFICANT regression (p<0.025 one-sided)' :
-              'NOT SIGNIFICANT — CI overlaps zero';
-  console.log(`  verdict: ${sig}`);
-}
-
-console.log(`\nBootstrap took ${((performance.now() - t0) / 1000).toFixed(1)}s.`);
+#!/usr/bin/env node
+// beir-bootstrap-significance.mjs — paired bootstrap significance test for
+// BEIR retrieval results (ADR-086).
+//
+// Given a run JSON with per-query nDCG@10 and a baseline mean nDCG@10,
+// estimates whether our result is significantly above the baseline.
+//
+// Method: paired bootstrap — resample N=10000 with replacement from our
+// per-query nDCG scores, compute mean each time, take 2.5%-97.5% percentile
+// for 95% CI. Compare baseline to our CI: if baseline < CI lower bound,
+// the difference is significant at p < 0.05.
+//
+// For a paired test against ANOTHER per-query run (e.g. our pure-BM25 vs
+// our BGE-base), resample paired DIFFERENCES instead. Pass --paired
+// path/to/baseline-run.json.
+//
+// Usage:
+//   node scripts/beir-bootstrap-significance.mjs <run-json-path>
+//   node scripts/beir-bootstrap-significance.mjs <ours> --paired <baseline>
+//
+// Example:
+//   node scripts/beir-bootstrap-significance.mjs \
+//     docs/benchmarks/runs/beir-nfcorpus-bge-latest.json
+
+import { readFileSync } from 'node:fs';
+import { performance } from 'node:perf_hooks';
+
+const args = process.argv.slice(2);
+if (args.length === 0) {
+  console.error('Usage: node beir-bootstrap-significance.mjs <run-json> [--paired <baseline-run-json>]');
+  process.exit(1);
+}
+const RUN_PATH = args[0];
+const PAIRED_IDX = args.indexOf('--paired');
+const PAIRED_PATH = PAIRED_IDX > 0 ? args[PAIRED_IDX + 1] : null;
+const ITERATIONS = Number(process.env.BOOTSTRAP_ITERATIONS) || 10000;
+
+// Deterministic PRNG (Mulberry32) — so re-runs are reproducible.
+const SEED = Number(process.env.BOOTSTRAP_SEED) || 42;
+function mulberry32(a) {
+  return function () {
+    a |= 0; a = (a + 0x6D2B79F5) | 0;
+    let t = a;
+    t = Math.imul(t ^ (t >>> 15), t | 1);
+    t ^= t + Math.imul(t ^ (t >>> 7), t | 61);
+    return ((t ^ (t >>> 14)) >>> 0) / 4294967296;
+  };
+}
+
+function mean(arr) { let s = 0; for (const v of arr) s += v; return s / arr.length; }
+
+function percentile(sortedArr, p) {
+  const idx = Math.max(0, Math.min(sortedArr.length - 1, Math.floor(p * sortedArr.length)));
+  return sortedArr[idx];
+}
+
+function pointEstimate(scores) {
+  return {
+    mean: mean(scores),
+    n: scores.length,
+  };
+}
+
+function bootstrapCI(scores, iterations, rng, percentiles = [0.025, 0.975]) {
+  const means = new Float64Array(iterations);
+  const n = scores.length;
+  for (let i = 0; i < iterations; i++) {
+    let s = 0;
+    for (let j = 0; j < n; j++) s += scores[Math.floor(rng() * n)];
+    means[i] = s / n;
+  }
+  const sorted = Array.from(means).sort((a, b) => a - b);
+  return {
+    point: mean(scores),
+    lower: percentile(sorted, percentiles[0]),
+    upper: percentile(sorted, percentiles[1]),
+    p_value_vs_baseline_zero: null,
+  };
+}
+
+// Paired bootstrap: compare our per-query scores to a baseline's per-query
+// scores. Resample (idx) and average paired differences. The 95% CI on the
+// mean difference tells us if it's significantly non-zero.
+function pairedBootstrap(oursByQid, baselineByQid, iterations, rng) {
+  const sharedIds = [...oursByQid.keys()].filter((id) => baselineByQid.has(id));
+  const diffs = sharedIds.map((id) => oursByQid.get(id) - baselineByQid.get(id));
+  const n = diffs.length;
+  const means = new Float64Array(iterations);
+  for (let i = 0; i < iterations; i++) {
+    let s = 0;
+    for (let j = 0; j < n; j++) s += diffs[Math.floor(rng() * n)];
+    means[i] = s / n;
+  }
+  const sorted = Array.from(means).sort((a, b) => a - b);
+  // One-sided p ≈ fraction of resamples where mean diff ≤ 0
+  let negCount = 0;
+  for (const m of means) if (m <= 0) negCount++;
+  return {
+    n_paired: n,
+    point_diff: mean(diffs),
+    lower_95: percentile(sorted, 0.025),
+    upper_95: percentile(sorted, 0.975),
+    p_one_sided: negCount / iterations,
+  };
+}
+
+function loadRun(path) {
+  const data = JSON.parse(readFileSync(path, 'utf-8'));
+  if (!data.perQuery || !Array.isArray(data.perQuery)) {
+    console.error(`Run JSON at ${path} has no perQuery array — re-run the bench with the updated harness.`);
+    process.exit(2);
+  }
+  return data;
+}
+
+const ours = loadRun(RUN_PATH);
+console.log(`# BEIR significance test (ADR-086)`);
+console.log(`Run: ${RUN_PATH}`);
+console.log(`Dataset: ${ours.dataset ?? '?'}  Model: ${ours.model ?? '?'}  Queries: ${ours.perQuery.length}`);
+console.log(`Bootstrap: ${ITERATIONS} iterations, seed=${SEED}`);
+
+const rng = mulberry32(SEED);
+const ourScores = ours.perQuery.map((q) => q.ndcg10);
+
+const t0 = performance.now();
+const oursCI = bootstrapCI(ourScores, ITERATIONS, rng);
+console.log(`\n=== Our nDCG@10 (1-sample bootstrap CI) ===`);
+console.log(`  point:    ${oursCI.point.toFixed(4)}`);
+console.log(`  95% CI:   [${oursCI.lower.toFixed(4)}, ${oursCI.upper.toFixed(4)}]`);
+
+if (ours.baselines) {
+  console.log(`\n=== vs each published baseline (CI overlap) ===`);
+  for (const [name, score] of Object.entries(ours.baselines)) {
+    const ciDiff = oursCI.point - score;
+    const lowerDiff = oursCI.lower - score;
+    const upperDiff = oursCI.upper - score;
+    const direction = ciDiff > 0 ? '↑ above' : '↓ below';
+    const sig = (lowerDiff > 0 && upperDiff > 0) ? 'p<0.05 (95% CI excludes baseline)'
+              : (lowerDiff < 0 && upperDiff < 0) ? 'p<0.05 (95% CI below baseline)'
+              : 'n.s. (95% CI overlaps baseline)';
+    console.log(`  ${score.toFixed(3)}  ${name.padEnd(28)}  Δ=${ciDiff >= 0 ? '+' : ''}${ciDiff.toFixed(4)}  ${direction}  [${sig}]`);
+  }
+}
+
+if (PAIRED_PATH) {
+  const baseline = loadRun(PAIRED_PATH);
+  console.log(`\n=== Paired bootstrap vs ${PAIRED_PATH} ===`);
+  const oursByQid = new Map(ours.perQuery.map((q) => [q.qid, q.ndcg10]));
+  const baseByQid = new Map(baseline.perQuery.map((q) => [q.qid, q.ndcg10]));
+  const rng2 = mulberry32(SEED + 1);
+  const paired = pairedBootstrap(oursByQid, baseByQid, ITERATIONS, rng2);
+  console.log(`  paired queries: ${paired.n_paired}`);
+  console.log(`  Δ (ours - baseline): ${paired.point_diff >= 0 ? '+' : ''}${paired.point_diff.toFixed(4)}`);
+  console.log(`  95% CI: [${paired.lower_95.toFixed(4)}, ${paired.upper_95.toFixed(4)}]`);
+  console.log(`  one-sided p (Δ ≤ 0): ${paired.p_one_sided.toFixed(4)}`);
+  const sig = paired.lower_95 > 0 ? 'SIGNIFICANT improvement (p<0.025 one-sided)' :
+              paired.upper_95 < 0 ? 'SIGNIFICANT regression (p<0.025 one-sided)' :
+              'NOT SIGNIFICANT — CI overlaps zero';
+  console.log(`  verdict: ${sig}`);
+}
+
+console.log(`\nBootstrap took ${((performance.now() - t0) / 1000).toFixed(1)}s.`);