@staticn0va/wigolo 0.1.0 → 0.1.2

package/dist/search/reranker/authority-boost.js

@@ -0,0 +1,179 @@
+const STOPWORDS = /* @__PURE__ */ new Set([
+  "the",
+  "a",
+  "an",
+  "what",
+  "is",
+  "are",
+  "was",
+  "were",
+  "how",
+  "why",
+  "when",
+  "where",
+  "who",
+  "do",
+  "does",
+  "did",
+  "for",
+  "of",
+  "to",
+  "in",
+  "on",
+  "with",
+  "and",
+  "or",
+  "but",
+  "as",
+  "at",
+  "by",
+  "from",
+  "into",
+  "about",
+  "than",
+  "this",
+  "that",
+  "these",
+  "those",
+  "it",
+  "its",
+  "be",
+  "been",
+  "has",
+  "have",
+  "had",
+  "can",
+  "could",
+  "should",
+  "would",
+  "may",
+  "might",
+  "must",
+  "will",
+  "shall",
+  "i",
+  "you",
+  "we",
+  "they",
+  "he",
+  "she",
+  "them",
+  "my",
+  "your",
+  "our",
+  "their",
+  "latest",
+  "current",
+  "newest",
+  "recent",
+  "best",
+  "top",
+  "most"
+]);
+const AUTHORITATIVE_TLD = /\.(io|org|dev|edu|gov)$/;
+const KNOWN_DOCS_HOSTS = /* @__PURE__ */ new Set([
+  "docs.python.org",
+  "developer.mozilla.org",
+  "kubernetes.io",
+  "cloud.google.com",
+  "aws.amazon.com",
+  "docs.aws.amazon.com",
+  "learn.microsoft.com",
+  "docs.microsoft.com",
+  "developer.apple.com",
+  "docs.docker.com",
+  "docs.npmjs.com",
+  "docs.github.com",
+  "docs.anthropic.com"
+]);
+const KNOWN_SUBJECT_DOMAIN = {
+  redis: ["redis.io", "redis.com"],
+  postgres: ["postgresql.org"],
+  postgresql: ["postgresql.org"],
+  mysql: ["mysql.com", "dev.mysql.com"],
+  python: ["python.org", "docs.python.org"],
+  react: ["react.dev", "reactjs.org"],
+  nextjs: ["nextjs.org"],
+  vue: ["vuejs.org"],
+  angular: ["angular.io", "angular.dev"],
+  node: ["nodejs.org"],
+  nodejs: ["nodejs.org"],
+  rust: ["rust-lang.org", "doc.rust-lang.org"],
+  go: ["go.dev", "golang.org"],
+  golang: ["go.dev", "golang.org"],
+  typescript: ["typescriptlang.org"],
+  javascript: ["developer.mozilla.org"],
+  anthropic: ["anthropic.com", "docs.anthropic.com"],
+  openai: ["openai.com", "platform.openai.com"],
+  google: ["google.com", "cloud.google.com"],
+  microsoft: ["microsoft.com", "learn.microsoft.com"],
+  apple: ["apple.com", "developer.apple.com"],
+  github: ["github.com", "docs.github.com"],
+  gitlab: ["gitlab.com"],
+  docker: ["docker.com", "docs.docker.com"],
+  kubernetes: ["kubernetes.io"],
+  k8s: ["kubernetes.io"],
+  aws: ["aws.amazon.com", "docs.aws.amazon.com"],
+  azure: ["azure.microsoft.com", "learn.microsoft.com"],
+  gcp: ["cloud.google.com"],
+  npm: ["npmjs.com", "docs.npmjs.com"],
+  pnpm: ["pnpm.io"],
+  yarn: ["yarnpkg.com"],
+  mcp: ["modelcontextprotocol.io", "spec.modelcontextprotocol.io", "docs.anthropic.com"]
+};
+function extractSubjects(query) {
+  const tokens = query.toLowerCase().replace(/[^a-z0-9\s]/g, " ").split(/\s+/).filter((t) => t.length >= 2 && t.length <= 16 && !STOPWORDS.has(t));
+  return [...new Set(tokens)];
+}
+function hostOf(url) {
+  try {
+    return new URL(url).hostname.toLowerCase();
+  } catch {
+    return null;
+  }
+}
+function applyAuthorityBoost(query, results) {
+  if (results.length === 0) return results;
+  const subjects = extractSubjects(query);
+  const knownDomains = /* @__PURE__ */ new Set();
+  for (const s of subjects) {
+    const mapped = KNOWN_SUBJECT_DOMAIN[s];
+    if (mapped) for (const d of mapped) knownDomains.add(d);
+  }
+  return results.map((r) => {
+    const host = hostOf(r.url);
+    if (!host) return r;
+    let boost = 0;
+    if (knownDomains.has(host)) boost += 0.2;
+    else for (const dom of knownDomains) {
+      if (host.endsWith(`.${dom}`)) {
+        boost += 0.18;
+        break;
+      }
+    }
+    if (boost === 0) {
+      for (const subj of subjects) {
+        if (host === `${subj}.io` || host === `${subj}.com` || host === `${subj}.org` || host === `${subj}.dev`) {
+          boost += 0.15;
+          break;
+        }
+        if (host.startsWith(`${subj}.`) || host.includes(`.${subj}.`)) {
+          boost += 0.1;
+          break;
+        }
+      }
+    }
+    if (KNOWN_DOCS_HOSTS.has(host)) boost = Math.max(boost, 0.18);
+    else if (host.startsWith("docs.")) boost += 0.08;
+    if (boost === 0 && AUTHORITATIVE_TLD.test(host)) boost += 0.04;
+    if (boost === 0) return r;
+    return {
+      ...r,
+      relevance_score: Math.min(1, r.relevance_score + boost)
+    };
+  });
+}
+export {
+  applyAuthorityBoost
+};
+//# sourceMappingURL=authority-boost.js.map

package/dist/search/reranker/authority-boost.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../src/search/reranker/authority-boost.ts"],"sourcesContent":["import type { MergedSearchResult } from '../dedup.js';\n\nconst STOPWORDS = new Set([\n  'the', 'a', 'an', 'what', 'is', 'are', 'was', 'were', 'how', 'why', 'when', 'where', 'who',\n  'do', 'does', 'did', 'for', 'of', 'to', 'in', 'on', 'with', 'and', 'or', 'but', 'as', 'at',\n  'by', 'from', 'into', 'about', 'than', 'this', 'that', 'these', 'those', 'it', 'its', 'be',\n  'been', 'has', 'have', 'had', 'can', 'could', 'should', 'would', 'may', 'might', 'must',\n  'will', 'shall', 'i', 'you', 'we', 'they', 'he', 'she', 'them', 'my', 'your', 'our', 'their',\n  'latest', 'current', 'newest', 'recent', 'best', 'top', 'most',\n]);\n\nconst AUTHORITATIVE_TLD = /\\.(io|org|dev|edu|gov)$/;\nconst KNOWN_DOCS_HOSTS = new Set([\n  'docs.python.org', 'developer.mozilla.org', 'kubernetes.io', 'cloud.google.com',\n  'aws.amazon.com', 'docs.aws.amazon.com', 'learn.microsoft.com', 'docs.microsoft.com',\n  'developer.apple.com', 'docs.docker.com', 'docs.npmjs.com', 'docs.github.com',\n  'docs.anthropic.com',\n]);\n\nconst KNOWN_SUBJECT_DOMAIN: Record<string, string[]> = {\n  redis: ['redis.io', 'redis.com'],\n  postgres: ['postgresql.org'],\n  postgresql: ['postgresql.org'],\n  mysql: ['mysql.com', 'dev.mysql.com'],\n  python: ['python.org', 'docs.python.org'],\n  react: ['react.dev', 'reactjs.org'],\n  nextjs: ['nextjs.org'],\n  vue: ['vuejs.org'],\n  angular: ['angular.io', 'angular.dev'],\n  node: ['nodejs.org'],\n  nodejs: ['nodejs.org'],\n  rust: ['rust-lang.org', 'doc.rust-lang.org'],\n  go: ['go.dev', 'golang.org'],\n  golang: ['go.dev', 'golang.org'],\n  typescript: ['typescriptlang.org'],\n  javascript: ['developer.mozilla.org'],\n  anthropic: ['anthropic.com', 'docs.anthropic.com'],\n  openai: ['openai.com', 'platform.openai.com'],\n  google: ['google.com', 'cloud.google.com'],\n  microsoft: ['microsoft.com', 'learn.microsoft.com'],\n  apple: ['apple.com', 'developer.apple.com'],\n  github: ['github.com', 'docs.github.com'],\n  gitlab: ['gitlab.com'],\n  docker: ['docker.com', 'docs.docker.com'],\n  kubernetes: ['kubernetes.io'],\n  k8s: ['kubernetes.io'],\n  aws: ['aws.amazon.com', 'docs.aws.amazon.com'],\n  azure: ['azure.microsoft.com', 'learn.microsoft.com'],\n  gcp: ['cloud.google.com'],\n  npm: ['npmjs.com', 'docs.npmjs.com'],\n  pnpm: ['pnpm.io'],\n  yarn: ['yarnpkg.com'],\n  mcp: ['modelcontextprotocol.io', 'spec.modelcontextprotocol.io', 'docs.anthropic.com'],\n};\n\nfunction extractSubjects(query: string): string[] {\n  const tokens = query\n    .toLowerCase()\n    .replace(/[^a-z0-9\\s]/g, ' ')\n    .split(/\\s+/)\n    .filter((t) => t.length >= 2 && t.length <= 16 && !STOPWORDS.has(t));\n  return [...new Set(tokens)];\n}\n\nfunction hostOf(url: string): string | null {\n  try {\n    return new URL(url).hostname.toLowerCase();\n  } catch {\n    return null;\n  }\n}\n\nexport function applyAuthorityBoost(\n  query: string,\n  results: MergedSearchResult[],\n): MergedSearchResult[] {\n  if (results.length === 0) return results;\n  const subjects = extractSubjects(query);\n  const knownDomains = new Set<string>();\n  for (const s of subjects) {\n    const mapped = KNOWN_SUBJECT_DOMAIN[s];\n    if (mapped) for (const d of mapped) knownDomains.add(d);\n  }\n\n  return results.map((r) => {\n    const host = hostOf(r.url);\n    if (!host) return r;\n\n    let boost = 0;\n\n    if (knownDomains.has(host)) boost += 0.20;\n    else for (const dom of knownDomains) {\n      if (host.endsWith(`.${dom}`)) { boost += 0.18; break; }\n    }\n\n    if (boost === 0) {\n      for (const subj of subjects) {\n        if (host === `${subj}.io` || host === `${subj}.com` || host === `${subj}.org` || host === `${subj}.dev`) {\n          boost += 0.15;\n          break;\n        }\n        if (host.startsWith(`${subj}.`) || host.includes(`.${subj}.`)) {\n          boost += 0.10;\n          break;\n        }\n      }\n    }\n\n    if (KNOWN_DOCS_HOSTS.has(host)) boost = Math.max(boost, 0.18);\n    else if (host.startsWith('docs.')) boost += 0.08;\n\n    if (boost === 0 && AUTHORITATIVE_TLD.test(host)) boost += 0.04;\n\n    if (boost === 0) return r;\n\n    return {\n      ...r,\n      relevance_score: Math.min(1, r.relevance_score + boost),\n    };\n  });\n}\n"],"mappings":"AAEA,MAAM,YAAY,oBAAI,IAAI;AAAA,EACxB;AAAA,EAAO;AAAA,EAAK;AAAA,EAAM;AAAA,EAAQ;AAAA,EAAM;AAAA,EAAO;AAAA,EAAO;AAAA,EAAQ;AAAA,EAAO;AAAA,EAAO;AAAA,EAAQ;AAAA,EAAS;AAAA,EACrF;AAAA,EAAM;AAAA,EAAQ;AAAA,EAAO;AAAA,EAAO;AAAA,EAAM;AAAA,EAAM;AAAA,EAAM;AAAA,EAAM;AAAA,EAAQ;AAAA,EAAO;AAAA,EAAM;AAAA,EAAO;AAAA,EAAM;AAAA,EACtF;AAAA,EAAM;AAAA,EAAQ;AAAA,EAAQ;AAAA,EAAS;AAAA,EAAQ;AAAA,EAAQ;AAAA,EAAQ;AAAA,EAAS;AAAA,EAAS;AAAA,EAAM;AAAA,EAAO;AAAA,EACtF;AAAA,EAAQ;AAAA,EAAO;AAAA,EAAQ;AAAA,EAAO;AAAA,EAAO;AAAA,EAAS;AAAA,EAAU;AAAA,EAAS;AAAA,EAAO;AAAA,EAAS;AAAA,EACjF;AAAA,EAAQ;AAAA,EAAS;AAAA,EAAK;AAAA,EAAO;AAAA,EAAM;AAAA,EAAQ;AAAA,EAAM;AAAA,EAAO;AAAA,EAAQ;AAAA,EAAM;AAAA,EAAQ;AAAA,EAAO;AAAA,EACrF;AAAA,EAAU;AAAA,EAAW;AAAA,EAAU;AAAA,EAAU;AAAA,EAAQ;AAAA,EAAO;AAC1D,CAAC;AAED,MAAM,oBAAoB;AAC1B,MAAM,mBAAmB,oBAAI,IAAI;AAAA,EAC/B;AAAA,EAAmB;AAAA,EAAyB;AAAA,EAAiB;AAAA,EAC7D;AAAA,EAAkB;AAAA,EAAuB;AAAA,EAAuB;AAAA,EAChE;AAAA,EAAuB;AAAA,EAAmB;AAAA,EAAkB;AAAA,EAC5D;AACF,CAAC;AAED,MAAM,uBAAiD;AAAA,EACrD,OAAO,CAAC,YAAY,WAAW;AAAA,EAC/B,UAAU,CAAC,gBAAgB;AAAA,EAC3B,YAAY,CAAC,gBAAgB;AAAA,EAC7B,OAAO,CAAC,aAAa,eAAe;AAAA,EACpC,QAAQ,CAAC,cAAc,iBAAiB;AAAA,EACxC,OAAO,CAAC,aAAa,aAAa;AAAA,EAClC,QAAQ,CAAC,YAAY;AAAA,EACrB,KAAK,CAAC,WAAW;AAAA,EACjB,SAAS,CAAC,cAAc,aAAa;AAAA,EACrC,MAAM,CAAC,YAAY;AAAA,EACnB,QAAQ,CAAC,YAAY;AAAA,EACrB,MAAM,CAAC,iBAAiB,mBAAmB;AAAA,EAC3C,IAAI,CAAC,UAAU,YAAY;AAAA,EAC3B,QAAQ,CAAC,UAAU,YAAY;AAAA,EAC/B,YAAY,CAAC,oBAAoB;AAAA,EACjC,YAAY,CAAC,uBAAuB;AAAA,EACpC,WAAW,CAAC,iBAAiB,oBAAoB;AAAA,EACjD,QAAQ,CAAC,cAAc,qBAAqB;AAAA,EAC5C,QAAQ,CAAC,cAAc,kBAAkB;AAAA,EACzC,WAAW,CAAC,iBAAiB,qBAAqB;AAAA,EAClD,OAAO,CAAC,aAAa,qBAAqB;AAAA,EAC1C,QAAQ,CAAC,cAAc,iBAAiB;AAAA,EACxC,QAAQ,CAAC,YAAY;AAAA,EACrB,QAAQ,CAAC,cAAc,iBAAiB;AAAA,EACxC,YAAY,CAAC,eAAe;AAAA,EAC5B,KAAK,CAAC,eAAe;AAAA,EACrB,KAAK,CAAC,kBAAkB,qBAAqB;AAAA,EAC7C,OAAO,CAAC,uBAAuB,qBAAqB;AAAA,EACpD,KAAK,CAAC,kBAAkB;AAAA,EACxB,KAAK,CAAC,aAAa,gBAAgB;AAAA,EACnC,MAAM,CAAC,SAAS;AAAA,EAChB,MAAM,CAAC,aAAa;AAAA,EACpB,KAAK,CAAC,2BAA2B,gCAAgC,oBAAoB;AACvF;AAEA,SAAS,gBAAgB,OAAyB;AAChD,QAAM,SAAS,MACZ,YAAY,EACZ,QAAQ,gBAAgB,GAAG,EAC3B,MAAM,KAAK,EACX,OAAO,CAAC,MAAM,EAAE,UAAU,KAAK,EAAE,UAAU,MAAM,CAAC,UAAU,IAAI,CAAC,CAAC;AACrE,SAAO,CAAC,GAAG,IAAI,IAAI,MAAM,CAAC;AAC5B;AAEA,SAAS,OAAO,KAA4B;AAC1C,MAAI;AACF,WAAO,IAAI,IAAI,GAAG,EAAE,SAAS,YAAY;AAAA,EAC3C,QAAQ;AACN,WAAO;AAAA,EACT;AACF;AAEO,SAAS,oBACd,OACA,SACsB;AACtB,MAAI,QAAQ,WAAW,EAAG,QAAO;AACjC,QAAM,WAAW,gBAAgB,KAAK;AACtC,QAAM,eAAe,oBAAI,IAAY;AACrC,aAAW,KAAK,UAAU;AACxB,UAAM,SAAS,qBAAqB,CAAC;AACrC,QAAI,OAAQ,YAAW,KAAK,OAAQ,cAAa,IAAI,CAAC;AAAA,EACxD;AAEA,SAAO,QAAQ,IAAI,CAAC,MAAM;AACxB,UAAM,OAAO,OAAO,EAAE,GAAG;AACzB,QAAI,CAAC,KAAM,QAAO;AAElB,QAAI,QAAQ;AAEZ,QAAI,aAAa,IAAI,IAAI,EAAG,UAAS;AAAA,QAChC,YAAW,OAAO,cAAc;AACnC,UAAI,KAAK,SAAS,IAAI,GAAG,EAAE,GAAG;AAAE,iBAAS;AAAM;AAAA,MAAO;AAAA,IACxD;AAEA,QAAI,UAAU,GAAG;AACf,iBAAW,QAAQ,UAAU;AAC3B,YAAI,SAAS,GAAG,IAAI,SAAS,SAAS,GAAG,IAAI,UAAU,SAAS,GAAG,IAAI,UAAU,SAAS,GAAG,IAAI,QAAQ;AACvG,mBAAS;AACT;AAAA,QACF;AACA,YAAI,KAAK,WAAW,GAAG,IAAI,GAAG,KAAK,KAAK,SAAS,IAAI,IAAI,GAAG,GAAG;AAC7D,mBAAS;AACT;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,QAAI,iBAAiB,IAAI,IAAI,EAAG,SAAQ,KAAK,IAAI,OAAO,IAAI;AAAA,aACnD,KAAK,WAAW,OAAO,EAAG,UAAS;AAE5C,QAAI,UAAU,KAAK,kBAAkB,KAAK,IAAI,EAAG,UAAS;AAE1D,QAAI,UAAU,EAAG,QAAO;AAExB,WAAO;AAAA,MACL,GAAG;AAAA,MACH,iBAAiB,KAAK,IAAI,GAAG,EAAE,kBAAkB,KAAK;AAAA,IACxD;AAAA,EACF,CAAC;AACH;","names":[]}

package/dist/search/reranker/consensus-boost.d.ts

@@ -0,0 +1,3 @@
+import type { MergedSearchResult } from '../dedup.js';
+export declare function applyConsensusBoost(results: MergedSearchResult[]): MergedSearchResult[];
+//# sourceMappingURL=consensus-boost.d.ts.map

package/dist/search/reranker/consensus-boost.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"consensus-boost.d.ts","sourceRoot":"","sources":["../../../src/search/reranker/consensus-boost.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,kBAAkB,EAAE,MAAM,aAAa,CAAC;AAoBtD,wBAAgB,mBAAmB,CAAC,OAAO,EAAE,kBAAkB,EAAE,GAAG,kBAAkB,EAAE,CAavF"}

package/dist/search/reranker/consensus-boost.js

@@ -0,0 +1,27 @@
+const CONSENSUS_BOOST_BY_ENGINE_COUNT = {
+  1: 0,
+  2: 0.05,
+  3: 0.1
+};
+const CAP_AT_OR_ABOVE_4 = 0.12;
+function applyConsensusBoost(results) {
+  if (results.length === 0) return results;
+  return results.map((r) => {
+    const n = uniqueEngines(r.engines);
+    if (n <= 1) return r;
+    const boost = n >= 4 ? CAP_AT_OR_ABOVE_4 : CONSENSUS_BOOST_BY_ENGINE_COUNT[n] ?? 0;
+    if (boost === 0) return r;
+    return {
+      ...r,
+      relevance_score: Math.min(1, r.relevance_score + boost)
+    };
+  });
+}
+function uniqueEngines(engines) {
+  if (!engines || engines.length === 0) return 0;
+  return new Set(engines.map((e) => e.toLowerCase())).size;
+}
+export {
+  applyConsensusBoost
+};
+//# sourceMappingURL=consensus-boost.js.map

package/dist/search/reranker/consensus-boost.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../src/search/reranker/consensus-boost.ts"],"sourcesContent":["import type { MergedSearchResult } from '../dedup.js';\n\n// Engine-consensus boost (Q7). Replaces per-subject authority enumeration with a\n// data-driven signal: URLs returned by multiple independent search engines are\n// likely authoritative regardless of query topic. Generalizes to all subjects;\n// no hand-curated map.\n//\n// Provenance: Reciprocal Rank Fusion's consensus property — documents agreed on\n// by multiple rankers tend to be relevant (Cormack et al., SIGIR 2009;\n// OpenSearch 2.19 RRF documentation; Azure AI Search RRF). We surface engine\n// agreement as a first-class re-rank signal in addition to the existing RRF\n// merge in dedup.ts.\n\nconst CONSENSUS_BOOST_BY_ENGINE_COUNT: Record<number, number> = {\n  1: 0,\n  2: 0.05,\n  3: 0.10,\n};\nconst CAP_AT_OR_ABOVE_4 = 0.12;\n\nexport function applyConsensusBoost(results: MergedSearchResult[]): MergedSearchResult[] {\n  if (results.length === 0) return results;\n\n  return results.map((r) => {\n    const n = uniqueEngines(r.engines);\n    if (n <= 1) return r;\n    const boost = n >= 4 ? CAP_AT_OR_ABOVE_4 : CONSENSUS_BOOST_BY_ENGINE_COUNT[n] ?? 0;\n    if (boost === 0) return r;\n    return {\n      ...r,\n      relevance_score: Math.min(1, r.relevance_score + boost),\n    };\n  });\n}\n\nfunction uniqueEngines(engines: string[] | undefined): number {\n  if (!engines || engines.length === 0) return 0;\n  return new Set(engines.map((e) => e.toLowerCase())).size;\n}\n"],"mappings":"AAaA,MAAM,kCAA0D;AAAA,EAC9D,GAAG;AAAA,EACH,GAAG;AAAA,EACH,GAAG;AACL;AACA,MAAM,oBAAoB;AAEnB,SAAS,oBAAoB,SAAqD;AACvF,MAAI,QAAQ,WAAW,EAAG,QAAO;AAEjC,SAAO,QAAQ,IAAI,CAAC,MAAM;AACxB,UAAM,IAAI,cAAc,EAAE,OAAO;AACjC,QAAI,KAAK,EAAG,QAAO;AACnB,UAAM,QAAQ,KAAK,IAAI,oBAAoB,gCAAgC,CAAC,KAAK;AACjF,QAAI,UAAU,EAAG,QAAO;AACxB,WAAO;AAAA,MACL,GAAG;AAAA,MACH,iBAAiB,KAAK,IAAI,GAAG,EAAE,kBAAkB,KAAK;AAAA,IACxD;AAAA,EACF,CAAC;AACH;AAEA,SAAS,cAAc,SAAuC;AAC5D,MAAI,CAAC,WAAW,QAAQ,WAAW,EAAG,QAAO;AAC7C,SAAO,IAAI,IAAI,QAAQ,IAAI,CAAC,MAAM,EAAE,YAAY,CAAC,CAAC,EAAE;AACtD;","names":[]}

package/dist/search/reranker/recency-boost.d.ts

@@ -0,0 +1,3 @@
+import type { MergedSearchResult } from '../dedup.js';
+export declare function applyRecencyBoost(query: string, results: MergedSearchResult[], now?: Date): MergedSearchResult[];
+//# sourceMappingURL=recency-boost.d.ts.map

package/dist/search/reranker/recency-boost.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"recency-boost.d.ts","sourceRoot":"","sources":["../../../src/search/reranker/recency-boost.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,kBAAkB,EAAE,MAAM,aAAa,CAAC;AAGtD,wBAAgB,iBAAiB,CAC/B,KAAK,EAAE,MAAM,EACb,OAAO,EAAE,kBAAkB,EAAE,EAC7B,GAAG,GAAE,IAAiB,GACrB,kBAAkB,EAAE,CAOtB"}

package/dist/search/reranker/recency-boost.js

@@ -0,0 +1,13 @@
+import { hasRecencyIntent, recencyFactor } from "./recency.js";
+function applyRecencyBoost(query, results, now = /* @__PURE__ */ new Date()) {
+  if (!hasRecencyIntent(query, now)) return results;
+  return results.map((r) => {
+    const factor = recencyFactor(r.published_date, now);
+    if (factor === 1) return r;
+    return { ...r, relevance_score: r.relevance_score * factor };
+  });
+}
+export {
+  applyRecencyBoost
+};
+//# sourceMappingURL=recency-boost.js.map

package/dist/search/reranker/recency-boost.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../src/search/reranker/recency-boost.ts"],"sourcesContent":["import type { MergedSearchResult } from '../dedup.js';\nimport { hasRecencyIntent, recencyFactor } from './recency.js';\n\nexport function applyRecencyBoost(\n  query: string,\n  results: MergedSearchResult[],\n  now: Date = new Date(),\n): MergedSearchResult[] {\n  if (!hasRecencyIntent(query, now)) return results;\n  return results.map((r) => {\n    const factor = recencyFactor(r.published_date, now);\n    if (factor === 1.0) return r;\n    return { ...r, relevance_score: r.relevance_score * factor };\n  });\n}\n"],"mappings":"AACA,SAAS,kBAAkB,qBAAqB;AAEzC,SAAS,kBACd,OACA,SACA,MAAY,oBAAI,KAAK,GACC;AACtB,MAAI,CAAC,iBAAiB,OAAO,GAAG,EAAG,QAAO;AAC1C,SAAO,QAAQ,IAAI,CAAC,MAAM;AACxB,UAAM,SAAS,cAAc,EAAE,gBAAgB,GAAG;AAClD,QAAI,WAAW,EAAK,QAAO;AAC3B,WAAO,EAAE,GAAG,GAAG,iBAAiB,EAAE,kBAAkB,OAAO;AAAA,EAC7D,CAAC;AACH;","names":[]}

package/dist/search/reranker/recency.d.ts

@@ -0,0 +1,3 @@
+export declare function hasRecencyIntent(query: string, now?: Date): boolean;
+export declare function recencyFactor(publishedDate: string | undefined, now?: Date): number;
+//# sourceMappingURL=recency.d.ts.map

package/dist/search/reranker/recency.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"recency.d.ts","sourceRoot":"","sources":["../../../src/search/reranker/recency.ts"],"names":[],"mappings":"AAEA,wBAAgB,gBAAgB,CAAC,KAAK,EAAE,MAAM,EAAE,GAAG,GAAE,IAAiB,GAAG,OAAO,CAM/E;AAED,wBAAgB,aAAa,CAAC,aAAa,EAAE,MAAM,GAAG,SAAS,EAAE,GAAG,GAAE,IAAiB,GAAG,MAAM,CAS/F"}

package/dist/search/reranker/recency.js

@@ -0,0 +1,23 @@
+const RECENCY_TOKENS = /\b(recent|latest|new|just released|today|this week)\b/i;
+function hasRecencyIntent(query, now = /* @__PURE__ */ new Date()) {
+  if (RECENCY_TOKENS.test(query)) return true;
+  const yearMatches = query.match(/\b(20\d{2})\b/g);
+  if (!yearMatches) return false;
+  const currentYear = now.getUTCFullYear();
+  return yearMatches.some((y) => parseInt(y, 10) >= currentYear);
+}
+function recencyFactor(publishedDate, now = /* @__PURE__ */ new Date()) {
+  if (!publishedDate) return 1;
+  const ts = new Date(publishedDate).getTime();
+  if (isNaN(ts)) return 1;
+  const ageDays = (now.getTime() - ts) / (1e3 * 60 * 60 * 24);
+  if (ageDays < 7) return 1.5;
+  if (ageDays < 30) return 1.3;
+  if (ageDays < 90) return 1.1;
+  return 1;
+}
+export {
+  hasRecencyIntent,
+  recencyFactor
+};
+//# sourceMappingURL=recency.js.map

package/dist/search/reranker/recency.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../src/search/reranker/recency.ts"],"sourcesContent":["const RECENCY_TOKENS = /\\b(recent|latest|new|just released|today|this week)\\b/i;\n\nexport function hasRecencyIntent(query: string, now: Date = new Date()): boolean {\n  if (RECENCY_TOKENS.test(query)) return true;\n  const yearMatches = query.match(/\\b(20\\d{2})\\b/g);\n  if (!yearMatches) return false;\n  const currentYear = now.getUTCFullYear();\n  return yearMatches.some((y) => parseInt(y, 10) >= currentYear);\n}\n\nexport function recencyFactor(publishedDate: string | undefined, now: Date = new Date()): number {\n  if (!publishedDate) return 1.0;\n  const ts = new Date(publishedDate).getTime();\n  if (isNaN(ts)) return 1.0;\n  const ageDays = (now.getTime() - ts) / (1000 * 60 * 60 * 24);\n  if (ageDays < 7) return 1.5;\n  if (ageDays < 30) return 1.3;\n  if (ageDays < 90) return 1.1;\n  return 1.0;\n}\n"],"mappings":"AAAA,MAAM,iBAAiB;AAEhB,SAAS,iBAAiB,OAAe,MAAY,oBAAI,KAAK,GAAY;AAC/E,MAAI,eAAe,KAAK,KAAK,EAAG,QAAO;AACvC,QAAM,cAAc,MAAM,MAAM,gBAAgB;AAChD,MAAI,CAAC,YAAa,QAAO;AACzB,QAAM,cAAc,IAAI,eAAe;AACvC,SAAO,YAAY,KAAK,CAAC,MAAM,SAAS,GAAG,EAAE,KAAK,WAAW;AAC/D;AAEO,SAAS,cAAc,eAAmC,MAAY,oBAAI,KAAK,GAAW;AAC/F,MAAI,CAAC,cAAe,QAAO;AAC3B,QAAM,KAAK,IAAI,KAAK,aAAa,EAAE,QAAQ;AAC3C,MAAI,MAAM,EAAE,EAAG,QAAO;AACtB,QAAM,WAAW,IAAI,QAAQ,IAAI,OAAO,MAAO,KAAK,KAAK;AACzD,MAAI,UAAU,EAAG,QAAO;AACxB,MAAI,UAAU,GAAI,QAAO;AACzB,MAAI,UAAU,GAAI,QAAO;AACzB,SAAO;AACT;","names":[]}

package/dist/search/reranker/transformers-rerank-provider.d.ts

@@ -0,0 +1,13 @@
+import type { RerankProvider, RerankCandidate, RerankResult } from '../../providers/rerank-provider.js';
+export declare class TransformersRerankProvider implements RerankProvider {
+    private tokenizer;
+    private model;
+    private loadPromise;
+    readonly modelId: string;
+    constructor();
+    warmup(): Promise<void>;
+    private load;
+    rerank(query: string, candidates: RerankCandidate[], topK?: number): Promise<RerankResult[]>;
+    dispose(): Promise<void>;
+}
+//# sourceMappingURL=transformers-rerank-provider.d.ts.map

package/dist/search/reranker/transformers-rerank-provider.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"transformers-rerank-provider.d.ts","sourceRoot":"","sources":["../../../src/search/reranker/transformers-rerank-provider.ts"],"names":[],"mappings":"AAMA,OAAO,KAAK,EACV,cAAc,EACd,eAAe,EACf,YAAY,EACb,MAAM,oCAAoC,CAAC;AAwC5C,qBAAa,0BAA2B,YAAW,cAAc;IAC/D,OAAO,CAAC,SAAS,CAA0B;IAC3C,OAAO,CAAC,KAAK,CAAsB;IACnC,OAAO,CAAC,WAAW,CAAgE;IACnF,QAAQ,CAAC,OAAO,EAAE,MAAM,CAAC;;IAMnB,MAAM,IAAI,OAAO,CAAC,IAAI,CAAC;IAI7B,OAAO,CAAC,IAAI;IA6BN,MAAM,CACV,KAAK,EAAE,MAAM,EACb,UAAU,EAAE,eAAe,EAAE,EAC7B,IAAI,SAAoB,GACvB,OAAO,CAAC,YAAY,EAAE,CAAC;IAkCpB,OAAO,IAAI,OAAO,CAAC,IAAI,CAAC;CAa/B"}

package/dist/search/reranker/transformers-rerank-provider.js

@@ -0,0 +1,94 @@
+import { join } from "node:path";
+import {
+  AutoTokenizer,
+  AutoModelForSequenceClassification,
+  env
+} from "@huggingface/transformers";
+import { createLogger } from "../../logger.js";
+import { getConfig } from "../../config.js";
+const log = createLogger("reranker");
+function wrapLoadError(err) {
+  const message = err instanceof Error ? err.message : String(err);
+  const looksLikeMissingModel = /tokenizer_class|tokenizer_config|preprocessor_config|fetch failed|ENOTFOUND|ECONNREFUSED|ETIMEDOUT|ENETUNREACH/i.test(
+    message
+  );
+  if (looksLikeMissingModel) {
+    return new Error(
+      `Reranker model not downloaded \u2014 run \`wigolo warmup\` (cause: ${message})`
+    );
+  }
+  return new Error(`Failed to load reranker model: ${message}`);
+}
+class TransformersRerankProvider {
+  tokenizer = null;
+  model = null;
+  loadPromise = null;
+  modelId;
+  constructor() {
+    this.modelId = "Xenova/ms-marco-MiniLM-L-6-v2";
+  }
+  async warmup() {
+    await this.load();
+  }
+  load() {
+    if (this.tokenizer && this.model) {
+      return Promise.resolve({ tokenizer: this.tokenizer, model: this.model });
+    }
+    if (this.loadPromise) return this.loadPromise;
+    log.info("Loading rerank model", { modelId: this.modelId });
+    const cacheDir = join(getConfig().dataDir, "transformers");
+    env.cacheDir = cacheDir;
+    this.loadPromise = Promise.all([
+      AutoTokenizer.from_pretrained(this.modelId),
+      AutoModelForSequenceClassification.from_pretrained(this.modelId)
+    ]).then(([tokenizer, model]) => {
+      this.tokenizer = tokenizer;
+      this.model = model;
+      return { tokenizer, model };
+    }).catch((err) => {
+      this.loadPromise = null;
+      throw wrapLoadError(err);
+    });
+    return this.loadPromise;
+  }
+  async rerank(query, candidates, topK = candidates.length) {
+    if (candidates.length === 0) return [];
+    const { tokenizer, model } = await this.load();
+    const queries = candidates.map(() => query);
+    const docs = candidates.map((c) => c.text);
+    const inputs = tokenizer(queries, {
+      text_pair: docs,
+      padding: true,
+      truncation: true
+    });
+    const outputs = await model(inputs);
+    const logits = outputs.logits;
+    const stride = logits.dims.length >= 2 ? logits.dims[1] : 1;
+    const data = logits.data;
+    const scored = candidates.map((c, i) => ({
+      id: c.id,
+      score: Number(data[i * stride])
+    }));
+    return scored.sort((a, b) => b.score - a.score).slice(0, topK);
+  }
+  // Release the underlying ONNX session before process exit. Without this,
+  // the runtime's worker threads race during C++ destructor teardown and
+  // surface as `mutex lock failed: Invalid argument` on macOS.
+  async dispose() {
+    const model = this.model;
+    this.model = null;
+    this.tokenizer = null;
+    this.loadPromise = null;
+    if (!model) return;
+    try {
+      const m = model;
+      if (typeof m.dispose === "function") await m.dispose();
+    } catch (err) {
+      log.debug("reranker dispose failed", { error: err instanceof Error ? err.message : String(err) });
+    }
+  }
+}
+export {
+  TransformersRerankProvider
+};
+//# sourceMappingURL=transformers-rerank-provider.js.map

package/dist/search/reranker/transformers-rerank-provider.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../src/search/reranker/transformers-rerank-provider.ts"],"sourcesContent":["import { join } from 'node:path';\nimport {\n  AutoTokenizer,\n  AutoModelForSequenceClassification,\n  env,\n} from '@huggingface/transformers';\nimport type {\n  RerankProvider,\n  RerankCandidate,\n  RerankResult,\n} from '../../providers/rerank-provider.js';\nimport { createLogger } from '../../logger.js';\nimport { getConfig } from '../../config.js';\n\nconst log = createLogger('reranker');\n\n// Cross-encoder reranker via Transformers.js.\n//\n// The high-level `pipeline('text-classification', ...)` API does not pass\n// `text_pair`, so it can't drive a cross-encoder properly. We therefore\n// load the tokenizer + sequence-classification model directly: feed\n// (query, document) pairs to the tokenizer and read raw logits from the\n// model. ms-marco-MiniLM-L-6-v2 is a single-output regressor (num_labels=1)\n// where higher logit = more relevant, so the logit is used as the rerank\n// score with no further transform.\ntype Tokenizer = Awaited<ReturnType<typeof AutoTokenizer.from_pretrained>>;\ntype Model = Awaited<ReturnType<typeof AutoModelForSequenceClassification.from_pretrained>>;\n\ninterface LogitsTensor {\n  data: ArrayLike<number>;\n  dims: number[];\n}\n\n// Recognize the noisy huggingface fetch failure signature and replace it\n// with an actionable instruction. Transformers.js parses a config that\n// failed to download, then dereferences `tokenizer_class` on undefined.\nfunction wrapLoadError(err: unknown): Error {\n  const message = err instanceof Error ? err.message : String(err);\n  const looksLikeMissingModel =\n    /tokenizer_class|tokenizer_config|preprocessor_config|fetch failed|ENOTFOUND|ECONNREFUSED|ETIMEDOUT|ENETUNREACH/i.test(\n      message,\n    );\n  if (looksLikeMissingModel) {\n    return new Error(\n      `Reranker model not downloaded — run \\`wigolo warmup\\` (cause: ${message})`,\n    );\n  }\n  return new Error(`Failed to load reranker model: ${message}`);\n}\n\nexport class TransformersRerankProvider implements RerankProvider {\n  private tokenizer: Tokenizer | null = null;\n  private model: Model | null = null;\n  private loadPromise: Promise<{ tokenizer: Tokenizer; model: Model }> | null = null;\n  readonly modelId: string;\n\n  constructor() {\n    this.modelId = 'Xenova/ms-marco-MiniLM-L-6-v2';\n  }\n\n  async warmup(): Promise<void> {\n    await this.load();\n  }\n\n  private load(): Promise<{ tokenizer: Tokenizer; model: Model }> {\n    if (this.tokenizer && this.model) {\n      return Promise.resolve({ tokenizer: this.tokenizer, model: this.model });\n    }\n    if (this.loadPromise) return this.loadPromise;\n\n    log.info('Loading rerank model', { modelId: this.modelId });\n    const cacheDir = join(getConfig().dataDir, 'transformers');\n    // Direct the library at a writable cache under the wigolo data dir so\n    // models don't end up in a user home cache the daemon can't manage.\n    env.cacheDir = cacheDir;\n\n    this.loadPromise = Promise.all([\n      AutoTokenizer.from_pretrained(this.modelId),\n      AutoModelForSequenceClassification.from_pretrained(this.modelId),\n    ])\n      .then(([tokenizer, model]) => {\n        this.tokenizer = tokenizer;\n        this.model = model;\n        return { tokenizer, model };\n      })\n      .catch((err: unknown) => {\n        this.loadPromise = null;\n        throw wrapLoadError(err);\n      });\n\n    return this.loadPromise;\n  }\n\n  async rerank(\n    query: string,\n    candidates: RerankCandidate[],\n    topK = candidates.length,\n  ): Promise<RerankResult[]> {\n    if (candidates.length === 0) return [];\n\n    const { tokenizer, model } = await this.load();\n\n    // Build batch: query repeated against each document.\n    const queries = candidates.map(() => query);\n    const docs = candidates.map((c) => c.text);\n\n    const inputs = tokenizer(queries, {\n      text_pair: docs,\n      padding: true,\n      truncation: true,\n    });\n\n    const outputs = (await model(inputs)) as { logits: LogitsTensor };\n    const logits = outputs.logits;\n    // logits shape is [batch, 1] for single-label regression rerankers.\n    // For multi-label heads (rare for rerankers) we still take the first\n    // value as the relevance score.\n    const stride = logits.dims.length >= 2 ? logits.dims[1] : 1;\n    const data = logits.data;\n\n    const scored: RerankResult[] = candidates.map((c, i) => ({\n      id: c.id,\n      score: Number(data[i * stride]),\n    }));\n\n    return scored.sort((a, b) => b.score - a.score).slice(0, topK);\n  }\n\n  // Release the underlying ONNX session before process exit. Without this,\n  // the runtime's worker threads race during C++ destructor teardown and\n  // surface as `mutex lock failed: Invalid argument` on macOS.\n  async dispose(): Promise<void> {\n    const model = this.model;\n    this.model = null;\n    this.tokenizer = null;\n    this.loadPromise = null;\n    if (!model) return;\n    try {\n      const m = model as unknown as { dispose?: () => Promise<unknown> };\n      if (typeof m.dispose === 'function') await m.dispose();\n    } catch (err) {\n      log.debug('reranker dispose failed', { error: err instanceof Error ? err.message : String(err) });\n    }\n  }\n}\n"],"mappings":"AAAA,SAAS,YAAY;AACrB;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAMP,SAAS,oBAAoB;AAC7B,SAAS,iBAAiB;AAE1B,MAAM,MAAM,aAAa,UAAU;AAsBnC,SAAS,cAAc,KAAqB;AAC1C,QAAM,UAAU,eAAe,QAAQ,IAAI,UAAU,OAAO,GAAG;AAC/D,QAAM,wBACJ,kHAAkH;AAAA,IAChH;AAAA,EACF;AACF,MAAI,uBAAuB;AACzB,WAAO,IAAI;AAAA,MACT,sEAAiE,OAAO;AAAA,IAC1E;AAAA,EACF;AACA,SAAO,IAAI,MAAM,kCAAkC,OAAO,EAAE;AAC9D;AAEO,MAAM,2BAAqD;AAAA,EACxD,YAA8B;AAAA,EAC9B,QAAsB;AAAA,EACtB,cAAsE;AAAA,EACrE;AAAA,EAET,cAAc;AACZ,SAAK,UAAU;AAAA,EACjB;AAAA,EAEA,MAAM,SAAwB;AAC5B,UAAM,KAAK,KAAK;AAAA,EAClB;AAAA,EAEQ,OAAwD;AAC9D,QAAI,KAAK,aAAa,KAAK,OAAO;AAChC,aAAO,QAAQ,QAAQ,EAAE,WAAW,KAAK,WAAW,OAAO,KAAK,MAAM,CAAC;AAAA,IACzE;AACA,QAAI,KAAK,YAAa,QAAO,KAAK;AAElC,QAAI,KAAK,wBAAwB,EAAE,SAAS,KAAK,QAAQ,CAAC;AAC1D,UAAM,WAAW,KAAK,UAAU,EAAE,SAAS,cAAc;AAGzD,QAAI,WAAW;AAEf,SAAK,cAAc,QAAQ,IAAI;AAAA,MAC7B,cAAc,gBAAgB,KAAK,OAAO;AAAA,MAC1C,mCAAmC,gBAAgB,KAAK,OAAO;AAAA,IACjE,CAAC,EACE,KAAK,CAAC,CAAC,WAAW,KAAK,MAAM;AAC5B,WAAK,YAAY;AACjB,WAAK,QAAQ;AACb,aAAO,EAAE,WAAW,MAAM;AAAA,IAC5B,CAAC,EACA,MAAM,CAAC,QAAiB;AACvB,WAAK,cAAc;AACnB,YAAM,cAAc,GAAG;AAAA,IACzB,CAAC;AAEH,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,OACJ,OACA,YACA,OAAO,WAAW,QACO;AACzB,QAAI,WAAW,WAAW,EAAG,QAAO,CAAC;AAErC,UAAM,EAAE,WAAW,MAAM,IAAI,MAAM,KAAK,KAAK;AAG7C,UAAM,UAAU,WAAW,IAAI,MAAM,KAAK;AAC1C,UAAM,OAAO,WAAW,IAAI,CAAC,MAAM,EAAE,IAAI;AAEzC,UAAM,SAAS,UAAU,SAAS;AAAA,MAChC,WAAW;AAAA,MACX,SAAS;AAAA,MACT,YAAY;AAAA,IACd,CAAC;AAED,UAAM,UAAW,MAAM,MAAM,MAAM;AACnC,UAAM,SAAS,QAAQ;AAIvB,UAAM,SAAS,OAAO,KAAK,UAAU,IAAI,OAAO,KAAK,CAAC,IAAI;AAC1D,UAAM,OAAO,OAAO;AAEpB,UAAM,SAAyB,WAAW,IAAI,CAAC,GAAG,OAAO;AAAA,MACvD,IAAI,EAAE;AAAA,MACN,OAAO,OAAO,KAAK,IAAI,MAAM,CAAC;AAAA,IAChC,EAAE;AAEF,WAAO,OAAO,KAAK,CAAC,GAAG,MAAM,EAAE,QAAQ,EAAE,KAAK,EAAE,MAAM,GAAG,IAAI;AAAA,EAC/D;AAAA;AAAA;AAAA;AAAA,EAKA,MAAM,UAAyB;AAC7B,UAAM,QAAQ,KAAK;AACnB,SAAK,QAAQ;AACb,SAAK,YAAY;AACjB,SAAK,cAAc;AACnB,QAAI,CAAC,MAAO;AACZ,QAAI;AACF,YAAM,IAAI;AACV,UAAI,OAAO,EAAE,YAAY,WAAY,OAAM,EAAE,QAAQ;AAAA,IACvD,SAAS,KAAK;AACZ,UAAI,MAAM,2BAA2B,EAAE,OAAO,eAAe,QAAQ,IAAI,UAAU,OAAO,GAAG,EAAE,CAAC;AAAA,IAClG;AAAA,EACF;AACF;","names":[]}

package/dist/search/rrf.d.ts

@@ -0,0 +1,17 @@
+/**
+ * Reciprocal Rank Fusion -- merges multiple ranked lists into a single
+ * score map. Each list maps URL -> rank (1-based). The fused score for
+ * a URL is the sum of 1/(k + rank) across all lists it appears in.
+ */
+export declare function reciprocalRankFusion(lists: Map<string, number>[], k?: number): Map<string, number>;
+/**
+ * Convert a fused score map into a sorted array of [url, score] pairs,
+ * descending by score.
+ */
+export declare function sortByRRFScore(scores: Map<string, number>): Array<[string, number]>;
+/**
+ * Build a rank map from an ordered array of URLs.
+ * First URL gets rank 1, second gets rank 2, etc.
+ */
+export declare function buildRankMap(urls: string[]): Map<string, number>;
+//# sourceMappingURL=rrf.d.ts.map

package/dist/search/rrf.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"rrf.d.ts","sourceRoot":"","sources":["../../src/search/rrf.ts"],"names":[],"mappings":"AAIA;;;;GAIG;AACH,wBAAgB,oBAAoB,CAClC,KAAK,EAAE,GAAG,CAAC,MAAM,EAAE,MAAM,CAAC,EAAE,EAC5B,CAAC,GAAE,MAAW,GACb,GAAG,CAAC,MAAM,EAAE,MAAM,CAAC,CAqBrB;AAED;;;GAGG;AACH,wBAAgB,cAAc,CAC5B,MAAM,EAAE,GAAG,CAAC,MAAM,EAAE,MAAM,CAAC,GAC1B,KAAK,CAAC,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC,CAEzB;AAED;;;GAGG;AACH,wBAAgB,YAAY,CAAC,IAAI,EAAE,MAAM,EAAE,GAAG,GAAG,CAAC,MAAM,EAAE,MAAM,CAAC,CAQhE"}

package/dist/search/rrf.js

@@ -0,0 +1,39 @@
+import { createLogger } from "../logger.js";
+const log = createLogger("search");
+function reciprocalRankFusion(lists, k = 60) {
+  try {
+    const scores = /* @__PURE__ */ new Map();
+    for (const list of lists) {
+      for (const [url, rank] of list) {
+        const contribution = 1 / (k + rank);
+        scores.set(url, (scores.get(url) ?? 0) + contribution);
+      }
+    }
+    log.debug("RRF fusion complete", {
+      inputLists: lists.length,
+      uniqueUrls: scores.size
+    });
+    return scores;
+  } catch (err) {
+    log.error("RRF fusion failed", { error: String(err) });
+    return /* @__PURE__ */ new Map();
+  }
+}
+function sortByRRFScore(scores) {
+  return [...scores.entries()].sort((a, b) => b[1] - a[1]);
+}
+function buildRankMap(urls) {
+  const rankMap = /* @__PURE__ */ new Map();
+  for (let i = 0; i < urls.length; i++) {
+    if (!rankMap.has(urls[i])) {
+      rankMap.set(urls[i], i + 1);
+    }
+  }
+  return rankMap;
+}
+export {
+  buildRankMap,
+  reciprocalRankFusion,
+  sortByRRFScore
+};
+//# sourceMappingURL=rrf.js.map

package/dist/search/rrf.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../src/search/rrf.ts"],"sourcesContent":["import { createLogger } from '../logger.js';\n\nconst log = createLogger('search');\n\n/**\n * Reciprocal Rank Fusion -- merges multiple ranked lists into a single\n * score map. Each list maps URL -> rank (1-based). The fused score for\n * a URL is the sum of 1/(k + rank) across all lists it appears in.\n */\nexport function reciprocalRankFusion(\n  lists: Map<string, number>[],\n  k: number = 60,\n): Map<string, number> {\n  try {\n    const scores = new Map<string, number>();\n\n    for (const list of lists) {\n      for (const [url, rank] of list) {\n        const contribution = 1 / (k + rank);\n        scores.set(url, (scores.get(url) ?? 0) + contribution);\n      }\n    }\n\n    log.debug('RRF fusion complete', {\n      inputLists: lists.length,\n      uniqueUrls: scores.size,\n    });\n\n    return scores;\n  } catch (err) {\n    log.error('RRF fusion failed', { error: String(err) });\n    return new Map();\n  }\n}\n\n/**\n * Convert a fused score map into a sorted array of [url, score] pairs,\n * descending by score.\n */\nexport function sortByRRFScore(\n  scores: Map<string, number>,\n): Array<[string, number]> {\n  return [...scores.entries()].sort((a, b) => b[1] - a[1]);\n}\n\n/**\n * Build a rank map from an ordered array of URLs.\n * First URL gets rank 1, second gets rank 2, etc.\n */\nexport function buildRankMap(urls: string[]): Map<string, number> {\n  const rankMap = new Map<string, number>();\n  for (let i = 0; i < urls.length; i++) {\n    if (!rankMap.has(urls[i])) {\n      rankMap.set(urls[i], i + 1);\n    }\n  }\n  return rankMap;\n}\n"],"mappings":"AAAA,SAAS,oBAAoB;AAE7B,MAAM,MAAM,aAAa,QAAQ;AAO1B,SAAS,qBACd,OACA,IAAY,IACS;AACrB,MAAI;AACF,UAAM,SAAS,oBAAI,IAAoB;AAEvC,eAAW,QAAQ,OAAO;AACxB,iBAAW,CAAC,KAAK,IAAI,KAAK,MAAM;AAC9B,cAAM,eAAe,KAAK,IAAI;AAC9B,eAAO,IAAI,MAAM,OAAO,IAAI,GAAG,KAAK,KAAK,YAAY;AAAA,MACvD;AAAA,IACF;AAEA,QAAI,MAAM,uBAAuB;AAAA,MAC/B,YAAY,MAAM;AAAA,MAClB,YAAY,OAAO;AAAA,IACrB,CAAC;AAED,WAAO;AAAA,EACT,SAAS,KAAK;AACZ,QAAI,MAAM,qBAAqB,EAAE,OAAO,OAAO,GAAG,EAAE,CAAC;AACrD,WAAO,oBAAI,IAAI;AAAA,EACjB;AACF;AAMO,SAAS,eACd,QACyB;AACzB,SAAO,CAAC,GAAG,OAAO,QAAQ,CAAC,EAAE,KAAK,CAAC,GAAG,MAAM,EAAE,CAAC,IAAI,EAAE,CAAC,CAAC;AACzD;AAMO,SAAS,aAAa,MAAqC;AAChE,QAAM,UAAU,oBAAI,IAAoB;AACxC,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK;AACpC,QAAI,CAAC,QAAQ,IAAI,KAAK,CAAC,CAAC,GAAG;AACzB,cAAQ,IAAI,KAAK,CAAC,GAAG,IAAI,CAAC;AAAA,IAC5B;AAAA,EACF;AACA,SAAO;AACT;","names":[]}

package/dist/search/sampling.d.ts

@@ -0,0 +1,25 @@
+export interface SamplingMessage {
+    role: 'user' | 'assistant';
+    content: {
+        type: 'text';
+        text: string;
+    };
+}
+export interface SamplingResponse {
+    model: string;
+    content: {
+        type: string;
+        text: string;
+    };
+}
+export interface SamplingCapableServer {
+    getClientCapabilities: () => Record<string, unknown> | null | undefined;
+    createMessage: (params: {
+        messages: SamplingMessage[];
+        maxTokens: number;
+    }) => Promise<SamplingResponse>;
+}
+export declare function checkSamplingSupport(server: SamplingCapableServer): boolean;
+export declare function requestSampling(server: SamplingCapableServer, messages: SamplingMessage[], maxTokens: number): Promise<SamplingResponse>;
+export declare function extractTextFromSamplingResponse(response: unknown): string;
+//# sourceMappingURL=sampling.d.ts.map

package/dist/search/sampling.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"sampling.d.ts","sourceRoot":"","sources":["../../src/search/sampling.ts"],"names":[],"mappings":"AAIA,MAAM,WAAW,eAAe;IAC9B,IAAI,EAAE,MAAM,GAAG,WAAW,CAAC;IAC3B,OAAO,EAAE;QAAE,IAAI,EAAE,MAAM,CAAC;QAAC,IAAI,EAAE,MAAM,CAAA;KAAE,CAAC;CACzC;AAED,MAAM,WAAW,gBAAgB;IAC/B,KAAK,EAAE,MAAM,CAAC;IACd,OAAO,EAAE;QAAE,IAAI,EAAE,MAAM,CAAC;QAAC,IAAI,EAAE,MAAM,CAAA;KAAE,CAAC;CACzC;AAED,MAAM,WAAW,qBAAqB;IACpC,qBAAqB,EAAE,MAAM,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAAG,IAAI,GAAG,SAAS,CAAC;IACxE,aAAa,EAAE,CAAC,MAAM,EAAE;QACtB,QAAQ,EAAE,eAAe,EAAE,CAAC;QAC5B,SAAS,EAAE,MAAM,CAAC;KACnB,KAAK,OAAO,CAAC,gBAAgB,CAAC,CAAC;CACjC;AAED,wBAAgB,oBAAoB,CAAC,MAAM,EAAE,qBAAqB,GAAG,OAAO,CAS3E;AAED,wBAAsB,eAAe,CACnC,MAAM,EAAE,qBAAqB,EAC7B,QAAQ,EAAE,eAAe,EAAE,EAC3B,SAAS,EAAE,MAAM,GAChB,OAAO,CAAC,gBAAgB,CAAC,CAsB3B;AAED,wBAAgB,+BAA+B,CAC7C,QAAQ,EAAE,OAAO,GAChB,MAAM,CAiBR"}

package/dist/search/sampling.js

@@ -0,0 +1,52 @@
+import { createLogger } from "../logger.js";
+const log = createLogger("search");
+function checkSamplingSupport(server) {
+  try {
+    const capabilities = server.getClientCapabilities();
+    if (!capabilities) return false;
+    return capabilities.sampling !== void 0;
+  } catch (err) {
+    log.debug("sampling support check failed", { error: String(err) });
+    return false;
+  }
+}
+async function requestSampling(server, messages, maxTokens) {
+  if (!checkSamplingSupport(server)) {
+    throw new Error("Client does not support MCP sampling");
+  }
+  log.debug("sending sampling request", {
+    messageCount: messages.length,
+    maxTokens
+  });
+  const response = await server.createMessage({
+    messages,
+    maxTokens
+  });
+  log.debug("sampling response received", {
+    model: response?.model,
+    contentType: response?.content?.type,
+    textLength: response?.content?.text?.length ?? 0
+  });
+  return response;
+}
+function extractTextFromSamplingResponse(response) {
+  try {
+    if (!response || typeof response !== "object") return "";
+    const resp = response;
+    const content = resp.content;
+    if (!content) return "";
+    if (content.type === "text" && typeof content.text === "string") {
+      return content.text.trim();
+    }
+    return "";
+  } catch (err) {
+    log.debug("failed to extract text from sampling response", { error: String(err) });
+    return "";
+  }
+}
+export {
+  checkSamplingSupport,
+  extractTextFromSamplingResponse,
+  requestSampling
+};
+//# sourceMappingURL=sampling.js.map

package/dist/search/sampling.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../src/search/sampling.ts"],"sourcesContent":["import { createLogger } from '../logger.js';\n\nconst log = createLogger('search');\n\nexport interface SamplingMessage {\n  role: 'user' | 'assistant';\n  content: { type: 'text'; text: string };\n}\n\nexport interface SamplingResponse {\n  model: string;\n  content: { type: string; text: string };\n}\n\nexport interface SamplingCapableServer {\n  getClientCapabilities: () => Record<string, unknown> | null | undefined;\n  createMessage: (params: {\n    messages: SamplingMessage[];\n    maxTokens: number;\n  }) => Promise<SamplingResponse>;\n}\n\nexport function checkSamplingSupport(server: SamplingCapableServer): boolean {\n  try {\n    const capabilities = server.getClientCapabilities();\n    if (!capabilities) return false;\n    return capabilities.sampling !== undefined;\n  } catch (err) {\n    log.debug('sampling support check failed', { error: String(err) });\n    return false;\n  }\n}\n\nexport async function requestSampling(\n  server: SamplingCapableServer,\n  messages: SamplingMessage[],\n  maxTokens: number,\n): Promise<SamplingResponse> {\n  if (!checkSamplingSupport(server)) {\n    throw new Error('Client does not support MCP sampling');\n  }\n\n  log.debug('sending sampling request', {\n    messageCount: messages.length,\n    maxTokens,\n  });\n\n  const response = await server.createMessage({\n    messages,\n    maxTokens,\n  });\n\n  log.debug('sampling response received', {\n    model: response?.model,\n    contentType: response?.content?.type,\n    textLength: response?.content?.text?.length ?? 0,\n  });\n\n  return response;\n}\n\nexport function extractTextFromSamplingResponse(\n  response: unknown,\n): string {\n  try {\n    if (!response || typeof response !== 'object') return '';\n\n    const resp = response as Record<string, unknown>;\n    const content = resp.content as Record<string, unknown> | null | undefined;\n    if (!content) return '';\n\n    if (content.type === 'text' && typeof content.text === 'string') {\n      return content.text.trim();\n    }\n\n    return '';\n  } catch (err) {\n    log.debug('failed to extract text from sampling response', { error: String(err) });\n    return '';\n  }\n}\n"],"mappings":"AAAA,SAAS,oBAAoB;AAE7B,MAAM,MAAM,aAAa,QAAQ;AAoB1B,SAAS,qBAAqB,QAAwC;AAC3E,MAAI;AACF,UAAM,eAAe,OAAO,sBAAsB;AAClD,QAAI,CAAC,aAAc,QAAO;AAC1B,WAAO,aAAa,aAAa;AAAA,EACnC,SAAS,KAAK;AACZ,QAAI,MAAM,iCAAiC,EAAE,OAAO,OAAO,GAAG,EAAE,CAAC;AACjE,WAAO;AAAA,EACT;AACF;AAEA,eAAsB,gBACpB,QACA,UACA,WAC2B;AAC3B,MAAI,CAAC,qBAAqB,MAAM,GAAG;AACjC,UAAM,IAAI,MAAM,sCAAsC;AAAA,EACxD;AAEA,MAAI,MAAM,4BAA4B;AAAA,IACpC,cAAc,SAAS;AAAA,IACvB;AAAA,EACF,CAAC;AAED,QAAM,WAAW,MAAM,OAAO,cAAc;AAAA,IAC1C;AAAA,IACA;AAAA,EACF,CAAC;AAED,MAAI,MAAM,8BAA8B;AAAA,IACtC,OAAO,UAAU;AAAA,IACjB,aAAa,UAAU,SAAS;AAAA,IAChC,YAAY,UAAU,SAAS,MAAM,UAAU;AAAA,EACjD,CAAC;AAED,SAAO;AACT;AAEO,SAAS,gCACd,UACQ;AACR,MAAI;AACF,QAAI,CAAC,YAAY,OAAO,aAAa,SAAU,QAAO;AAEtD,UAAM,OAAO;AACb,UAAM,UAAU,KAAK;AACrB,QAAI,CAAC,QAAS,QAAO;AAErB,QAAI,QAAQ,SAAS,UAAU,OAAO,QAAQ,SAAS,UAAU;AAC/D,aAAO,QAAQ,KAAK,KAAK;AAAA,IAC3B;AAEA,WAAO;AAAA,EACT,SAAS,KAAK;AACZ,QAAI,MAAM,iDAAiD,EAAE,OAAO,OAAO,GAAG,EAAE,CAAC;AACjF,WAAO;AAAA,EACT;AACF;","names":[]}

package/dist/search/searxng.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"searxng.d.ts","sourceRoot":"","sources":["../../src/search/searxng.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,YAAY,EAAE,mBAAmB,EAAE,eAAe,EAAE,MAAM,aAAa,CAAC;AA2CtF,qBAAa,aAAc,YAAW,YAAY;IAGpC,OAAO,CAAC,QAAQ,CAAC,OAAO;IAFpC,IAAI,SAAa;gBAEY,OAAO,EAAE,MAAM;IAEtC,MAAM,CAAC,KAAK,EAAE,MAAM,EAAE,OAAO,GAAE,mBAAwB,GAAG,OAAO,CAAC,eAAe,EAAE,CAAC;CAyD3F"}
+{"version":3,"file":"searxng.d.ts","sourceRoot":"","sources":["../../src/search/searxng.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,YAAY,EAAE,mBAAmB,EAAE,eAAe,EAAE,MAAM,aAAa,CAAC;AAkDtF,qBAAa,aAAc,YAAW,YAAY;IAGpC,OAAO,CAAC,QAAQ,CAAC,OAAO;IAFpC,IAAI,SAAa;gBAEY,OAAO,EAAE,MAAM;IAEtC,MAAM,CAAC,KAAK,EAAE,MAAM,EAAE,OAAO,GAAE,mBAAwB,GAAG,OAAO,CAAC,eAAe,EAAE,CAAC;CA6D3F"}