@wcs-colab/plugin-fuzzy-phrase 3.1.16-custom.newbase.7 โ†’ 3.1.16-custom.newbase.9

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (5) hide show
  1. package/dist/index.cjs +5 -6
  2. package/dist/index.cjs.map +1 -1
  3. package/dist/index.js +5 -6
  4. package/dist/index.js.map +1 -1
  5. package/package.json +1 -1
package/dist/index.cjs CHANGED
@@ -309,7 +309,8 @@ function calculatePhraseScore(phraseWords, queryTokens, config, documentFrequenc
309
309
  const weightedDensity = densityScore * weights.density;
310
310
  const weightedSemantic = semanticScore * weights.semantic;
311
311
  const totalScore = weightedBase + weightedOrder + weightedProximity + weightedDensity + weightedSemantic;
312
- const maxBaseWeight = Math.max(weights.exact, weights.fuzzy);
312
+ const canHaveFuzzyMatches = config.tolerance > 0 && weights.fuzzy > 0;
313
+ const maxBaseWeight = canHaveFuzzyMatches ? Math.max(weights.exact, weights.fuzzy) : weights.exact;
313
314
  const effectiveProximityWeight = config.maxGap > 0 && weights.proximity > 0 && queryTokens.length > 1 ? weights.proximity : 0;
314
315
  const maxPossibleScore = maxBaseWeight + weights.order + effectiveProximityWeight + weights.density + weights.semantic;
315
316
  const normalizedScore = totalScore / maxPossibleScore;
@@ -521,10 +522,7 @@ async function searchWithFuzzyPhrase(orama, params, language) {
521
522
  state.config.enableSynonyms ? state.synonymMap : void 0,
522
523
  state.config.synonymMatchScore
523
524
  );
524
- const filteredCandidates = filterCandidatesByScore(
525
- candidatesMap,
526
- state.config.minScore
527
- );
525
+ const filteredCandidates = tolerance === 0 ? candidatesMap : filterCandidatesByScore(candidatesMap, state.config.minScore);
528
526
  console.log(`\u{1F3AF} Found candidates: ${Array.from(filteredCandidates.values()).reduce((sum, c) => sum + c.length, 0)} total`);
529
527
  const documentMatches = [];
530
528
  console.log("\u{1F50D} DEBUG orama.data structure:", {
@@ -564,7 +562,8 @@ async function searchWithFuzzyPhrase(orama, params, language) {
564
562
  {
565
563
  weights: state.config.weights,
566
564
  maxGap: state.config.maxGap,
567
- proximitySpanMultiplier: state.config.proximitySpanMultiplier
565
+ proximitySpanMultiplier: state.config.proximitySpanMultiplier,
566
+ tolerance
568
567
  },
569
568
  state.documentFrequency,
570
569
  state.totalDocuments
package/dist/index.cjs.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/fuzzy.ts","../src/candidates.ts","../src/scoring.ts","../src/index.ts"],"names":[],"mappings":";AA4BO,SAAS,mBACd,GACA,GACA,OAC0B;AAE1B,MAAI,MAAM,GAAG;AACX,WAAO,EAAE,WAAW,MAAM,UAAU,EAAE;AAAA,EACxC;AAEA,QAAM,OAAO,EAAE;AACf,QAAM,OAAO,EAAE;AAGf,MAAI,KAAK,IAAI,OAAO,IAAI,IAAI,OAAO;AACjC,WAAO,EAAE,WAAW,OAAO,UAAU,QAAQ,EAAE;AAAA,EACjD;AAGA,MAAI,OAAO,MAAM;AACf,KAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC;AAAA,EAChB;AAEA,QAAM,IAAI,EAAE;AACZ,QAAM,IAAI,EAAE;AAGZ,MAAI,UAAU,IAAI,MAAM,IAAI,CAAC;AAC7B,MAAI,UAAU,IAAI,MAAM,IAAI,CAAC;AAG7B,WAAS,IAAI,GAAG,KAAK,GAAG,KAAK;AAC3B,YAAQ,CAAC,IAAI;AAAA,EACf;AAEA,WAAS,IAAI,GAAG,KAAK,GAAG,KAAK;AAC3B,YAAQ,CAAC,IAAI;AACb,QAAI,WAAW;AAEf,aAAS,IAAI,GAAG,KAAK,GAAG,KAAK;AAC3B,YAAM,OAAO,EAAE,IAAI,CAAC,MAAM,EAAE,IAAI,CAAC,IAAI,IAAI;AAEzC,cAAQ,CAAC,IAAI,KAAK;AAAA,QAChB,QAAQ,CAAC,IAAI;AAAA;AAAA,QACb,QAAQ,IAAI,CAAC,IAAI;AAAA;AAAA,QACjB,QAAQ,IAAI,CAAC,IAAI;AAAA;AAAA,MACnB;AAEA,iBAAW,KAAK,IAAI,UAAU,QAAQ,CAAC,CAAC;AAAA,IAC1C;AAGA,QAAI,WAAW,OAAO;AACpB,aAAO,EAAE,WAAW,OAAO,UAAU,QAAQ,EAAE;AAAA,IACjD;AAGA,KAAC,SAAS,OAAO,IAAI,CAAC,SAAS,OAAO;AAAA,EACxC;AAEA,QAAM,WAAW,QAAQ,CAAC;AAC1B,SAAO;AAAA,IACL,WAAW,YAAY;AAAA,IACvB;AAAA,EACF;AACF;AAUO,SAAS,WACd,MACA,YACA,WACuD;AAEvD,MAAI,SAAS,YAAY;AACvB,WAAO,EAAE,SAAS,MAAM,UAAU,GAAG,OAAO,EAAI;AAAA,EAClD;AAGA,MAAI,KAAK,WAAW,UAAU,GAAG;AAC/B,WAAO,EAAE,SAAS,MAAM,UAAU,GAAG,OAAO,KAAK;AAAA,EACnD;AAGA,QAAM,SAAS,mBAAmB,MAAM,YAAY,SAAS;AAE7D,MAAI,OAAO,WAAW;AAGpB,UAAM,QAAQ,IAAO,OAAO,WAAW;AACvC,WAAO;AAAA,MACL,SAAS;AAAA,MACT,UAAU,OAAO;AAAA,MACjB,OAAO,KAAK,IAAI,KAAK,KAAK;AAAA;AAAA,IAC5B;AAAA,EACF;AAEA,SAAO,EAAE,SAAS,OAAO,UAAU,YAAY,GAAG,OAAO,EAAE;AAC7D;AAWO,SAAS,2BACd,aACA,eACQ;AACR,QAAM,cAAc,YAAY;AAEhC,MAAI,eAAe,GAAG;AACpB,WAAO;AAAA,EACT,WAAW,eAAe,GAAG;AAC3B,WAAO,gBAAgB;AAAA,EACzB,WAAW,eAAe,GAAG;AAC3B,WAAO,gBAAgB;AAAA,EACzB,OAAO;AACL,WAAO,gBAAgB;AAAA,EACzB;AACF;;;ACjJO,SAAS,+BAA+B,WAA6B;AAC1E,QAAM,aAAa,oBAAI,IAAY;AACnC,MAAI,eAAe;AACnB,MAAI,aAAa;AAEjB,WAAS,SAAS,MAAW,QAAgB,GAAG;AAC9C,QAAI,CAAC,MAAM;AACT;AAAA,IACF;AAEA;AAIA,QAAI,KAAK,KAAK,KAAK,KAAK,OAAO,KAAK,MAAM,YAAY,KAAK,EAAE,SAAS,GAAG;AACvE,iBAAW,IAAI,KAAK,CAAC;AACrB;AAAA,IACF;AAGA,QAAI,KAAK,GAAG;AACV,UAAI,KAAK,aAAa,KAAK;AAEzB,mBAAW,CAAC,MAAM,SAAS,KAAK,KAAK,GAAG;AACtC,mBAAS,WAAW,QAAQ,CAAC;AAAA,QAC/B;AAAA,MACF,WAAW,MAAM,QAAQ,KAAK,CAAC,GAAG;AAEhC,mBAAW,CAAC,MAAM,SAAS,KAAK,KAAK,GAAG;AACtC,mBAAS,WAAW,QAAQ,CAAC;AAAA,QAC/B;AAAA,MACF,WAAW,OAAO,KAAK,MAAM,UAAU;AAErC,mBAAW,aAAa,OAAO,OAAO,KAAK,CAAC,GAAG;AAC7C,mBAAS,WAAW,QAAQ,CAAC;AAAA,QAC/B;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAEA,WAAS,SAAS;AAClB,UAAQ,IAAI,uBAAgB,WAAW,IAAI,eAAe,YAAY,gBAAgB;AACtF,SAAO;AACT;AAYO,SAAS,uBACd,YACA,YACA,WACA,UACA,eAAuB,KACV;AACb,QAAM,aAA0B,CAAC;AACjC,QAAM,OAAO,oBAAI,IAAY;AAG7B,MAAI,WAAW,IAAI,UAAU,GAAG;AAC9B,eAAW,KAAK;AAAA,MACd,MAAM;AAAA,MACN,MAAM;AAAA,MACN;AAAA,MACA,UAAU;AAAA,MACV,OAAO;AAAA,IACT,CAAC;AACD,SAAK,IAAI,UAAU;AAAA,EACrB;AAGA,aAAW,QAAQ,YAAY;AAC7B,QAAI,KAAK,IAAI,IAAI;AAAG;AAEpB,UAAM,QAAQ,WAAW,MAAM,YAAY,SAAS;AACpD,QAAI,MAAM,SAAS;AACjB,iBAAW,KAAK;AAAA,QACd;AAAA,QACA,MAAM;AAAA,QACN;AAAA,QACA,UAAU,MAAM;AAAA,QAChB,OAAO,MAAM;AAAA,MACf,CAAC;AACD,WAAK,IAAI,IAAI;AAAA,IACf;AAAA,EACF;AAGA,MAAI,YAAY,SAAS,UAAU,GAAG;AACpC,eAAW,WAAW,SAAS,UAAU,GAAG;AAC1C,UAAI,KAAK,IAAI,OAAO;AAAG;AACvB,UAAI,WAAW,IAAI,OAAO,GAAG;AAC3B,mBAAW,KAAK;AAAA,UACd,MAAM;AAAA,UACN,MAAM;AAAA,UACN;AAAA,UACA,UAAU;AAAA,UACV,OAAO;AAAA,QACT,CAAC;AACD,aAAK,IAAI,OAAO;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAEA,SAAO;AACT;AAYO,SAAS,kBACd,aACA,YACA,WACA,UACA,eAAuB,KACG;AAC1B,QAAM,gBAAgB,oBAAI,IAAyB;AAEnD,aAAW,SAAS,aAAa;AAC/B,UAAM,kBAAkB;AAAA,MACtB;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AACA,kBAAc,IAAI,OAAO,eAAe;AAAA,EAC1C;AAEA,SAAO;AACT;AAyBO,SAAS,wBACd,eACA,UAC0B;AAC1B,QAAM,WAAW,oBAAI,IAAyB;AAE9C,aAAW,CAAC,OAAO,UAAU,KAAK,cAAc,QAAQ,GAAG;AACzD,UAAM,qBAAqB,WAAW,OAAO,OAAK,EAAE,SAAS,QAAQ;AACrE,QAAI,mBAAmB,SAAS,GAAG;AACjC,eAAS,IAAI,OAAO,kBAAkB;AAAA,IACxC;AAAA,EACF;AAEA,SAAO;AACT;;;AClKO,SAAS,sBACd,gBACA,eACA,QACA,mBACA,gBACe;AACf,QAAM,UAAyB,CAAC;AAChC,QAAM,cAAc,MAAM,KAAK,cAAc,KAAK,CAAC;AAGnD,QAAM,cAA2B,CAAC;AAElC,WAAS,IAAI,GAAG,IAAI,eAAe,QAAQ,KAAK;AAC9C,UAAM,UAAU,eAAe,CAAC;AAGhC,eAAW,CAAC,YAAY,UAAU,KAAK,cAAc,QAAQ,GAAG;AAC9D,iBAAW,aAAa,YAAY;AAClC,YAAI,UAAU,SAAS,SAAS;AAC9B,sBAAY,KAAK;AAAA,YACf,MAAM;AAAA,YACN;AAAA,YACA,UAAU;AAAA,YACV,MAAM,UAAU;AAAA,YAChB,UAAU,UAAU;AAAA,YACpB,OAAO,UAAU;AAAA,UACnB,CAAC;AAAA,QACH;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAGA,WAAS,IAAI,GAAG,IAAI,YAAY,QAAQ,KAAK;AAC3C,UAAM,SAAS;AAAA,MACb;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA;AAAA,IACF;AAEA,QAAI,UAAU,OAAO,MAAM,SAAS,GAAG;AACrC,cAAQ,KAAK,MAAM;AAAA,IACrB;AAAA,EACF;AAGA,SAAO,mBAAmB,OAAO;AACnC;AAeA,SAAS,wBACP,aACA,YACA,aACA,QACA,mBACA,gBACA,gBACA,gBACoB;AACpB,QAAM,aAAa,YAAY,UAAU;AACzC,QAAM,cAA2B,CAAC,UAAU;AAC5C,QAAM,gBAAgB,oBAAI,IAAI,CAAC,WAAW,UAAU,CAAC;AACrD,QAAM,WAAsB,CAAC;AAC7B,MAAI,eAAe;AAGnB,WAAS,IAAI,aAAa,GAAG,IAAI,YAAY,QAAQ,KAAK;AACxD,UAAM,QAAQ,YAAY,CAAC;AAC3B,UAAM,UAAU,YAAY,YAAY,SAAS,CAAC,EAAE;AACpD,UAAM,MAAM,MAAM,WAAW,UAAU;AAGvC,QAAI,MAAM,OAAO,QAAQ;AACvB;AAAA,IACF;AAGA,QAAI,CAAC,cAAc,IAAI,MAAM,UAAU,GAAG;AAExC,eAAS,MAAM,UAAU,GAAG,MAAM,MAAM,UAAU,OAAO;AACvD;AACA,iBAAS,KAAK;AAAA,UACZ,MAAM,eAAe,GAAG;AAAA,UACxB,UAAU;AAAA,UACV,UAAU;AAAA,QACZ,CAAC;AAAA,MACH;AAEA,kBAAY,KAAK,KAAK;AACtB,oBAAc,IAAI,MAAM,UAAU;AAGlC,UAAI,cAAc,SAAS,YAAY,QAAQ;AAC7C;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAGA,MAAI,YAAY,SAAS,GAAG;AAC1B,UAAM,WAAW,YAAY,SAAS,YAAY;AAClD,UAAM,OAAO,YAAY,YAAY,SAAS,CAAC,EAAE,WAAW,YAAY,CAAC,EAAE,WAAW;AAEtF,UAAM,EAAE,OAAO,UAAU,IAAI;AAAA,MAC3B;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAEA,WAAO;AAAA,MACL,OAAO;AAAA,MACP;AAAA,MACA,SAAS;AAAA,MACT;AAAA,MACA,eAAe,YAAY,CAAC,EAAE;AAAA,MAC9B,aAAa,YAAY,YAAY,SAAS,CAAC,EAAE;AAAA,MACjD;AAAA,MACA,SAAS,UAAU,aAAa,WAAW;AAAA,MAC3C;AAAA,MACA,gBAAgB;AAAA,IAClB;AAAA,EACF;AAEA,SAAO;AACT;AAcA,SAAS,qBACP,aACA,aACA,QACA,mBACA,gBACA,gBACA,UACuI;AAGvI,MAAI,YAAY;AAChB,aAAW,QAAQ,aAAa;AAC9B,UAAM,SAAS,KAAK,SAAS,UAAU,OAAO,QAAQ,QACvC,KAAK,SAAS,UAAU,OAAO,QAAQ,QACvC,OAAO,QAAQ,QAAQ;AACtC,iBAAa,KAAK,QAAQ;AAAA,EAC5B;AACA,eAAa,YAAY;AAGzB,QAAM,UAAU,UAAU,aAAa,WAAW;AAClD,QAAM,aAAa,UAAU,IAAM;AAInC,MAAI,iBAAiB;AACrB,MAAI,OAAO,SAAS,KAAK,OAAO,QAAQ,YAAY,KAAK,YAAY,SAAS,GAAG;AAC/E,UAAM,OAAO,YAAY,YAAY,SAAS,CAAC,EAAE,WAAW,YAAY,CAAC,EAAE,WAAW;AACtF,UAAM,kBAAkB,YAAY,SAAS,OAAO;AACpD,qBAAiB,KAAK,IAAI,GAAG,IAAO,OAAO,eAAgB;AAAA,EAC7D;AAIA,MAAI,eAAe;AAEnB,MAAI,YAAY,WAAW,GAAG;AAE5B,UAAM,mBAAmB,eAAe;AAExC,mBAAe,KAAK,IAAI,GAAK,mBAAmB,EAAE;AAAA,EACpD;AAKA,QAAM,gBAAgB;AAAA,IACpB;AAAA,IACA;AAAA,IACA;AAAA,EACF;AAGA,QAAM,UAAU,OAAO;AAGvB,QAAM,eAAe;AACrB,QAAM,gBAAgB,aAAa,QAAQ;AAC3C,QAAM,oBAAoB,iBAAiB,QAAQ;AACnD,QAAM,kBAAkB,eAAe,QAAQ;AAC/C,QAAM,mBAAmB,gBAAgB,QAAQ;AAEjD,QAAM,aAAa,eAAe,gBAAgB,oBAAoB,kBAAkB;AAIxF,QAAM,gBAAgB,KAAK,IAAI,QAAQ,OAAO,QAAQ,KAAK;AAE3D,QAAM,2BAA4B,OAAO,SAAS,KAAK,QAAQ,YAAY,KAAK,YAAY,SAAS,IAAK,QAAQ,YAAY;AAC9H,QAAM,mBAAmB,gBAAgB,QAAQ,QAAQ,2BAA2B,QAAQ,UAAU,QAAQ;AAG9G,QAAM,kBAAkB,aAAa;AAIrC,QAAM,qBAAqB,YAAY,SAAS,IAAI,WAAW;AAC/D,QAAM,QAAQ,kBAAkB;AAGhC,QAAM,OAAO,eAAe;AAC5B,QAAM,QAAQ,gBAAgB;AAC9B,QAAM,YAAY,oBAAoB;AACtC,QAAM,UAAU,kBAAkB;AAClC,QAAM,WAAW,mBAAmB;AAEpC,SAAO;AAAA,IACL;AAAA,IACA,WAAW;AAAA,MACT;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,UAAU;AAAA;AAAA,IACZ;AAAA,EACF;AACF;AASA,SAAS,UAAU,aAA0B,aAAgC;AAC3E,QAAM,aAAa,IAAI,IAAI,YAAY,IAAI,CAAC,OAAO,UAAU,CAAC,OAAO,KAAK,CAAC,CAAC;AAE5E,WAAS,IAAI,GAAG,IAAI,YAAY,QAAQ,KAAK;AAC3C,UAAM,YAAY,WAAW,IAAI,YAAY,IAAI,CAAC,EAAE,UAAU,KAAK;AACnE,UAAM,YAAY,WAAW,IAAI,YAAY,CAAC,EAAE,UAAU,KAAK;AAE/D,QAAI,YAAY,WAAW;AACzB,aAAO;AAAA,IACT;AAAA,EACF;AAEA,SAAO;AACT;AAUA,SAAS,uBACP,aACA,mBACA,gBACQ;AAER,MAAI,mBAAmB,GAAG;AACxB,WAAO;AAAA,EACT;AAEA,MAAI,WAAW;AAEf,aAAW,QAAQ,aAAa;AAC9B,UAAM,KAAK,kBAAkB,IAAI,KAAK,IAAI,KAAK;AAC/C,UAAM,MAAM,KAAK,IAAI,iBAAiB,EAAE;AACxC,gBAAY;AAAA,EACd;AAGA,QAAM,WAAW,WAAW,YAAY;AAGxC,SAAO,KAAK,IAAI,GAAK,WAAW,EAAE;AACpC;AAQA,SAAS,mBAAmB,SAAuC;AACjE,MAAI,QAAQ,WAAW;AAAG,WAAO,CAAC;AAGlC,QAAM,SAAS,QAAQ,MAAM,EAAE,KAAK,CAAC,GAAG,MAAM,EAAE,QAAQ,EAAE,KAAK;AAC/D,QAAM,SAAwB,CAAC;AAC/B,QAAM,UAAU,oBAAI,IAAY;AAEhC,aAAW,UAAU,QAAQ;AAE3B,QAAI,WAAW;AACf,aAAS,MAAM,OAAO,eAAe,OAAO,OAAO,aAAa,OAAO;AACrE,UAAI,QAAQ,IAAI,GAAG,GAAG;AACpB,mBAAW;AACX;AAAA,MACF;AAAA,IACF;AAEA,QAAI,CAAC,UAAU;AACb,aAAO,KAAK,MAAM;AAElB,eAAS,MAAM,OAAO,eAAe,OAAO,OAAO,aAAa,OAAO;AACrE,gBAAQ,IAAI,GAAG;AAAA,MACjB;AAAA,IACF;AAAA,EACF;AAEA,SAAO,OAAO,KAAK,CAAC,GAAG,MAAM,EAAE,QAAQ,EAAE,KAAK;AAChD;;;AC9WA,IAAM,iBAA8C;AAAA,EAClD,cAAc;AAAA;AAAA,EACd,WAAW;AAAA,EACX,mBAAmB;AAAA,EACnB,gBAAgB;AAAA,EAChB,UAAU;AAAA,EACV,mBAAmB;AAAA,EACnB,SAAS;AAAA,IACP,OAAO;AAAA,IACP,OAAO;AAAA,IACP,OAAO;AAAA,IACP,WAAW;AAAA,IACX,SAAS;AAAA,IACT,UAAU;AAAA,EACZ;AAAA,EACA,QAAQ;AAAA,EACR,UAAU;AAAA,EACV,yBAAyB;AAAA,EACzB,mBAAmB;AAAA,EACnB,yBAAyB;AAC3B;AAKA,IAAM,eAAe,oBAAI,QAA+B;AAQjD,SAAS,kBAAkB,aAAgC,CAAC,GAAgB;AAEjF,QAAM,SAAsC;AAAA,IAC1C,cAAc,WAAW,gBAAgB,eAAe;AAAA,IACxD,WAAW,WAAW,aAAa,eAAe;AAAA,IAClD,mBAAmB,WAAW,qBAAqB,eAAe;AAAA,IAClE,gBAAgB,WAAW,kBAAkB,eAAe;AAAA,IAC5D,UAAU,WAAW,YAAY,eAAe;AAAA,IAChD,mBAAmB,WAAW,qBAAqB,eAAe;AAAA,IAClE,SAAS;AAAA,MACP,OAAO,WAAW,SAAS,SAAS,eAAe,QAAQ;AAAA,MAC3D,OAAO,WAAW,SAAS,SAAS,eAAe,QAAQ;AAAA,MAC3D,OAAO,WAAW,SAAS,SAAS,eAAe,QAAQ;AAAA,MAC3D,WAAW,WAAW,SAAS,aAAa,eAAe,QAAQ;AAAA,MACnE,SAAS,WAAW,SAAS,WAAW,eAAe,QAAQ;AAAA,MAC/D,UAAU,WAAW,SAAS,YAAY,eAAe,QAAQ;AAAA,IACnE;AAAA,IACA,QAAQ,WAAW,UAAU,eAAe;AAAA,IAC5C,UAAU,WAAW,YAAY,eAAe;AAAA,IAChD,yBAAyB,WAAW,2BAA2B,eAAe;AAAA,IAC9E,mBAAmB,WAAW,qBAAqB,eAAe;AAAA,IAClE,yBAAyB,WAAW,2BAA2B,eAAe;AAAA,EAChF;AAEA,QAAM,SAAsB;AAAA,IAC1B,MAAM;AAAA;AAAA;AAAA;AAAA,IAKN,aAAa,OAAO,UAAoB;AACtC,cAAQ,IAAI,+CAAwC;AAGpD,YAAM,QAAqB;AAAA,QACzB,YAAY,CAAC;AAAA,QACb;AAAA,QACA,mBAAmB,oBAAI,IAAI;AAAA,QAC3B,gBAAgB;AAAA,MAClB;AAGA,UAAI,OAAO,kBAAkB,OAAO,UAAU;AAC5C,YAAI;AACF,kBAAQ,IAAI,6CAAsC;AAClD,gBAAM,aAAa,MAAM,yBAAyB,OAAO,QAAQ;AACjE,kBAAQ,IAAI,iBAAY,OAAO,KAAK,MAAM,UAAU,EAAE,MAAM,sBAAsB;AAAA,QACpF,SAAS,OAAO;AACd,kBAAQ,MAAM,0CAAgC,KAAK;AAAA,QAErD;AAAA,MACF;AAGA,YAAM,OAAQ,MAAM,MAAc,MAAM;AACxC,UAAI,MAAM;AACR,cAAM,iBAAiB,OAAO,KAAK,IAAI,EAAE;AACzC,cAAM,oBAAoB,6BAA6B,MAAM,OAAO,YAAY;AAChF,gBAAQ,IAAI,iDAA0C,MAAM,cAAc,YAAY;AAAA,MACxF;AAGA,mBAAa,IAAI,OAAO,KAAK;AAC7B,cAAQ,IAAI,wCAAmC;AAI/C,mBAAa,MAAM;AACjB,YAAI,OAAQ,WAAmB,2BAA2B,YAAY;AACpE,kBAAQ,IAAI,qCAA8B;AAC1C,UAAC,WAAmB,uBAAuB;AAAA,QAC7C,OAAO;AACL,kBAAQ,KAAK,yDAA+C;AAAA,QAC9D;AAAA,MACF,CAAC;AAAA,IACH;AAAA,EACF;AAEA,SAAO;AACT;AAQA,eAAsB,sBACpB,OACA,QACA,UACoC;AACpC,QAAM,YAAY,YAAY,IAAI;AAGlC,QAAM,QAAQ,aAAa,IAAI,KAAK;AAEpC,MAAI,CAAC,OAAO;AACV,YAAQ,MAAM,qCAAgC;AAC9C,UAAM,IAAI,MAAM,8CAA8C;AAAA,EAChE;AAEA,QAAM,EAAE,MAAM,WAAW,IAAI;AAE7B,MAAI,CAAC,QAAQ,OAAO,SAAS,UAAU;AACrC,WAAO,EAAE,SAAS,EAAE,WAAW,OAAO,KAAK,EAAE,GAAG,MAAM,CAAC,GAAG,OAAO,EAAE;AAAA,EACrE;AAGA,QAAM,eAAgB,cAAc,WAAW,CAAC,KAAM,MAAM,OAAO;AAGnE,QAAM,cAAc,SAAS,IAAI;AAEjC,MAAI,YAAY,WAAW,GAAG;AAC5B,WAAO,EAAE,SAAS,EAAE,WAAW,OAAO,KAAK,EAAE,GAAG,MAAM,CAAC,GAAG,OAAO,EAAE;AAAA,EACrE;AAGA,QAAM,YAAY,MAAM,OAAO,oBAC3B,2BAA2B,aAAa,MAAM,OAAO,SAAS,IAC9D,MAAM,OAAO;AAEjB,UAAQ,IAAI,mCAA4B,IAAI,MAAM,YAAY,MAAM,uBAAuB,SAAS,GAAG;AAGvG,MAAI;AAEJ,MAAI;AAGF,UAAM,YAAa,MAAc,MAAM;AAEvC,QAAI,CAAC,WAAW;AACd,cAAQ,MAAM,gDAA2C;AACzD,aAAO,EAAE,SAAS,EAAE,WAAW,OAAO,KAAK,EAAE,GAAG,MAAM,CAAC,GAAG,OAAO,EAAE;AAAA,IACrE;AAEA,YAAQ,IAAI,qCAA8B,OAAO,KAAK,aAAa,CAAC,CAAC,CAAC;AAGtE,QAAI,YAAY;AAGhB,QAAI,UAAU,UAAU,YAAY,GAAG,MAAM;AAC3C,kBAAY,UAAU,QAAQ,YAAY,EAAE;AAC5C,cAAQ,IAAI,4DAAuD;AAAA,IACrE,WAES,UAAU,YAAY,GAAG,MAAM;AACtC,kBAAY,UAAU,YAAY,EAAE;AACpC,cAAQ,IAAI,6DAAwD;AAAA,IACtE;AAEA,QAAI,CAAC,WAAW;AACd,cAAQ,MAAM,6CAAwC,YAAY;AAClE,cAAQ,MAAM,qCAAqC,OAAO,KAAK,SAAS,CAAC;AACzE,aAAO,EAAE,SAAS,EAAE,WAAW,OAAO,KAAK,EAAE,GAAG,MAAM,CAAC,GAAG,OAAO,EAAE;AAAA,IACrE;AAEA,iBAAa,+BAA+B,SAAS;AACrD,YAAQ,IAAI,uBAAgB,WAAW,IAAI,0BAA0B;AAAA,EACvE,SAAS,OAAO;AACd,YAAQ,MAAM,wCAAmC,KAAK;AACtD,WAAO,EAAE,SAAS,EAAE,WAAW,OAAO,KAAK,EAAE,GAAG,MAAM,CAAC,GAAG,OAAO,EAAE;AAAA,EACrE;AAGA,QAAM,gBAAgB;AAAA,IACpB;AAAA,IACA;AAAA,IACA;AAAA,IACA,MAAM,OAAO,iBAAiB,MAAM,aAAa;AAAA,IACjD,MAAM,OAAO;AAAA,EACf;AAGA,QAAM,qBAAqB;AAAA,IACzB;AAAA,IACA,MAAM,OAAO;AAAA,EACf;AAEA,UAAQ,IAAI,+BAAwB,MAAM,KAAK,mBAAmB,OAAO,CAAC,EAAE,OAAO,CAAC,KAAK,MAAM,MAAM,EAAE,QAAQ,CAAC,CAAC,QAAQ;AAGzH,QAAM,kBAAmC,CAAC;AAE1C,UAAQ,IAAI,yCAAkC;AAAA,IAC5C,UAAU,OAAO,KAAM,MAAc,QAAQ,CAAC,CAAC;AAAA,IAC/C,SAAS,CAAC,CAAG,MAAc,MAAM;AAAA,IACjC,UAAW,MAAc,MAAM,OAAO,OAAQ,MAAc,KAAK,OAAO;AAAA,EAC1E,CAAC;AAGD,MAAI,OAA4B,CAAC;AAGjC,MAAK,MAAc,MAAM,MAAM,MAAM;AACnC,WAAQ,MAAc,KAAK,KAAK;AAChC,YAAQ,IAAI,2CAAsC;AAAA,EACpD,WAEU,MAAc,MAAM,QAAQ,OAAQ,MAAc,KAAK,SAAS,UAAU;AAElF,UAAM,WAAW,OAAO,KAAM,MAAc,KAAK,IAAI,EAAE,CAAC;AACxD,QAAI,YAAY,aAAa,iCAAiC,aAAa,SAAS;AAClF,aAAQ,MAAc,KAAK;AAC3B,cAAQ,IAAI,+CAA0C;AAAA,IACxD;AAAA,EACF;AAEA,MAAI,OAAO,KAAK,IAAI,EAAE,WAAW,GAAG;AAClC,YAAQ,IAAI,0DAAqD;AAAA,MAC/D,aAAa,CAAC,CAAG,MAAc,MAAM;AAAA,MACrC,cAAe,MAAc,MAAM,OAAO,OAAO,KAAM,MAAc,KAAK,IAAI,IAAI;AAAA,MAClF,iBAAiB,CAAC,CAAG,MAAc,MAAM,MAAM;AAAA,MAC/C,mBAAoB,MAAc,MAAM,MAAM,OAAO,OAAO,KAAM,MAAc,KAAK,KAAK,IAAI,EAAE,SAAS;AAAA,IAC3G,CAAC;AAAA,EACH;AAEA,UAAQ,IAAI,+BAAwB,OAAO,KAAK,IAAI,EAAE,MAAM,YAAY;AAExE,aAAW,CAAC,OAAO,GAAG,KAAK,OAAO,QAAQ,IAAI,GAAG;AAC/C,UAAM,OAAO,IAAI,YAAY;AAE7B,QAAI,CAAC,QAAQ,OAAO,SAAS,UAAU;AACrC;AAAA,IACF;AAGA,UAAM,YAAY,SAAS,IAAI;AAI/B,UAAM,UAAU;AAAA,MACd;AAAA,MACA;AAAA,MACA;AAAA,QACE,SAAS,MAAM,OAAO;AAAA,QACtB,QAAQ,MAAM,OAAO;AAAA,QACrB,yBAAyB,MAAM,OAAO;AAAA,MACxC;AAAA,MACA,MAAM;AAAA,MACN,MAAM;AAAA,IACR;AAEA,QAAI,QAAQ,SAAS,GAAG;AAEtB,YAAM,WAAW,KAAK,IAAI,GAAG,QAAQ,IAAI,OAAK,EAAE,KAAK,CAAC;AAEtD,sBAAgB,KAAK;AAAA,QACnB,IAAI;AAAA,QACJ;AAAA,QACA,OAAO;AAAA,QACP,UAAU;AAAA,MACZ,CAAC;AAAA,IACH;AAAA,EACF;AAGA,kBAAgB,KAAK,CAAC,GAAG,MAAM,EAAE,QAAQ,EAAE,KAAK;AAGhD,MAAI,kBAAkB;AACtB,MAAI,MAAM,OAAO,2BAA2B,MAAM,OAAO,oBAAoB,GAAG;AAC9E,UAAM,YAAY,MAAM,OAAO;AAC/B,UAAM,cAAc,gBAAgB;AACpC,sBAAkB,gBAAgB,OAAO,OAAK,EAAE,SAAS,SAAS;AAClE,YAAQ,IAAI,uCAA2B,WAAW,WAAM,gBAAgB,MAAM,gBAAgB,SAAS,GAAG;AAAA,EAC5G;AAGA,QAAM,QAAQ,OAAO,SAAS,gBAAgB;AAC9C,QAAM,iBAAiB,gBAAgB,MAAM,GAAG,KAAK;AAGrD,QAAM,OAAO,eAAe,IAAI,YAAU;AAAA,IACxC,IAAI,MAAM;AAAA,IACV,OAAO,MAAM;AAAA,IACb,UAAU,MAAM;AAAA;AAAA,IAEhB,UAAU,MAAM;AAAA,EAClB,EAAE;AAEF,QAAM,UAAU,YAAY,IAAI,IAAI;AAEpC,UAAQ,IAAI,gBAAW,KAAK,MAAM,eAAe,QAAQ,QAAQ,CAAC,CAAC,cAAc,KAAK,GAAG;AAEzF,SAAO;AAAA,IACL,SAAS;AAAA,MACP,WAAW,GAAG,QAAQ,QAAQ,CAAC,CAAC;AAAA,MAChC,KAAK,KAAK,MAAM,UAAU,GAAO;AAAA;AAAA,IACnC;AAAA,IACA;AAAA,IACA,OAAO,KAAK;AAAA,EACd;AACF;AAKA,eAAe,yBACb,gBACqB;AACrB,MAAI;AACF,YAAQ,IAAI,0DAAmD;AAG/D,UAAM,EAAE,aAAa,IAAI,MAAM,OAAO,uBAAuB;AAE7D,UAAM,WAAW,aAAa,eAAe,KAAK,eAAe,UAAU;AAG3E,UAAM,EAAE,MAAM,MAAM,IAAI,MAAM,SAAS,IAAI,iBAAiB;AAE5D,YAAQ,IAAI,2CAAoC;AAAA,MAC9C,UAAU,CAAC,CAAC;AAAA,MACZ,cAAc,OAAO;AAAA,MACrB,SAAS,CAAC,CAAC;AAAA,MACX,UAAU,OAAO;AAAA,MACjB,UAAU,OAAO,OAAO,KAAK,IAAI,EAAE,SAAS;AAAA,IAC9C,CAAC;AAED,QAAI,OAAO;AACT,YAAM,IAAI,MAAM,mBAAmB,MAAM,OAAO,EAAE;AAAA,IACpD;AAEA,UAAM,aAAa,QAAQ,CAAC;AAC5B,YAAQ,IAAI,oBAAa,OAAO,KAAK,UAAU,EAAE,MAAM,gCAAgC;AAEvF,WAAO;AAAA,EACT,SAAS,OAAO;AACd,YAAQ,MAAM,iDAA4C,KAAK;AAC/D,UAAM;AAAA,EACR;AACF;AAKA,SAAS,6BACP,MACA,cACqB;AACrB,QAAM,KAAK,oBAAI,IAAoB;AAEnC,aAAW,OAAO,OAAO,OAAO,IAAI,GAAG;AACrC,UAAM,OAAO,IAAI,YAAY;AAE7B,QAAI,CAAC,QAAQ,OAAO,SAAS,UAAU;AACrC;AAAA,IACF;AAGA,UAAM,QAAQ,IAAI,IAAI,SAAS,IAAI,CAAC;AAGpC,eAAW,QAAQ,OAAO;AACxB,SAAG,IAAI,OAAO,GAAG,IAAI,IAAI,KAAK,KAAK,CAAC;AAAA,IACtC;AAAA,EACF;AAEA,SAAO;AACT;AAQA,SAAS,cAAc,MAAsB;AAC3C,SAAO,KACJ,YAAY,EACZ,UAAU,KAAK,EACf,QAAQ,oBAAoB,EAAE,EAE9B,QAAQ,gFAAgF,GAAG,EAC3F,QAAQ,6DAA6D,EAAE,EACvE,QAAQ,mBAAmB,GAAG,EAC9B,QAAQ,4BAA4B,GAAG,EACvC,QAAQ,QAAQ,GAAG,EACnB,KAAK;AACV;AAQA,SAAS,SAAS,MAAwB;AAExC,SAAO,cAAc,IAAI,EACtB,MAAM,KAAK,EACX,OAAO,WAAS,MAAM,SAAS,CAAC;AACrC","sourcesContent":["/**\n * Fuzzy matching utilities using bounded Levenshtein distance\n * \n * This is the same algorithm used by Orama's match-highlight plugin\n * for consistent fuzzy matching behavior.\n */\n\n/**\n * Result of bounded Levenshtein distance calculation\n */\nexport interface BoundedLevenshteinResult {\n /** Whether the distance is within bounds */\n isBounded: boolean;\n /** The actual distance (only valid if isBounded is true) */\n distance: number;\n}\n\n/**\n * Calculate bounded Levenshtein distance between two strings\n * \n * Stops early if distance exceeds the bound for better performance.\n * This is the same algorithm as Orama's internal boundedLevenshtein.\n * \n * @param a - First string\n * @param b - Second string\n * @param bound - Maximum allowed distance\n * @returns Result indicating if strings are within bound and the distance\n */\nexport function boundedLevenshtein(\n a: string,\n b: string,\n bound: number\n): BoundedLevenshteinResult {\n // Quick checks\n if (a === b) {\n return { isBounded: true, distance: 0 };\n }\n\n const aLen = a.length;\n const bLen = b.length;\n\n // If length difference exceeds bound, no need to calculate\n if (Math.abs(aLen - bLen) > bound) {\n return { isBounded: false, distance: bound + 1 };\n }\n\n // Swap to ensure a is shorter (optimization)\n if (aLen > bLen) {\n [a, b] = [b, a];\n }\n\n const m = a.length;\n const n = b.length;\n\n // Use single array instead of matrix (memory optimization)\n let prevRow = new Array(n + 1);\n let currRow = new Array(n + 1);\n\n // Initialize first row\n for (let j = 0; j <= n; j++) {\n prevRow[j] = j;\n }\n\n for (let i = 1; i <= m; i++) {\n currRow[0] = i;\n let minInRow = i;\n\n for (let j = 1; j <= n; j++) {\n const cost = a[i - 1] === b[j - 1] ? 0 : 1;\n\n currRow[j] = Math.min(\n prevRow[j] + 1, // deletion\n currRow[j - 1] + 1, // insertion\n prevRow[j - 1] + cost // substitution\n );\n\n minInRow = Math.min(minInRow, currRow[j]);\n }\n\n // Early termination: if all values in row exceed bound, we're done\n if (minInRow > bound) {\n return { isBounded: false, distance: bound + 1 };\n }\n\n // Swap rows for next iteration\n [prevRow, currRow] = [currRow, prevRow];\n }\n\n const distance = prevRow[n];\n return {\n isBounded: distance <= bound,\n distance\n };\n}\n\n/**\n * Check if a word matches a query token with fuzzy matching\n * \n * @param word - Word from document\n * @param queryToken - Token from search query\n * @param tolerance - Maximum edit distance allowed\n * @returns Match result with score\n */\nexport function fuzzyMatch(\n word: string,\n queryToken: string,\n tolerance: number\n): { matches: boolean; distance: number; score: number } {\n // Exact match\n if (word === queryToken) {\n return { matches: true, distance: 0, score: 1.0 };\n }\n\n // Prefix match (high score, no distance)\n if (word.startsWith(queryToken)) {\n return { matches: true, distance: 0, score: 0.95 };\n }\n\n // Fuzzy match with tolerance\n const result = boundedLevenshtein(word, queryToken, tolerance);\n \n if (result.isBounded) {\n // Score decreases with distance\n // distance 1 = 0.8, distance 2 = 0.6, etc.\n const score = 1.0 - (result.distance * 0.2);\n return {\n matches: true,\n distance: result.distance,\n score: Math.max(0.1, score) // Minimum score of 0.1\n };\n }\n\n return { matches: false, distance: tolerance + 1, score: 0 };\n}\n\n/**\n * Calculate adaptive tolerance based on query length\n * \n * Longer queries get higher tolerance for better fuzzy matching.\n * \n * @param queryTokens - Array of query tokens\n * @param baseTolerance - Base tolerance value\n * @returns Calculated tolerance (always an integer)\n */\nexport function calculateAdaptiveTolerance(\n queryTokens: string[],\n baseTolerance: number\n): number {\n const queryLength = queryTokens.length;\n \n if (queryLength <= 2) {\n return baseTolerance;\n } else if (queryLength <= 4) {\n return baseTolerance + 1;\n } else if (queryLength <= 6) {\n return baseTolerance + 2;\n } else {\n return baseTolerance + 3;\n }\n}\n","/**\n * Candidate expansion: Find all possible matches for query tokens\n * including exact matches, fuzzy matches, and synonyms\n */\n\nimport { fuzzyMatch } from './fuzzy.js';\nimport type { Candidate, SynonymMap } from './types.js';\n\n/**\n * Extract all unique words from the radix tree index\n * \n * @param radixNode - Root node of the radix tree\n * @returns Set of all unique words in the index\n */\nexport function extractVocabularyFromRadixTree(radixNode: any): Set<string> {\n const vocabulary = new Set<string>();\n let nodesVisited = 0;\n let wordsFound = 0;\n \n function traverse(node: any, depth: number = 0) {\n if (!node) {\n return;\n }\n \n nodesVisited++;\n \n // Check if this node represents a complete word\n // e = true means it's an end of a word\n if (node.e && node.w && typeof node.w === 'string' && node.w.length > 0) {\n vocabulary.add(node.w);\n wordsFound++;\n }\n \n // Children can be Map, Array, or Object\n if (node.c) {\n if (node.c instanceof Map) {\n // Map format\n for (const [_key, childNode] of node.c) {\n traverse(childNode, depth + 1);\n }\n } else if (Array.isArray(node.c)) {\n // Array format: [[key, childNode], ...]\n for (const [_key, childNode] of node.c) {\n traverse(childNode, depth + 1);\n }\n } else if (typeof node.c === 'object') {\n // Object format: {key: childNode, ...}\n for (const childNode of Object.values(node.c)) {\n traverse(childNode, depth + 1);\n }\n }\n }\n }\n \n traverse(radixNode);\n console.log(`๐Ÿ“š Extracted ${vocabulary.size} words from ${nodesVisited} nodes visited`);\n return vocabulary;\n}\n\n/**\n * Find all candidate matches for a single query token\n * \n * @param queryToken - Token from search query\n * @param vocabulary - Set of all words in the index\n * @param tolerance - Fuzzy matching tolerance\n * @param synonyms - Synonym map (optional)\n * @param synonymScore - Score multiplier for synonym matches\n * @returns Array of candidate matches\n */\nexport function findCandidatesForToken(\n queryToken: string,\n vocabulary: Set<string>,\n tolerance: number,\n synonyms?: SynonymMap,\n synonymScore: number = 0.8\n): Candidate[] {\n const candidates: Candidate[] = [];\n const seen = new Set<string>();\n\n // 1. Check for exact match\n if (vocabulary.has(queryToken)) {\n candidates.push({\n word: queryToken,\n type: 'exact',\n queryToken,\n distance: 0,\n score: 1.0\n });\n seen.add(queryToken);\n }\n\n // 2. Check for fuzzy matches\n for (const word of vocabulary) {\n if (seen.has(word)) continue;\n\n const match = fuzzyMatch(word, queryToken, tolerance);\n if (match.matches) {\n candidates.push({\n word,\n type: 'fuzzy',\n queryToken,\n distance: match.distance,\n score: match.score\n });\n seen.add(word);\n }\n }\n\n // 3. Check for synonym matches\n if (synonyms && synonyms[queryToken]) {\n for (const synonym of synonyms[queryToken]) {\n if (seen.has(synonym)) continue;\n if (vocabulary.has(synonym)) {\n candidates.push({\n word: synonym,\n type: 'synonym',\n queryToken,\n distance: 0,\n score: synonymScore\n });\n seen.add(synonym);\n }\n }\n }\n\n return candidates;\n}\n\n/**\n * Find candidates for all query tokens\n * \n * @param queryTokens - Array of tokens from search query\n * @param vocabulary - Set of all words in the index\n * @param tolerance - Fuzzy matching tolerance\n * @param synonyms - Synonym map (optional)\n * @param synonymScore - Score multiplier for synonym matches\n * @returns Map of query tokens to their candidate matches\n */\nexport function findAllCandidates(\n queryTokens: string[],\n vocabulary: Set<string>,\n tolerance: number,\n synonyms?: SynonymMap,\n synonymScore: number = 0.8\n): Map<string, Candidate[]> {\n const candidatesMap = new Map<string, Candidate[]>();\n\n for (const token of queryTokens) {\n const tokenCandidates = findCandidatesForToken(\n token,\n vocabulary,\n tolerance,\n synonyms,\n synonymScore\n );\n candidatesMap.set(token, tokenCandidates);\n }\n\n return candidatesMap;\n}\n\n/**\n * Get total number of candidates across all tokens\n * \n * @param candidatesMap - Map of token to candidates\n * @returns Total count of all candidates\n */\nexport function getTotalCandidateCount(\n candidatesMap: Map<string, Candidate[]>\n): number {\n let total = 0;\n for (const candidates of candidatesMap.values()) {\n total += candidates.length;\n }\n return total;\n}\n\n/**\n * Filter candidates by minimum score threshold\n * \n * @param candidatesMap - Map of token to candidates\n * @param minScore - Minimum score threshold\n * @returns Filtered candidates map\n */\nexport function filterCandidatesByScore(\n candidatesMap: Map<string, Candidate[]>,\n minScore: number\n): Map<string, Candidate[]> {\n const filtered = new Map<string, Candidate[]>();\n\n for (const [token, candidates] of candidatesMap.entries()) {\n const filteredCandidates = candidates.filter(c => c.score >= minScore);\n if (filteredCandidates.length > 0) {\n filtered.set(token, filteredCandidates);\n }\n }\n\n return filtered;\n}\n","/**\n * Phrase scoring algorithm with semantic weighting\n */\n\nimport type { WordMatch, PhraseMatch, Candidate, GapWord } from './types.js';\n\n/**\n * Configuration for phrase scoring\n */\nexport interface ScoringConfig {\n weights: {\n exact: number;\n fuzzy: number;\n order: number;\n proximity: number;\n density: number;\n semantic: number;\n };\n maxGap: number;\n /** \n * Multiplier for proximity window calculation.\n * proximityWindow = queryTokens.length ร— proximitySpanMultiplier\n */\n proximitySpanMultiplier: number;\n}\n\n/**\n * Find all phrase matches in a document\n * \n * @param documentTokens - Tokenized document content (needed to extract gap words)\n * @param candidatesMap - Map of query tokens to their candidates\n * @param config - Scoring configuration\n * @param documentFrequency - Document frequency map for TF-IDF\n * @param totalDocuments - Total number of documents\n * @returns Array of phrase matches\n */\nexport function findPhrasesInDocument(\n documentTokens: string[],\n candidatesMap: Map<string, Candidate[]>,\n config: ScoringConfig,\n documentFrequency: Map<string, number>,\n totalDocuments: number\n): PhraseMatch[] {\n const phrases: PhraseMatch[] = [];\n const queryTokens = Array.from(candidatesMap.keys());\n\n // Find all word matches in document\n const wordMatches: WordMatch[] = [];\n \n for (let i = 0; i < documentTokens.length; i++) {\n const docWord = documentTokens[i];\n \n // Check if this word matches any query token\n for (const [queryToken, candidates] of candidatesMap.entries()) {\n for (const candidate of candidates) {\n if (candidate.word === docWord) {\n wordMatches.push({\n word: docWord,\n queryToken,\n position: i,\n type: candidate.type,\n distance: candidate.distance,\n score: candidate.score\n });\n }\n }\n }\n }\n\n // Build phrases from word matches using sliding window\n for (let i = 0; i < wordMatches.length; i++) {\n const phrase = buildPhraseFromPosition(\n wordMatches,\n i,\n queryTokens,\n config,\n documentFrequency,\n totalDocuments,\n wordMatches,\n documentTokens // Pass document tokens to extract gap words\n );\n \n if (phrase && phrase.words.length > 0) {\n phrases.push(phrase);\n }\n }\n\n // Deduplicate and sort by score\n return deduplicatePhrases(phrases);\n}\n\n/**\n * Build a phrase starting from a specific word match position\n * \n * @param wordMatches - All word matches in document\n * @param startIndex - Starting index in wordMatches array\n * @param queryTokens - Original query tokens\n * @param config - Scoring configuration\n * @param documentFrequency - Document frequency map\n * @param totalDocuments - Total document count\n * @param allWordMatches - All word matches in document (for density calculation)\n * @param documentTokens - Original document tokens (for gap word extraction)\n * @returns Phrase match or null\n */\nfunction buildPhraseFromPosition(\n wordMatches: WordMatch[],\n startIndex: number,\n queryTokens: string[],\n config: ScoringConfig,\n documentFrequency: Map<string, number>,\n totalDocuments: number,\n allWordMatches: WordMatch[],\n documentTokens: string[]\n): PhraseMatch | null {\n const startMatch = wordMatches[startIndex];\n const phraseWords: WordMatch[] = [startMatch];\n const coveredTokens = new Set([startMatch.queryToken]);\n const gapWords: GapWord[] = [];\n let totalGapUsed = 0;\n\n // Look for nearby matches to complete the phrase\n for (let i = startIndex + 1; i < wordMatches.length; i++) {\n const match = wordMatches[i];\n const lastPos = phraseWords[phraseWords.length - 1].position;\n const gap = match.position - lastPos - 1;\n\n // Stop if gap exceeds maximum\n if (gap > config.maxGap) {\n break;\n }\n\n // Only process if it's a different query token we need\n if (!coveredTokens.has(match.queryToken)) {\n // Track gap words between last match and current match\n for (let pos = lastPos + 1; pos < match.position; pos++) {\n totalGapUsed++;\n gapWords.push({\n word: documentTokens[pos],\n position: pos,\n gapIndex: totalGapUsed\n });\n }\n\n phraseWords.push(match);\n coveredTokens.add(match.queryToken);\n\n // Stop if we have all query tokens\n if (coveredTokens.size === queryTokens.length) {\n break;\n }\n }\n }\n\n // Calculate phrase score\n if (phraseWords.length > 0) {\n const coverage = phraseWords.length / queryTokens.length;\n const span = phraseWords[phraseWords.length - 1].position - phraseWords[0].position + 1;\n \n const { score, breakdown } = calculatePhraseScore(\n phraseWords,\n queryTokens,\n config,\n documentFrequency,\n totalDocuments,\n allWordMatches,\n coverage\n );\n\n return {\n words: phraseWords,\n gapWords,\n gapUsed: totalGapUsed,\n coverage,\n startPosition: phraseWords[0].position,\n endPosition: phraseWords[phraseWords.length - 1].position,\n span,\n inOrder: isInOrder(phraseWords, queryTokens),\n score,\n scoreBreakdown: breakdown\n };\n }\n\n return null;\n}\n\n/**\n * Calculate overall phrase score\n * \n * @param phraseWords - Words in the phrase\n * @param queryTokens - Original query tokens\n * @param config - Scoring configuration\n * @param documentFrequency - Document frequency map\n * @param totalDocuments - Total document count\n * @param allWordMatches - All word matches in document (for density calculation)\n * @param coverage - Pre-calculated coverage ratio (phraseWords.length / queryTokens.length)\n * @returns Phrase score (0-1) and detailed component breakdown\n */\nfunction calculatePhraseScore(\n phraseWords: WordMatch[],\n queryTokens: string[],\n config: ScoringConfig,\n documentFrequency: Map<string, number>,\n totalDocuments: number,\n allWordMatches: WordMatch[],\n coverage: number\n): { score: number; breakdown: { base: number; order: number; proximity: number; density: number; semantic: number; coverage: number } } {\n // Base score from word matches\n // Each word contributes: matchScore ร— typeWeight\n let baseScore = 0;\n for (const word of phraseWords) {\n const weight = word.type === 'exact' ? config.weights.exact :\n word.type === 'fuzzy' ? config.weights.fuzzy : \n config.weights.fuzzy * 0.8; // synonym gets 80% of fuzzy weight\n baseScore += word.score * weight;\n }\n baseScore /= phraseWords.length;\n\n // Order bonus: 1.0 if words appear in query order, 0.5 otherwise\n const inOrder = isInOrder(phraseWords, queryTokens);\n const orderScore = inOrder ? 1.0 : 0.5;\n\n // Proximity bonus (closer words score higher)\n // Short-circuit: skip if maxGap=0, proximity weight is 0, or single-word query (proximity meaningless)\n let proximityScore = 0;\n if (config.maxGap > 0 && config.weights.proximity > 0 && queryTokens.length > 1) {\n const span = phraseWords[phraseWords.length - 1].position - phraseWords[0].position + 1;\n const proximityWindow = queryTokens.length * config.proximitySpanMultiplier;\n proximityScore = Math.max(0, 1.0 - (span / proximityWindow));\n }\n\n // Density: Only applies to single-word queries (measures word repetition in document)\n // For multi-word phrase queries, density is 0 (coverage handles completeness separately)\n let densityScore = 0;\n \n if (queryTokens.length === 1) {\n // Single-word query: reward repetition\n const totalOccurrences = allWordMatches.length;\n // Cap at reasonable maximum to avoid runaway scores\n densityScore = Math.min(1.0, totalOccurrences / 10);\n }\n // For multi-word queries: densityScore stays 0\n // Coverage is applied as a multiplier at the end instead\n\n // Semantic score (TF-IDF based)\n const semanticScore = calculateSemanticScore(\n phraseWords,\n documentFrequency,\n totalDocuments\n );\n\n // Weighted combination\n const weights = config.weights;\n \n // Calculate weighted components\n const weightedBase = baseScore;\n const weightedOrder = orderScore * weights.order;\n const weightedProximity = proximityScore * weights.proximity;\n const weightedDensity = densityScore * weights.density;\n const weightedSemantic = semanticScore * weights.semantic;\n \n const totalScore = weightedBase + weightedOrder + weightedProximity + weightedDensity + weightedSemantic;\n\n // Calculate max possible score\n // FIX: Use actual max base weight (highest of exact/fuzzy) instead of hardcoded 1.0\n const maxBaseWeight = Math.max(weights.exact, weights.fuzzy);\n // Only include proximity in max if it can actually contribute (avoids penalizing scores when maxGap=0 or single-word)\n const effectiveProximityWeight = (config.maxGap > 0 && weights.proximity > 0 && queryTokens.length > 1) ? weights.proximity : 0;\n const maxPossibleScore = maxBaseWeight + weights.order + effectiveProximityWeight + weights.density + weights.semantic;\n \n // Normalize to 0-1 range\n const normalizedScore = totalScore / maxPossibleScore;\n \n // FIX: Apply coverage as a MULTIPLIER for multi-word queries\n // This ensures incomplete matches (2/3) can never outscore complete matches (3/3)\n const coverageMultiplier = queryTokens.length > 1 ? coverage : 1.0;\n const score = normalizedScore * coverageMultiplier;\n\n // Component contributions to the final normalized score (before coverage multiplier)\n const base = weightedBase / maxPossibleScore;\n const order = weightedOrder / maxPossibleScore;\n const proximity = weightedProximity / maxPossibleScore;\n const density = weightedDensity / maxPossibleScore;\n const semantic = weightedSemantic / maxPossibleScore;\n\n return {\n score,\n breakdown: {\n base,\n order,\n proximity,\n density,\n semantic,\n coverage: coverageMultiplier // Show coverage multiplier in breakdown\n }\n };\n}\n\n/**\n * Check if words are in the same order as query tokens\n * \n * @param phraseWords - Words in the phrase\n * @param queryTokens - Original query tokens\n * @returns True if in order\n */\nfunction isInOrder(phraseWords: WordMatch[], queryTokens: string[]): boolean {\n const tokenOrder = new Map(queryTokens.map((token, index) => [token, index]));\n \n for (let i = 1; i < phraseWords.length; i++) {\n const prevOrder = tokenOrder.get(phraseWords[i - 1].queryToken) ?? -1;\n const currOrder = tokenOrder.get(phraseWords[i].queryToken) ?? -1;\n \n if (currOrder < prevOrder) {\n return false;\n }\n }\n \n return true;\n}\n\n/**\n * Calculate semantic score using TF-IDF\n * \n * @param phraseWords - Words in the phrase\n * @param documentFrequency - Document frequency map\n * @param totalDocuments - Total document count\n * @returns Semantic score (0-1)\n */\nfunction calculateSemanticScore(\n phraseWords: WordMatch[],\n documentFrequency: Map<string, number>,\n totalDocuments: number\n): number {\n // Handle edge case: no documents\n if (totalDocuments === 0) {\n return 0;\n }\n \n let tfidfSum = 0;\n \n for (const word of phraseWords) {\n const df = documentFrequency.get(word.word) || 1;\n const idf = Math.log(totalDocuments / df);\n tfidfSum += idf;\n }\n \n // Normalize by phrase length\n const avgTfidf = tfidfSum / phraseWords.length;\n \n // Normalize to 0-1 range (assuming max IDF of ~10)\n return Math.min(1.0, avgTfidf / 10);\n}\n\n/**\n * Deduplicate overlapping phrases, keeping highest scoring ones\n * \n * @param phrases - Array of phrase matches\n * @returns Deduplicated phrases sorted by score\n */\nfunction deduplicatePhrases(phrases: PhraseMatch[]): PhraseMatch[] {\n if (phrases.length === 0) return [];\n\n // Sort by score descending\n const sorted = phrases.slice().sort((a, b) => b.score - a.score);\n const result: PhraseMatch[] = [];\n const covered = new Set<number>();\n\n for (const phrase of sorted) {\n // Check if this phrase overlaps with already selected phrases\n let overlaps = false;\n for (let pos = phrase.startPosition; pos <= phrase.endPosition; pos++) {\n if (covered.has(pos)) {\n overlaps = true;\n break;\n }\n }\n\n if (!overlaps) {\n result.push(phrase);\n // Mark positions as covered\n for (let pos = phrase.startPosition; pos <= phrase.endPosition; pos++) {\n covered.add(pos);\n }\n }\n }\n\n return result.sort((a, b) => b.score - a.score);\n}\n","/**\n * Fuzzy Phrase Plugin for Orama\n * \n * Advanced fuzzy phrase matching with semantic weighting and synonym expansion.\n * Completely independent from QPS - accesses Orama's radix tree directly.\n */\n\nimport type { AnyOrama, OramaPlugin, Results, TypedDocument } from '@wcs-colab/orama';\nimport type { FuzzyPhraseConfig, PluginState, SynonymMap, DocumentMatch } from './types.js';\nimport { calculateAdaptiveTolerance } from './fuzzy.js';\nimport { \n extractVocabularyFromRadixTree, \n findAllCandidates,\n filterCandidatesByScore \n} from './candidates.js';\nimport { findPhrasesInDocument } from './scoring.js';\n\n/**\n * Default configuration\n */\nconst DEFAULT_CONFIG: Required<FuzzyPhraseConfig> = {\n textProperty: 'normalized_content', // Must match server's field name\n tolerance: 1,\n adaptiveTolerance: true,\n enableSynonyms: false,\n supabase: undefined as any,\n synonymMatchScore: 0.8,\n weights: {\n exact: 1.0,\n fuzzy: 0.8,\n order: 0.3,\n proximity: 0.2,\n density: 0.2,\n semantic: 0.15\n },\n maxGap: 5,\n minScore: 0.1,\n enableFinalScoreMinimum: false,\n finalScoreMinimum: 0.3,\n proximitySpanMultiplier: 5\n};\n\n/**\n * Plugin state storage (keyed by Orama instance)\n */\nconst pluginStates = new WeakMap<AnyOrama, PluginState>();\n\n/**\n * Create the Fuzzy Phrase Plugin\n * \n * @param userConfig - User configuration options\n * @returns Orama plugin instance\n */\nexport function pluginFuzzyPhrase(userConfig: FuzzyPhraseConfig = {}): OramaPlugin {\n // Merge user config with defaults\n const config: Required<FuzzyPhraseConfig> = {\n textProperty: userConfig.textProperty ?? DEFAULT_CONFIG.textProperty,\n tolerance: userConfig.tolerance ?? DEFAULT_CONFIG.tolerance,\n adaptiveTolerance: userConfig.adaptiveTolerance ?? DEFAULT_CONFIG.adaptiveTolerance,\n enableSynonyms: userConfig.enableSynonyms ?? DEFAULT_CONFIG.enableSynonyms,\n supabase: userConfig.supabase || DEFAULT_CONFIG.supabase,\n synonymMatchScore: userConfig.synonymMatchScore ?? DEFAULT_CONFIG.synonymMatchScore,\n weights: {\n exact: userConfig.weights?.exact ?? DEFAULT_CONFIG.weights.exact,\n fuzzy: userConfig.weights?.fuzzy ?? DEFAULT_CONFIG.weights.fuzzy,\n order: userConfig.weights?.order ?? DEFAULT_CONFIG.weights.order,\n proximity: userConfig.weights?.proximity ?? DEFAULT_CONFIG.weights.proximity,\n density: userConfig.weights?.density ?? DEFAULT_CONFIG.weights.density,\n semantic: userConfig.weights?.semantic ?? DEFAULT_CONFIG.weights.semantic\n },\n maxGap: userConfig.maxGap ?? DEFAULT_CONFIG.maxGap,\n minScore: userConfig.minScore ?? DEFAULT_CONFIG.minScore,\n enableFinalScoreMinimum: userConfig.enableFinalScoreMinimum ?? DEFAULT_CONFIG.enableFinalScoreMinimum,\n finalScoreMinimum: userConfig.finalScoreMinimum ?? DEFAULT_CONFIG.finalScoreMinimum,\n proximitySpanMultiplier: userConfig.proximitySpanMultiplier ?? DEFAULT_CONFIG.proximitySpanMultiplier\n };\n\n const plugin: OramaPlugin = {\n name: 'fuzzy-phrase',\n\n /**\n * Initialize plugin after index is created\n */\n afterCreate: async (orama: AnyOrama) => {\n console.log('๐Ÿ”ฎ Initializing Fuzzy Phrase Plugin...');\n\n // Initialize state\n const state: PluginState = {\n synonymMap: {},\n config,\n documentFrequency: new Map(),\n totalDocuments: 0\n };\n\n // Load synonyms from Supabase if enabled\n if (config.enableSynonyms && config.supabase) {\n try {\n console.log('๐Ÿ“– Loading synonyms from Supabase...');\n state.synonymMap = await loadSynonymsFromSupabase(config.supabase);\n console.log(`โœ… Loaded ${Object.keys(state.synonymMap).length} words with synonyms`);\n } catch (error) {\n console.error('โš ๏ธ Failed to load synonyms:', error);\n // Continue without synonyms\n }\n }\n\n // Calculate document frequencies for TF-IDF from document store\n const docs = (orama.data as any)?.docs?.docs;\n if (docs) {\n state.totalDocuments = Object.keys(docs).length;\n state.documentFrequency = calculateDocumentFrequencies(docs, config.textProperty);\n console.log(`๐Ÿ“Š Calculated document frequencies for ${state.totalDocuments} documents`);\n }\n\n // Store state\n pluginStates.set(orama, state);\n console.log('โœ… Fuzzy Phrase Plugin initialized');\n \n // Signal ready - emit a custom event that can be listened to\n // Use setImmediate to ensure this runs after the afterCreate hook completes\n setImmediate(() => {\n if (typeof (globalThis as any).fuzzyPhrasePluginReady === 'function') {\n console.log('๐Ÿ“ก Signaling plugin ready...');\n (globalThis as any).fuzzyPhrasePluginReady();\n } else {\n console.warn('โš ๏ธ fuzzyPhrasePluginReady callback not found');\n }\n });\n }\n };\n\n return plugin;\n}\n\n/**\n * Search with fuzzy phrase matching\n * \n * This function should be called instead of the regular search() function\n * to enable fuzzy phrase matching.\n */\nexport async function searchWithFuzzyPhrase<T extends AnyOrama>(\n orama: T, \n params: { term?: string; properties?: string[]; limit?: number },\n language?: string\n): Promise<Results<TypedDocument<T>>> {\n const startTime = performance.now();\n \n // Get plugin state\n const state = pluginStates.get(orama);\n \n if (!state) {\n console.error('โŒ Plugin state not initialized');\n throw new Error('Fuzzy Phrase Plugin not properly initialized');\n }\n\n const { term, properties } = params;\n \n if (!term || typeof term !== 'string') {\n return { elapsed: { formatted: '0ms', raw: 0 }, hits: [], count: 0 };\n }\n\n // Use specified property or default\n const textProperty = (properties && properties[0]) || state.config.textProperty;\n\n // Tokenize query\n const queryTokens = tokenize(term);\n \n if (queryTokens.length === 0) {\n return { elapsed: { formatted: '0ms', raw: 0 }, hits: [], count: 0 };\n }\n\n // Calculate tolerance (adaptive or fixed)\n const tolerance = state.config.adaptiveTolerance\n ? calculateAdaptiveTolerance(queryTokens, state.config.tolerance)\n : state.config.tolerance;\n\n console.log(`๐Ÿ” Fuzzy phrase search: \"${term}\" (${queryTokens.length} tokens, tolerance: ${tolerance})`);\n\n // Extract vocabulary from radix tree\n let vocabulary: Set<string>;\n \n try {\n // Access radix tree - the actual index data is in orama.data.index, not orama.index\n // orama.index is just the component interface (methods)\n const indexData = (orama as any).data?.index;\n \n if (!indexData) {\n console.error('โŒ No index data found in orama.data.index');\n return { elapsed: { formatted: '0ms', raw: 0 }, hits: [], count: 0 };\n }\n \n console.log('๐Ÿ” DEBUG: Index data keys:', Object.keys(indexData || {}));\n \n // Try different paths to find the radix tree\n let radixNode = null;\n \n // Path 1: QPS-style (orama.data.index.indexes[property].node)\n if (indexData.indexes?.[textProperty]?.node) {\n radixNode = indexData.indexes[textProperty].node;\n console.log('โœ… Found radix via QPS-style path (data.index.indexes)');\n }\n // Path 2: Standard Orama (orama.data.index[property].node)\n else if (indexData[textProperty]?.node) {\n radixNode = indexData[textProperty].node;\n console.log('โœ… Found radix via standard path (data.index[property])');\n }\n \n if (!radixNode) {\n console.error('โŒ Radix tree not found for property:', textProperty);\n console.error(' Available properties in index:', Object.keys(indexData));\n return { elapsed: { formatted: '0ms', raw: 0 }, hits: [], count: 0 };\n }\n\n vocabulary = extractVocabularyFromRadixTree(radixNode);\n console.log(`๐Ÿ“š Extracted ${vocabulary.size} unique words from index`);\n } catch (error) {\n console.error('โŒ Failed to extract vocabulary:', error);\n return { elapsed: { formatted: '0ms', raw: 0 }, hits: [], count: 0 };\n }\n\n // Find candidates for all query tokens\n const candidatesMap = findAllCandidates(\n queryTokens,\n vocabulary,\n tolerance,\n state.config.enableSynonyms ? state.synonymMap : undefined,\n state.config.synonymMatchScore\n );\n\n // Filter by minimum score\n const filteredCandidates = filterCandidatesByScore(\n candidatesMap,\n state.config.minScore\n );\n\n console.log(`๐ŸŽฏ Found candidates: ${Array.from(filteredCandidates.values()).reduce((sum, c) => sum + c.length, 0)} total`);\n\n // Search through all documents\n const documentMatches: DocumentMatch[] = [];\n \n console.log('๐Ÿ” DEBUG orama.data structure:', {\n dataKeys: Object.keys((orama as any).data || {}),\n hasDocs: !!((orama as any).data?.docs),\n docsType: (orama as any).data?.docs ? typeof (orama as any).data.docs : 'undefined'\n });\n \n // Try multiple possible document storage locations\n let docs: Record<string, any> = {};\n \n // Access the actual documents - they're nested in orama.data.docs.docs\n if ((orama as any).data?.docs?.docs) {\n docs = (orama as any).data.docs.docs;\n console.log('โœ… Found docs at orama.data.docs.docs');\n }\n // Fallback: orama.data.docs (might be the correct structure in some cases)\n else if ((orama as any).data?.docs && typeof (orama as any).data.docs === 'object') {\n // Check if it has document-like properties (not sharedInternalDocumentStore, etc.)\n const firstKey = Object.keys((orama as any).data.docs)[0];\n if (firstKey && firstKey !== 'sharedInternalDocumentStore' && firstKey !== 'count') {\n docs = (orama as any).data.docs;\n console.log('โœ… Found docs at orama.data.docs (direct)');\n }\n }\n \n if (Object.keys(docs).length === 0) {\n console.log('โŒ Could not find documents - available structure:', {\n hasDataDocs: !!((orama as any).data?.docs),\n dataDocsKeys: (orama as any).data?.docs ? Object.keys((orama as any).data.docs) : 'none',\n hasDataDocsDocs: !!((orama as any).data?.docs?.docs),\n dataDocsDocsCount: (orama as any).data?.docs?.docs ? Object.keys((orama as any).data.docs.docs).length : 0\n });\n }\n \n console.log(`๐Ÿ“„ Searching through ${Object.keys(docs).length} documents`);\n\n for (const [docId, doc] of Object.entries(docs)) {\n const text = doc[textProperty];\n \n if (!text || typeof text !== 'string') {\n continue;\n }\n\n // Tokenize document\n const docTokens = tokenize(text);\n\n // Find phrases in this document\n // Note: state.config.weights is guaranteed to have all properties from default merge\n const phrases = findPhrasesInDocument(\n docTokens,\n filteredCandidates,\n {\n weights: state.config.weights as { exact: number; fuzzy: number; order: number; proximity: number; density: number; semantic: number },\n maxGap: state.config.maxGap,\n proximitySpanMultiplier: state.config.proximitySpanMultiplier\n },\n state.documentFrequency,\n state.totalDocuments\n );\n\n if (phrases.length > 0) {\n // Calculate overall document score (highest phrase score)\n const docScore = Math.max(...phrases.map(p => p.score));\n\n documentMatches.push({\n id: docId,\n phrases,\n score: docScore,\n document: doc\n });\n }\n }\n\n // Sort by score descending\n documentMatches.sort((a, b) => b.score - a.score);\n\n // Apply final score minimum filter if enabled\n let filteredMatches = documentMatches;\n if (state.config.enableFinalScoreMinimum && state.config.finalScoreMinimum > 0) {\n const threshold = state.config.finalScoreMinimum;\n const beforeCount = filteredMatches.length;\n filteredMatches = filteredMatches.filter(m => m.score >= threshold);\n console.log(`๐ŸŽš๏ธ Final score filter: ${beforeCount} โ†’ ${filteredMatches.length} (threshold: ${threshold})`);\n }\n\n // Apply limit if specified\n const limit = params.limit ?? filteredMatches.length;\n const limitedMatches = filteredMatches.slice(0, limit);\n\n // Convert to Orama results format\n const hits = limitedMatches.map(match => ({\n id: match.id,\n score: match.score,\n document: match.document,\n // Store phrases for highlighting\n _phrases: match.phrases\n })) as any[];\n\n const elapsed = performance.now() - startTime;\n\n console.log(`โœ… Found ${hits.length} results in ${elapsed.toFixed(2)}ms (limit: ${limit})`);\n\n return {\n elapsed: {\n formatted: `${elapsed.toFixed(2)}ms`,\n raw: Math.floor(elapsed * 1000000) // nanoseconds\n },\n hits,\n count: hits.length\n } as any;\n}\n\n/**\n * Load synonyms from Supabase\n */\nasync function loadSynonymsFromSupabase(\n supabaseConfig: { url: string; serviceKey: string }\n): Promise<SynonymMap> {\n try {\n console.log('๐Ÿ” DEBUG: Calling Supabase RPC get_synonym_map...');\n \n // Dynamic import to avoid bundling Supabase client if not needed\n const { createClient } = await import('@supabase/supabase-js');\n \n const supabase = createClient(supabaseConfig.url, supabaseConfig.serviceKey);\n \n // Call the get_synonym_map function\n const { data, error } = await supabase.rpc('get_synonym_map');\n \n console.log('๐Ÿ” DEBUG: Supabase RPC response:', {\n hasError: !!error,\n errorMessage: error?.message,\n hasData: !!data,\n dataType: typeof data,\n dataKeys: data ? Object.keys(data).length : 0\n });\n \n if (error) {\n throw new Error(`Supabase error: ${error.message}`);\n }\n \n const synonymMap = data || {};\n console.log(`๐Ÿ“š Loaded ${Object.keys(synonymMap).length} synonym entries from Supabase`);\n \n return synonymMap;\n } catch (error) {\n console.error('โŒ Failed to load synonyms from Supabase:', error);\n throw error;\n }\n}\n\n/**\n * Calculate document frequencies for TF-IDF\n */\nfunction calculateDocumentFrequencies(\n docs: Record<string, any>,\n textProperty: string\n): Map<string, number> {\n const df = new Map<string, number>();\n\n for (const doc of Object.values(docs)) {\n const text = doc[textProperty];\n \n if (!text || typeof text !== 'string') {\n continue;\n }\n\n // Get unique words in this document\n const words = new Set(tokenize(text));\n\n // Increment document frequency for each unique word\n for (const word of words) {\n df.set(word, (df.get(word) || 0) + 1);\n }\n }\n\n return df;\n}\n\n/**\n * Normalize text using the same rules as server-side\n * \n * CRITICAL: This must match the normalizeText() function in server/index.js exactly\n * PLUS we remove all punctuation to match Orama's French tokenizer behavior\n */\nfunction normalizeText(text: string): string {\n return text\n .toLowerCase()\n .normalize('NFD')\n .replace(/[\\u0300-\\u036f]/g, '') // Remove diacritics\n // Replace French elisions (l', d', etc.) with space to preserve word boundaries\n .replace(/\\b[ldcjmnst][\\u2018\\u2019\\u201A\\u201B\\u2032\\u2035\\u0027\\u0060\\u00B4](?=\\w)/gi, ' ')\n .replace(/[\\u2018\\u2019\\u201A\\u201B\\u2032\\u2035\\u0027\\u0060\\u00B4]/g, '') // Remove remaining apostrophes\n .replace(/[\\u201c\\u201d]/g, '\"') // Normalize curly quotes to straight quotes\n .replace(/[.,;:!?()[\\]{}\\-โ€”โ€“ยซยป\"\"]/g, ' ') // Remove punctuation (replace with space to preserve word boundaries)\n .replace(/\\s+/g, ' ') // Normalize multiple spaces to single space\n .trim();\n}\n\n/**\n * Tokenization matching normalized text behavior\n * \n * Note: Text should already be normalized before indexing, so we normalize again\n * to ensure plugin tokenization matches index tokenization\n */\nfunction tokenize(text: string): string[] {\n // Normalize first (same as indexing), then split by whitespace\n return normalizeText(text)\n .split(/\\s+/)\n .filter(token => token.length > 0);\n}\n\n/**\n * Export types for external use\n */\nexport type {\n FuzzyPhraseConfig,\n WordMatch,\n PhraseMatch,\n DocumentMatch,\n SynonymMap,\n Candidate\n} from './types.js';\n"]}
1
+ {"version":3,"sources":["../src/fuzzy.ts","../src/candidates.ts","../src/scoring.ts","../src/index.ts"],"names":[],"mappings":";AA4BO,SAAS,mBACd,GACA,GACA,OAC0B;AAE1B,MAAI,MAAM,GAAG;AACX,WAAO,EAAE,WAAW,MAAM,UAAU,EAAE;AAAA,EACxC;AAEA,QAAM,OAAO,EAAE;AACf,QAAM,OAAO,EAAE;AAGf,MAAI,KAAK,IAAI,OAAO,IAAI,IAAI,OAAO;AACjC,WAAO,EAAE,WAAW,OAAO,UAAU,QAAQ,EAAE;AAAA,EACjD;AAGA,MAAI,OAAO,MAAM;AACf,KAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC;AAAA,EAChB;AAEA,QAAM,IAAI,EAAE;AACZ,QAAM,IAAI,EAAE;AAGZ,MAAI,UAAU,IAAI,MAAM,IAAI,CAAC;AAC7B,MAAI,UAAU,IAAI,MAAM,IAAI,CAAC;AAG7B,WAAS,IAAI,GAAG,KAAK,GAAG,KAAK;AAC3B,YAAQ,CAAC,IAAI;AAAA,EACf;AAEA,WAAS,IAAI,GAAG,KAAK,GAAG,KAAK;AAC3B,YAAQ,CAAC,IAAI;AACb,QAAI,WAAW;AAEf,aAAS,IAAI,GAAG,KAAK,GAAG,KAAK;AAC3B,YAAM,OAAO,EAAE,IAAI,CAAC,MAAM,EAAE,IAAI,CAAC,IAAI,IAAI;AAEzC,cAAQ,CAAC,IAAI,KAAK;AAAA,QAChB,QAAQ,CAAC,IAAI;AAAA;AAAA,QACb,QAAQ,IAAI,CAAC,IAAI;AAAA;AAAA,QACjB,QAAQ,IAAI,CAAC,IAAI;AAAA;AAAA,MACnB;AAEA,iBAAW,KAAK,IAAI,UAAU,QAAQ,CAAC,CAAC;AAAA,IAC1C;AAGA,QAAI,WAAW,OAAO;AACpB,aAAO,EAAE,WAAW,OAAO,UAAU,QAAQ,EAAE;AAAA,IACjD;AAGA,KAAC,SAAS,OAAO,IAAI,CAAC,SAAS,OAAO;AAAA,EACxC;AAEA,QAAM,WAAW,QAAQ,CAAC;AAC1B,SAAO;AAAA,IACL,WAAW,YAAY;AAAA,IACvB;AAAA,EACF;AACF;AAUO,SAAS,WACd,MACA,YACA,WACuD;AAEvD,MAAI,SAAS,YAAY;AACvB,WAAO,EAAE,SAAS,MAAM,UAAU,GAAG,OAAO,EAAI;AAAA,EAClD;AAGA,MAAI,KAAK,WAAW,UAAU,GAAG;AAC/B,WAAO,EAAE,SAAS,MAAM,UAAU,GAAG,OAAO,KAAK;AAAA,EACnD;AAGA,QAAM,SAAS,mBAAmB,MAAM,YAAY,SAAS;AAE7D,MAAI,OAAO,WAAW;AAGpB,UAAM,QAAQ,IAAO,OAAO,WAAW;AACvC,WAAO;AAAA,MACL,SAAS;AAAA,MACT,UAAU,OAAO;AAAA,MACjB,OAAO,KAAK,IAAI,KAAK,KAAK;AAAA;AAAA,IAC5B;AAAA,EACF;AAEA,SAAO,EAAE,SAAS,OAAO,UAAU,YAAY,GAAG,OAAO,EAAE;AAC7D;AAWO,SAAS,2BACd,aACA,eACQ;AACR,QAAM,cAAc,YAAY;AAEhC,MAAI,eAAe,GAAG;AACpB,WAAO;AAAA,EACT,WAAW,eAAe,GAAG;AAC3B,WAAO,gBAAgB;AAAA,EACzB,WAAW,eAAe,GAAG;AAC3B,WAAO,gBAAgB;AAAA,EACzB,OAAO;AACL,WAAO,gBAAgB;AAAA,EACzB;AACF;;;ACjJO,SAAS,+BAA+B,WAA6B;AAC1E,QAAM,aAAa,oBAAI,IAAY;AACnC,MAAI,eAAe;AACnB,MAAI,aAAa;AAEjB,WAAS,SAAS,MAAW,QAAgB,GAAG;AAC9C,QAAI,CAAC,MAAM;AACT;AAAA,IACF;AAEA;AAIA,QAAI,KAAK,KAAK,KAAK,KAAK,OAAO,KAAK,MAAM,YAAY,KAAK,EAAE,SAAS,GAAG;AACvE,iBAAW,IAAI,KAAK,CAAC;AACrB;AAAA,IACF;AAGA,QAAI,KAAK,GAAG;AACV,UAAI,KAAK,aAAa,KAAK;AAEzB,mBAAW,CAAC,MAAM,SAAS,KAAK,KAAK,GAAG;AACtC,mBAAS,WAAW,QAAQ,CAAC;AAAA,QAC/B;AAAA,MACF,WAAW,MAAM,QAAQ,KAAK,CAAC,GAAG;AAEhC,mBAAW,CAAC,MAAM,SAAS,KAAK,KAAK,GAAG;AACtC,mBAAS,WAAW,QAAQ,CAAC;AAAA,QAC/B;AAAA,MACF,WAAW,OAAO,KAAK,MAAM,UAAU;AAErC,mBAAW,aAAa,OAAO,OAAO,KAAK,CAAC,GAAG;AAC7C,mBAAS,WAAW,QAAQ,CAAC;AAAA,QAC/B;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAEA,WAAS,SAAS;AAClB,UAAQ,IAAI,uBAAgB,WAAW,IAAI,eAAe,YAAY,gBAAgB;AACtF,SAAO;AACT;AAYO,SAAS,uBACd,YACA,YACA,WACA,UACA,eAAuB,KACV;AACb,QAAM,aAA0B,CAAC;AACjC,QAAM,OAAO,oBAAI,IAAY;AAG7B,MAAI,WAAW,IAAI,UAAU,GAAG;AAC9B,eAAW,KAAK;AAAA,MACd,MAAM;AAAA,MACN,MAAM;AAAA,MACN;AAAA,MACA,UAAU;AAAA,MACV,OAAO;AAAA,IACT,CAAC;AACD,SAAK,IAAI,UAAU;AAAA,EACrB;AAGA,aAAW,QAAQ,YAAY;AAC7B,QAAI,KAAK,IAAI,IAAI;AAAG;AAEpB,UAAM,QAAQ,WAAW,MAAM,YAAY,SAAS;AACpD,QAAI,MAAM,SAAS;AACjB,iBAAW,KAAK;AAAA,QACd;AAAA,QACA,MAAM;AAAA,QACN;AAAA,QACA,UAAU,MAAM;AAAA,QAChB,OAAO,MAAM;AAAA,MACf,CAAC;AACD,WAAK,IAAI,IAAI;AAAA,IACf;AAAA,EACF;AAGA,MAAI,YAAY,SAAS,UAAU,GAAG;AACpC,eAAW,WAAW,SAAS,UAAU,GAAG;AAC1C,UAAI,KAAK,IAAI,OAAO;AAAG;AACvB,UAAI,WAAW,IAAI,OAAO,GAAG;AAC3B,mBAAW,KAAK;AAAA,UACd,MAAM;AAAA,UACN,MAAM;AAAA,UACN;AAAA,UACA,UAAU;AAAA,UACV,OAAO;AAAA,QACT,CAAC;AACD,aAAK,IAAI,OAAO;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAEA,SAAO;AACT;AAYO,SAAS,kBACd,aACA,YACA,WACA,UACA,eAAuB,KACG;AAC1B,QAAM,gBAAgB,oBAAI,IAAyB;AAEnD,aAAW,SAAS,aAAa;AAC/B,UAAM,kBAAkB;AAAA,MACtB;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AACA,kBAAc,IAAI,OAAO,eAAe;AAAA,EAC1C;AAEA,SAAO;AACT;AAyBO,SAAS,wBACd,eACA,UAC0B;AAC1B,QAAM,WAAW,oBAAI,IAAyB;AAE9C,aAAW,CAAC,OAAO,UAAU,KAAK,cAAc,QAAQ,GAAG;AACzD,UAAM,qBAAqB,WAAW,OAAO,OAAK,EAAE,SAAS,QAAQ;AACrE,QAAI,mBAAmB,SAAS,GAAG;AACjC,eAAS,IAAI,OAAO,kBAAkB;AAAA,IACxC;AAAA,EACF;AAEA,SAAO;AACT;;;AC9JO,SAAS,sBACd,gBACA,eACA,QACA,mBACA,gBACe;AACf,QAAM,UAAyB,CAAC;AAChC,QAAM,cAAc,MAAM,KAAK,cAAc,KAAK,CAAC;AAGnD,QAAM,cAA2B,CAAC;AAElC,WAAS,IAAI,GAAG,IAAI,eAAe,QAAQ,KAAK;AAC9C,UAAM,UAAU,eAAe,CAAC;AAGhC,eAAW,CAAC,YAAY,UAAU,KAAK,cAAc,QAAQ,GAAG;AAC9D,iBAAW,aAAa,YAAY;AAClC,YAAI,UAAU,SAAS,SAAS;AAC9B,sBAAY,KAAK;AAAA,YACf,MAAM;AAAA,YACN;AAAA,YACA,UAAU;AAAA,YACV,MAAM,UAAU;AAAA,YAChB,UAAU,UAAU;AAAA,YACpB,OAAO,UAAU;AAAA,UACnB,CAAC;AAAA,QACH;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAGA,WAAS,IAAI,GAAG,IAAI,YAAY,QAAQ,KAAK;AAC3C,UAAM,SAAS;AAAA,MACb;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA;AAAA,IACF;AAEA,QAAI,UAAU,OAAO,MAAM,SAAS,GAAG;AACrC,cAAQ,KAAK,MAAM;AAAA,IACrB;AAAA,EACF;AAGA,SAAO,mBAAmB,OAAO;AACnC;AAeA,SAAS,wBACP,aACA,YACA,aACA,QACA,mBACA,gBACA,gBACA,gBACoB;AACpB,QAAM,aAAa,YAAY,UAAU;AACzC,QAAM,cAA2B,CAAC,UAAU;AAC5C,QAAM,gBAAgB,oBAAI,IAAI,CAAC,WAAW,UAAU,CAAC;AACrD,QAAM,WAAsB,CAAC;AAC7B,MAAI,eAAe;AAGnB,WAAS,IAAI,aAAa,GAAG,IAAI,YAAY,QAAQ,KAAK;AACxD,UAAM,QAAQ,YAAY,CAAC;AAC3B,UAAM,UAAU,YAAY,YAAY,SAAS,CAAC,EAAE;AACpD,UAAM,MAAM,MAAM,WAAW,UAAU;AAGvC,QAAI,MAAM,OAAO,QAAQ;AACvB;AAAA,IACF;AAGA,QAAI,CAAC,cAAc,IAAI,MAAM,UAAU,GAAG;AAExC,eAAS,MAAM,UAAU,GAAG,MAAM,MAAM,UAAU,OAAO;AACvD;AACA,iBAAS,KAAK;AAAA,UACZ,MAAM,eAAe,GAAG;AAAA,UACxB,UAAU;AAAA,UACV,UAAU;AAAA,QACZ,CAAC;AAAA,MACH;AAEA,kBAAY,KAAK,KAAK;AACtB,oBAAc,IAAI,MAAM,UAAU;AAGlC,UAAI,cAAc,SAAS,YAAY,QAAQ;AAC7C;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAGA,MAAI,YAAY,SAAS,GAAG;AAC1B,UAAM,WAAW,YAAY,SAAS,YAAY;AAClD,UAAM,OAAO,YAAY,YAAY,SAAS,CAAC,EAAE,WAAW,YAAY,CAAC,EAAE,WAAW;AAEtF,UAAM,EAAE,OAAO,UAAU,IAAI;AAAA,MAC3B;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAEA,WAAO;AAAA,MACL,OAAO;AAAA,MACP;AAAA,MACA,SAAS;AAAA,MACT;AAAA,MACA,eAAe,YAAY,CAAC,EAAE;AAAA,MAC9B,aAAa,YAAY,YAAY,SAAS,CAAC,EAAE;AAAA,MACjD;AAAA,MACA,SAAS,UAAU,aAAa,WAAW;AAAA,MAC3C;AAAA,MACA,gBAAgB;AAAA,IAClB;AAAA,EACF;AAEA,SAAO;AACT;AAcA,SAAS,qBACP,aACA,aACA,QACA,mBACA,gBACA,gBACA,UACuI;AAGvI,MAAI,YAAY;AAChB,aAAW,QAAQ,aAAa;AAC9B,UAAM,SAAS,KAAK,SAAS,UAAU,OAAO,QAAQ,QACvC,KAAK,SAAS,UAAU,OAAO,QAAQ,QACvC,OAAO,QAAQ,QAAQ;AACtC,iBAAa,KAAK,QAAQ;AAAA,EAC5B;AACA,eAAa,YAAY;AAGzB,QAAM,UAAU,UAAU,aAAa,WAAW;AAClD,QAAM,aAAa,UAAU,IAAM;AAInC,MAAI,iBAAiB;AACrB,MAAI,OAAO,SAAS,KAAK,OAAO,QAAQ,YAAY,KAAK,YAAY,SAAS,GAAG;AAC/E,UAAM,OAAO,YAAY,YAAY,SAAS,CAAC,EAAE,WAAW,YAAY,CAAC,EAAE,WAAW;AACtF,UAAM,kBAAkB,YAAY,SAAS,OAAO;AACpD,qBAAiB,KAAK,IAAI,GAAG,IAAO,OAAO,eAAgB;AAAA,EAC7D;AAIA,MAAI,eAAe;AAEnB,MAAI,YAAY,WAAW,GAAG;AAE5B,UAAM,mBAAmB,eAAe;AAExC,mBAAe,KAAK,IAAI,GAAK,mBAAmB,EAAE;AAAA,EACpD;AAKA,QAAM,gBAAgB;AAAA,IACpB;AAAA,IACA;AAAA,IACA;AAAA,EACF;AAGA,QAAM,UAAU,OAAO;AAGvB,QAAM,eAAe;AACrB,QAAM,gBAAgB,aAAa,QAAQ;AAC3C,QAAM,oBAAoB,iBAAiB,QAAQ;AACnD,QAAM,kBAAkB,eAAe,QAAQ;AAC/C,QAAM,mBAAmB,gBAAgB,QAAQ;AAEjD,QAAM,aAAa,eAAe,gBAAgB,oBAAoB,kBAAkB;AAKxF,QAAM,sBAAsB,OAAO,YAAY,KAAK,QAAQ,QAAQ;AACpE,QAAM,gBAAgB,sBAAsB,KAAK,IAAI,QAAQ,OAAO,QAAQ,KAAK,IAAI,QAAQ;AAE7F,QAAM,2BAA4B,OAAO,SAAS,KAAK,QAAQ,YAAY,KAAK,YAAY,SAAS,IAAK,QAAQ,YAAY;AAC9H,QAAM,mBAAmB,gBAAgB,QAAQ,QAAQ,2BAA2B,QAAQ,UAAU,QAAQ;AAG9G,QAAM,kBAAkB,aAAa;AAIrC,QAAM,qBAAqB,YAAY,SAAS,IAAI,WAAW;AAC/D,QAAM,QAAQ,kBAAkB;AAGhC,QAAM,OAAO,eAAe;AAC5B,QAAM,QAAQ,gBAAgB;AAC9B,QAAM,YAAY,oBAAoB;AACtC,QAAM,UAAU,kBAAkB;AAClC,QAAM,WAAW,mBAAmB;AAEpC,SAAO;AAAA,IACL;AAAA,IACA,WAAW;AAAA,MACT;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,UAAU;AAAA;AAAA,IACZ;AAAA,EACF;AACF;AASA,SAAS,UAAU,aAA0B,aAAgC;AAC3E,QAAM,aAAa,IAAI,IAAI,YAAY,IAAI,CAAC,OAAO,UAAU,CAAC,OAAO,KAAK,CAAC,CAAC;AAE5E,WAAS,IAAI,GAAG,IAAI,YAAY,QAAQ,KAAK;AAC3C,UAAM,YAAY,WAAW,IAAI,YAAY,IAAI,CAAC,EAAE,UAAU,KAAK;AACnE,UAAM,YAAY,WAAW,IAAI,YAAY,CAAC,EAAE,UAAU,KAAK;AAE/D,QAAI,YAAY,WAAW;AACzB,aAAO;AAAA,IACT;AAAA,EACF;AAEA,SAAO;AACT;AAUA,SAAS,uBACP,aACA,mBACA,gBACQ;AAER,MAAI,mBAAmB,GAAG;AACxB,WAAO;AAAA,EACT;AAEA,MAAI,WAAW;AAEf,aAAW,QAAQ,aAAa;AAC9B,UAAM,KAAK,kBAAkB,IAAI,KAAK,IAAI,KAAK;AAC/C,UAAM,MAAM,KAAK,IAAI,iBAAiB,EAAE;AACxC,gBAAY;AAAA,EACd;AAGA,QAAM,WAAW,WAAW,YAAY;AAGxC,SAAO,KAAK,IAAI,GAAK,WAAW,EAAE;AACpC;AAQA,SAAS,mBAAmB,SAAuC;AACjE,MAAI,QAAQ,WAAW;AAAG,WAAO,CAAC;AAGlC,QAAM,SAAS,QAAQ,MAAM,EAAE,KAAK,CAAC,GAAG,MAAM,EAAE,QAAQ,EAAE,KAAK;AAC/D,QAAM,SAAwB,CAAC;AAC/B,QAAM,UAAU,oBAAI,IAAY;AAEhC,aAAW,UAAU,QAAQ;AAE3B,QAAI,WAAW;AACf,aAAS,MAAM,OAAO,eAAe,OAAO,OAAO,aAAa,OAAO;AACrE,UAAI,QAAQ,IAAI,GAAG,GAAG;AACpB,mBAAW;AACX;AAAA,MACF;AAAA,IACF;AAEA,QAAI,CAAC,UAAU;AACb,aAAO,KAAK,MAAM;AAElB,eAAS,MAAM,OAAO,eAAe,OAAO,OAAO,aAAa,OAAO;AACrE,gBAAQ,IAAI,GAAG;AAAA,MACjB;AAAA,IACF;AAAA,EACF;AAEA,SAAO,OAAO,KAAK,CAAC,GAAG,MAAM,EAAE,QAAQ,EAAE,KAAK;AAChD;;;ACpXA,IAAM,iBAA8C;AAAA,EAClD,cAAc;AAAA;AAAA,EACd,WAAW;AAAA,EACX,mBAAmB;AAAA,EACnB,gBAAgB;AAAA,EAChB,UAAU;AAAA,EACV,mBAAmB;AAAA,EACnB,SAAS;AAAA,IACP,OAAO;AAAA,IACP,OAAO;AAAA,IACP,OAAO;AAAA,IACP,WAAW;AAAA,IACX,SAAS;AAAA,IACT,UAAU;AAAA,EACZ;AAAA,EACA,QAAQ;AAAA,EACR,UAAU;AAAA,EACV,yBAAyB;AAAA,EACzB,mBAAmB;AAAA,EACnB,yBAAyB;AAC3B;AAKA,IAAM,eAAe,oBAAI,QAA+B;AAQjD,SAAS,kBAAkB,aAAgC,CAAC,GAAgB;AAEjF,QAAM,SAAsC;AAAA,IAC1C,cAAc,WAAW,gBAAgB,eAAe;AAAA,IACxD,WAAW,WAAW,aAAa,eAAe;AAAA,IAClD,mBAAmB,WAAW,qBAAqB,eAAe;AAAA,IAClE,gBAAgB,WAAW,kBAAkB,eAAe;AAAA,IAC5D,UAAU,WAAW,YAAY,eAAe;AAAA,IAChD,mBAAmB,WAAW,qBAAqB,eAAe;AAAA,IAClE,SAAS;AAAA,MACP,OAAO,WAAW,SAAS,SAAS,eAAe,QAAQ;AAAA,MAC3D,OAAO,WAAW,SAAS,SAAS,eAAe,QAAQ;AAAA,MAC3D,OAAO,WAAW,SAAS,SAAS,eAAe,QAAQ;AAAA,MAC3D,WAAW,WAAW,SAAS,aAAa,eAAe,QAAQ;AAAA,MACnE,SAAS,WAAW,SAAS,WAAW,eAAe,QAAQ;AAAA,MAC/D,UAAU,WAAW,SAAS,YAAY,eAAe,QAAQ;AAAA,IACnE;AAAA,IACA,QAAQ,WAAW,UAAU,eAAe;AAAA,IAC5C,UAAU,WAAW,YAAY,eAAe;AAAA,IAChD,yBAAyB,WAAW,2BAA2B,eAAe;AAAA,IAC9E,mBAAmB,WAAW,qBAAqB,eAAe;AAAA,IAClE,yBAAyB,WAAW,2BAA2B,eAAe;AAAA,EAChF;AAEA,QAAM,SAAsB;AAAA,IAC1B,MAAM;AAAA;AAAA;AAAA;AAAA,IAKN,aAAa,OAAO,UAAoB;AACtC,cAAQ,IAAI,+CAAwC;AAGpD,YAAM,QAAqB;AAAA,QACzB,YAAY,CAAC;AAAA,QACb;AAAA,QACA,mBAAmB,oBAAI,IAAI;AAAA,QAC3B,gBAAgB;AAAA,MAClB;AAGA,UAAI,OAAO,kBAAkB,OAAO,UAAU;AAC5C,YAAI;AACF,kBAAQ,IAAI,6CAAsC;AAClD,gBAAM,aAAa,MAAM,yBAAyB,OAAO,QAAQ;AACjE,kBAAQ,IAAI,iBAAY,OAAO,KAAK,MAAM,UAAU,EAAE,MAAM,sBAAsB;AAAA,QACpF,SAAS,OAAO;AACd,kBAAQ,MAAM,0CAAgC,KAAK;AAAA,QAErD;AAAA,MACF;AAGA,YAAM,OAAQ,MAAM,MAAc,MAAM;AACxC,UAAI,MAAM;AACR,cAAM,iBAAiB,OAAO,KAAK,IAAI,EAAE;AACzC,cAAM,oBAAoB,6BAA6B,MAAM,OAAO,YAAY;AAChF,gBAAQ,IAAI,iDAA0C,MAAM,cAAc,YAAY;AAAA,MACxF;AAGA,mBAAa,IAAI,OAAO,KAAK;AAC7B,cAAQ,IAAI,wCAAmC;AAI/C,mBAAa,MAAM;AACjB,YAAI,OAAQ,WAAmB,2BAA2B,YAAY;AACpE,kBAAQ,IAAI,qCAA8B;AAC1C,UAAC,WAAmB,uBAAuB;AAAA,QAC7C,OAAO;AACL,kBAAQ,KAAK,yDAA+C;AAAA,QAC9D;AAAA,MACF,CAAC;AAAA,IACH;AAAA,EACF;AAEA,SAAO;AACT;AAQA,eAAsB,sBACpB,OACA,QACA,UACoC;AACpC,QAAM,YAAY,YAAY,IAAI;AAGlC,QAAM,QAAQ,aAAa,IAAI,KAAK;AAEpC,MAAI,CAAC,OAAO;AACV,YAAQ,MAAM,qCAAgC;AAC9C,UAAM,IAAI,MAAM,8CAA8C;AAAA,EAChE;AAEA,QAAM,EAAE,MAAM,WAAW,IAAI;AAE7B,MAAI,CAAC,QAAQ,OAAO,SAAS,UAAU;AACrC,WAAO,EAAE,SAAS,EAAE,WAAW,OAAO,KAAK,EAAE,GAAG,MAAM,CAAC,GAAG,OAAO,EAAE;AAAA,EACrE;AAGA,QAAM,eAAgB,cAAc,WAAW,CAAC,KAAM,MAAM,OAAO;AAGnE,QAAM,cAAc,SAAS,IAAI;AAEjC,MAAI,YAAY,WAAW,GAAG;AAC5B,WAAO,EAAE,SAAS,EAAE,WAAW,OAAO,KAAK,EAAE,GAAG,MAAM,CAAC,GAAG,OAAO,EAAE;AAAA,EACrE;AAGA,QAAM,YAAY,MAAM,OAAO,oBAC3B,2BAA2B,aAAa,MAAM,OAAO,SAAS,IAC9D,MAAM,OAAO;AAEjB,UAAQ,IAAI,mCAA4B,IAAI,MAAM,YAAY,MAAM,uBAAuB,SAAS,GAAG;AAGvG,MAAI;AAEJ,MAAI;AAGF,UAAM,YAAa,MAAc,MAAM;AAEvC,QAAI,CAAC,WAAW;AACd,cAAQ,MAAM,gDAA2C;AACzD,aAAO,EAAE,SAAS,EAAE,WAAW,OAAO,KAAK,EAAE,GAAG,MAAM,CAAC,GAAG,OAAO,EAAE;AAAA,IACrE;AAEA,YAAQ,IAAI,qCAA8B,OAAO,KAAK,aAAa,CAAC,CAAC,CAAC;AAGtE,QAAI,YAAY;AAGhB,QAAI,UAAU,UAAU,YAAY,GAAG,MAAM;AAC3C,kBAAY,UAAU,QAAQ,YAAY,EAAE;AAC5C,cAAQ,IAAI,4DAAuD;AAAA,IACrE,WAES,UAAU,YAAY,GAAG,MAAM;AACtC,kBAAY,UAAU,YAAY,EAAE;AACpC,cAAQ,IAAI,6DAAwD;AAAA,IACtE;AAEA,QAAI,CAAC,WAAW;AACd,cAAQ,MAAM,6CAAwC,YAAY;AAClE,cAAQ,MAAM,qCAAqC,OAAO,KAAK,SAAS,CAAC;AACzE,aAAO,EAAE,SAAS,EAAE,WAAW,OAAO,KAAK,EAAE,GAAG,MAAM,CAAC,GAAG,OAAO,EAAE;AAAA,IACrE;AAEA,iBAAa,+BAA+B,SAAS;AACrD,YAAQ,IAAI,uBAAgB,WAAW,IAAI,0BAA0B;AAAA,EACvE,SAAS,OAAO;AACd,YAAQ,MAAM,wCAAmC,KAAK;AACtD,WAAO,EAAE,SAAS,EAAE,WAAW,OAAO,KAAK,EAAE,GAAG,MAAM,CAAC,GAAG,OAAO,EAAE;AAAA,EACrE;AAGA,QAAM,gBAAgB;AAAA,IACpB;AAAA,IACA;AAAA,IACA;AAAA,IACA,MAAM,OAAO,iBAAiB,MAAM,aAAa;AAAA,IACjD,MAAM,OAAO;AAAA,EACf;AAGA,QAAM,qBAAqB,cAAc,IACrC,gBACA,wBAAwB,eAAe,MAAM,OAAO,QAAQ;AAEhE,UAAQ,IAAI,+BAAwB,MAAM,KAAK,mBAAmB,OAAO,CAAC,EAAE,OAAO,CAAC,KAAK,MAAM,MAAM,EAAE,QAAQ,CAAC,CAAC,QAAQ;AAGzH,QAAM,kBAAmC,CAAC;AAE1C,UAAQ,IAAI,yCAAkC;AAAA,IAC5C,UAAU,OAAO,KAAM,MAAc,QAAQ,CAAC,CAAC;AAAA,IAC/C,SAAS,CAAC,CAAG,MAAc,MAAM;AAAA,IACjC,UAAW,MAAc,MAAM,OAAO,OAAQ,MAAc,KAAK,OAAO;AAAA,EAC1E,CAAC;AAGD,MAAI,OAA4B,CAAC;AAGjC,MAAK,MAAc,MAAM,MAAM,MAAM;AACnC,WAAQ,MAAc,KAAK,KAAK;AAChC,YAAQ,IAAI,2CAAsC;AAAA,EACpD,WAEU,MAAc,MAAM,QAAQ,OAAQ,MAAc,KAAK,SAAS,UAAU;AAElF,UAAM,WAAW,OAAO,KAAM,MAAc,KAAK,IAAI,EAAE,CAAC;AACxD,QAAI,YAAY,aAAa,iCAAiC,aAAa,SAAS;AAClF,aAAQ,MAAc,KAAK;AAC3B,cAAQ,IAAI,+CAA0C;AAAA,IACxD;AAAA,EACF;AAEA,MAAI,OAAO,KAAK,IAAI,EAAE,WAAW,GAAG;AAClC,YAAQ,IAAI,0DAAqD;AAAA,MAC/D,aAAa,CAAC,CAAG,MAAc,MAAM;AAAA,MACrC,cAAe,MAAc,MAAM,OAAO,OAAO,KAAM,MAAc,KAAK,IAAI,IAAI;AAAA,MAClF,iBAAiB,CAAC,CAAG,MAAc,MAAM,MAAM;AAAA,MAC/C,mBAAoB,MAAc,MAAM,MAAM,OAAO,OAAO,KAAM,MAAc,KAAK,KAAK,IAAI,EAAE,SAAS;AAAA,IAC3G,CAAC;AAAA,EACH;AAEA,UAAQ,IAAI,+BAAwB,OAAO,KAAK,IAAI,EAAE,MAAM,YAAY;AAExE,aAAW,CAAC,OAAO,GAAG,KAAK,OAAO,QAAQ,IAAI,GAAG;AAC/C,UAAM,OAAO,IAAI,YAAY;AAE7B,QAAI,CAAC,QAAQ,OAAO,SAAS,UAAU;AACrC;AAAA,IACF;AAGA,UAAM,YAAY,SAAS,IAAI;AAI/B,UAAM,UAAU;AAAA,MACd;AAAA,MACA;AAAA,MACA;AAAA,QACE,SAAS,MAAM,OAAO;AAAA,QACtB,QAAQ,MAAM,OAAO;AAAA,QACrB,yBAAyB,MAAM,OAAO;AAAA,QACtC;AAAA,MACF;AAAA,MACA,MAAM;AAAA,MACN,MAAM;AAAA,IACR;AAEA,QAAI,QAAQ,SAAS,GAAG;AAEtB,YAAM,WAAW,KAAK,IAAI,GAAG,QAAQ,IAAI,OAAK,EAAE,KAAK,CAAC;AAEtD,sBAAgB,KAAK;AAAA,QACnB,IAAI;AAAA,QACJ;AAAA,QACA,OAAO;AAAA,QACP,UAAU;AAAA,MACZ,CAAC;AAAA,IACH;AAAA,EACF;AAGA,kBAAgB,KAAK,CAAC,GAAG,MAAM,EAAE,QAAQ,EAAE,KAAK;AAGhD,MAAI,kBAAkB;AACtB,MAAI,MAAM,OAAO,2BAA2B,MAAM,OAAO,oBAAoB,GAAG;AAC9E,UAAM,YAAY,MAAM,OAAO;AAC/B,UAAM,cAAc,gBAAgB;AACpC,sBAAkB,gBAAgB,OAAO,OAAK,EAAE,SAAS,SAAS;AAClE,YAAQ,IAAI,uCAA2B,WAAW,WAAM,gBAAgB,MAAM,gBAAgB,SAAS,GAAG;AAAA,EAC5G;AAGA,QAAM,QAAQ,OAAO,SAAS,gBAAgB;AAC9C,QAAM,iBAAiB,gBAAgB,MAAM,GAAG,KAAK;AAGrD,QAAM,OAAO,eAAe,IAAI,YAAU;AAAA,IACxC,IAAI,MAAM;AAAA,IACV,OAAO,MAAM;AAAA,IACb,UAAU,MAAM;AAAA;AAAA,IAEhB,UAAU,MAAM;AAAA,EAClB,EAAE;AAEF,QAAM,UAAU,YAAY,IAAI,IAAI;AAEpC,UAAQ,IAAI,gBAAW,KAAK,MAAM,eAAe,QAAQ,QAAQ,CAAC,CAAC,cAAc,KAAK,GAAG;AAEzF,SAAO;AAAA,IACL,SAAS;AAAA,MACP,WAAW,GAAG,QAAQ,QAAQ,CAAC,CAAC;AAAA,MAChC,KAAK,KAAK,MAAM,UAAU,GAAO;AAAA;AAAA,IACnC;AAAA,IACA;AAAA,IACA,OAAO,KAAK;AAAA,EACd;AACF;AAKA,eAAe,yBACb,gBACqB;AACrB,MAAI;AACF,YAAQ,IAAI,0DAAmD;AAG/D,UAAM,EAAE,aAAa,IAAI,MAAM,OAAO,uBAAuB;AAE7D,UAAM,WAAW,aAAa,eAAe,KAAK,eAAe,UAAU;AAG3E,UAAM,EAAE,MAAM,MAAM,IAAI,MAAM,SAAS,IAAI,iBAAiB;AAE5D,YAAQ,IAAI,2CAAoC;AAAA,MAC9C,UAAU,CAAC,CAAC;AAAA,MACZ,cAAc,OAAO;AAAA,MACrB,SAAS,CAAC,CAAC;AAAA,MACX,UAAU,OAAO;AAAA,MACjB,UAAU,OAAO,OAAO,KAAK,IAAI,EAAE,SAAS;AAAA,IAC9C,CAAC;AAED,QAAI,OAAO;AACT,YAAM,IAAI,MAAM,mBAAmB,MAAM,OAAO,EAAE;AAAA,IACpD;AAEA,UAAM,aAAa,QAAQ,CAAC;AAC5B,YAAQ,IAAI,oBAAa,OAAO,KAAK,UAAU,EAAE,MAAM,gCAAgC;AAEvF,WAAO;AAAA,EACT,SAAS,OAAO;AACd,YAAQ,MAAM,iDAA4C,KAAK;AAC/D,UAAM;AAAA,EACR;AACF;AAKA,SAAS,6BACP,MACA,cACqB;AACrB,QAAM,KAAK,oBAAI,IAAoB;AAEnC,aAAW,OAAO,OAAO,OAAO,IAAI,GAAG;AACrC,UAAM,OAAO,IAAI,YAAY;AAE7B,QAAI,CAAC,QAAQ,OAAO,SAAS,UAAU;AACrC;AAAA,IACF;AAGA,UAAM,QAAQ,IAAI,IAAI,SAAS,IAAI,CAAC;AAGpC,eAAW,QAAQ,OAAO;AACxB,SAAG,IAAI,OAAO,GAAG,IAAI,IAAI,KAAK,KAAK,CAAC;AAAA,IACtC;AAAA,EACF;AAEA,SAAO;AACT;AAQA,SAAS,cAAc,MAAsB;AAC3C,SAAO,KACJ,YAAY,EACZ,UAAU,KAAK,EACf,QAAQ,oBAAoB,EAAE,EAE9B,QAAQ,gFAAgF,GAAG,EAC3F,QAAQ,6DAA6D,EAAE,EACvE,QAAQ,mBAAmB,GAAG,EAC9B,QAAQ,4BAA4B,GAAG,EACvC,QAAQ,QAAQ,GAAG,EACnB,KAAK;AACV;AAQA,SAAS,SAAS,MAAwB;AAExC,SAAO,cAAc,IAAI,EACtB,MAAM,KAAK,EACX,OAAO,WAAS,MAAM,SAAS,CAAC;AACrC","sourcesContent":["/**\n * Fuzzy matching utilities using bounded Levenshtein distance\n * \n * This is the same algorithm used by Orama's match-highlight plugin\n * for consistent fuzzy matching behavior.\n */\n\n/**\n * Result of bounded Levenshtein distance calculation\n */\nexport interface BoundedLevenshteinResult {\n /** Whether the distance is within bounds */\n isBounded: boolean;\n /** The actual distance (only valid if isBounded is true) */\n distance: number;\n}\n\n/**\n * Calculate bounded Levenshtein distance between two strings\n * \n * Stops early if distance exceeds the bound for better performance.\n * This is the same algorithm as Orama's internal boundedLevenshtein.\n * \n * @param a - First string\n * @param b - Second string\n * @param bound - Maximum allowed distance\n * @returns Result indicating if strings are within bound and the distance\n */\nexport function boundedLevenshtein(\n a: string,\n b: string,\n bound: number\n): BoundedLevenshteinResult {\n // Quick checks\n if (a === b) {\n return { isBounded: true, distance: 0 };\n }\n\n const aLen = a.length;\n const bLen = b.length;\n\n // If length difference exceeds bound, no need to calculate\n if (Math.abs(aLen - bLen) > bound) {\n return { isBounded: false, distance: bound + 1 };\n }\n\n // Swap to ensure a is shorter (optimization)\n if (aLen > bLen) {\n [a, b] = [b, a];\n }\n\n const m = a.length;\n const n = b.length;\n\n // Use single array instead of matrix (memory optimization)\n let prevRow = new Array(n + 1);\n let currRow = new Array(n + 1);\n\n // Initialize first row\n for (let j = 0; j <= n; j++) {\n prevRow[j] = j;\n }\n\n for (let i = 1; i <= m; i++) {\n currRow[0] = i;\n let minInRow = i;\n\n for (let j = 1; j <= n; j++) {\n const cost = a[i - 1] === b[j - 1] ? 0 : 1;\n\n currRow[j] = Math.min(\n prevRow[j] + 1, // deletion\n currRow[j - 1] + 1, // insertion\n prevRow[j - 1] + cost // substitution\n );\n\n minInRow = Math.min(minInRow, currRow[j]);\n }\n\n // Early termination: if all values in row exceed bound, we're done\n if (minInRow > bound) {\n return { isBounded: false, distance: bound + 1 };\n }\n\n // Swap rows for next iteration\n [prevRow, currRow] = [currRow, prevRow];\n }\n\n const distance = prevRow[n];\n return {\n isBounded: distance <= bound,\n distance\n };\n}\n\n/**\n * Check if a word matches a query token with fuzzy matching\n * \n * @param word - Word from document\n * @param queryToken - Token from search query\n * @param tolerance - Maximum edit distance allowed\n * @returns Match result with score\n */\nexport function fuzzyMatch(\n word: string,\n queryToken: string,\n tolerance: number\n): { matches: boolean; distance: number; score: number } {\n // Exact match\n if (word === queryToken) {\n return { matches: true, distance: 0, score: 1.0 };\n }\n\n // Prefix match (high score, no distance)\n if (word.startsWith(queryToken)) {\n return { matches: true, distance: 0, score: 0.95 };\n }\n\n // Fuzzy match with tolerance\n const result = boundedLevenshtein(word, queryToken, tolerance);\n \n if (result.isBounded) {\n // Score decreases with distance\n // distance 1 = 0.8, distance 2 = 0.6, etc.\n const score = 1.0 - (result.distance * 0.2);\n return {\n matches: true,\n distance: result.distance,\n score: Math.max(0.1, score) // Minimum score of 0.1\n };\n }\n\n return { matches: false, distance: tolerance + 1, score: 0 };\n}\n\n/**\n * Calculate adaptive tolerance based on query length\n * \n * Longer queries get higher tolerance for better fuzzy matching.\n * \n * @param queryTokens - Array of query tokens\n * @param baseTolerance - Base tolerance value\n * @returns Calculated tolerance (always an integer)\n */\nexport function calculateAdaptiveTolerance(\n queryTokens: string[],\n baseTolerance: number\n): number {\n const queryLength = queryTokens.length;\n \n if (queryLength <= 2) {\n return baseTolerance;\n } else if (queryLength <= 4) {\n return baseTolerance + 1;\n } else if (queryLength <= 6) {\n return baseTolerance + 2;\n } else {\n return baseTolerance + 3;\n }\n}\n","/**\n * Candidate expansion: Find all possible matches for query tokens\n * including exact matches, fuzzy matches, and synonyms\n */\n\nimport { fuzzyMatch } from './fuzzy.js';\nimport type { Candidate, SynonymMap } from './types.js';\n\n/**\n * Extract all unique words from the radix tree index\n * \n * @param radixNode - Root node of the radix tree\n * @returns Set of all unique words in the index\n */\nexport function extractVocabularyFromRadixTree(radixNode: any): Set<string> {\n const vocabulary = new Set<string>();\n let nodesVisited = 0;\n let wordsFound = 0;\n \n function traverse(node: any, depth: number = 0) {\n if (!node) {\n return;\n }\n \n nodesVisited++;\n \n // Check if this node represents a complete word\n // e = true means it's an end of a word\n if (node.e && node.w && typeof node.w === 'string' && node.w.length > 0) {\n vocabulary.add(node.w);\n wordsFound++;\n }\n \n // Children can be Map, Array, or Object\n if (node.c) {\n if (node.c instanceof Map) {\n // Map format\n for (const [_key, childNode] of node.c) {\n traverse(childNode, depth + 1);\n }\n } else if (Array.isArray(node.c)) {\n // Array format: [[key, childNode], ...]\n for (const [_key, childNode] of node.c) {\n traverse(childNode, depth + 1);\n }\n } else if (typeof node.c === 'object') {\n // Object format: {key: childNode, ...}\n for (const childNode of Object.values(node.c)) {\n traverse(childNode, depth + 1);\n }\n }\n }\n }\n \n traverse(radixNode);\n console.log(`๐Ÿ“š Extracted ${vocabulary.size} words from ${nodesVisited} nodes visited`);\n return vocabulary;\n}\n\n/**\n * Find all candidate matches for a single query token\n * \n * @param queryToken - Token from search query\n * @param vocabulary - Set of all words in the index\n * @param tolerance - Fuzzy matching tolerance\n * @param synonyms - Synonym map (optional)\n * @param synonymScore - Score multiplier for synonym matches\n * @returns Array of candidate matches\n */\nexport function findCandidatesForToken(\n queryToken: string,\n vocabulary: Set<string>,\n tolerance: number,\n synonyms?: SynonymMap,\n synonymScore: number = 0.8\n): Candidate[] {\n const candidates: Candidate[] = [];\n const seen = new Set<string>();\n\n // 1. Check for exact match\n if (vocabulary.has(queryToken)) {\n candidates.push({\n word: queryToken,\n type: 'exact',\n queryToken,\n distance: 0,\n score: 1.0\n });\n seen.add(queryToken);\n }\n\n // 2. Check for fuzzy matches\n for (const word of vocabulary) {\n if (seen.has(word)) continue;\n\n const match = fuzzyMatch(word, queryToken, tolerance);\n if (match.matches) {\n candidates.push({\n word,\n type: 'fuzzy',\n queryToken,\n distance: match.distance,\n score: match.score\n });\n seen.add(word);\n }\n }\n\n // 3. Check for synonym matches\n if (synonyms && synonyms[queryToken]) {\n for (const synonym of synonyms[queryToken]) {\n if (seen.has(synonym)) continue;\n if (vocabulary.has(synonym)) {\n candidates.push({\n word: synonym,\n type: 'synonym',\n queryToken,\n distance: 0,\n score: synonymScore\n });\n seen.add(synonym);\n }\n }\n }\n\n return candidates;\n}\n\n/**\n * Find candidates for all query tokens\n * \n * @param queryTokens - Array of tokens from search query\n * @param vocabulary - Set of all words in the index\n * @param tolerance - Fuzzy matching tolerance\n * @param synonyms - Synonym map (optional)\n * @param synonymScore - Score multiplier for synonym matches\n * @returns Map of query tokens to their candidate matches\n */\nexport function findAllCandidates(\n queryTokens: string[],\n vocabulary: Set<string>,\n tolerance: number,\n synonyms?: SynonymMap,\n synonymScore: number = 0.8\n): Map<string, Candidate[]> {\n const candidatesMap = new Map<string, Candidate[]>();\n\n for (const token of queryTokens) {\n const tokenCandidates = findCandidatesForToken(\n token,\n vocabulary,\n tolerance,\n synonyms,\n synonymScore\n );\n candidatesMap.set(token, tokenCandidates);\n }\n\n return candidatesMap;\n}\n\n/**\n * Get total number of candidates across all tokens\n * \n * @param candidatesMap - Map of token to candidates\n * @returns Total count of all candidates\n */\nexport function getTotalCandidateCount(\n candidatesMap: Map<string, Candidate[]>\n): number {\n let total = 0;\n for (const candidates of candidatesMap.values()) {\n total += candidates.length;\n }\n return total;\n}\n\n/**\n * Filter candidates by minimum score threshold\n * \n * @param candidatesMap - Map of token to candidates\n * @param minScore - Minimum score threshold\n * @returns Filtered candidates map\n */\nexport function filterCandidatesByScore(\n candidatesMap: Map<string, Candidate[]>,\n minScore: number\n): Map<string, Candidate[]> {\n const filtered = new Map<string, Candidate[]>();\n\n for (const [token, candidates] of candidatesMap.entries()) {\n const filteredCandidates = candidates.filter(c => c.score >= minScore);\n if (filteredCandidates.length > 0) {\n filtered.set(token, filteredCandidates);\n }\n }\n\n return filtered;\n}\n","/**\n * Phrase scoring algorithm with semantic weighting\n */\n\nimport type { WordMatch, PhraseMatch, Candidate, GapWord } from './types.js';\n\n/**\n * Configuration for phrase scoring\n */\nexport interface ScoringConfig {\n weights: {\n exact: number;\n fuzzy: number;\n order: number;\n proximity: number;\n density: number;\n semantic: number;\n };\n maxGap: number;\n /** \n * Multiplier for proximity window calculation.\n * proximityWindow = queryTokens.length ร— proximitySpanMultiplier\n */\n proximitySpanMultiplier: number;\n /**\n * Fuzzy tolerance (Levenshtein distance). When 0, only exact matches.\n */\n tolerance: number;\n}\n\n/**\n * Find all phrase matches in a document\n * \n * @param documentTokens - Tokenized document content (needed to extract gap words)\n * @param candidatesMap - Map of query tokens to their candidates\n * @param config - Scoring configuration\n * @param documentFrequency - Document frequency map for TF-IDF\n * @param totalDocuments - Total number of documents\n * @returns Array of phrase matches\n */\nexport function findPhrasesInDocument(\n documentTokens: string[],\n candidatesMap: Map<string, Candidate[]>,\n config: ScoringConfig,\n documentFrequency: Map<string, number>,\n totalDocuments: number\n): PhraseMatch[] {\n const phrases: PhraseMatch[] = [];\n const queryTokens = Array.from(candidatesMap.keys());\n\n // Find all word matches in document\n const wordMatches: WordMatch[] = [];\n \n for (let i = 0; i < documentTokens.length; i++) {\n const docWord = documentTokens[i];\n \n // Check if this word matches any query token\n for (const [queryToken, candidates] of candidatesMap.entries()) {\n for (const candidate of candidates) {\n if (candidate.word === docWord) {\n wordMatches.push({\n word: docWord,\n queryToken,\n position: i,\n type: candidate.type,\n distance: candidate.distance,\n score: candidate.score\n });\n }\n }\n }\n }\n\n // Build phrases from word matches using sliding window\n for (let i = 0; i < wordMatches.length; i++) {\n const phrase = buildPhraseFromPosition(\n wordMatches,\n i,\n queryTokens,\n config,\n documentFrequency,\n totalDocuments,\n wordMatches,\n documentTokens // Pass document tokens to extract gap words\n );\n \n if (phrase && phrase.words.length > 0) {\n phrases.push(phrase);\n }\n }\n\n // Deduplicate and sort by score\n return deduplicatePhrases(phrases);\n}\n\n/**\n * Build a phrase starting from a specific word match position\n * \n * @param wordMatches - All word matches in document\n * @param startIndex - Starting index in wordMatches array\n * @param queryTokens - Original query tokens\n * @param config - Scoring configuration\n * @param documentFrequency - Document frequency map\n * @param totalDocuments - Total document count\n * @param allWordMatches - All word matches in document (for density calculation)\n * @param documentTokens - Original document tokens (for gap word extraction)\n * @returns Phrase match or null\n */\nfunction buildPhraseFromPosition(\n wordMatches: WordMatch[],\n startIndex: number,\n queryTokens: string[],\n config: ScoringConfig,\n documentFrequency: Map<string, number>,\n totalDocuments: number,\n allWordMatches: WordMatch[],\n documentTokens: string[]\n): PhraseMatch | null {\n const startMatch = wordMatches[startIndex];\n const phraseWords: WordMatch[] = [startMatch];\n const coveredTokens = new Set([startMatch.queryToken]);\n const gapWords: GapWord[] = [];\n let totalGapUsed = 0;\n\n // Look for nearby matches to complete the phrase\n for (let i = startIndex + 1; i < wordMatches.length; i++) {\n const match = wordMatches[i];\n const lastPos = phraseWords[phraseWords.length - 1].position;\n const gap = match.position - lastPos - 1;\n\n // Stop if gap exceeds maximum\n if (gap > config.maxGap) {\n break;\n }\n\n // Only process if it's a different query token we need\n if (!coveredTokens.has(match.queryToken)) {\n // Track gap words between last match and current match\n for (let pos = lastPos + 1; pos < match.position; pos++) {\n totalGapUsed++;\n gapWords.push({\n word: documentTokens[pos],\n position: pos,\n gapIndex: totalGapUsed\n });\n }\n\n phraseWords.push(match);\n coveredTokens.add(match.queryToken);\n\n // Stop if we have all query tokens\n if (coveredTokens.size === queryTokens.length) {\n break;\n }\n }\n }\n\n // Calculate phrase score\n if (phraseWords.length > 0) {\n const coverage = phraseWords.length / queryTokens.length;\n const span = phraseWords[phraseWords.length - 1].position - phraseWords[0].position + 1;\n \n const { score, breakdown } = calculatePhraseScore(\n phraseWords,\n queryTokens,\n config,\n documentFrequency,\n totalDocuments,\n allWordMatches,\n coverage\n );\n\n return {\n words: phraseWords,\n gapWords,\n gapUsed: totalGapUsed,\n coverage,\n startPosition: phraseWords[0].position,\n endPosition: phraseWords[phraseWords.length - 1].position,\n span,\n inOrder: isInOrder(phraseWords, queryTokens),\n score,\n scoreBreakdown: breakdown\n };\n }\n\n return null;\n}\n\n/**\n * Calculate overall phrase score\n * \n * @param phraseWords - Words in the phrase\n * @param queryTokens - Original query tokens\n * @param config - Scoring configuration\n * @param documentFrequency - Document frequency map\n * @param totalDocuments - Total document count\n * @param allWordMatches - All word matches in document (for density calculation)\n * @param coverage - Pre-calculated coverage ratio (phraseWords.length / queryTokens.length)\n * @returns Phrase score (0-1) and detailed component breakdown\n */\nfunction calculatePhraseScore(\n phraseWords: WordMatch[],\n queryTokens: string[],\n config: ScoringConfig,\n documentFrequency: Map<string, number>,\n totalDocuments: number,\n allWordMatches: WordMatch[],\n coverage: number\n): { score: number; breakdown: { base: number; order: number; proximity: number; density: number; semantic: number; coverage: number } } {\n // Base score from word matches\n // Each word contributes: matchScore ร— typeWeight\n let baseScore = 0;\n for (const word of phraseWords) {\n const weight = word.type === 'exact' ? config.weights.exact :\n word.type === 'fuzzy' ? config.weights.fuzzy : \n config.weights.fuzzy * 0.8; // synonym gets 80% of fuzzy weight\n baseScore += word.score * weight;\n }\n baseScore /= phraseWords.length;\n\n // Order bonus: 1.0 if words appear in query order, 0.5 otherwise\n const inOrder = isInOrder(phraseWords, queryTokens);\n const orderScore = inOrder ? 1.0 : 0.5;\n\n // Proximity bonus (closer words score higher)\n // Short-circuit: skip if maxGap=0, proximity weight is 0, or single-word query (proximity meaningless)\n let proximityScore = 0;\n if (config.maxGap > 0 && config.weights.proximity > 0 && queryTokens.length > 1) {\n const span = phraseWords[phraseWords.length - 1].position - phraseWords[0].position + 1;\n const proximityWindow = queryTokens.length * config.proximitySpanMultiplier;\n proximityScore = Math.max(0, 1.0 - (span / proximityWindow));\n }\n\n // Density: Only applies to single-word queries (measures word repetition in document)\n // For multi-word phrase queries, density is 0 (coverage handles completeness separately)\n let densityScore = 0;\n \n if (queryTokens.length === 1) {\n // Single-word query: reward repetition\n const totalOccurrences = allWordMatches.length;\n // Cap at reasonable maximum to avoid runaway scores\n densityScore = Math.min(1.0, totalOccurrences / 10);\n }\n // For multi-word queries: densityScore stays 0\n // Coverage is applied as a multiplier at the end instead\n\n // Semantic score (TF-IDF based)\n const semanticScore = calculateSemanticScore(\n phraseWords,\n documentFrequency,\n totalDocuments\n );\n\n // Weighted combination\n const weights = config.weights;\n \n // Calculate weighted components\n const weightedBase = baseScore;\n const weightedOrder = orderScore * weights.order;\n const weightedProximity = proximityScore * weights.proximity;\n const weightedDensity = densityScore * weights.density;\n const weightedSemantic = semanticScore * weights.semantic;\n \n const totalScore = weightedBase + weightedOrder + weightedProximity + weightedDensity + weightedSemantic;\n\n // Calculate max possible score\n // FIX: Use actual max base weight (highest of exact/fuzzy) instead of hardcoded 1.0\n // When tolerance=0 or fuzzy weight=0, only exact matches are possible\n const canHaveFuzzyMatches = config.tolerance > 0 && weights.fuzzy > 0;\n const maxBaseWeight = canHaveFuzzyMatches ? Math.max(weights.exact, weights.fuzzy) : weights.exact;\n // Only include proximity in max if it can actually contribute (avoids penalizing scores when maxGap=0 or single-word)\n const effectiveProximityWeight = (config.maxGap > 0 && weights.proximity > 0 && queryTokens.length > 1) ? weights.proximity : 0;\n const maxPossibleScore = maxBaseWeight + weights.order + effectiveProximityWeight + weights.density + weights.semantic;\n \n // Normalize to 0-1 range\n const normalizedScore = totalScore / maxPossibleScore;\n \n // FIX: Apply coverage as a MULTIPLIER for multi-word queries\n // This ensures incomplete matches (2/3) can never outscore complete matches (3/3)\n const coverageMultiplier = queryTokens.length > 1 ? coverage : 1.0;\n const score = normalizedScore * coverageMultiplier;\n\n // Component contributions to the final normalized score (before coverage multiplier)\n const base = weightedBase / maxPossibleScore;\n const order = weightedOrder / maxPossibleScore;\n const proximity = weightedProximity / maxPossibleScore;\n const density = weightedDensity / maxPossibleScore;\n const semantic = weightedSemantic / maxPossibleScore;\n\n return {\n score,\n breakdown: {\n base,\n order,\n proximity,\n density,\n semantic,\n coverage: coverageMultiplier // Show coverage multiplier in breakdown\n }\n };\n}\n\n/**\n * Check if words are in the same order as query tokens\n * \n * @param phraseWords - Words in the phrase\n * @param queryTokens - Original query tokens\n * @returns True if in order\n */\nfunction isInOrder(phraseWords: WordMatch[], queryTokens: string[]): boolean {\n const tokenOrder = new Map(queryTokens.map((token, index) => [token, index]));\n \n for (let i = 1; i < phraseWords.length; i++) {\n const prevOrder = tokenOrder.get(phraseWords[i - 1].queryToken) ?? -1;\n const currOrder = tokenOrder.get(phraseWords[i].queryToken) ?? -1;\n \n if (currOrder < prevOrder) {\n return false;\n }\n }\n \n return true;\n}\n\n/**\n * Calculate semantic score using TF-IDF\n * \n * @param phraseWords - Words in the phrase\n * @param documentFrequency - Document frequency map\n * @param totalDocuments - Total document count\n * @returns Semantic score (0-1)\n */\nfunction calculateSemanticScore(\n phraseWords: WordMatch[],\n documentFrequency: Map<string, number>,\n totalDocuments: number\n): number {\n // Handle edge case: no documents\n if (totalDocuments === 0) {\n return 0;\n }\n \n let tfidfSum = 0;\n \n for (const word of phraseWords) {\n const df = documentFrequency.get(word.word) || 1;\n const idf = Math.log(totalDocuments / df);\n tfidfSum += idf;\n }\n \n // Normalize by phrase length\n const avgTfidf = tfidfSum / phraseWords.length;\n \n // Normalize to 0-1 range (assuming max IDF of ~10)\n return Math.min(1.0, avgTfidf / 10);\n}\n\n/**\n * Deduplicate overlapping phrases, keeping highest scoring ones\n * \n * @param phrases - Array of phrase matches\n * @returns Deduplicated phrases sorted by score\n */\nfunction deduplicatePhrases(phrases: PhraseMatch[]): PhraseMatch[] {\n if (phrases.length === 0) return [];\n\n // Sort by score descending\n const sorted = phrases.slice().sort((a, b) => b.score - a.score);\n const result: PhraseMatch[] = [];\n const covered = new Set<number>();\n\n for (const phrase of sorted) {\n // Check if this phrase overlaps with already selected phrases\n let overlaps = false;\n for (let pos = phrase.startPosition; pos <= phrase.endPosition; pos++) {\n if (covered.has(pos)) {\n overlaps = true;\n break;\n }\n }\n\n if (!overlaps) {\n result.push(phrase);\n // Mark positions as covered\n for (let pos = phrase.startPosition; pos <= phrase.endPosition; pos++) {\n covered.add(pos);\n }\n }\n }\n\n return result.sort((a, b) => b.score - a.score);\n}\n","/**\n * Fuzzy Phrase Plugin for Orama\n * \n * Advanced fuzzy phrase matching with semantic weighting and synonym expansion.\n * Completely independent from QPS - accesses Orama's radix tree directly.\n */\n\nimport type { AnyOrama, OramaPlugin, Results, TypedDocument } from '@wcs-colab/orama';\nimport type { FuzzyPhraseConfig, PluginState, SynonymMap, DocumentMatch } from './types.js';\nimport { calculateAdaptiveTolerance } from './fuzzy.js';\nimport { \n extractVocabularyFromRadixTree, \n findAllCandidates,\n filterCandidatesByScore \n} from './candidates.js';\nimport { findPhrasesInDocument } from './scoring.js';\n\n/**\n * Default configuration\n */\nconst DEFAULT_CONFIG: Required<FuzzyPhraseConfig> = {\n textProperty: 'normalized_content', // Must match server's field name\n tolerance: 1,\n adaptiveTolerance: true,\n enableSynonyms: false,\n supabase: undefined as any,\n synonymMatchScore: 0.8,\n weights: {\n exact: 1.0,\n fuzzy: 0.8,\n order: 0.3,\n proximity: 0.2,\n density: 0.2,\n semantic: 0.15\n },\n maxGap: 5,\n minScore: 0.1,\n enableFinalScoreMinimum: false,\n finalScoreMinimum: 0.3,\n proximitySpanMultiplier: 5\n};\n\n/**\n * Plugin state storage (keyed by Orama instance)\n */\nconst pluginStates = new WeakMap<AnyOrama, PluginState>();\n\n/**\n * Create the Fuzzy Phrase Plugin\n * \n * @param userConfig - User configuration options\n * @returns Orama plugin instance\n */\nexport function pluginFuzzyPhrase(userConfig: FuzzyPhraseConfig = {}): OramaPlugin {\n // Merge user config with defaults\n const config: Required<FuzzyPhraseConfig> = {\n textProperty: userConfig.textProperty ?? DEFAULT_CONFIG.textProperty,\n tolerance: userConfig.tolerance ?? DEFAULT_CONFIG.tolerance,\n adaptiveTolerance: userConfig.adaptiveTolerance ?? DEFAULT_CONFIG.adaptiveTolerance,\n enableSynonyms: userConfig.enableSynonyms ?? DEFAULT_CONFIG.enableSynonyms,\n supabase: userConfig.supabase || DEFAULT_CONFIG.supabase,\n synonymMatchScore: userConfig.synonymMatchScore ?? DEFAULT_CONFIG.synonymMatchScore,\n weights: {\n exact: userConfig.weights?.exact ?? DEFAULT_CONFIG.weights.exact,\n fuzzy: userConfig.weights?.fuzzy ?? DEFAULT_CONFIG.weights.fuzzy,\n order: userConfig.weights?.order ?? DEFAULT_CONFIG.weights.order,\n proximity: userConfig.weights?.proximity ?? DEFAULT_CONFIG.weights.proximity,\n density: userConfig.weights?.density ?? DEFAULT_CONFIG.weights.density,\n semantic: userConfig.weights?.semantic ?? DEFAULT_CONFIG.weights.semantic\n },\n maxGap: userConfig.maxGap ?? DEFAULT_CONFIG.maxGap,\n minScore: userConfig.minScore ?? DEFAULT_CONFIG.minScore,\n enableFinalScoreMinimum: userConfig.enableFinalScoreMinimum ?? DEFAULT_CONFIG.enableFinalScoreMinimum,\n finalScoreMinimum: userConfig.finalScoreMinimum ?? DEFAULT_CONFIG.finalScoreMinimum,\n proximitySpanMultiplier: userConfig.proximitySpanMultiplier ?? DEFAULT_CONFIG.proximitySpanMultiplier\n };\n\n const plugin: OramaPlugin = {\n name: 'fuzzy-phrase',\n\n /**\n * Initialize plugin after index is created\n */\n afterCreate: async (orama: AnyOrama) => {\n console.log('๐Ÿ”ฎ Initializing Fuzzy Phrase Plugin...');\n\n // Initialize state\n const state: PluginState = {\n synonymMap: {},\n config,\n documentFrequency: new Map(),\n totalDocuments: 0\n };\n\n // Load synonyms from Supabase if enabled\n if (config.enableSynonyms && config.supabase) {\n try {\n console.log('๐Ÿ“– Loading synonyms from Supabase...');\n state.synonymMap = await loadSynonymsFromSupabase(config.supabase);\n console.log(`โœ… Loaded ${Object.keys(state.synonymMap).length} words with synonyms`);\n } catch (error) {\n console.error('โš ๏ธ Failed to load synonyms:', error);\n // Continue without synonyms\n }\n }\n\n // Calculate document frequencies for TF-IDF from document store\n const docs = (orama.data as any)?.docs?.docs;\n if (docs) {\n state.totalDocuments = Object.keys(docs).length;\n state.documentFrequency = calculateDocumentFrequencies(docs, config.textProperty);\n console.log(`๐Ÿ“Š Calculated document frequencies for ${state.totalDocuments} documents`);\n }\n\n // Store state\n pluginStates.set(orama, state);\n console.log('โœ… Fuzzy Phrase Plugin initialized');\n \n // Signal ready - emit a custom event that can be listened to\n // Use setImmediate to ensure this runs after the afterCreate hook completes\n setImmediate(() => {\n if (typeof (globalThis as any).fuzzyPhrasePluginReady === 'function') {\n console.log('๐Ÿ“ก Signaling plugin ready...');\n (globalThis as any).fuzzyPhrasePluginReady();\n } else {\n console.warn('โš ๏ธ fuzzyPhrasePluginReady callback not found');\n }\n });\n }\n };\n\n return plugin;\n}\n\n/**\n * Search with fuzzy phrase matching\n * \n * This function should be called instead of the regular search() function\n * to enable fuzzy phrase matching.\n */\nexport async function searchWithFuzzyPhrase<T extends AnyOrama>(\n orama: T, \n params: { term?: string; properties?: string[]; limit?: number },\n language?: string\n): Promise<Results<TypedDocument<T>>> {\n const startTime = performance.now();\n \n // Get plugin state\n const state = pluginStates.get(orama);\n \n if (!state) {\n console.error('โŒ Plugin state not initialized');\n throw new Error('Fuzzy Phrase Plugin not properly initialized');\n }\n\n const { term, properties } = params;\n \n if (!term || typeof term !== 'string') {\n return { elapsed: { formatted: '0ms', raw: 0 }, hits: [], count: 0 };\n }\n\n // Use specified property or default\n const textProperty = (properties && properties[0]) || state.config.textProperty;\n\n // Tokenize query\n const queryTokens = tokenize(term);\n \n if (queryTokens.length === 0) {\n return { elapsed: { formatted: '0ms', raw: 0 }, hits: [], count: 0 };\n }\n\n // Calculate tolerance (adaptive or fixed)\n const tolerance = state.config.adaptiveTolerance\n ? calculateAdaptiveTolerance(queryTokens, state.config.tolerance)\n : state.config.tolerance;\n\n console.log(`๐Ÿ” Fuzzy phrase search: \"${term}\" (${queryTokens.length} tokens, tolerance: ${tolerance})`);\n\n // Extract vocabulary from radix tree\n let vocabulary: Set<string>;\n \n try {\n // Access radix tree - the actual index data is in orama.data.index, not orama.index\n // orama.index is just the component interface (methods)\n const indexData = (orama as any).data?.index;\n \n if (!indexData) {\n console.error('โŒ No index data found in orama.data.index');\n return { elapsed: { formatted: '0ms', raw: 0 }, hits: [], count: 0 };\n }\n \n console.log('๐Ÿ” DEBUG: Index data keys:', Object.keys(indexData || {}));\n \n // Try different paths to find the radix tree\n let radixNode = null;\n \n // Path 1: QPS-style (orama.data.index.indexes[property].node)\n if (indexData.indexes?.[textProperty]?.node) {\n radixNode = indexData.indexes[textProperty].node;\n console.log('โœ… Found radix via QPS-style path (data.index.indexes)');\n }\n // Path 2: Standard Orama (orama.data.index[property].node)\n else if (indexData[textProperty]?.node) {\n radixNode = indexData[textProperty].node;\n console.log('โœ… Found radix via standard path (data.index[property])');\n }\n \n if (!radixNode) {\n console.error('โŒ Radix tree not found for property:', textProperty);\n console.error(' Available properties in index:', Object.keys(indexData));\n return { elapsed: { formatted: '0ms', raw: 0 }, hits: [], count: 0 };\n }\n\n vocabulary = extractVocabularyFromRadixTree(radixNode);\n console.log(`๐Ÿ“š Extracted ${vocabulary.size} unique words from index`);\n } catch (error) {\n console.error('โŒ Failed to extract vocabulary:', error);\n return { elapsed: { formatted: '0ms', raw: 0 }, hits: [], count: 0 };\n }\n\n // Find candidates for all query tokens\n const candidatesMap = findAllCandidates(\n queryTokens,\n vocabulary,\n tolerance,\n state.config.enableSynonyms ? state.synonymMap : undefined,\n state.config.synonymMatchScore\n );\n\n // Filter by minimum score (skip when tolerance=0 since all matches are exact with score 1.0)\n const filteredCandidates = tolerance === 0\n ? candidatesMap // Skip filtering - all matches are exact\n : filterCandidatesByScore(candidatesMap, state.config.minScore);\n\n console.log(`๐ŸŽฏ Found candidates: ${Array.from(filteredCandidates.values()).reduce((sum, c) => sum + c.length, 0)} total`);\n\n // Search through all documents\n const documentMatches: DocumentMatch[] = [];\n \n console.log('๐Ÿ” DEBUG orama.data structure:', {\n dataKeys: Object.keys((orama as any).data || {}),\n hasDocs: !!((orama as any).data?.docs),\n docsType: (orama as any).data?.docs ? typeof (orama as any).data.docs : 'undefined'\n });\n \n // Try multiple possible document storage locations\n let docs: Record<string, any> = {};\n \n // Access the actual documents - they're nested in orama.data.docs.docs\n if ((orama as any).data?.docs?.docs) {\n docs = (orama as any).data.docs.docs;\n console.log('โœ… Found docs at orama.data.docs.docs');\n }\n // Fallback: orama.data.docs (might be the correct structure in some cases)\n else if ((orama as any).data?.docs && typeof (orama as any).data.docs === 'object') {\n // Check if it has document-like properties (not sharedInternalDocumentStore, etc.)\n const firstKey = Object.keys((orama as any).data.docs)[0];\n if (firstKey && firstKey !== 'sharedInternalDocumentStore' && firstKey !== 'count') {\n docs = (orama as any).data.docs;\n console.log('โœ… Found docs at orama.data.docs (direct)');\n }\n }\n \n if (Object.keys(docs).length === 0) {\n console.log('โŒ Could not find documents - available structure:', {\n hasDataDocs: !!((orama as any).data?.docs),\n dataDocsKeys: (orama as any).data?.docs ? Object.keys((orama as any).data.docs) : 'none',\n hasDataDocsDocs: !!((orama as any).data?.docs?.docs),\n dataDocsDocsCount: (orama as any).data?.docs?.docs ? Object.keys((orama as any).data.docs.docs).length : 0\n });\n }\n \n console.log(`๐Ÿ“„ Searching through ${Object.keys(docs).length} documents`);\n\n for (const [docId, doc] of Object.entries(docs)) {\n const text = doc[textProperty];\n \n if (!text || typeof text !== 'string') {\n continue;\n }\n\n // Tokenize document\n const docTokens = tokenize(text);\n\n // Find phrases in this document\n // Note: state.config.weights is guaranteed to have all properties from default merge\n const phrases = findPhrasesInDocument(\n docTokens,\n filteredCandidates,\n {\n weights: state.config.weights as { exact: number; fuzzy: number; order: number; proximity: number; density: number; semantic: number },\n maxGap: state.config.maxGap,\n proximitySpanMultiplier: state.config.proximitySpanMultiplier,\n tolerance\n },\n state.documentFrequency,\n state.totalDocuments\n );\n\n if (phrases.length > 0) {\n // Calculate overall document score (highest phrase score)\n const docScore = Math.max(...phrases.map(p => p.score));\n\n documentMatches.push({\n id: docId,\n phrases,\n score: docScore,\n document: doc\n });\n }\n }\n\n // Sort by score descending\n documentMatches.sort((a, b) => b.score - a.score);\n\n // Apply final score minimum filter if enabled\n let filteredMatches = documentMatches;\n if (state.config.enableFinalScoreMinimum && state.config.finalScoreMinimum > 0) {\n const threshold = state.config.finalScoreMinimum;\n const beforeCount = filteredMatches.length;\n filteredMatches = filteredMatches.filter(m => m.score >= threshold);\n console.log(`๐ŸŽš๏ธ Final score filter: ${beforeCount} โ†’ ${filteredMatches.length} (threshold: ${threshold})`);\n }\n\n // Apply limit if specified\n const limit = params.limit ?? filteredMatches.length;\n const limitedMatches = filteredMatches.slice(0, limit);\n\n // Convert to Orama results format\n const hits = limitedMatches.map(match => ({\n id: match.id,\n score: match.score,\n document: match.document,\n // Store phrases for highlighting\n _phrases: match.phrases\n })) as any[];\n\n const elapsed = performance.now() - startTime;\n\n console.log(`โœ… Found ${hits.length} results in ${elapsed.toFixed(2)}ms (limit: ${limit})`);\n\n return {\n elapsed: {\n formatted: `${elapsed.toFixed(2)}ms`,\n raw: Math.floor(elapsed * 1000000) // nanoseconds\n },\n hits,\n count: hits.length\n } as any;\n}\n\n/**\n * Load synonyms from Supabase\n */\nasync function loadSynonymsFromSupabase(\n supabaseConfig: { url: string; serviceKey: string }\n): Promise<SynonymMap> {\n try {\n console.log('๐Ÿ” DEBUG: Calling Supabase RPC get_synonym_map...');\n \n // Dynamic import to avoid bundling Supabase client if not needed\n const { createClient } = await import('@supabase/supabase-js');\n \n const supabase = createClient(supabaseConfig.url, supabaseConfig.serviceKey);\n \n // Call the get_synonym_map function\n const { data, error } = await supabase.rpc('get_synonym_map');\n \n console.log('๐Ÿ” DEBUG: Supabase RPC response:', {\n hasError: !!error,\n errorMessage: error?.message,\n hasData: !!data,\n dataType: typeof data,\n dataKeys: data ? Object.keys(data).length : 0\n });\n \n if (error) {\n throw new Error(`Supabase error: ${error.message}`);\n }\n \n const synonymMap = data || {};\n console.log(`๐Ÿ“š Loaded ${Object.keys(synonymMap).length} synonym entries from Supabase`);\n \n return synonymMap;\n } catch (error) {\n console.error('โŒ Failed to load synonyms from Supabase:', error);\n throw error;\n }\n}\n\n/**\n * Calculate document frequencies for TF-IDF\n */\nfunction calculateDocumentFrequencies(\n docs: Record<string, any>,\n textProperty: string\n): Map<string, number> {\n const df = new Map<string, number>();\n\n for (const doc of Object.values(docs)) {\n const text = doc[textProperty];\n \n if (!text || typeof text !== 'string') {\n continue;\n }\n\n // Get unique words in this document\n const words = new Set(tokenize(text));\n\n // Increment document frequency for each unique word\n for (const word of words) {\n df.set(word, (df.get(word) || 0) + 1);\n }\n }\n\n return df;\n}\n\n/**\n * Normalize text using the same rules as server-side\n * \n * CRITICAL: This must match the normalizeText() function in server/index.js exactly\n * PLUS we remove all punctuation to match Orama's French tokenizer behavior\n */\nfunction normalizeText(text: string): string {\n return text\n .toLowerCase()\n .normalize('NFD')\n .replace(/[\\u0300-\\u036f]/g, '') // Remove diacritics\n // Replace French elisions (l', d', etc.) with space to preserve word boundaries\n .replace(/\\b[ldcjmnst][\\u2018\\u2019\\u201A\\u201B\\u2032\\u2035\\u0027\\u0060\\u00B4](?=\\w)/gi, ' ')\n .replace(/[\\u2018\\u2019\\u201A\\u201B\\u2032\\u2035\\u0027\\u0060\\u00B4]/g, '') // Remove remaining apostrophes\n .replace(/[\\u201c\\u201d]/g, '\"') // Normalize curly quotes to straight quotes\n .replace(/[.,;:!?()[\\]{}\\-โ€”โ€“ยซยป\"\"]/g, ' ') // Remove punctuation (replace with space to preserve word boundaries)\n .replace(/\\s+/g, ' ') // Normalize multiple spaces to single space\n .trim();\n}\n\n/**\n * Tokenization matching normalized text behavior\n * \n * Note: Text should already be normalized before indexing, so we normalize again\n * to ensure plugin tokenization matches index tokenization\n */\nfunction tokenize(text: string): string[] {\n // Normalize first (same as indexing), then split by whitespace\n return normalizeText(text)\n .split(/\\s+/)\n .filter(token => token.length > 0);\n}\n\n/**\n * Export types for external use\n */\nexport type {\n FuzzyPhraseConfig,\n WordMatch,\n PhraseMatch,\n DocumentMatch,\n SynonymMap,\n Candidate\n} from './types.js';\n"]}
package/dist/index.js CHANGED
@@ -307,7 +307,8 @@ function calculatePhraseScore(phraseWords, queryTokens, config, documentFrequenc
307
307
  const weightedDensity = densityScore * weights.density;
308
308
  const weightedSemantic = semanticScore * weights.semantic;
309
309
  const totalScore = weightedBase + weightedOrder + weightedProximity + weightedDensity + weightedSemantic;
310
- const maxBaseWeight = Math.max(weights.exact, weights.fuzzy);
310
+ const canHaveFuzzyMatches = config.tolerance > 0 && weights.fuzzy > 0;
311
+ const maxBaseWeight = canHaveFuzzyMatches ? Math.max(weights.exact, weights.fuzzy) : weights.exact;
311
312
  const effectiveProximityWeight = config.maxGap > 0 && weights.proximity > 0 && queryTokens.length > 1 ? weights.proximity : 0;
312
313
  const maxPossibleScore = maxBaseWeight + weights.order + effectiveProximityWeight + weights.density + weights.semantic;
313
314
  const normalizedScore = totalScore / maxPossibleScore;
@@ -519,10 +520,7 @@ async function searchWithFuzzyPhrase(orama, params, language) {
519
520
  state.config.enableSynonyms ? state.synonymMap : void 0,
520
521
  state.config.synonymMatchScore
521
522
  );
522
- const filteredCandidates = filterCandidatesByScore(
523
- candidatesMap,
524
- state.config.minScore
525
- );
523
+ const filteredCandidates = tolerance === 0 ? candidatesMap : filterCandidatesByScore(candidatesMap, state.config.minScore);
526
524
  console.log(`\u{1F3AF} Found candidates: ${Array.from(filteredCandidates.values()).reduce((sum, c) => sum + c.length, 0)} total`);
527
525
  const documentMatches = [];
528
526
  console.log("\u{1F50D} DEBUG orama.data structure:", {
@@ -562,7 +560,8 @@ async function searchWithFuzzyPhrase(orama, params, language) {
562
560
  {
563
561
  weights: state.config.weights,
564
562
  maxGap: state.config.maxGap,
565
- proximitySpanMultiplier: state.config.proximitySpanMultiplier
563
+ proximitySpanMultiplier: state.config.proximitySpanMultiplier,
564
+ tolerance
566
565
  },
567
566
  state.documentFrequency,
568
567
  state.totalDocuments
package/dist/index.js.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/fuzzy.ts","../src/candidates.ts","../src/scoring.ts","../src/index.ts"],"names":[],"mappings":";AA4BO,SAAS,mBACd,GACA,GACA,OAC0B;AAE1B,MAAI,MAAM,GAAG;AACX,WAAO,EAAE,WAAW,MAAM,UAAU,EAAE;AAAA,EACxC;AAEA,QAAM,OAAO,EAAE;AACf,QAAM,OAAO,EAAE;AAGf,MAAI,KAAK,IAAI,OAAO,IAAI,IAAI,OAAO;AACjC,WAAO,EAAE,WAAW,OAAO,UAAU,QAAQ,EAAE;AAAA,EACjD;AAGA,MAAI,OAAO,MAAM;AACf,KAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC;AAAA,EAChB;AAEA,QAAM,IAAI,EAAE;AACZ,QAAM,IAAI,EAAE;AAGZ,MAAI,UAAU,IAAI,MAAM,IAAI,CAAC;AAC7B,MAAI,UAAU,IAAI,MAAM,IAAI,CAAC;AAG7B,WAAS,IAAI,GAAG,KAAK,GAAG,KAAK;AAC3B,YAAQ,CAAC,IAAI;AAAA,EACf;AAEA,WAAS,IAAI,GAAG,KAAK,GAAG,KAAK;AAC3B,YAAQ,CAAC,IAAI;AACb,QAAI,WAAW;AAEf,aAAS,IAAI,GAAG,KAAK,GAAG,KAAK;AAC3B,YAAM,OAAO,EAAE,IAAI,CAAC,MAAM,EAAE,IAAI,CAAC,IAAI,IAAI;AAEzC,cAAQ,CAAC,IAAI,KAAK;AAAA,QAChB,QAAQ,CAAC,IAAI;AAAA;AAAA,QACb,QAAQ,IAAI,CAAC,IAAI;AAAA;AAAA,QACjB,QAAQ,IAAI,CAAC,IAAI;AAAA;AAAA,MACnB;AAEA,iBAAW,KAAK,IAAI,UAAU,QAAQ,CAAC,CAAC;AAAA,IAC1C;AAGA,QAAI,WAAW,OAAO;AACpB,aAAO,EAAE,WAAW,OAAO,UAAU,QAAQ,EAAE;AAAA,IACjD;AAGA,KAAC,SAAS,OAAO,IAAI,CAAC,SAAS,OAAO;AAAA,EACxC;AAEA,QAAM,WAAW,QAAQ,CAAC;AAC1B,SAAO;AAAA,IACL,WAAW,YAAY;AAAA,IACvB;AAAA,EACF;AACF;AAUO,SAAS,WACd,MACA,YACA,WACuD;AAEvD,MAAI,SAAS,YAAY;AACvB,WAAO,EAAE,SAAS,MAAM,UAAU,GAAG,OAAO,EAAI;AAAA,EAClD;AAGA,MAAI,KAAK,WAAW,UAAU,GAAG;AAC/B,WAAO,EAAE,SAAS,MAAM,UAAU,GAAG,OAAO,KAAK;AAAA,EACnD;AAGA,QAAM,SAAS,mBAAmB,MAAM,YAAY,SAAS;AAE7D,MAAI,OAAO,WAAW;AAGpB,UAAM,QAAQ,IAAO,OAAO,WAAW;AACvC,WAAO;AAAA,MACL,SAAS;AAAA,MACT,UAAU,OAAO;AAAA,MACjB,OAAO,KAAK,IAAI,KAAK,KAAK;AAAA;AAAA,IAC5B;AAAA,EACF;AAEA,SAAO,EAAE,SAAS,OAAO,UAAU,YAAY,GAAG,OAAO,EAAE;AAC7D;AAWO,SAAS,2BACd,aACA,eACQ;AACR,QAAM,cAAc,YAAY;AAEhC,MAAI,eAAe,GAAG;AACpB,WAAO;AAAA,EACT,WAAW,eAAe,GAAG;AAC3B,WAAO,gBAAgB;AAAA,EACzB,WAAW,eAAe,GAAG;AAC3B,WAAO,gBAAgB;AAAA,EACzB,OAAO;AACL,WAAO,gBAAgB;AAAA,EACzB;AACF;;;ACjJO,SAAS,+BAA+B,WAA6B;AAC1E,QAAM,aAAa,oBAAI,IAAY;AACnC,MAAI,eAAe;AACnB,MAAI,aAAa;AAEjB,WAAS,SAAS,MAAW,QAAgB,GAAG;AAC9C,QAAI,CAAC,MAAM;AACT;AAAA,IACF;AAEA;AAIA,QAAI,KAAK,KAAK,KAAK,KAAK,OAAO,KAAK,MAAM,YAAY,KAAK,EAAE,SAAS,GAAG;AACvE,iBAAW,IAAI,KAAK,CAAC;AACrB;AAAA,IACF;AAGA,QAAI,KAAK,GAAG;AACV,UAAI,KAAK,aAAa,KAAK;AAEzB,mBAAW,CAAC,MAAM,SAAS,KAAK,KAAK,GAAG;AACtC,mBAAS,WAAW,QAAQ,CAAC;AAAA,QAC/B;AAAA,MACF,WAAW,MAAM,QAAQ,KAAK,CAAC,GAAG;AAEhC,mBAAW,CAAC,MAAM,SAAS,KAAK,KAAK,GAAG;AACtC,mBAAS,WAAW,QAAQ,CAAC;AAAA,QAC/B;AAAA,MACF,WAAW,OAAO,KAAK,MAAM,UAAU;AAErC,mBAAW,aAAa,OAAO,OAAO,KAAK,CAAC,GAAG;AAC7C,mBAAS,WAAW,QAAQ,CAAC;AAAA,QAC/B;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAEA,WAAS,SAAS;AAClB,UAAQ,IAAI,uBAAgB,WAAW,IAAI,eAAe,YAAY,gBAAgB;AACtF,SAAO;AACT;AAYO,SAAS,uBACd,YACA,YACA,WACA,UACA,eAAuB,KACV;AACb,QAAM,aAA0B,CAAC;AACjC,QAAM,OAAO,oBAAI,IAAY;AAG7B,MAAI,WAAW,IAAI,UAAU,GAAG;AAC9B,eAAW,KAAK;AAAA,MACd,MAAM;AAAA,MACN,MAAM;AAAA,MACN;AAAA,MACA,UAAU;AAAA,MACV,OAAO;AAAA,IACT,CAAC;AACD,SAAK,IAAI,UAAU;AAAA,EACrB;AAGA,aAAW,QAAQ,YAAY;AAC7B,QAAI,KAAK,IAAI,IAAI;AAAG;AAEpB,UAAM,QAAQ,WAAW,MAAM,YAAY,SAAS;AACpD,QAAI,MAAM,SAAS;AACjB,iBAAW,KAAK;AAAA,QACd;AAAA,QACA,MAAM;AAAA,QACN;AAAA,QACA,UAAU,MAAM;AAAA,QAChB,OAAO,MAAM;AAAA,MACf,CAAC;AACD,WAAK,IAAI,IAAI;AAAA,IACf;AAAA,EACF;AAGA,MAAI,YAAY,SAAS,UAAU,GAAG;AACpC,eAAW,WAAW,SAAS,UAAU,GAAG;AAC1C,UAAI,KAAK,IAAI,OAAO;AAAG;AACvB,UAAI,WAAW,IAAI,OAAO,GAAG;AAC3B,mBAAW,KAAK;AAAA,UACd,MAAM;AAAA,UACN,MAAM;AAAA,UACN;AAAA,UACA,UAAU;AAAA,UACV,OAAO;AAAA,QACT,CAAC;AACD,aAAK,IAAI,OAAO;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAEA,SAAO;AACT;AAYO,SAAS,kBACd,aACA,YACA,WACA,UACA,eAAuB,KACG;AAC1B,QAAM,gBAAgB,oBAAI,IAAyB;AAEnD,aAAW,SAAS,aAAa;AAC/B,UAAM,kBAAkB;AAAA,MACtB;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AACA,kBAAc,IAAI,OAAO,eAAe;AAAA,EAC1C;AAEA,SAAO;AACT;AAyBO,SAAS,wBACd,eACA,UAC0B;AAC1B,QAAM,WAAW,oBAAI,IAAyB;AAE9C,aAAW,CAAC,OAAO,UAAU,KAAK,cAAc,QAAQ,GAAG;AACzD,UAAM,qBAAqB,WAAW,OAAO,OAAK,EAAE,SAAS,QAAQ;AACrE,QAAI,mBAAmB,SAAS,GAAG;AACjC,eAAS,IAAI,OAAO,kBAAkB;AAAA,IACxC;AAAA,EACF;AAEA,SAAO;AACT;;;AClKO,SAAS,sBACd,gBACA,eACA,QACA,mBACA,gBACe;AACf,QAAM,UAAyB,CAAC;AAChC,QAAM,cAAc,MAAM,KAAK,cAAc,KAAK,CAAC;AAGnD,QAAM,cAA2B,CAAC;AAElC,WAAS,IAAI,GAAG,IAAI,eAAe,QAAQ,KAAK;AAC9C,UAAM,UAAU,eAAe,CAAC;AAGhC,eAAW,CAAC,YAAY,UAAU,KAAK,cAAc,QAAQ,GAAG;AAC9D,iBAAW,aAAa,YAAY;AAClC,YAAI,UAAU,SAAS,SAAS;AAC9B,sBAAY,KAAK;AAAA,YACf,MAAM;AAAA,YACN;AAAA,YACA,UAAU;AAAA,YACV,MAAM,UAAU;AAAA,YAChB,UAAU,UAAU;AAAA,YACpB,OAAO,UAAU;AAAA,UACnB,CAAC;AAAA,QACH;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAGA,WAAS,IAAI,GAAG,IAAI,YAAY,QAAQ,KAAK;AAC3C,UAAM,SAAS;AAAA,MACb;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA;AAAA,IACF;AAEA,QAAI,UAAU,OAAO,MAAM,SAAS,GAAG;AACrC,cAAQ,KAAK,MAAM;AAAA,IACrB;AAAA,EACF;AAGA,SAAO,mBAAmB,OAAO;AACnC;AAeA,SAAS,wBACP,aACA,YACA,aACA,QACA,mBACA,gBACA,gBACA,gBACoB;AACpB,QAAM,aAAa,YAAY,UAAU;AACzC,QAAM,cAA2B,CAAC,UAAU;AAC5C,QAAM,gBAAgB,oBAAI,IAAI,CAAC,WAAW,UAAU,CAAC;AACrD,QAAM,WAAsB,CAAC;AAC7B,MAAI,eAAe;AAGnB,WAAS,IAAI,aAAa,GAAG,IAAI,YAAY,QAAQ,KAAK;AACxD,UAAM,QAAQ,YAAY,CAAC;AAC3B,UAAM,UAAU,YAAY,YAAY,SAAS,CAAC,EAAE;AACpD,UAAM,MAAM,MAAM,WAAW,UAAU;AAGvC,QAAI,MAAM,OAAO,QAAQ;AACvB;AAAA,IACF;AAGA,QAAI,CAAC,cAAc,IAAI,MAAM,UAAU,GAAG;AAExC,eAAS,MAAM,UAAU,GAAG,MAAM,MAAM,UAAU,OAAO;AACvD;AACA,iBAAS,KAAK;AAAA,UACZ,MAAM,eAAe,GAAG;AAAA,UACxB,UAAU;AAAA,UACV,UAAU;AAAA,QACZ,CAAC;AAAA,MACH;AAEA,kBAAY,KAAK,KAAK;AACtB,oBAAc,IAAI,MAAM,UAAU;AAGlC,UAAI,cAAc,SAAS,YAAY,QAAQ;AAC7C;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAGA,MAAI,YAAY,SAAS,GAAG;AAC1B,UAAM,WAAW,YAAY,SAAS,YAAY;AAClD,UAAM,OAAO,YAAY,YAAY,SAAS,CAAC,EAAE,WAAW,YAAY,CAAC,EAAE,WAAW;AAEtF,UAAM,EAAE,OAAO,UAAU,IAAI;AAAA,MAC3B;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAEA,WAAO;AAAA,MACL,OAAO;AAAA,MACP;AAAA,MACA,SAAS;AAAA,MACT;AAAA,MACA,eAAe,YAAY,CAAC,EAAE;AAAA,MAC9B,aAAa,YAAY,YAAY,SAAS,CAAC,EAAE;AAAA,MACjD;AAAA,MACA,SAAS,UAAU,aAAa,WAAW;AAAA,MAC3C;AAAA,MACA,gBAAgB;AAAA,IAClB;AAAA,EACF;AAEA,SAAO;AACT;AAcA,SAAS,qBACP,aACA,aACA,QACA,mBACA,gBACA,gBACA,UACuI;AAGvI,MAAI,YAAY;AAChB,aAAW,QAAQ,aAAa;AAC9B,UAAM,SAAS,KAAK,SAAS,UAAU,OAAO,QAAQ,QACvC,KAAK,SAAS,UAAU,OAAO,QAAQ,QACvC,OAAO,QAAQ,QAAQ;AACtC,iBAAa,KAAK,QAAQ;AAAA,EAC5B;AACA,eAAa,YAAY;AAGzB,QAAM,UAAU,UAAU,aAAa,WAAW;AAClD,QAAM,aAAa,UAAU,IAAM;AAInC,MAAI,iBAAiB;AACrB,MAAI,OAAO,SAAS,KAAK,OAAO,QAAQ,YAAY,KAAK,YAAY,SAAS,GAAG;AAC/E,UAAM,OAAO,YAAY,YAAY,SAAS,CAAC,EAAE,WAAW,YAAY,CAAC,EAAE,WAAW;AACtF,UAAM,kBAAkB,YAAY,SAAS,OAAO;AACpD,qBAAiB,KAAK,IAAI,GAAG,IAAO,OAAO,eAAgB;AAAA,EAC7D;AAIA,MAAI,eAAe;AAEnB,MAAI,YAAY,WAAW,GAAG;AAE5B,UAAM,mBAAmB,eAAe;AAExC,mBAAe,KAAK,IAAI,GAAK,mBAAmB,EAAE;AAAA,EACpD;AAKA,QAAM,gBAAgB;AAAA,IACpB;AAAA,IACA;AAAA,IACA;AAAA,EACF;AAGA,QAAM,UAAU,OAAO;AAGvB,QAAM,eAAe;AACrB,QAAM,gBAAgB,aAAa,QAAQ;AAC3C,QAAM,oBAAoB,iBAAiB,QAAQ;AACnD,QAAM,kBAAkB,eAAe,QAAQ;AAC/C,QAAM,mBAAmB,gBAAgB,QAAQ;AAEjD,QAAM,aAAa,eAAe,gBAAgB,oBAAoB,kBAAkB;AAIxF,QAAM,gBAAgB,KAAK,IAAI,QAAQ,OAAO,QAAQ,KAAK;AAE3D,QAAM,2BAA4B,OAAO,SAAS,KAAK,QAAQ,YAAY,KAAK,YAAY,SAAS,IAAK,QAAQ,YAAY;AAC9H,QAAM,mBAAmB,gBAAgB,QAAQ,QAAQ,2BAA2B,QAAQ,UAAU,QAAQ;AAG9G,QAAM,kBAAkB,aAAa;AAIrC,QAAM,qBAAqB,YAAY,SAAS,IAAI,WAAW;AAC/D,QAAM,QAAQ,kBAAkB;AAGhC,QAAM,OAAO,eAAe;AAC5B,QAAM,QAAQ,gBAAgB;AAC9B,QAAM,YAAY,oBAAoB;AACtC,QAAM,UAAU,kBAAkB;AAClC,QAAM,WAAW,mBAAmB;AAEpC,SAAO;AAAA,IACL;AAAA,IACA,WAAW;AAAA,MACT;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,UAAU;AAAA;AAAA,IACZ;AAAA,EACF;AACF;AASA,SAAS,UAAU,aAA0B,aAAgC;AAC3E,QAAM,aAAa,IAAI,IAAI,YAAY,IAAI,CAAC,OAAO,UAAU,CAAC,OAAO,KAAK,CAAC,CAAC;AAE5E,WAAS,IAAI,GAAG,IAAI,YAAY,QAAQ,KAAK;AAC3C,UAAM,YAAY,WAAW,IAAI,YAAY,IAAI,CAAC,EAAE,UAAU,KAAK;AACnE,UAAM,YAAY,WAAW,IAAI,YAAY,CAAC,EAAE,UAAU,KAAK;AAE/D,QAAI,YAAY,WAAW;AACzB,aAAO;AAAA,IACT;AAAA,EACF;AAEA,SAAO;AACT;AAUA,SAAS,uBACP,aACA,mBACA,gBACQ;AAER,MAAI,mBAAmB,GAAG;AACxB,WAAO;AAAA,EACT;AAEA,MAAI,WAAW;AAEf,aAAW,QAAQ,aAAa;AAC9B,UAAM,KAAK,kBAAkB,IAAI,KAAK,IAAI,KAAK;AAC/C,UAAM,MAAM,KAAK,IAAI,iBAAiB,EAAE;AACxC,gBAAY;AAAA,EACd;AAGA,QAAM,WAAW,WAAW,YAAY;AAGxC,SAAO,KAAK,IAAI,GAAK,WAAW,EAAE;AACpC;AAQA,SAAS,mBAAmB,SAAuC;AACjE,MAAI,QAAQ,WAAW;AAAG,WAAO,CAAC;AAGlC,QAAM,SAAS,QAAQ,MAAM,EAAE,KAAK,CAAC,GAAG,MAAM,EAAE,QAAQ,EAAE,KAAK;AAC/D,QAAM,SAAwB,CAAC;AAC/B,QAAM,UAAU,oBAAI,IAAY;AAEhC,aAAW,UAAU,QAAQ;AAE3B,QAAI,WAAW;AACf,aAAS,MAAM,OAAO,eAAe,OAAO,OAAO,aAAa,OAAO;AACrE,UAAI,QAAQ,IAAI,GAAG,GAAG;AACpB,mBAAW;AACX;AAAA,MACF;AAAA,IACF;AAEA,QAAI,CAAC,UAAU;AACb,aAAO,KAAK,MAAM;AAElB,eAAS,MAAM,OAAO,eAAe,OAAO,OAAO,aAAa,OAAO;AACrE,gBAAQ,IAAI,GAAG;AAAA,MACjB;AAAA,IACF;AAAA,EACF;AAEA,SAAO,OAAO,KAAK,CAAC,GAAG,MAAM,EAAE,QAAQ,EAAE,KAAK;AAChD;;;AC9WA,IAAM,iBAA8C;AAAA,EAClD,cAAc;AAAA;AAAA,EACd,WAAW;AAAA,EACX,mBAAmB;AAAA,EACnB,gBAAgB;AAAA,EAChB,UAAU;AAAA,EACV,mBAAmB;AAAA,EACnB,SAAS;AAAA,IACP,OAAO;AAAA,IACP,OAAO;AAAA,IACP,OAAO;AAAA,IACP,WAAW;AAAA,IACX,SAAS;AAAA,IACT,UAAU;AAAA,EACZ;AAAA,EACA,QAAQ;AAAA,EACR,UAAU;AAAA,EACV,yBAAyB;AAAA,EACzB,mBAAmB;AAAA,EACnB,yBAAyB;AAC3B;AAKA,IAAM,eAAe,oBAAI,QAA+B;AAQjD,SAAS,kBAAkB,aAAgC,CAAC,GAAgB;AAEjF,QAAM,SAAsC;AAAA,IAC1C,cAAc,WAAW,gBAAgB,eAAe;AAAA,IACxD,WAAW,WAAW,aAAa,eAAe;AAAA,IAClD,mBAAmB,WAAW,qBAAqB,eAAe;AAAA,IAClE,gBAAgB,WAAW,kBAAkB,eAAe;AAAA,IAC5D,UAAU,WAAW,YAAY,eAAe;AAAA,IAChD,mBAAmB,WAAW,qBAAqB,eAAe;AAAA,IAClE,SAAS;AAAA,MACP,OAAO,WAAW,SAAS,SAAS,eAAe,QAAQ;AAAA,MAC3D,OAAO,WAAW,SAAS,SAAS,eAAe,QAAQ;AAAA,MAC3D,OAAO,WAAW,SAAS,SAAS,eAAe,QAAQ;AAAA,MAC3D,WAAW,WAAW,SAAS,aAAa,eAAe,QAAQ;AAAA,MACnE,SAAS,WAAW,SAAS,WAAW,eAAe,QAAQ;AAAA,MAC/D,UAAU,WAAW,SAAS,YAAY,eAAe,QAAQ;AAAA,IACnE;AAAA,IACA,QAAQ,WAAW,UAAU,eAAe;AAAA,IAC5C,UAAU,WAAW,YAAY,eAAe;AAAA,IAChD,yBAAyB,WAAW,2BAA2B,eAAe;AAAA,IAC9E,mBAAmB,WAAW,qBAAqB,eAAe;AAAA,IAClE,yBAAyB,WAAW,2BAA2B,eAAe;AAAA,EAChF;AAEA,QAAM,SAAsB;AAAA,IAC1B,MAAM;AAAA;AAAA;AAAA;AAAA,IAKN,aAAa,OAAO,UAAoB;AACtC,cAAQ,IAAI,+CAAwC;AAGpD,YAAM,QAAqB;AAAA,QACzB,YAAY,CAAC;AAAA,QACb;AAAA,QACA,mBAAmB,oBAAI,IAAI;AAAA,QAC3B,gBAAgB;AAAA,MAClB;AAGA,UAAI,OAAO,kBAAkB,OAAO,UAAU;AAC5C,YAAI;AACF,kBAAQ,IAAI,6CAAsC;AAClD,gBAAM,aAAa,MAAM,yBAAyB,OAAO,QAAQ;AACjE,kBAAQ,IAAI,iBAAY,OAAO,KAAK,MAAM,UAAU,EAAE,MAAM,sBAAsB;AAAA,QACpF,SAAS,OAAO;AACd,kBAAQ,MAAM,0CAAgC,KAAK;AAAA,QAErD;AAAA,MACF;AAGA,YAAM,OAAQ,MAAM,MAAc,MAAM;AACxC,UAAI,MAAM;AACR,cAAM,iBAAiB,OAAO,KAAK,IAAI,EAAE;AACzC,cAAM,oBAAoB,6BAA6B,MAAM,OAAO,YAAY;AAChF,gBAAQ,IAAI,iDAA0C,MAAM,cAAc,YAAY;AAAA,MACxF;AAGA,mBAAa,IAAI,OAAO,KAAK;AAC7B,cAAQ,IAAI,wCAAmC;AAI/C,mBAAa,MAAM;AACjB,YAAI,OAAQ,WAAmB,2BAA2B,YAAY;AACpE,kBAAQ,IAAI,qCAA8B;AAC1C,UAAC,WAAmB,uBAAuB;AAAA,QAC7C,OAAO;AACL,kBAAQ,KAAK,yDAA+C;AAAA,QAC9D;AAAA,MACF,CAAC;AAAA,IACH;AAAA,EACF;AAEA,SAAO;AACT;AAQA,eAAsB,sBACpB,OACA,QACA,UACoC;AACpC,QAAM,YAAY,YAAY,IAAI;AAGlC,QAAM,QAAQ,aAAa,IAAI,KAAK;AAEpC,MAAI,CAAC,OAAO;AACV,YAAQ,MAAM,qCAAgC;AAC9C,UAAM,IAAI,MAAM,8CAA8C;AAAA,EAChE;AAEA,QAAM,EAAE,MAAM,WAAW,IAAI;AAE7B,MAAI,CAAC,QAAQ,OAAO,SAAS,UAAU;AACrC,WAAO,EAAE,SAAS,EAAE,WAAW,OAAO,KAAK,EAAE,GAAG,MAAM,CAAC,GAAG,OAAO,EAAE;AAAA,EACrE;AAGA,QAAM,eAAgB,cAAc,WAAW,CAAC,KAAM,MAAM,OAAO;AAGnE,QAAM,cAAc,SAAS,IAAI;AAEjC,MAAI,YAAY,WAAW,GAAG;AAC5B,WAAO,EAAE,SAAS,EAAE,WAAW,OAAO,KAAK,EAAE,GAAG,MAAM,CAAC,GAAG,OAAO,EAAE;AAAA,EACrE;AAGA,QAAM,YAAY,MAAM,OAAO,oBAC3B,2BAA2B,aAAa,MAAM,OAAO,SAAS,IAC9D,MAAM,OAAO;AAEjB,UAAQ,IAAI,mCAA4B,IAAI,MAAM,YAAY,MAAM,uBAAuB,SAAS,GAAG;AAGvG,MAAI;AAEJ,MAAI;AAGF,UAAM,YAAa,MAAc,MAAM;AAEvC,QAAI,CAAC,WAAW;AACd,cAAQ,MAAM,gDAA2C;AACzD,aAAO,EAAE,SAAS,EAAE,WAAW,OAAO,KAAK,EAAE,GAAG,MAAM,CAAC,GAAG,OAAO,EAAE;AAAA,IACrE;AAEA,YAAQ,IAAI,qCAA8B,OAAO,KAAK,aAAa,CAAC,CAAC,CAAC;AAGtE,QAAI,YAAY;AAGhB,QAAI,UAAU,UAAU,YAAY,GAAG,MAAM;AAC3C,kBAAY,UAAU,QAAQ,YAAY,EAAE;AAC5C,cAAQ,IAAI,4DAAuD;AAAA,IACrE,WAES,UAAU,YAAY,GAAG,MAAM;AACtC,kBAAY,UAAU,YAAY,EAAE;AACpC,cAAQ,IAAI,6DAAwD;AAAA,IACtE;AAEA,QAAI,CAAC,WAAW;AACd,cAAQ,MAAM,6CAAwC,YAAY;AAClE,cAAQ,MAAM,qCAAqC,OAAO,KAAK,SAAS,CAAC;AACzE,aAAO,EAAE,SAAS,EAAE,WAAW,OAAO,KAAK,EAAE,GAAG,MAAM,CAAC,GAAG,OAAO,EAAE;AAAA,IACrE;AAEA,iBAAa,+BAA+B,SAAS;AACrD,YAAQ,IAAI,uBAAgB,WAAW,IAAI,0BAA0B;AAAA,EACvE,SAAS,OAAO;AACd,YAAQ,MAAM,wCAAmC,KAAK;AACtD,WAAO,EAAE,SAAS,EAAE,WAAW,OAAO,KAAK,EAAE,GAAG,MAAM,CAAC,GAAG,OAAO,EAAE;AAAA,EACrE;AAGA,QAAM,gBAAgB;AAAA,IACpB;AAAA,IACA;AAAA,IACA;AAAA,IACA,MAAM,OAAO,iBAAiB,MAAM,aAAa;AAAA,IACjD,MAAM,OAAO;AAAA,EACf;AAGA,QAAM,qBAAqB;AAAA,IACzB;AAAA,IACA,MAAM,OAAO;AAAA,EACf;AAEA,UAAQ,IAAI,+BAAwB,MAAM,KAAK,mBAAmB,OAAO,CAAC,EAAE,OAAO,CAAC,KAAK,MAAM,MAAM,EAAE,QAAQ,CAAC,CAAC,QAAQ;AAGzH,QAAM,kBAAmC,CAAC;AAE1C,UAAQ,IAAI,yCAAkC;AAAA,IAC5C,UAAU,OAAO,KAAM,MAAc,QAAQ,CAAC,CAAC;AAAA,IAC/C,SAAS,CAAC,CAAG,MAAc,MAAM;AAAA,IACjC,UAAW,MAAc,MAAM,OAAO,OAAQ,MAAc,KAAK,OAAO;AAAA,EAC1E,CAAC;AAGD,MAAI,OAA4B,CAAC;AAGjC,MAAK,MAAc,MAAM,MAAM,MAAM;AACnC,WAAQ,MAAc,KAAK,KAAK;AAChC,YAAQ,IAAI,2CAAsC;AAAA,EACpD,WAEU,MAAc,MAAM,QAAQ,OAAQ,MAAc,KAAK,SAAS,UAAU;AAElF,UAAM,WAAW,OAAO,KAAM,MAAc,KAAK,IAAI,EAAE,CAAC;AACxD,QAAI,YAAY,aAAa,iCAAiC,aAAa,SAAS;AAClF,aAAQ,MAAc,KAAK;AAC3B,cAAQ,IAAI,+CAA0C;AAAA,IACxD;AAAA,EACF;AAEA,MAAI,OAAO,KAAK,IAAI,EAAE,WAAW,GAAG;AAClC,YAAQ,IAAI,0DAAqD;AAAA,MAC/D,aAAa,CAAC,CAAG,MAAc,MAAM;AAAA,MACrC,cAAe,MAAc,MAAM,OAAO,OAAO,KAAM,MAAc,KAAK,IAAI,IAAI;AAAA,MAClF,iBAAiB,CAAC,CAAG,MAAc,MAAM,MAAM;AAAA,MAC/C,mBAAoB,MAAc,MAAM,MAAM,OAAO,OAAO,KAAM,MAAc,KAAK,KAAK,IAAI,EAAE,SAAS;AAAA,IAC3G,CAAC;AAAA,EACH;AAEA,UAAQ,IAAI,+BAAwB,OAAO,KAAK,IAAI,EAAE,MAAM,YAAY;AAExE,aAAW,CAAC,OAAO,GAAG,KAAK,OAAO,QAAQ,IAAI,GAAG;AAC/C,UAAM,OAAO,IAAI,YAAY;AAE7B,QAAI,CAAC,QAAQ,OAAO,SAAS,UAAU;AACrC;AAAA,IACF;AAGA,UAAM,YAAY,SAAS,IAAI;AAI/B,UAAM,UAAU;AAAA,MACd;AAAA,MACA;AAAA,MACA;AAAA,QACE,SAAS,MAAM,OAAO;AAAA,QACtB,QAAQ,MAAM,OAAO;AAAA,QACrB,yBAAyB,MAAM,OAAO;AAAA,MACxC;AAAA,MACA,MAAM;AAAA,MACN,MAAM;AAAA,IACR;AAEA,QAAI,QAAQ,SAAS,GAAG;AAEtB,YAAM,WAAW,KAAK,IAAI,GAAG,QAAQ,IAAI,OAAK,EAAE,KAAK,CAAC;AAEtD,sBAAgB,KAAK;AAAA,QACnB,IAAI;AAAA,QACJ;AAAA,QACA,OAAO;AAAA,QACP,UAAU;AAAA,MACZ,CAAC;AAAA,IACH;AAAA,EACF;AAGA,kBAAgB,KAAK,CAAC,GAAG,MAAM,EAAE,QAAQ,EAAE,KAAK;AAGhD,MAAI,kBAAkB;AACtB,MAAI,MAAM,OAAO,2BAA2B,MAAM,OAAO,oBAAoB,GAAG;AAC9E,UAAM,YAAY,MAAM,OAAO;AAC/B,UAAM,cAAc,gBAAgB;AACpC,sBAAkB,gBAAgB,OAAO,OAAK,EAAE,SAAS,SAAS;AAClE,YAAQ,IAAI,uCAA2B,WAAW,WAAM,gBAAgB,MAAM,gBAAgB,SAAS,GAAG;AAAA,EAC5G;AAGA,QAAM,QAAQ,OAAO,SAAS,gBAAgB;AAC9C,QAAM,iBAAiB,gBAAgB,MAAM,GAAG,KAAK;AAGrD,QAAM,OAAO,eAAe,IAAI,YAAU;AAAA,IACxC,IAAI,MAAM;AAAA,IACV,OAAO,MAAM;AAAA,IACb,UAAU,MAAM;AAAA;AAAA,IAEhB,UAAU,MAAM;AAAA,EAClB,EAAE;AAEF,QAAM,UAAU,YAAY,IAAI,IAAI;AAEpC,UAAQ,IAAI,gBAAW,KAAK,MAAM,eAAe,QAAQ,QAAQ,CAAC,CAAC,cAAc,KAAK,GAAG;AAEzF,SAAO;AAAA,IACL,SAAS;AAAA,MACP,WAAW,GAAG,QAAQ,QAAQ,CAAC,CAAC;AAAA,MAChC,KAAK,KAAK,MAAM,UAAU,GAAO;AAAA;AAAA,IACnC;AAAA,IACA;AAAA,IACA,OAAO,KAAK;AAAA,EACd;AACF;AAKA,eAAe,yBACb,gBACqB;AACrB,MAAI;AACF,YAAQ,IAAI,0DAAmD;AAG/D,UAAM,EAAE,aAAa,IAAI,MAAM,OAAO,uBAAuB;AAE7D,UAAM,WAAW,aAAa,eAAe,KAAK,eAAe,UAAU;AAG3E,UAAM,EAAE,MAAM,MAAM,IAAI,MAAM,SAAS,IAAI,iBAAiB;AAE5D,YAAQ,IAAI,2CAAoC;AAAA,MAC9C,UAAU,CAAC,CAAC;AAAA,MACZ,cAAc,OAAO;AAAA,MACrB,SAAS,CAAC,CAAC;AAAA,MACX,UAAU,OAAO;AAAA,MACjB,UAAU,OAAO,OAAO,KAAK,IAAI,EAAE,SAAS;AAAA,IAC9C,CAAC;AAED,QAAI,OAAO;AACT,YAAM,IAAI,MAAM,mBAAmB,MAAM,OAAO,EAAE;AAAA,IACpD;AAEA,UAAM,aAAa,QAAQ,CAAC;AAC5B,YAAQ,IAAI,oBAAa,OAAO,KAAK,UAAU,EAAE,MAAM,gCAAgC;AAEvF,WAAO;AAAA,EACT,SAAS,OAAO;AACd,YAAQ,MAAM,iDAA4C,KAAK;AAC/D,UAAM;AAAA,EACR;AACF;AAKA,SAAS,6BACP,MACA,cACqB;AACrB,QAAM,KAAK,oBAAI,IAAoB;AAEnC,aAAW,OAAO,OAAO,OAAO,IAAI,GAAG;AACrC,UAAM,OAAO,IAAI,YAAY;AAE7B,QAAI,CAAC,QAAQ,OAAO,SAAS,UAAU;AACrC;AAAA,IACF;AAGA,UAAM,QAAQ,IAAI,IAAI,SAAS,IAAI,CAAC;AAGpC,eAAW,QAAQ,OAAO;AACxB,SAAG,IAAI,OAAO,GAAG,IAAI,IAAI,KAAK,KAAK,CAAC;AAAA,IACtC;AAAA,EACF;AAEA,SAAO;AACT;AAQA,SAAS,cAAc,MAAsB;AAC3C,SAAO,KACJ,YAAY,EACZ,UAAU,KAAK,EACf,QAAQ,oBAAoB,EAAE,EAE9B,QAAQ,gFAAgF,GAAG,EAC3F,QAAQ,6DAA6D,EAAE,EACvE,QAAQ,mBAAmB,GAAG,EAC9B,QAAQ,4BAA4B,GAAG,EACvC,QAAQ,QAAQ,GAAG,EACnB,KAAK;AACV;AAQA,SAAS,SAAS,MAAwB;AAExC,SAAO,cAAc,IAAI,EACtB,MAAM,KAAK,EACX,OAAO,WAAS,MAAM,SAAS,CAAC;AACrC","sourcesContent":["/**\n * Fuzzy matching utilities using bounded Levenshtein distance\n * \n * This is the same algorithm used by Orama's match-highlight plugin\n * for consistent fuzzy matching behavior.\n */\n\n/**\n * Result of bounded Levenshtein distance calculation\n */\nexport interface BoundedLevenshteinResult {\n /** Whether the distance is within bounds */\n isBounded: boolean;\n /** The actual distance (only valid if isBounded is true) */\n distance: number;\n}\n\n/**\n * Calculate bounded Levenshtein distance between two strings\n * \n * Stops early if distance exceeds the bound for better performance.\n * This is the same algorithm as Orama's internal boundedLevenshtein.\n * \n * @param a - First string\n * @param b - Second string\n * @param bound - Maximum allowed distance\n * @returns Result indicating if strings are within bound and the distance\n */\nexport function boundedLevenshtein(\n a: string,\n b: string,\n bound: number\n): BoundedLevenshteinResult {\n // Quick checks\n if (a === b) {\n return { isBounded: true, distance: 0 };\n }\n\n const aLen = a.length;\n const bLen = b.length;\n\n // If length difference exceeds bound, no need to calculate\n if (Math.abs(aLen - bLen) > bound) {\n return { isBounded: false, distance: bound + 1 };\n }\n\n // Swap to ensure a is shorter (optimization)\n if (aLen > bLen) {\n [a, b] = [b, a];\n }\n\n const m = a.length;\n const n = b.length;\n\n // Use single array instead of matrix (memory optimization)\n let prevRow = new Array(n + 1);\n let currRow = new Array(n + 1);\n\n // Initialize first row\n for (let j = 0; j <= n; j++) {\n prevRow[j] = j;\n }\n\n for (let i = 1; i <= m; i++) {\n currRow[0] = i;\n let minInRow = i;\n\n for (let j = 1; j <= n; j++) {\n const cost = a[i - 1] === b[j - 1] ? 0 : 1;\n\n currRow[j] = Math.min(\n prevRow[j] + 1, // deletion\n currRow[j - 1] + 1, // insertion\n prevRow[j - 1] + cost // substitution\n );\n\n minInRow = Math.min(minInRow, currRow[j]);\n }\n\n // Early termination: if all values in row exceed bound, we're done\n if (minInRow > bound) {\n return { isBounded: false, distance: bound + 1 };\n }\n\n // Swap rows for next iteration\n [prevRow, currRow] = [currRow, prevRow];\n }\n\n const distance = prevRow[n];\n return {\n isBounded: distance <= bound,\n distance\n };\n}\n\n/**\n * Check if a word matches a query token with fuzzy matching\n * \n * @param word - Word from document\n * @param queryToken - Token from search query\n * @param tolerance - Maximum edit distance allowed\n * @returns Match result with score\n */\nexport function fuzzyMatch(\n word: string,\n queryToken: string,\n tolerance: number\n): { matches: boolean; distance: number; score: number } {\n // Exact match\n if (word === queryToken) {\n return { matches: true, distance: 0, score: 1.0 };\n }\n\n // Prefix match (high score, no distance)\n if (word.startsWith(queryToken)) {\n return { matches: true, distance: 0, score: 0.95 };\n }\n\n // Fuzzy match with tolerance\n const result = boundedLevenshtein(word, queryToken, tolerance);\n \n if (result.isBounded) {\n // Score decreases with distance\n // distance 1 = 0.8, distance 2 = 0.6, etc.\n const score = 1.0 - (result.distance * 0.2);\n return {\n matches: true,\n distance: result.distance,\n score: Math.max(0.1, score) // Minimum score of 0.1\n };\n }\n\n return { matches: false, distance: tolerance + 1, score: 0 };\n}\n\n/**\n * Calculate adaptive tolerance based on query length\n * \n * Longer queries get higher tolerance for better fuzzy matching.\n * \n * @param queryTokens - Array of query tokens\n * @param baseTolerance - Base tolerance value\n * @returns Calculated tolerance (always an integer)\n */\nexport function calculateAdaptiveTolerance(\n queryTokens: string[],\n baseTolerance: number\n): number {\n const queryLength = queryTokens.length;\n \n if (queryLength <= 2) {\n return baseTolerance;\n } else if (queryLength <= 4) {\n return baseTolerance + 1;\n } else if (queryLength <= 6) {\n return baseTolerance + 2;\n } else {\n return baseTolerance + 3;\n }\n}\n","/**\n * Candidate expansion: Find all possible matches for query tokens\n * including exact matches, fuzzy matches, and synonyms\n */\n\nimport { fuzzyMatch } from './fuzzy.js';\nimport type { Candidate, SynonymMap } from './types.js';\n\n/**\n * Extract all unique words from the radix tree index\n * \n * @param radixNode - Root node of the radix tree\n * @returns Set of all unique words in the index\n */\nexport function extractVocabularyFromRadixTree(radixNode: any): Set<string> {\n const vocabulary = new Set<string>();\n let nodesVisited = 0;\n let wordsFound = 0;\n \n function traverse(node: any, depth: number = 0) {\n if (!node) {\n return;\n }\n \n nodesVisited++;\n \n // Check if this node represents a complete word\n // e = true means it's an end of a word\n if (node.e && node.w && typeof node.w === 'string' && node.w.length > 0) {\n vocabulary.add(node.w);\n wordsFound++;\n }\n \n // Children can be Map, Array, or Object\n if (node.c) {\n if (node.c instanceof Map) {\n // Map format\n for (const [_key, childNode] of node.c) {\n traverse(childNode, depth + 1);\n }\n } else if (Array.isArray(node.c)) {\n // Array format: [[key, childNode], ...]\n for (const [_key, childNode] of node.c) {\n traverse(childNode, depth + 1);\n }\n } else if (typeof node.c === 'object') {\n // Object format: {key: childNode, ...}\n for (const childNode of Object.values(node.c)) {\n traverse(childNode, depth + 1);\n }\n }\n }\n }\n \n traverse(radixNode);\n console.log(`๐Ÿ“š Extracted ${vocabulary.size} words from ${nodesVisited} nodes visited`);\n return vocabulary;\n}\n\n/**\n * Find all candidate matches for a single query token\n * \n * @param queryToken - Token from search query\n * @param vocabulary - Set of all words in the index\n * @param tolerance - Fuzzy matching tolerance\n * @param synonyms - Synonym map (optional)\n * @param synonymScore - Score multiplier for synonym matches\n * @returns Array of candidate matches\n */\nexport function findCandidatesForToken(\n queryToken: string,\n vocabulary: Set<string>,\n tolerance: number,\n synonyms?: SynonymMap,\n synonymScore: number = 0.8\n): Candidate[] {\n const candidates: Candidate[] = [];\n const seen = new Set<string>();\n\n // 1. Check for exact match\n if (vocabulary.has(queryToken)) {\n candidates.push({\n word: queryToken,\n type: 'exact',\n queryToken,\n distance: 0,\n score: 1.0\n });\n seen.add(queryToken);\n }\n\n // 2. Check for fuzzy matches\n for (const word of vocabulary) {\n if (seen.has(word)) continue;\n\n const match = fuzzyMatch(word, queryToken, tolerance);\n if (match.matches) {\n candidates.push({\n word,\n type: 'fuzzy',\n queryToken,\n distance: match.distance,\n score: match.score\n });\n seen.add(word);\n }\n }\n\n // 3. Check for synonym matches\n if (synonyms && synonyms[queryToken]) {\n for (const synonym of synonyms[queryToken]) {\n if (seen.has(synonym)) continue;\n if (vocabulary.has(synonym)) {\n candidates.push({\n word: synonym,\n type: 'synonym',\n queryToken,\n distance: 0,\n score: synonymScore\n });\n seen.add(synonym);\n }\n }\n }\n\n return candidates;\n}\n\n/**\n * Find candidates for all query tokens\n * \n * @param queryTokens - Array of tokens from search query\n * @param vocabulary - Set of all words in the index\n * @param tolerance - Fuzzy matching tolerance\n * @param synonyms - Synonym map (optional)\n * @param synonymScore - Score multiplier for synonym matches\n * @returns Map of query tokens to their candidate matches\n */\nexport function findAllCandidates(\n queryTokens: string[],\n vocabulary: Set<string>,\n tolerance: number,\n synonyms?: SynonymMap,\n synonymScore: number = 0.8\n): Map<string, Candidate[]> {\n const candidatesMap = new Map<string, Candidate[]>();\n\n for (const token of queryTokens) {\n const tokenCandidates = findCandidatesForToken(\n token,\n vocabulary,\n tolerance,\n synonyms,\n synonymScore\n );\n candidatesMap.set(token, tokenCandidates);\n }\n\n return candidatesMap;\n}\n\n/**\n * Get total number of candidates across all tokens\n * \n * @param candidatesMap - Map of token to candidates\n * @returns Total count of all candidates\n */\nexport function getTotalCandidateCount(\n candidatesMap: Map<string, Candidate[]>\n): number {\n let total = 0;\n for (const candidates of candidatesMap.values()) {\n total += candidates.length;\n }\n return total;\n}\n\n/**\n * Filter candidates by minimum score threshold\n * \n * @param candidatesMap - Map of token to candidates\n * @param minScore - Minimum score threshold\n * @returns Filtered candidates map\n */\nexport function filterCandidatesByScore(\n candidatesMap: Map<string, Candidate[]>,\n minScore: number\n): Map<string, Candidate[]> {\n const filtered = new Map<string, Candidate[]>();\n\n for (const [token, candidates] of candidatesMap.entries()) {\n const filteredCandidates = candidates.filter(c => c.score >= minScore);\n if (filteredCandidates.length > 0) {\n filtered.set(token, filteredCandidates);\n }\n }\n\n return filtered;\n}\n","/**\n * Phrase scoring algorithm with semantic weighting\n */\n\nimport type { WordMatch, PhraseMatch, Candidate, GapWord } from './types.js';\n\n/**\n * Configuration for phrase scoring\n */\nexport interface ScoringConfig {\n weights: {\n exact: number;\n fuzzy: number;\n order: number;\n proximity: number;\n density: number;\n semantic: number;\n };\n maxGap: number;\n /** \n * Multiplier for proximity window calculation.\n * proximityWindow = queryTokens.length ร— proximitySpanMultiplier\n */\n proximitySpanMultiplier: number;\n}\n\n/**\n * Find all phrase matches in a document\n * \n * @param documentTokens - Tokenized document content (needed to extract gap words)\n * @param candidatesMap - Map of query tokens to their candidates\n * @param config - Scoring configuration\n * @param documentFrequency - Document frequency map for TF-IDF\n * @param totalDocuments - Total number of documents\n * @returns Array of phrase matches\n */\nexport function findPhrasesInDocument(\n documentTokens: string[],\n candidatesMap: Map<string, Candidate[]>,\n config: ScoringConfig,\n documentFrequency: Map<string, number>,\n totalDocuments: number\n): PhraseMatch[] {\n const phrases: PhraseMatch[] = [];\n const queryTokens = Array.from(candidatesMap.keys());\n\n // Find all word matches in document\n const wordMatches: WordMatch[] = [];\n \n for (let i = 0; i < documentTokens.length; i++) {\n const docWord = documentTokens[i];\n \n // Check if this word matches any query token\n for (const [queryToken, candidates] of candidatesMap.entries()) {\n for (const candidate of candidates) {\n if (candidate.word === docWord) {\n wordMatches.push({\n word: docWord,\n queryToken,\n position: i,\n type: candidate.type,\n distance: candidate.distance,\n score: candidate.score\n });\n }\n }\n }\n }\n\n // Build phrases from word matches using sliding window\n for (let i = 0; i < wordMatches.length; i++) {\n const phrase = buildPhraseFromPosition(\n wordMatches,\n i,\n queryTokens,\n config,\n documentFrequency,\n totalDocuments,\n wordMatches,\n documentTokens // Pass document tokens to extract gap words\n );\n \n if (phrase && phrase.words.length > 0) {\n phrases.push(phrase);\n }\n }\n\n // Deduplicate and sort by score\n return deduplicatePhrases(phrases);\n}\n\n/**\n * Build a phrase starting from a specific word match position\n * \n * @param wordMatches - All word matches in document\n * @param startIndex - Starting index in wordMatches array\n * @param queryTokens - Original query tokens\n * @param config - Scoring configuration\n * @param documentFrequency - Document frequency map\n * @param totalDocuments - Total document count\n * @param allWordMatches - All word matches in document (for density calculation)\n * @param documentTokens - Original document tokens (for gap word extraction)\n * @returns Phrase match or null\n */\nfunction buildPhraseFromPosition(\n wordMatches: WordMatch[],\n startIndex: number,\n queryTokens: string[],\n config: ScoringConfig,\n documentFrequency: Map<string, number>,\n totalDocuments: number,\n allWordMatches: WordMatch[],\n documentTokens: string[]\n): PhraseMatch | null {\n const startMatch = wordMatches[startIndex];\n const phraseWords: WordMatch[] = [startMatch];\n const coveredTokens = new Set([startMatch.queryToken]);\n const gapWords: GapWord[] = [];\n let totalGapUsed = 0;\n\n // Look for nearby matches to complete the phrase\n for (let i = startIndex + 1; i < wordMatches.length; i++) {\n const match = wordMatches[i];\n const lastPos = phraseWords[phraseWords.length - 1].position;\n const gap = match.position - lastPos - 1;\n\n // Stop if gap exceeds maximum\n if (gap > config.maxGap) {\n break;\n }\n\n // Only process if it's a different query token we need\n if (!coveredTokens.has(match.queryToken)) {\n // Track gap words between last match and current match\n for (let pos = lastPos + 1; pos < match.position; pos++) {\n totalGapUsed++;\n gapWords.push({\n word: documentTokens[pos],\n position: pos,\n gapIndex: totalGapUsed\n });\n }\n\n phraseWords.push(match);\n coveredTokens.add(match.queryToken);\n\n // Stop if we have all query tokens\n if (coveredTokens.size === queryTokens.length) {\n break;\n }\n }\n }\n\n // Calculate phrase score\n if (phraseWords.length > 0) {\n const coverage = phraseWords.length / queryTokens.length;\n const span = phraseWords[phraseWords.length - 1].position - phraseWords[0].position + 1;\n \n const { score, breakdown } = calculatePhraseScore(\n phraseWords,\n queryTokens,\n config,\n documentFrequency,\n totalDocuments,\n allWordMatches,\n coverage\n );\n\n return {\n words: phraseWords,\n gapWords,\n gapUsed: totalGapUsed,\n coverage,\n startPosition: phraseWords[0].position,\n endPosition: phraseWords[phraseWords.length - 1].position,\n span,\n inOrder: isInOrder(phraseWords, queryTokens),\n score,\n scoreBreakdown: breakdown\n };\n }\n\n return null;\n}\n\n/**\n * Calculate overall phrase score\n * \n * @param phraseWords - Words in the phrase\n * @param queryTokens - Original query tokens\n * @param config - Scoring configuration\n * @param documentFrequency - Document frequency map\n * @param totalDocuments - Total document count\n * @param allWordMatches - All word matches in document (for density calculation)\n * @param coverage - Pre-calculated coverage ratio (phraseWords.length / queryTokens.length)\n * @returns Phrase score (0-1) and detailed component breakdown\n */\nfunction calculatePhraseScore(\n phraseWords: WordMatch[],\n queryTokens: string[],\n config: ScoringConfig,\n documentFrequency: Map<string, number>,\n totalDocuments: number,\n allWordMatches: WordMatch[],\n coverage: number\n): { score: number; breakdown: { base: number; order: number; proximity: number; density: number; semantic: number; coverage: number } } {\n // Base score from word matches\n // Each word contributes: matchScore ร— typeWeight\n let baseScore = 0;\n for (const word of phraseWords) {\n const weight = word.type === 'exact' ? config.weights.exact :\n word.type === 'fuzzy' ? config.weights.fuzzy : \n config.weights.fuzzy * 0.8; // synonym gets 80% of fuzzy weight\n baseScore += word.score * weight;\n }\n baseScore /= phraseWords.length;\n\n // Order bonus: 1.0 if words appear in query order, 0.5 otherwise\n const inOrder = isInOrder(phraseWords, queryTokens);\n const orderScore = inOrder ? 1.0 : 0.5;\n\n // Proximity bonus (closer words score higher)\n // Short-circuit: skip if maxGap=0, proximity weight is 0, or single-word query (proximity meaningless)\n let proximityScore = 0;\n if (config.maxGap > 0 && config.weights.proximity > 0 && queryTokens.length > 1) {\n const span = phraseWords[phraseWords.length - 1].position - phraseWords[0].position + 1;\n const proximityWindow = queryTokens.length * config.proximitySpanMultiplier;\n proximityScore = Math.max(0, 1.0 - (span / proximityWindow));\n }\n\n // Density: Only applies to single-word queries (measures word repetition in document)\n // For multi-word phrase queries, density is 0 (coverage handles completeness separately)\n let densityScore = 0;\n \n if (queryTokens.length === 1) {\n // Single-word query: reward repetition\n const totalOccurrences = allWordMatches.length;\n // Cap at reasonable maximum to avoid runaway scores\n densityScore = Math.min(1.0, totalOccurrences / 10);\n }\n // For multi-word queries: densityScore stays 0\n // Coverage is applied as a multiplier at the end instead\n\n // Semantic score (TF-IDF based)\n const semanticScore = calculateSemanticScore(\n phraseWords,\n documentFrequency,\n totalDocuments\n );\n\n // Weighted combination\n const weights = config.weights;\n \n // Calculate weighted components\n const weightedBase = baseScore;\n const weightedOrder = orderScore * weights.order;\n const weightedProximity = proximityScore * weights.proximity;\n const weightedDensity = densityScore * weights.density;\n const weightedSemantic = semanticScore * weights.semantic;\n \n const totalScore = weightedBase + weightedOrder + weightedProximity + weightedDensity + weightedSemantic;\n\n // Calculate max possible score\n // FIX: Use actual max base weight (highest of exact/fuzzy) instead of hardcoded 1.0\n const maxBaseWeight = Math.max(weights.exact, weights.fuzzy);\n // Only include proximity in max if it can actually contribute (avoids penalizing scores when maxGap=0 or single-word)\n const effectiveProximityWeight = (config.maxGap > 0 && weights.proximity > 0 && queryTokens.length > 1) ? weights.proximity : 0;\n const maxPossibleScore = maxBaseWeight + weights.order + effectiveProximityWeight + weights.density + weights.semantic;\n \n // Normalize to 0-1 range\n const normalizedScore = totalScore / maxPossibleScore;\n \n // FIX: Apply coverage as a MULTIPLIER for multi-word queries\n // This ensures incomplete matches (2/3) can never outscore complete matches (3/3)\n const coverageMultiplier = queryTokens.length > 1 ? coverage : 1.0;\n const score = normalizedScore * coverageMultiplier;\n\n // Component contributions to the final normalized score (before coverage multiplier)\n const base = weightedBase / maxPossibleScore;\n const order = weightedOrder / maxPossibleScore;\n const proximity = weightedProximity / maxPossibleScore;\n const density = weightedDensity / maxPossibleScore;\n const semantic = weightedSemantic / maxPossibleScore;\n\n return {\n score,\n breakdown: {\n base,\n order,\n proximity,\n density,\n semantic,\n coverage: coverageMultiplier // Show coverage multiplier in breakdown\n }\n };\n}\n\n/**\n * Check if words are in the same order as query tokens\n * \n * @param phraseWords - Words in the phrase\n * @param queryTokens - Original query tokens\n * @returns True if in order\n */\nfunction isInOrder(phraseWords: WordMatch[], queryTokens: string[]): boolean {\n const tokenOrder = new Map(queryTokens.map((token, index) => [token, index]));\n \n for (let i = 1; i < phraseWords.length; i++) {\n const prevOrder = tokenOrder.get(phraseWords[i - 1].queryToken) ?? -1;\n const currOrder = tokenOrder.get(phraseWords[i].queryToken) ?? -1;\n \n if (currOrder < prevOrder) {\n return false;\n }\n }\n \n return true;\n}\n\n/**\n * Calculate semantic score using TF-IDF\n * \n * @param phraseWords - Words in the phrase\n * @param documentFrequency - Document frequency map\n * @param totalDocuments - Total document count\n * @returns Semantic score (0-1)\n */\nfunction calculateSemanticScore(\n phraseWords: WordMatch[],\n documentFrequency: Map<string, number>,\n totalDocuments: number\n): number {\n // Handle edge case: no documents\n if (totalDocuments === 0) {\n return 0;\n }\n \n let tfidfSum = 0;\n \n for (const word of phraseWords) {\n const df = documentFrequency.get(word.word) || 1;\n const idf = Math.log(totalDocuments / df);\n tfidfSum += idf;\n }\n \n // Normalize by phrase length\n const avgTfidf = tfidfSum / phraseWords.length;\n \n // Normalize to 0-1 range (assuming max IDF of ~10)\n return Math.min(1.0, avgTfidf / 10);\n}\n\n/**\n * Deduplicate overlapping phrases, keeping highest scoring ones\n * \n * @param phrases - Array of phrase matches\n * @returns Deduplicated phrases sorted by score\n */\nfunction deduplicatePhrases(phrases: PhraseMatch[]): PhraseMatch[] {\n if (phrases.length === 0) return [];\n\n // Sort by score descending\n const sorted = phrases.slice().sort((a, b) => b.score - a.score);\n const result: PhraseMatch[] = [];\n const covered = new Set<number>();\n\n for (const phrase of sorted) {\n // Check if this phrase overlaps with already selected phrases\n let overlaps = false;\n for (let pos = phrase.startPosition; pos <= phrase.endPosition; pos++) {\n if (covered.has(pos)) {\n overlaps = true;\n break;\n }\n }\n\n if (!overlaps) {\n result.push(phrase);\n // Mark positions as covered\n for (let pos = phrase.startPosition; pos <= phrase.endPosition; pos++) {\n covered.add(pos);\n }\n }\n }\n\n return result.sort((a, b) => b.score - a.score);\n}\n","/**\n * Fuzzy Phrase Plugin for Orama\n * \n * Advanced fuzzy phrase matching with semantic weighting and synonym expansion.\n * Completely independent from QPS - accesses Orama's radix tree directly.\n */\n\nimport type { AnyOrama, OramaPlugin, Results, TypedDocument } from '@wcs-colab/orama';\nimport type { FuzzyPhraseConfig, PluginState, SynonymMap, DocumentMatch } from './types.js';\nimport { calculateAdaptiveTolerance } from './fuzzy.js';\nimport { \n extractVocabularyFromRadixTree, \n findAllCandidates,\n filterCandidatesByScore \n} from './candidates.js';\nimport { findPhrasesInDocument } from './scoring.js';\n\n/**\n * Default configuration\n */\nconst DEFAULT_CONFIG: Required<FuzzyPhraseConfig> = {\n textProperty: 'normalized_content', // Must match server's field name\n tolerance: 1,\n adaptiveTolerance: true,\n enableSynonyms: false,\n supabase: undefined as any,\n synonymMatchScore: 0.8,\n weights: {\n exact: 1.0,\n fuzzy: 0.8,\n order: 0.3,\n proximity: 0.2,\n density: 0.2,\n semantic: 0.15\n },\n maxGap: 5,\n minScore: 0.1,\n enableFinalScoreMinimum: false,\n finalScoreMinimum: 0.3,\n proximitySpanMultiplier: 5\n};\n\n/**\n * Plugin state storage (keyed by Orama instance)\n */\nconst pluginStates = new WeakMap<AnyOrama, PluginState>();\n\n/**\n * Create the Fuzzy Phrase Plugin\n * \n * @param userConfig - User configuration options\n * @returns Orama plugin instance\n */\nexport function pluginFuzzyPhrase(userConfig: FuzzyPhraseConfig = {}): OramaPlugin {\n // Merge user config with defaults\n const config: Required<FuzzyPhraseConfig> = {\n textProperty: userConfig.textProperty ?? DEFAULT_CONFIG.textProperty,\n tolerance: userConfig.tolerance ?? DEFAULT_CONFIG.tolerance,\n adaptiveTolerance: userConfig.adaptiveTolerance ?? DEFAULT_CONFIG.adaptiveTolerance,\n enableSynonyms: userConfig.enableSynonyms ?? DEFAULT_CONFIG.enableSynonyms,\n supabase: userConfig.supabase || DEFAULT_CONFIG.supabase,\n synonymMatchScore: userConfig.synonymMatchScore ?? DEFAULT_CONFIG.synonymMatchScore,\n weights: {\n exact: userConfig.weights?.exact ?? DEFAULT_CONFIG.weights.exact,\n fuzzy: userConfig.weights?.fuzzy ?? DEFAULT_CONFIG.weights.fuzzy,\n order: userConfig.weights?.order ?? DEFAULT_CONFIG.weights.order,\n proximity: userConfig.weights?.proximity ?? DEFAULT_CONFIG.weights.proximity,\n density: userConfig.weights?.density ?? DEFAULT_CONFIG.weights.density,\n semantic: userConfig.weights?.semantic ?? DEFAULT_CONFIG.weights.semantic\n },\n maxGap: userConfig.maxGap ?? DEFAULT_CONFIG.maxGap,\n minScore: userConfig.minScore ?? DEFAULT_CONFIG.minScore,\n enableFinalScoreMinimum: userConfig.enableFinalScoreMinimum ?? DEFAULT_CONFIG.enableFinalScoreMinimum,\n finalScoreMinimum: userConfig.finalScoreMinimum ?? DEFAULT_CONFIG.finalScoreMinimum,\n proximitySpanMultiplier: userConfig.proximitySpanMultiplier ?? DEFAULT_CONFIG.proximitySpanMultiplier\n };\n\n const plugin: OramaPlugin = {\n name: 'fuzzy-phrase',\n\n /**\n * Initialize plugin after index is created\n */\n afterCreate: async (orama: AnyOrama) => {\n console.log('๐Ÿ”ฎ Initializing Fuzzy Phrase Plugin...');\n\n // Initialize state\n const state: PluginState = {\n synonymMap: {},\n config,\n documentFrequency: new Map(),\n totalDocuments: 0\n };\n\n // Load synonyms from Supabase if enabled\n if (config.enableSynonyms && config.supabase) {\n try {\n console.log('๐Ÿ“– Loading synonyms from Supabase...');\n state.synonymMap = await loadSynonymsFromSupabase(config.supabase);\n console.log(`โœ… Loaded ${Object.keys(state.synonymMap).length} words with synonyms`);\n } catch (error) {\n console.error('โš ๏ธ Failed to load synonyms:', error);\n // Continue without synonyms\n }\n }\n\n // Calculate document frequencies for TF-IDF from document store\n const docs = (orama.data as any)?.docs?.docs;\n if (docs) {\n state.totalDocuments = Object.keys(docs).length;\n state.documentFrequency = calculateDocumentFrequencies(docs, config.textProperty);\n console.log(`๐Ÿ“Š Calculated document frequencies for ${state.totalDocuments} documents`);\n }\n\n // Store state\n pluginStates.set(orama, state);\n console.log('โœ… Fuzzy Phrase Plugin initialized');\n \n // Signal ready - emit a custom event that can be listened to\n // Use setImmediate to ensure this runs after the afterCreate hook completes\n setImmediate(() => {\n if (typeof (globalThis as any).fuzzyPhrasePluginReady === 'function') {\n console.log('๐Ÿ“ก Signaling plugin ready...');\n (globalThis as any).fuzzyPhrasePluginReady();\n } else {\n console.warn('โš ๏ธ fuzzyPhrasePluginReady callback not found');\n }\n });\n }\n };\n\n return plugin;\n}\n\n/**\n * Search with fuzzy phrase matching\n * \n * This function should be called instead of the regular search() function\n * to enable fuzzy phrase matching.\n */\nexport async function searchWithFuzzyPhrase<T extends AnyOrama>(\n orama: T, \n params: { term?: string; properties?: string[]; limit?: number },\n language?: string\n): Promise<Results<TypedDocument<T>>> {\n const startTime = performance.now();\n \n // Get plugin state\n const state = pluginStates.get(orama);\n \n if (!state) {\n console.error('โŒ Plugin state not initialized');\n throw new Error('Fuzzy Phrase Plugin not properly initialized');\n }\n\n const { term, properties } = params;\n \n if (!term || typeof term !== 'string') {\n return { elapsed: { formatted: '0ms', raw: 0 }, hits: [], count: 0 };\n }\n\n // Use specified property or default\n const textProperty = (properties && properties[0]) || state.config.textProperty;\n\n // Tokenize query\n const queryTokens = tokenize(term);\n \n if (queryTokens.length === 0) {\n return { elapsed: { formatted: '0ms', raw: 0 }, hits: [], count: 0 };\n }\n\n // Calculate tolerance (adaptive or fixed)\n const tolerance = state.config.adaptiveTolerance\n ? calculateAdaptiveTolerance(queryTokens, state.config.tolerance)\n : state.config.tolerance;\n\n console.log(`๐Ÿ” Fuzzy phrase search: \"${term}\" (${queryTokens.length} tokens, tolerance: ${tolerance})`);\n\n // Extract vocabulary from radix tree\n let vocabulary: Set<string>;\n \n try {\n // Access radix tree - the actual index data is in orama.data.index, not orama.index\n // orama.index is just the component interface (methods)\n const indexData = (orama as any).data?.index;\n \n if (!indexData) {\n console.error('โŒ No index data found in orama.data.index');\n return { elapsed: { formatted: '0ms', raw: 0 }, hits: [], count: 0 };\n }\n \n console.log('๐Ÿ” DEBUG: Index data keys:', Object.keys(indexData || {}));\n \n // Try different paths to find the radix tree\n let radixNode = null;\n \n // Path 1: QPS-style (orama.data.index.indexes[property].node)\n if (indexData.indexes?.[textProperty]?.node) {\n radixNode = indexData.indexes[textProperty].node;\n console.log('โœ… Found radix via QPS-style path (data.index.indexes)');\n }\n // Path 2: Standard Orama (orama.data.index[property].node)\n else if (indexData[textProperty]?.node) {\n radixNode = indexData[textProperty].node;\n console.log('โœ… Found radix via standard path (data.index[property])');\n }\n \n if (!radixNode) {\n console.error('โŒ Radix tree not found for property:', textProperty);\n console.error(' Available properties in index:', Object.keys(indexData));\n return { elapsed: { formatted: '0ms', raw: 0 }, hits: [], count: 0 };\n }\n\n vocabulary = extractVocabularyFromRadixTree(radixNode);\n console.log(`๐Ÿ“š Extracted ${vocabulary.size} unique words from index`);\n } catch (error) {\n console.error('โŒ Failed to extract vocabulary:', error);\n return { elapsed: { formatted: '0ms', raw: 0 }, hits: [], count: 0 };\n }\n\n // Find candidates for all query tokens\n const candidatesMap = findAllCandidates(\n queryTokens,\n vocabulary,\n tolerance,\n state.config.enableSynonyms ? state.synonymMap : undefined,\n state.config.synonymMatchScore\n );\n\n // Filter by minimum score\n const filteredCandidates = filterCandidatesByScore(\n candidatesMap,\n state.config.minScore\n );\n\n console.log(`๐ŸŽฏ Found candidates: ${Array.from(filteredCandidates.values()).reduce((sum, c) => sum + c.length, 0)} total`);\n\n // Search through all documents\n const documentMatches: DocumentMatch[] = [];\n \n console.log('๐Ÿ” DEBUG orama.data structure:', {\n dataKeys: Object.keys((orama as any).data || {}),\n hasDocs: !!((orama as any).data?.docs),\n docsType: (orama as any).data?.docs ? typeof (orama as any).data.docs : 'undefined'\n });\n \n // Try multiple possible document storage locations\n let docs: Record<string, any> = {};\n \n // Access the actual documents - they're nested in orama.data.docs.docs\n if ((orama as any).data?.docs?.docs) {\n docs = (orama as any).data.docs.docs;\n console.log('โœ… Found docs at orama.data.docs.docs');\n }\n // Fallback: orama.data.docs (might be the correct structure in some cases)\n else if ((orama as any).data?.docs && typeof (orama as any).data.docs === 'object') {\n // Check if it has document-like properties (not sharedInternalDocumentStore, etc.)\n const firstKey = Object.keys((orama as any).data.docs)[0];\n if (firstKey && firstKey !== 'sharedInternalDocumentStore' && firstKey !== 'count') {\n docs = (orama as any).data.docs;\n console.log('โœ… Found docs at orama.data.docs (direct)');\n }\n }\n \n if (Object.keys(docs).length === 0) {\n console.log('โŒ Could not find documents - available structure:', {\n hasDataDocs: !!((orama as any).data?.docs),\n dataDocsKeys: (orama as any).data?.docs ? Object.keys((orama as any).data.docs) : 'none',\n hasDataDocsDocs: !!((orama as any).data?.docs?.docs),\n dataDocsDocsCount: (orama as any).data?.docs?.docs ? Object.keys((orama as any).data.docs.docs).length : 0\n });\n }\n \n console.log(`๐Ÿ“„ Searching through ${Object.keys(docs).length} documents`);\n\n for (const [docId, doc] of Object.entries(docs)) {\n const text = doc[textProperty];\n \n if (!text || typeof text !== 'string') {\n continue;\n }\n\n // Tokenize document\n const docTokens = tokenize(text);\n\n // Find phrases in this document\n // Note: state.config.weights is guaranteed to have all properties from default merge\n const phrases = findPhrasesInDocument(\n docTokens,\n filteredCandidates,\n {\n weights: state.config.weights as { exact: number; fuzzy: number; order: number; proximity: number; density: number; semantic: number },\n maxGap: state.config.maxGap,\n proximitySpanMultiplier: state.config.proximitySpanMultiplier\n },\n state.documentFrequency,\n state.totalDocuments\n );\n\n if (phrases.length > 0) {\n // Calculate overall document score (highest phrase score)\n const docScore = Math.max(...phrases.map(p => p.score));\n\n documentMatches.push({\n id: docId,\n phrases,\n score: docScore,\n document: doc\n });\n }\n }\n\n // Sort by score descending\n documentMatches.sort((a, b) => b.score - a.score);\n\n // Apply final score minimum filter if enabled\n let filteredMatches = documentMatches;\n if (state.config.enableFinalScoreMinimum && state.config.finalScoreMinimum > 0) {\n const threshold = state.config.finalScoreMinimum;\n const beforeCount = filteredMatches.length;\n filteredMatches = filteredMatches.filter(m => m.score >= threshold);\n console.log(`๐ŸŽš๏ธ Final score filter: ${beforeCount} โ†’ ${filteredMatches.length} (threshold: ${threshold})`);\n }\n\n // Apply limit if specified\n const limit = params.limit ?? filteredMatches.length;\n const limitedMatches = filteredMatches.slice(0, limit);\n\n // Convert to Orama results format\n const hits = limitedMatches.map(match => ({\n id: match.id,\n score: match.score,\n document: match.document,\n // Store phrases for highlighting\n _phrases: match.phrases\n })) as any[];\n\n const elapsed = performance.now() - startTime;\n\n console.log(`โœ… Found ${hits.length} results in ${elapsed.toFixed(2)}ms (limit: ${limit})`);\n\n return {\n elapsed: {\n formatted: `${elapsed.toFixed(2)}ms`,\n raw: Math.floor(elapsed * 1000000) // nanoseconds\n },\n hits,\n count: hits.length\n } as any;\n}\n\n/**\n * Load synonyms from Supabase\n */\nasync function loadSynonymsFromSupabase(\n supabaseConfig: { url: string; serviceKey: string }\n): Promise<SynonymMap> {\n try {\n console.log('๐Ÿ” DEBUG: Calling Supabase RPC get_synonym_map...');\n \n // Dynamic import to avoid bundling Supabase client if not needed\n const { createClient } = await import('@supabase/supabase-js');\n \n const supabase = createClient(supabaseConfig.url, supabaseConfig.serviceKey);\n \n // Call the get_synonym_map function\n const { data, error } = await supabase.rpc('get_synonym_map');\n \n console.log('๐Ÿ” DEBUG: Supabase RPC response:', {\n hasError: !!error,\n errorMessage: error?.message,\n hasData: !!data,\n dataType: typeof data,\n dataKeys: data ? Object.keys(data).length : 0\n });\n \n if (error) {\n throw new Error(`Supabase error: ${error.message}`);\n }\n \n const synonymMap = data || {};\n console.log(`๐Ÿ“š Loaded ${Object.keys(synonymMap).length} synonym entries from Supabase`);\n \n return synonymMap;\n } catch (error) {\n console.error('โŒ Failed to load synonyms from Supabase:', error);\n throw error;\n }\n}\n\n/**\n * Calculate document frequencies for TF-IDF\n */\nfunction calculateDocumentFrequencies(\n docs: Record<string, any>,\n textProperty: string\n): Map<string, number> {\n const df = new Map<string, number>();\n\n for (const doc of Object.values(docs)) {\n const text = doc[textProperty];\n \n if (!text || typeof text !== 'string') {\n continue;\n }\n\n // Get unique words in this document\n const words = new Set(tokenize(text));\n\n // Increment document frequency for each unique word\n for (const word of words) {\n df.set(word, (df.get(word) || 0) + 1);\n }\n }\n\n return df;\n}\n\n/**\n * Normalize text using the same rules as server-side\n * \n * CRITICAL: This must match the normalizeText() function in server/index.js exactly\n * PLUS we remove all punctuation to match Orama's French tokenizer behavior\n */\nfunction normalizeText(text: string): string {\n return text\n .toLowerCase()\n .normalize('NFD')\n .replace(/[\\u0300-\\u036f]/g, '') // Remove diacritics\n // Replace French elisions (l', d', etc.) with space to preserve word boundaries\n .replace(/\\b[ldcjmnst][\\u2018\\u2019\\u201A\\u201B\\u2032\\u2035\\u0027\\u0060\\u00B4](?=\\w)/gi, ' ')\n .replace(/[\\u2018\\u2019\\u201A\\u201B\\u2032\\u2035\\u0027\\u0060\\u00B4]/g, '') // Remove remaining apostrophes\n .replace(/[\\u201c\\u201d]/g, '\"') // Normalize curly quotes to straight quotes\n .replace(/[.,;:!?()[\\]{}\\-โ€”โ€“ยซยป\"\"]/g, ' ') // Remove punctuation (replace with space to preserve word boundaries)\n .replace(/\\s+/g, ' ') // Normalize multiple spaces to single space\n .trim();\n}\n\n/**\n * Tokenization matching normalized text behavior\n * \n * Note: Text should already be normalized before indexing, so we normalize again\n * to ensure plugin tokenization matches index tokenization\n */\nfunction tokenize(text: string): string[] {\n // Normalize first (same as indexing), then split by whitespace\n return normalizeText(text)\n .split(/\\s+/)\n .filter(token => token.length > 0);\n}\n\n/**\n * Export types for external use\n */\nexport type {\n FuzzyPhraseConfig,\n WordMatch,\n PhraseMatch,\n DocumentMatch,\n SynonymMap,\n Candidate\n} from './types.js';\n"]}
1
+ {"version":3,"sources":["../src/fuzzy.ts","../src/candidates.ts","../src/scoring.ts","../src/index.ts"],"names":[],"mappings":";AA4BO,SAAS,mBACd,GACA,GACA,OAC0B;AAE1B,MAAI,MAAM,GAAG;AACX,WAAO,EAAE,WAAW,MAAM,UAAU,EAAE;AAAA,EACxC;AAEA,QAAM,OAAO,EAAE;AACf,QAAM,OAAO,EAAE;AAGf,MAAI,KAAK,IAAI,OAAO,IAAI,IAAI,OAAO;AACjC,WAAO,EAAE,WAAW,OAAO,UAAU,QAAQ,EAAE;AAAA,EACjD;AAGA,MAAI,OAAO,MAAM;AACf,KAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC;AAAA,EAChB;AAEA,QAAM,IAAI,EAAE;AACZ,QAAM,IAAI,EAAE;AAGZ,MAAI,UAAU,IAAI,MAAM,IAAI,CAAC;AAC7B,MAAI,UAAU,IAAI,MAAM,IAAI,CAAC;AAG7B,WAAS,IAAI,GAAG,KAAK,GAAG,KAAK;AAC3B,YAAQ,CAAC,IAAI;AAAA,EACf;AAEA,WAAS,IAAI,GAAG,KAAK,GAAG,KAAK;AAC3B,YAAQ,CAAC,IAAI;AACb,QAAI,WAAW;AAEf,aAAS,IAAI,GAAG,KAAK,GAAG,KAAK;AAC3B,YAAM,OAAO,EAAE,IAAI,CAAC,MAAM,EAAE,IAAI,CAAC,IAAI,IAAI;AAEzC,cAAQ,CAAC,IAAI,KAAK;AAAA,QAChB,QAAQ,CAAC,IAAI;AAAA;AAAA,QACb,QAAQ,IAAI,CAAC,IAAI;AAAA;AAAA,QACjB,QAAQ,IAAI,CAAC,IAAI;AAAA;AAAA,MACnB;AAEA,iBAAW,KAAK,IAAI,UAAU,QAAQ,CAAC,CAAC;AAAA,IAC1C;AAGA,QAAI,WAAW,OAAO;AACpB,aAAO,EAAE,WAAW,OAAO,UAAU,QAAQ,EAAE;AAAA,IACjD;AAGA,KAAC,SAAS,OAAO,IAAI,CAAC,SAAS,OAAO;AAAA,EACxC;AAEA,QAAM,WAAW,QAAQ,CAAC;AAC1B,SAAO;AAAA,IACL,WAAW,YAAY;AAAA,IACvB;AAAA,EACF;AACF;AAUO,SAAS,WACd,MACA,YACA,WACuD;AAEvD,MAAI,SAAS,YAAY;AACvB,WAAO,EAAE,SAAS,MAAM,UAAU,GAAG,OAAO,EAAI;AAAA,EAClD;AAGA,MAAI,KAAK,WAAW,UAAU,GAAG;AAC/B,WAAO,EAAE,SAAS,MAAM,UAAU,GAAG,OAAO,KAAK;AAAA,EACnD;AAGA,QAAM,SAAS,mBAAmB,MAAM,YAAY,SAAS;AAE7D,MAAI,OAAO,WAAW;AAGpB,UAAM,QAAQ,IAAO,OAAO,WAAW;AACvC,WAAO;AAAA,MACL,SAAS;AAAA,MACT,UAAU,OAAO;AAAA,MACjB,OAAO,KAAK,IAAI,KAAK,KAAK;AAAA;AAAA,IAC5B;AAAA,EACF;AAEA,SAAO,EAAE,SAAS,OAAO,UAAU,YAAY,GAAG,OAAO,EAAE;AAC7D;AAWO,SAAS,2BACd,aACA,eACQ;AACR,QAAM,cAAc,YAAY;AAEhC,MAAI,eAAe,GAAG;AACpB,WAAO;AAAA,EACT,WAAW,eAAe,GAAG;AAC3B,WAAO,gBAAgB;AAAA,EACzB,WAAW,eAAe,GAAG;AAC3B,WAAO,gBAAgB;AAAA,EACzB,OAAO;AACL,WAAO,gBAAgB;AAAA,EACzB;AACF;;;ACjJO,SAAS,+BAA+B,WAA6B;AAC1E,QAAM,aAAa,oBAAI,IAAY;AACnC,MAAI,eAAe;AACnB,MAAI,aAAa;AAEjB,WAAS,SAAS,MAAW,QAAgB,GAAG;AAC9C,QAAI,CAAC,MAAM;AACT;AAAA,IACF;AAEA;AAIA,QAAI,KAAK,KAAK,KAAK,KAAK,OAAO,KAAK,MAAM,YAAY,KAAK,EAAE,SAAS,GAAG;AACvE,iBAAW,IAAI,KAAK,CAAC;AACrB;AAAA,IACF;AAGA,QAAI,KAAK,GAAG;AACV,UAAI,KAAK,aAAa,KAAK;AAEzB,mBAAW,CAAC,MAAM,SAAS,KAAK,KAAK,GAAG;AACtC,mBAAS,WAAW,QAAQ,CAAC;AAAA,QAC/B;AAAA,MACF,WAAW,MAAM,QAAQ,KAAK,CAAC,GAAG;AAEhC,mBAAW,CAAC,MAAM,SAAS,KAAK,KAAK,GAAG;AACtC,mBAAS,WAAW,QAAQ,CAAC;AAAA,QAC/B;AAAA,MACF,WAAW,OAAO,KAAK,MAAM,UAAU;AAErC,mBAAW,aAAa,OAAO,OAAO,KAAK,CAAC,GAAG;AAC7C,mBAAS,WAAW,QAAQ,CAAC;AAAA,QAC/B;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAEA,WAAS,SAAS;AAClB,UAAQ,IAAI,uBAAgB,WAAW,IAAI,eAAe,YAAY,gBAAgB;AACtF,SAAO;AACT;AAYO,SAAS,uBACd,YACA,YACA,WACA,UACA,eAAuB,KACV;AACb,QAAM,aAA0B,CAAC;AACjC,QAAM,OAAO,oBAAI,IAAY;AAG7B,MAAI,WAAW,IAAI,UAAU,GAAG;AAC9B,eAAW,KAAK;AAAA,MACd,MAAM;AAAA,MACN,MAAM;AAAA,MACN;AAAA,MACA,UAAU;AAAA,MACV,OAAO;AAAA,IACT,CAAC;AACD,SAAK,IAAI,UAAU;AAAA,EACrB;AAGA,aAAW,QAAQ,YAAY;AAC7B,QAAI,KAAK,IAAI,IAAI;AAAG;AAEpB,UAAM,QAAQ,WAAW,MAAM,YAAY,SAAS;AACpD,QAAI,MAAM,SAAS;AACjB,iBAAW,KAAK;AAAA,QACd;AAAA,QACA,MAAM;AAAA,QACN;AAAA,QACA,UAAU,MAAM;AAAA,QAChB,OAAO,MAAM;AAAA,MACf,CAAC;AACD,WAAK,IAAI,IAAI;AAAA,IACf;AAAA,EACF;AAGA,MAAI,YAAY,SAAS,UAAU,GAAG;AACpC,eAAW,WAAW,SAAS,UAAU,GAAG;AAC1C,UAAI,KAAK,IAAI,OAAO;AAAG;AACvB,UAAI,WAAW,IAAI,OAAO,GAAG;AAC3B,mBAAW,KAAK;AAAA,UACd,MAAM;AAAA,UACN,MAAM;AAAA,UACN;AAAA,UACA,UAAU;AAAA,UACV,OAAO;AAAA,QACT,CAAC;AACD,aAAK,IAAI,OAAO;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAEA,SAAO;AACT;AAYO,SAAS,kBACd,aACA,YACA,WACA,UACA,eAAuB,KACG;AAC1B,QAAM,gBAAgB,oBAAI,IAAyB;AAEnD,aAAW,SAAS,aAAa;AAC/B,UAAM,kBAAkB;AAAA,MACtB;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AACA,kBAAc,IAAI,OAAO,eAAe;AAAA,EAC1C;AAEA,SAAO;AACT;AAyBO,SAAS,wBACd,eACA,UAC0B;AAC1B,QAAM,WAAW,oBAAI,IAAyB;AAE9C,aAAW,CAAC,OAAO,UAAU,KAAK,cAAc,QAAQ,GAAG;AACzD,UAAM,qBAAqB,WAAW,OAAO,OAAK,EAAE,SAAS,QAAQ;AACrE,QAAI,mBAAmB,SAAS,GAAG;AACjC,eAAS,IAAI,OAAO,kBAAkB;AAAA,IACxC;AAAA,EACF;AAEA,SAAO;AACT;;;AC9JO,SAAS,sBACd,gBACA,eACA,QACA,mBACA,gBACe;AACf,QAAM,UAAyB,CAAC;AAChC,QAAM,cAAc,MAAM,KAAK,cAAc,KAAK,CAAC;AAGnD,QAAM,cAA2B,CAAC;AAElC,WAAS,IAAI,GAAG,IAAI,eAAe,QAAQ,KAAK;AAC9C,UAAM,UAAU,eAAe,CAAC;AAGhC,eAAW,CAAC,YAAY,UAAU,KAAK,cAAc,QAAQ,GAAG;AAC9D,iBAAW,aAAa,YAAY;AAClC,YAAI,UAAU,SAAS,SAAS;AAC9B,sBAAY,KAAK;AAAA,YACf,MAAM;AAAA,YACN;AAAA,YACA,UAAU;AAAA,YACV,MAAM,UAAU;AAAA,YAChB,UAAU,UAAU;AAAA,YACpB,OAAO,UAAU;AAAA,UACnB,CAAC;AAAA,QACH;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAGA,WAAS,IAAI,GAAG,IAAI,YAAY,QAAQ,KAAK;AAC3C,UAAM,SAAS;AAAA,MACb;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA;AAAA,IACF;AAEA,QAAI,UAAU,OAAO,MAAM,SAAS,GAAG;AACrC,cAAQ,KAAK,MAAM;AAAA,IACrB;AAAA,EACF;AAGA,SAAO,mBAAmB,OAAO;AACnC;AAeA,SAAS,wBACP,aACA,YACA,aACA,QACA,mBACA,gBACA,gBACA,gBACoB;AACpB,QAAM,aAAa,YAAY,UAAU;AACzC,QAAM,cAA2B,CAAC,UAAU;AAC5C,QAAM,gBAAgB,oBAAI,IAAI,CAAC,WAAW,UAAU,CAAC;AACrD,QAAM,WAAsB,CAAC;AAC7B,MAAI,eAAe;AAGnB,WAAS,IAAI,aAAa,GAAG,IAAI,YAAY,QAAQ,KAAK;AACxD,UAAM,QAAQ,YAAY,CAAC;AAC3B,UAAM,UAAU,YAAY,YAAY,SAAS,CAAC,EAAE;AACpD,UAAM,MAAM,MAAM,WAAW,UAAU;AAGvC,QAAI,MAAM,OAAO,QAAQ;AACvB;AAAA,IACF;AAGA,QAAI,CAAC,cAAc,IAAI,MAAM,UAAU,GAAG;AAExC,eAAS,MAAM,UAAU,GAAG,MAAM,MAAM,UAAU,OAAO;AACvD;AACA,iBAAS,KAAK;AAAA,UACZ,MAAM,eAAe,GAAG;AAAA,UACxB,UAAU;AAAA,UACV,UAAU;AAAA,QACZ,CAAC;AAAA,MACH;AAEA,kBAAY,KAAK,KAAK;AACtB,oBAAc,IAAI,MAAM,UAAU;AAGlC,UAAI,cAAc,SAAS,YAAY,QAAQ;AAC7C;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAGA,MAAI,YAAY,SAAS,GAAG;AAC1B,UAAM,WAAW,YAAY,SAAS,YAAY;AAClD,UAAM,OAAO,YAAY,YAAY,SAAS,CAAC,EAAE,WAAW,YAAY,CAAC,EAAE,WAAW;AAEtF,UAAM,EAAE,OAAO,UAAU,IAAI;AAAA,MAC3B;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAEA,WAAO;AAAA,MACL,OAAO;AAAA,MACP;AAAA,MACA,SAAS;AAAA,MACT;AAAA,MACA,eAAe,YAAY,CAAC,EAAE;AAAA,MAC9B,aAAa,YAAY,YAAY,SAAS,CAAC,EAAE;AAAA,MACjD;AAAA,MACA,SAAS,UAAU,aAAa,WAAW;AAAA,MAC3C;AAAA,MACA,gBAAgB;AAAA,IAClB;AAAA,EACF;AAEA,SAAO;AACT;AAcA,SAAS,qBACP,aACA,aACA,QACA,mBACA,gBACA,gBACA,UACuI;AAGvI,MAAI,YAAY;AAChB,aAAW,QAAQ,aAAa;AAC9B,UAAM,SAAS,KAAK,SAAS,UAAU,OAAO,QAAQ,QACvC,KAAK,SAAS,UAAU,OAAO,QAAQ,QACvC,OAAO,QAAQ,QAAQ;AACtC,iBAAa,KAAK,QAAQ;AAAA,EAC5B;AACA,eAAa,YAAY;AAGzB,QAAM,UAAU,UAAU,aAAa,WAAW;AAClD,QAAM,aAAa,UAAU,IAAM;AAInC,MAAI,iBAAiB;AACrB,MAAI,OAAO,SAAS,KAAK,OAAO,QAAQ,YAAY,KAAK,YAAY,SAAS,GAAG;AAC/E,UAAM,OAAO,YAAY,YAAY,SAAS,CAAC,EAAE,WAAW,YAAY,CAAC,EAAE,WAAW;AACtF,UAAM,kBAAkB,YAAY,SAAS,OAAO;AACpD,qBAAiB,KAAK,IAAI,GAAG,IAAO,OAAO,eAAgB;AAAA,EAC7D;AAIA,MAAI,eAAe;AAEnB,MAAI,YAAY,WAAW,GAAG;AAE5B,UAAM,mBAAmB,eAAe;AAExC,mBAAe,KAAK,IAAI,GAAK,mBAAmB,EAAE;AAAA,EACpD;AAKA,QAAM,gBAAgB;AAAA,IACpB;AAAA,IACA;AAAA,IACA;AAAA,EACF;AAGA,QAAM,UAAU,OAAO;AAGvB,QAAM,eAAe;AACrB,QAAM,gBAAgB,aAAa,QAAQ;AAC3C,QAAM,oBAAoB,iBAAiB,QAAQ;AACnD,QAAM,kBAAkB,eAAe,QAAQ;AAC/C,QAAM,mBAAmB,gBAAgB,QAAQ;AAEjD,QAAM,aAAa,eAAe,gBAAgB,oBAAoB,kBAAkB;AAKxF,QAAM,sBAAsB,OAAO,YAAY,KAAK,QAAQ,QAAQ;AACpE,QAAM,gBAAgB,sBAAsB,KAAK,IAAI,QAAQ,OAAO,QAAQ,KAAK,IAAI,QAAQ;AAE7F,QAAM,2BAA4B,OAAO,SAAS,KAAK,QAAQ,YAAY,KAAK,YAAY,SAAS,IAAK,QAAQ,YAAY;AAC9H,QAAM,mBAAmB,gBAAgB,QAAQ,QAAQ,2BAA2B,QAAQ,UAAU,QAAQ;AAG9G,QAAM,kBAAkB,aAAa;AAIrC,QAAM,qBAAqB,YAAY,SAAS,IAAI,WAAW;AAC/D,QAAM,QAAQ,kBAAkB;AAGhC,QAAM,OAAO,eAAe;AAC5B,QAAM,QAAQ,gBAAgB;AAC9B,QAAM,YAAY,oBAAoB;AACtC,QAAM,UAAU,kBAAkB;AAClC,QAAM,WAAW,mBAAmB;AAEpC,SAAO;AAAA,IACL;AAAA,IACA,WAAW;AAAA,MACT;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,UAAU;AAAA;AAAA,IACZ;AAAA,EACF;AACF;AASA,SAAS,UAAU,aAA0B,aAAgC;AAC3E,QAAM,aAAa,IAAI,IAAI,YAAY,IAAI,CAAC,OAAO,UAAU,CAAC,OAAO,KAAK,CAAC,CAAC;AAE5E,WAAS,IAAI,GAAG,IAAI,YAAY,QAAQ,KAAK;AAC3C,UAAM,YAAY,WAAW,IAAI,YAAY,IAAI,CAAC,EAAE,UAAU,KAAK;AACnE,UAAM,YAAY,WAAW,IAAI,YAAY,CAAC,EAAE,UAAU,KAAK;AAE/D,QAAI,YAAY,WAAW;AACzB,aAAO;AAAA,IACT;AAAA,EACF;AAEA,SAAO;AACT;AAUA,SAAS,uBACP,aACA,mBACA,gBACQ;AAER,MAAI,mBAAmB,GAAG;AACxB,WAAO;AAAA,EACT;AAEA,MAAI,WAAW;AAEf,aAAW,QAAQ,aAAa;AAC9B,UAAM,KAAK,kBAAkB,IAAI,KAAK,IAAI,KAAK;AAC/C,UAAM,MAAM,KAAK,IAAI,iBAAiB,EAAE;AACxC,gBAAY;AAAA,EACd;AAGA,QAAM,WAAW,WAAW,YAAY;AAGxC,SAAO,KAAK,IAAI,GAAK,WAAW,EAAE;AACpC;AAQA,SAAS,mBAAmB,SAAuC;AACjE,MAAI,QAAQ,WAAW;AAAG,WAAO,CAAC;AAGlC,QAAM,SAAS,QAAQ,MAAM,EAAE,KAAK,CAAC,GAAG,MAAM,EAAE,QAAQ,EAAE,KAAK;AAC/D,QAAM,SAAwB,CAAC;AAC/B,QAAM,UAAU,oBAAI,IAAY;AAEhC,aAAW,UAAU,QAAQ;AAE3B,QAAI,WAAW;AACf,aAAS,MAAM,OAAO,eAAe,OAAO,OAAO,aAAa,OAAO;AACrE,UAAI,QAAQ,IAAI,GAAG,GAAG;AACpB,mBAAW;AACX;AAAA,MACF;AAAA,IACF;AAEA,QAAI,CAAC,UAAU;AACb,aAAO,KAAK,MAAM;AAElB,eAAS,MAAM,OAAO,eAAe,OAAO,OAAO,aAAa,OAAO;AACrE,gBAAQ,IAAI,GAAG;AAAA,MACjB;AAAA,IACF;AAAA,EACF;AAEA,SAAO,OAAO,KAAK,CAAC,GAAG,MAAM,EAAE,QAAQ,EAAE,KAAK;AAChD;;;ACpXA,IAAM,iBAA8C;AAAA,EAClD,cAAc;AAAA;AAAA,EACd,WAAW;AAAA,EACX,mBAAmB;AAAA,EACnB,gBAAgB;AAAA,EAChB,UAAU;AAAA,EACV,mBAAmB;AAAA,EACnB,SAAS;AAAA,IACP,OAAO;AAAA,IACP,OAAO;AAAA,IACP,OAAO;AAAA,IACP,WAAW;AAAA,IACX,SAAS;AAAA,IACT,UAAU;AAAA,EACZ;AAAA,EACA,QAAQ;AAAA,EACR,UAAU;AAAA,EACV,yBAAyB;AAAA,EACzB,mBAAmB;AAAA,EACnB,yBAAyB;AAC3B;AAKA,IAAM,eAAe,oBAAI,QAA+B;AAQjD,SAAS,kBAAkB,aAAgC,CAAC,GAAgB;AAEjF,QAAM,SAAsC;AAAA,IAC1C,cAAc,WAAW,gBAAgB,eAAe;AAAA,IACxD,WAAW,WAAW,aAAa,eAAe;AAAA,IAClD,mBAAmB,WAAW,qBAAqB,eAAe;AAAA,IAClE,gBAAgB,WAAW,kBAAkB,eAAe;AAAA,IAC5D,UAAU,WAAW,YAAY,eAAe;AAAA,IAChD,mBAAmB,WAAW,qBAAqB,eAAe;AAAA,IAClE,SAAS;AAAA,MACP,OAAO,WAAW,SAAS,SAAS,eAAe,QAAQ;AAAA,MAC3D,OAAO,WAAW,SAAS,SAAS,eAAe,QAAQ;AAAA,MAC3D,OAAO,WAAW,SAAS,SAAS,eAAe,QAAQ;AAAA,MAC3D,WAAW,WAAW,SAAS,aAAa,eAAe,QAAQ;AAAA,MACnE,SAAS,WAAW,SAAS,WAAW,eAAe,QAAQ;AAAA,MAC/D,UAAU,WAAW,SAAS,YAAY,eAAe,QAAQ;AAAA,IACnE;AAAA,IACA,QAAQ,WAAW,UAAU,eAAe;AAAA,IAC5C,UAAU,WAAW,YAAY,eAAe;AAAA,IAChD,yBAAyB,WAAW,2BAA2B,eAAe;AAAA,IAC9E,mBAAmB,WAAW,qBAAqB,eAAe;AAAA,IAClE,yBAAyB,WAAW,2BAA2B,eAAe;AAAA,EAChF;AAEA,QAAM,SAAsB;AAAA,IAC1B,MAAM;AAAA;AAAA;AAAA;AAAA,IAKN,aAAa,OAAO,UAAoB;AACtC,cAAQ,IAAI,+CAAwC;AAGpD,YAAM,QAAqB;AAAA,QACzB,YAAY,CAAC;AAAA,QACb;AAAA,QACA,mBAAmB,oBAAI,IAAI;AAAA,QAC3B,gBAAgB;AAAA,MAClB;AAGA,UAAI,OAAO,kBAAkB,OAAO,UAAU;AAC5C,YAAI;AACF,kBAAQ,IAAI,6CAAsC;AAClD,gBAAM,aAAa,MAAM,yBAAyB,OAAO,QAAQ;AACjE,kBAAQ,IAAI,iBAAY,OAAO,KAAK,MAAM,UAAU,EAAE,MAAM,sBAAsB;AAAA,QACpF,SAAS,OAAO;AACd,kBAAQ,MAAM,0CAAgC,KAAK;AAAA,QAErD;AAAA,MACF;AAGA,YAAM,OAAQ,MAAM,MAAc,MAAM;AACxC,UAAI,MAAM;AACR,cAAM,iBAAiB,OAAO,KAAK,IAAI,EAAE;AACzC,cAAM,oBAAoB,6BAA6B,MAAM,OAAO,YAAY;AAChF,gBAAQ,IAAI,iDAA0C,MAAM,cAAc,YAAY;AAAA,MACxF;AAGA,mBAAa,IAAI,OAAO,KAAK;AAC7B,cAAQ,IAAI,wCAAmC;AAI/C,mBAAa,MAAM;AACjB,YAAI,OAAQ,WAAmB,2BAA2B,YAAY;AACpE,kBAAQ,IAAI,qCAA8B;AAC1C,UAAC,WAAmB,uBAAuB;AAAA,QAC7C,OAAO;AACL,kBAAQ,KAAK,yDAA+C;AAAA,QAC9D;AAAA,MACF,CAAC;AAAA,IACH;AAAA,EACF;AAEA,SAAO;AACT;AAQA,eAAsB,sBACpB,OACA,QACA,UACoC;AACpC,QAAM,YAAY,YAAY,IAAI;AAGlC,QAAM,QAAQ,aAAa,IAAI,KAAK;AAEpC,MAAI,CAAC,OAAO;AACV,YAAQ,MAAM,qCAAgC;AAC9C,UAAM,IAAI,MAAM,8CAA8C;AAAA,EAChE;AAEA,QAAM,EAAE,MAAM,WAAW,IAAI;AAE7B,MAAI,CAAC,QAAQ,OAAO,SAAS,UAAU;AACrC,WAAO,EAAE,SAAS,EAAE,WAAW,OAAO,KAAK,EAAE,GAAG,MAAM,CAAC,GAAG,OAAO,EAAE;AAAA,EACrE;AAGA,QAAM,eAAgB,cAAc,WAAW,CAAC,KAAM,MAAM,OAAO;AAGnE,QAAM,cAAc,SAAS,IAAI;AAEjC,MAAI,YAAY,WAAW,GAAG;AAC5B,WAAO,EAAE,SAAS,EAAE,WAAW,OAAO,KAAK,EAAE,GAAG,MAAM,CAAC,GAAG,OAAO,EAAE;AAAA,EACrE;AAGA,QAAM,YAAY,MAAM,OAAO,oBAC3B,2BAA2B,aAAa,MAAM,OAAO,SAAS,IAC9D,MAAM,OAAO;AAEjB,UAAQ,IAAI,mCAA4B,IAAI,MAAM,YAAY,MAAM,uBAAuB,SAAS,GAAG;AAGvG,MAAI;AAEJ,MAAI;AAGF,UAAM,YAAa,MAAc,MAAM;AAEvC,QAAI,CAAC,WAAW;AACd,cAAQ,MAAM,gDAA2C;AACzD,aAAO,EAAE,SAAS,EAAE,WAAW,OAAO,KAAK,EAAE,GAAG,MAAM,CAAC,GAAG,OAAO,EAAE;AAAA,IACrE;AAEA,YAAQ,IAAI,qCAA8B,OAAO,KAAK,aAAa,CAAC,CAAC,CAAC;AAGtE,QAAI,YAAY;AAGhB,QAAI,UAAU,UAAU,YAAY,GAAG,MAAM;AAC3C,kBAAY,UAAU,QAAQ,YAAY,EAAE;AAC5C,cAAQ,IAAI,4DAAuD;AAAA,IACrE,WAES,UAAU,YAAY,GAAG,MAAM;AACtC,kBAAY,UAAU,YAAY,EAAE;AACpC,cAAQ,IAAI,6DAAwD;AAAA,IACtE;AAEA,QAAI,CAAC,WAAW;AACd,cAAQ,MAAM,6CAAwC,YAAY;AAClE,cAAQ,MAAM,qCAAqC,OAAO,KAAK,SAAS,CAAC;AACzE,aAAO,EAAE,SAAS,EAAE,WAAW,OAAO,KAAK,EAAE,GAAG,MAAM,CAAC,GAAG,OAAO,EAAE;AAAA,IACrE;AAEA,iBAAa,+BAA+B,SAAS;AACrD,YAAQ,IAAI,uBAAgB,WAAW,IAAI,0BAA0B;AAAA,EACvE,SAAS,OAAO;AACd,YAAQ,MAAM,wCAAmC,KAAK;AACtD,WAAO,EAAE,SAAS,EAAE,WAAW,OAAO,KAAK,EAAE,GAAG,MAAM,CAAC,GAAG,OAAO,EAAE;AAAA,EACrE;AAGA,QAAM,gBAAgB;AAAA,IACpB;AAAA,IACA;AAAA,IACA;AAAA,IACA,MAAM,OAAO,iBAAiB,MAAM,aAAa;AAAA,IACjD,MAAM,OAAO;AAAA,EACf;AAGA,QAAM,qBAAqB,cAAc,IACrC,gBACA,wBAAwB,eAAe,MAAM,OAAO,QAAQ;AAEhE,UAAQ,IAAI,+BAAwB,MAAM,KAAK,mBAAmB,OAAO,CAAC,EAAE,OAAO,CAAC,KAAK,MAAM,MAAM,EAAE,QAAQ,CAAC,CAAC,QAAQ;AAGzH,QAAM,kBAAmC,CAAC;AAE1C,UAAQ,IAAI,yCAAkC;AAAA,IAC5C,UAAU,OAAO,KAAM,MAAc,QAAQ,CAAC,CAAC;AAAA,IAC/C,SAAS,CAAC,CAAG,MAAc,MAAM;AAAA,IACjC,UAAW,MAAc,MAAM,OAAO,OAAQ,MAAc,KAAK,OAAO;AAAA,EAC1E,CAAC;AAGD,MAAI,OAA4B,CAAC;AAGjC,MAAK,MAAc,MAAM,MAAM,MAAM;AACnC,WAAQ,MAAc,KAAK,KAAK;AAChC,YAAQ,IAAI,2CAAsC;AAAA,EACpD,WAEU,MAAc,MAAM,QAAQ,OAAQ,MAAc,KAAK,SAAS,UAAU;AAElF,UAAM,WAAW,OAAO,KAAM,MAAc,KAAK,IAAI,EAAE,CAAC;AACxD,QAAI,YAAY,aAAa,iCAAiC,aAAa,SAAS;AAClF,aAAQ,MAAc,KAAK;AAC3B,cAAQ,IAAI,+CAA0C;AAAA,IACxD;AAAA,EACF;AAEA,MAAI,OAAO,KAAK,IAAI,EAAE,WAAW,GAAG;AAClC,YAAQ,IAAI,0DAAqD;AAAA,MAC/D,aAAa,CAAC,CAAG,MAAc,MAAM;AAAA,MACrC,cAAe,MAAc,MAAM,OAAO,OAAO,KAAM,MAAc,KAAK,IAAI,IAAI;AAAA,MAClF,iBAAiB,CAAC,CAAG,MAAc,MAAM,MAAM;AAAA,MAC/C,mBAAoB,MAAc,MAAM,MAAM,OAAO,OAAO,KAAM,MAAc,KAAK,KAAK,IAAI,EAAE,SAAS;AAAA,IAC3G,CAAC;AAAA,EACH;AAEA,UAAQ,IAAI,+BAAwB,OAAO,KAAK,IAAI,EAAE,MAAM,YAAY;AAExE,aAAW,CAAC,OAAO,GAAG,KAAK,OAAO,QAAQ,IAAI,GAAG;AAC/C,UAAM,OAAO,IAAI,YAAY;AAE7B,QAAI,CAAC,QAAQ,OAAO,SAAS,UAAU;AACrC;AAAA,IACF;AAGA,UAAM,YAAY,SAAS,IAAI;AAI/B,UAAM,UAAU;AAAA,MACd;AAAA,MACA;AAAA,MACA;AAAA,QACE,SAAS,MAAM,OAAO;AAAA,QACtB,QAAQ,MAAM,OAAO;AAAA,QACrB,yBAAyB,MAAM,OAAO;AAAA,QACtC;AAAA,MACF;AAAA,MACA,MAAM;AAAA,MACN,MAAM;AAAA,IACR;AAEA,QAAI,QAAQ,SAAS,GAAG;AAEtB,YAAM,WAAW,KAAK,IAAI,GAAG,QAAQ,IAAI,OAAK,EAAE,KAAK,CAAC;AAEtD,sBAAgB,KAAK;AAAA,QACnB,IAAI;AAAA,QACJ;AAAA,QACA,OAAO;AAAA,QACP,UAAU;AAAA,MACZ,CAAC;AAAA,IACH;AAAA,EACF;AAGA,kBAAgB,KAAK,CAAC,GAAG,MAAM,EAAE,QAAQ,EAAE,KAAK;AAGhD,MAAI,kBAAkB;AACtB,MAAI,MAAM,OAAO,2BAA2B,MAAM,OAAO,oBAAoB,GAAG;AAC9E,UAAM,YAAY,MAAM,OAAO;AAC/B,UAAM,cAAc,gBAAgB;AACpC,sBAAkB,gBAAgB,OAAO,OAAK,EAAE,SAAS,SAAS;AAClE,YAAQ,IAAI,uCAA2B,WAAW,WAAM,gBAAgB,MAAM,gBAAgB,SAAS,GAAG;AAAA,EAC5G;AAGA,QAAM,QAAQ,OAAO,SAAS,gBAAgB;AAC9C,QAAM,iBAAiB,gBAAgB,MAAM,GAAG,KAAK;AAGrD,QAAM,OAAO,eAAe,IAAI,YAAU;AAAA,IACxC,IAAI,MAAM;AAAA,IACV,OAAO,MAAM;AAAA,IACb,UAAU,MAAM;AAAA;AAAA,IAEhB,UAAU,MAAM;AAAA,EAClB,EAAE;AAEF,QAAM,UAAU,YAAY,IAAI,IAAI;AAEpC,UAAQ,IAAI,gBAAW,KAAK,MAAM,eAAe,QAAQ,QAAQ,CAAC,CAAC,cAAc,KAAK,GAAG;AAEzF,SAAO;AAAA,IACL,SAAS;AAAA,MACP,WAAW,GAAG,QAAQ,QAAQ,CAAC,CAAC;AAAA,MAChC,KAAK,KAAK,MAAM,UAAU,GAAO;AAAA;AAAA,IACnC;AAAA,IACA;AAAA,IACA,OAAO,KAAK;AAAA,EACd;AACF;AAKA,eAAe,yBACb,gBACqB;AACrB,MAAI;AACF,YAAQ,IAAI,0DAAmD;AAG/D,UAAM,EAAE,aAAa,IAAI,MAAM,OAAO,uBAAuB;AAE7D,UAAM,WAAW,aAAa,eAAe,KAAK,eAAe,UAAU;AAG3E,UAAM,EAAE,MAAM,MAAM,IAAI,MAAM,SAAS,IAAI,iBAAiB;AAE5D,YAAQ,IAAI,2CAAoC;AAAA,MAC9C,UAAU,CAAC,CAAC;AAAA,MACZ,cAAc,OAAO;AAAA,MACrB,SAAS,CAAC,CAAC;AAAA,MACX,UAAU,OAAO;AAAA,MACjB,UAAU,OAAO,OAAO,KAAK,IAAI,EAAE,SAAS;AAAA,IAC9C,CAAC;AAED,QAAI,OAAO;AACT,YAAM,IAAI,MAAM,mBAAmB,MAAM,OAAO,EAAE;AAAA,IACpD;AAEA,UAAM,aAAa,QAAQ,CAAC;AAC5B,YAAQ,IAAI,oBAAa,OAAO,KAAK,UAAU,EAAE,MAAM,gCAAgC;AAEvF,WAAO;AAAA,EACT,SAAS,OAAO;AACd,YAAQ,MAAM,iDAA4C,KAAK;AAC/D,UAAM;AAAA,EACR;AACF;AAKA,SAAS,6BACP,MACA,cACqB;AACrB,QAAM,KAAK,oBAAI,IAAoB;AAEnC,aAAW,OAAO,OAAO,OAAO,IAAI,GAAG;AACrC,UAAM,OAAO,IAAI,YAAY;AAE7B,QAAI,CAAC,QAAQ,OAAO,SAAS,UAAU;AACrC;AAAA,IACF;AAGA,UAAM,QAAQ,IAAI,IAAI,SAAS,IAAI,CAAC;AAGpC,eAAW,QAAQ,OAAO;AACxB,SAAG,IAAI,OAAO,GAAG,IAAI,IAAI,KAAK,KAAK,CAAC;AAAA,IACtC;AAAA,EACF;AAEA,SAAO;AACT;AAQA,SAAS,cAAc,MAAsB;AAC3C,SAAO,KACJ,YAAY,EACZ,UAAU,KAAK,EACf,QAAQ,oBAAoB,EAAE,EAE9B,QAAQ,gFAAgF,GAAG,EAC3F,QAAQ,6DAA6D,EAAE,EACvE,QAAQ,mBAAmB,GAAG,EAC9B,QAAQ,4BAA4B,GAAG,EACvC,QAAQ,QAAQ,GAAG,EACnB,KAAK;AACV;AAQA,SAAS,SAAS,MAAwB;AAExC,SAAO,cAAc,IAAI,EACtB,MAAM,KAAK,EACX,OAAO,WAAS,MAAM,SAAS,CAAC;AACrC","sourcesContent":["/**\n * Fuzzy matching utilities using bounded Levenshtein distance\n * \n * This is the same algorithm used by Orama's match-highlight plugin\n * for consistent fuzzy matching behavior.\n */\n\n/**\n * Result of bounded Levenshtein distance calculation\n */\nexport interface BoundedLevenshteinResult {\n /** Whether the distance is within bounds */\n isBounded: boolean;\n /** The actual distance (only valid if isBounded is true) */\n distance: number;\n}\n\n/**\n * Calculate bounded Levenshtein distance between two strings\n * \n * Stops early if distance exceeds the bound for better performance.\n * This is the same algorithm as Orama's internal boundedLevenshtein.\n * \n * @param a - First string\n * @param b - Second string\n * @param bound - Maximum allowed distance\n * @returns Result indicating if strings are within bound and the distance\n */\nexport function boundedLevenshtein(\n a: string,\n b: string,\n bound: number\n): BoundedLevenshteinResult {\n // Quick checks\n if (a === b) {\n return { isBounded: true, distance: 0 };\n }\n\n const aLen = a.length;\n const bLen = b.length;\n\n // If length difference exceeds bound, no need to calculate\n if (Math.abs(aLen - bLen) > bound) {\n return { isBounded: false, distance: bound + 1 };\n }\n\n // Swap to ensure a is shorter (optimization)\n if (aLen > bLen) {\n [a, b] = [b, a];\n }\n\n const m = a.length;\n const n = b.length;\n\n // Use single array instead of matrix (memory optimization)\n let prevRow = new Array(n + 1);\n let currRow = new Array(n + 1);\n\n // Initialize first row\n for (let j = 0; j <= n; j++) {\n prevRow[j] = j;\n }\n\n for (let i = 1; i <= m; i++) {\n currRow[0] = i;\n let minInRow = i;\n\n for (let j = 1; j <= n; j++) {\n const cost = a[i - 1] === b[j - 1] ? 0 : 1;\n\n currRow[j] = Math.min(\n prevRow[j] + 1, // deletion\n currRow[j - 1] + 1, // insertion\n prevRow[j - 1] + cost // substitution\n );\n\n minInRow = Math.min(minInRow, currRow[j]);\n }\n\n // Early termination: if all values in row exceed bound, we're done\n if (minInRow > bound) {\n return { isBounded: false, distance: bound + 1 };\n }\n\n // Swap rows for next iteration\n [prevRow, currRow] = [currRow, prevRow];\n }\n\n const distance = prevRow[n];\n return {\n isBounded: distance <= bound,\n distance\n };\n}\n\n/**\n * Check if a word matches a query token with fuzzy matching\n * \n * @param word - Word from document\n * @param queryToken - Token from search query\n * @param tolerance - Maximum edit distance allowed\n * @returns Match result with score\n */\nexport function fuzzyMatch(\n word: string,\n queryToken: string,\n tolerance: number\n): { matches: boolean; distance: number; score: number } {\n // Exact match\n if (word === queryToken) {\n return { matches: true, distance: 0, score: 1.0 };\n }\n\n // Prefix match (high score, no distance)\n if (word.startsWith(queryToken)) {\n return { matches: true, distance: 0, score: 0.95 };\n }\n\n // Fuzzy match with tolerance\n const result = boundedLevenshtein(word, queryToken, tolerance);\n \n if (result.isBounded) {\n // Score decreases with distance\n // distance 1 = 0.8, distance 2 = 0.6, etc.\n const score = 1.0 - (result.distance * 0.2);\n return {\n matches: true,\n distance: result.distance,\n score: Math.max(0.1, score) // Minimum score of 0.1\n };\n }\n\n return { matches: false, distance: tolerance + 1, score: 0 };\n}\n\n/**\n * Calculate adaptive tolerance based on query length\n * \n * Longer queries get higher tolerance for better fuzzy matching.\n * \n * @param queryTokens - Array of query tokens\n * @param baseTolerance - Base tolerance value\n * @returns Calculated tolerance (always an integer)\n */\nexport function calculateAdaptiveTolerance(\n queryTokens: string[],\n baseTolerance: number\n): number {\n const queryLength = queryTokens.length;\n \n if (queryLength <= 2) {\n return baseTolerance;\n } else if (queryLength <= 4) {\n return baseTolerance + 1;\n } else if (queryLength <= 6) {\n return baseTolerance + 2;\n } else {\n return baseTolerance + 3;\n }\n}\n","/**\n * Candidate expansion: Find all possible matches for query tokens\n * including exact matches, fuzzy matches, and synonyms\n */\n\nimport { fuzzyMatch } from './fuzzy.js';\nimport type { Candidate, SynonymMap } from './types.js';\n\n/**\n * Extract all unique words from the radix tree index\n * \n * @param radixNode - Root node of the radix tree\n * @returns Set of all unique words in the index\n */\nexport function extractVocabularyFromRadixTree(radixNode: any): Set<string> {\n const vocabulary = new Set<string>();\n let nodesVisited = 0;\n let wordsFound = 0;\n \n function traverse(node: any, depth: number = 0) {\n if (!node) {\n return;\n }\n \n nodesVisited++;\n \n // Check if this node represents a complete word\n // e = true means it's an end of a word\n if (node.e && node.w && typeof node.w === 'string' && node.w.length > 0) {\n vocabulary.add(node.w);\n wordsFound++;\n }\n \n // Children can be Map, Array, or Object\n if (node.c) {\n if (node.c instanceof Map) {\n // Map format\n for (const [_key, childNode] of node.c) {\n traverse(childNode, depth + 1);\n }\n } else if (Array.isArray(node.c)) {\n // Array format: [[key, childNode], ...]\n for (const [_key, childNode] of node.c) {\n traverse(childNode, depth + 1);\n }\n } else if (typeof node.c === 'object') {\n // Object format: {key: childNode, ...}\n for (const childNode of Object.values(node.c)) {\n traverse(childNode, depth + 1);\n }\n }\n }\n }\n \n traverse(radixNode);\n console.log(`๐Ÿ“š Extracted ${vocabulary.size} words from ${nodesVisited} nodes visited`);\n return vocabulary;\n}\n\n/**\n * Find all candidate matches for a single query token\n * \n * @param queryToken - Token from search query\n * @param vocabulary - Set of all words in the index\n * @param tolerance - Fuzzy matching tolerance\n * @param synonyms - Synonym map (optional)\n * @param synonymScore - Score multiplier for synonym matches\n * @returns Array of candidate matches\n */\nexport function findCandidatesForToken(\n queryToken: string,\n vocabulary: Set<string>,\n tolerance: number,\n synonyms?: SynonymMap,\n synonymScore: number = 0.8\n): Candidate[] {\n const candidates: Candidate[] = [];\n const seen = new Set<string>();\n\n // 1. Check for exact match\n if (vocabulary.has(queryToken)) {\n candidates.push({\n word: queryToken,\n type: 'exact',\n queryToken,\n distance: 0,\n score: 1.0\n });\n seen.add(queryToken);\n }\n\n // 2. Check for fuzzy matches\n for (const word of vocabulary) {\n if (seen.has(word)) continue;\n\n const match = fuzzyMatch(word, queryToken, tolerance);\n if (match.matches) {\n candidates.push({\n word,\n type: 'fuzzy',\n queryToken,\n distance: match.distance,\n score: match.score\n });\n seen.add(word);\n }\n }\n\n // 3. Check for synonym matches\n if (synonyms && synonyms[queryToken]) {\n for (const synonym of synonyms[queryToken]) {\n if (seen.has(synonym)) continue;\n if (vocabulary.has(synonym)) {\n candidates.push({\n word: synonym,\n type: 'synonym',\n queryToken,\n distance: 0,\n score: synonymScore\n });\n seen.add(synonym);\n }\n }\n }\n\n return candidates;\n}\n\n/**\n * Find candidates for all query tokens\n * \n * @param queryTokens - Array of tokens from search query\n * @param vocabulary - Set of all words in the index\n * @param tolerance - Fuzzy matching tolerance\n * @param synonyms - Synonym map (optional)\n * @param synonymScore - Score multiplier for synonym matches\n * @returns Map of query tokens to their candidate matches\n */\nexport function findAllCandidates(\n queryTokens: string[],\n vocabulary: Set<string>,\n tolerance: number,\n synonyms?: SynonymMap,\n synonymScore: number = 0.8\n): Map<string, Candidate[]> {\n const candidatesMap = new Map<string, Candidate[]>();\n\n for (const token of queryTokens) {\n const tokenCandidates = findCandidatesForToken(\n token,\n vocabulary,\n tolerance,\n synonyms,\n synonymScore\n );\n candidatesMap.set(token, tokenCandidates);\n }\n\n return candidatesMap;\n}\n\n/**\n * Get total number of candidates across all tokens\n * \n * @param candidatesMap - Map of token to candidates\n * @returns Total count of all candidates\n */\nexport function getTotalCandidateCount(\n candidatesMap: Map<string, Candidate[]>\n): number {\n let total = 0;\n for (const candidates of candidatesMap.values()) {\n total += candidates.length;\n }\n return total;\n}\n\n/**\n * Filter candidates by minimum score threshold\n * \n * @param candidatesMap - Map of token to candidates\n * @param minScore - Minimum score threshold\n * @returns Filtered candidates map\n */\nexport function filterCandidatesByScore(\n candidatesMap: Map<string, Candidate[]>,\n minScore: number\n): Map<string, Candidate[]> {\n const filtered = new Map<string, Candidate[]>();\n\n for (const [token, candidates] of candidatesMap.entries()) {\n const filteredCandidates = candidates.filter(c => c.score >= minScore);\n if (filteredCandidates.length > 0) {\n filtered.set(token, filteredCandidates);\n }\n }\n\n return filtered;\n}\n","/**\n * Phrase scoring algorithm with semantic weighting\n */\n\nimport type { WordMatch, PhraseMatch, Candidate, GapWord } from './types.js';\n\n/**\n * Configuration for phrase scoring\n */\nexport interface ScoringConfig {\n weights: {\n exact: number;\n fuzzy: number;\n order: number;\n proximity: number;\n density: number;\n semantic: number;\n };\n maxGap: number;\n /** \n * Multiplier for proximity window calculation.\n * proximityWindow = queryTokens.length ร— proximitySpanMultiplier\n */\n proximitySpanMultiplier: number;\n /**\n * Fuzzy tolerance (Levenshtein distance). When 0, only exact matches.\n */\n tolerance: number;\n}\n\n/**\n * Find all phrase matches in a document\n * \n * @param documentTokens - Tokenized document content (needed to extract gap words)\n * @param candidatesMap - Map of query tokens to their candidates\n * @param config - Scoring configuration\n * @param documentFrequency - Document frequency map for TF-IDF\n * @param totalDocuments - Total number of documents\n * @returns Array of phrase matches\n */\nexport function findPhrasesInDocument(\n documentTokens: string[],\n candidatesMap: Map<string, Candidate[]>,\n config: ScoringConfig,\n documentFrequency: Map<string, number>,\n totalDocuments: number\n): PhraseMatch[] {\n const phrases: PhraseMatch[] = [];\n const queryTokens = Array.from(candidatesMap.keys());\n\n // Find all word matches in document\n const wordMatches: WordMatch[] = [];\n \n for (let i = 0; i < documentTokens.length; i++) {\n const docWord = documentTokens[i];\n \n // Check if this word matches any query token\n for (const [queryToken, candidates] of candidatesMap.entries()) {\n for (const candidate of candidates) {\n if (candidate.word === docWord) {\n wordMatches.push({\n word: docWord,\n queryToken,\n position: i,\n type: candidate.type,\n distance: candidate.distance,\n score: candidate.score\n });\n }\n }\n }\n }\n\n // Build phrases from word matches using sliding window\n for (let i = 0; i < wordMatches.length; i++) {\n const phrase = buildPhraseFromPosition(\n wordMatches,\n i,\n queryTokens,\n config,\n documentFrequency,\n totalDocuments,\n wordMatches,\n documentTokens // Pass document tokens to extract gap words\n );\n \n if (phrase && phrase.words.length > 0) {\n phrases.push(phrase);\n }\n }\n\n // Deduplicate and sort by score\n return deduplicatePhrases(phrases);\n}\n\n/**\n * Build a phrase starting from a specific word match position\n * \n * @param wordMatches - All word matches in document\n * @param startIndex - Starting index in wordMatches array\n * @param queryTokens - Original query tokens\n * @param config - Scoring configuration\n * @param documentFrequency - Document frequency map\n * @param totalDocuments - Total document count\n * @param allWordMatches - All word matches in document (for density calculation)\n * @param documentTokens - Original document tokens (for gap word extraction)\n * @returns Phrase match or null\n */\nfunction buildPhraseFromPosition(\n wordMatches: WordMatch[],\n startIndex: number,\n queryTokens: string[],\n config: ScoringConfig,\n documentFrequency: Map<string, number>,\n totalDocuments: number,\n allWordMatches: WordMatch[],\n documentTokens: string[]\n): PhraseMatch | null {\n const startMatch = wordMatches[startIndex];\n const phraseWords: WordMatch[] = [startMatch];\n const coveredTokens = new Set([startMatch.queryToken]);\n const gapWords: GapWord[] = [];\n let totalGapUsed = 0;\n\n // Look for nearby matches to complete the phrase\n for (let i = startIndex + 1; i < wordMatches.length; i++) {\n const match = wordMatches[i];\n const lastPos = phraseWords[phraseWords.length - 1].position;\n const gap = match.position - lastPos - 1;\n\n // Stop if gap exceeds maximum\n if (gap > config.maxGap) {\n break;\n }\n\n // Only process if it's a different query token we need\n if (!coveredTokens.has(match.queryToken)) {\n // Track gap words between last match and current match\n for (let pos = lastPos + 1; pos < match.position; pos++) {\n totalGapUsed++;\n gapWords.push({\n word: documentTokens[pos],\n position: pos,\n gapIndex: totalGapUsed\n });\n }\n\n phraseWords.push(match);\n coveredTokens.add(match.queryToken);\n\n // Stop if we have all query tokens\n if (coveredTokens.size === queryTokens.length) {\n break;\n }\n }\n }\n\n // Calculate phrase score\n if (phraseWords.length > 0) {\n const coverage = phraseWords.length / queryTokens.length;\n const span = phraseWords[phraseWords.length - 1].position - phraseWords[0].position + 1;\n \n const { score, breakdown } = calculatePhraseScore(\n phraseWords,\n queryTokens,\n config,\n documentFrequency,\n totalDocuments,\n allWordMatches,\n coverage\n );\n\n return {\n words: phraseWords,\n gapWords,\n gapUsed: totalGapUsed,\n coverage,\n startPosition: phraseWords[0].position,\n endPosition: phraseWords[phraseWords.length - 1].position,\n span,\n inOrder: isInOrder(phraseWords, queryTokens),\n score,\n scoreBreakdown: breakdown\n };\n }\n\n return null;\n}\n\n/**\n * Calculate overall phrase score\n * \n * @param phraseWords - Words in the phrase\n * @param queryTokens - Original query tokens\n * @param config - Scoring configuration\n * @param documentFrequency - Document frequency map\n * @param totalDocuments - Total document count\n * @param allWordMatches - All word matches in document (for density calculation)\n * @param coverage - Pre-calculated coverage ratio (phraseWords.length / queryTokens.length)\n * @returns Phrase score (0-1) and detailed component breakdown\n */\nfunction calculatePhraseScore(\n phraseWords: WordMatch[],\n queryTokens: string[],\n config: ScoringConfig,\n documentFrequency: Map<string, number>,\n totalDocuments: number,\n allWordMatches: WordMatch[],\n coverage: number\n): { score: number; breakdown: { base: number; order: number; proximity: number; density: number; semantic: number; coverage: number } } {\n // Base score from word matches\n // Each word contributes: matchScore ร— typeWeight\n let baseScore = 0;\n for (const word of phraseWords) {\n const weight = word.type === 'exact' ? config.weights.exact :\n word.type === 'fuzzy' ? config.weights.fuzzy : \n config.weights.fuzzy * 0.8; // synonym gets 80% of fuzzy weight\n baseScore += word.score * weight;\n }\n baseScore /= phraseWords.length;\n\n // Order bonus: 1.0 if words appear in query order, 0.5 otherwise\n const inOrder = isInOrder(phraseWords, queryTokens);\n const orderScore = inOrder ? 1.0 : 0.5;\n\n // Proximity bonus (closer words score higher)\n // Short-circuit: skip if maxGap=0, proximity weight is 0, or single-word query (proximity meaningless)\n let proximityScore = 0;\n if (config.maxGap > 0 && config.weights.proximity > 0 && queryTokens.length > 1) {\n const span = phraseWords[phraseWords.length - 1].position - phraseWords[0].position + 1;\n const proximityWindow = queryTokens.length * config.proximitySpanMultiplier;\n proximityScore = Math.max(0, 1.0 - (span / proximityWindow));\n }\n\n // Density: Only applies to single-word queries (measures word repetition in document)\n // For multi-word phrase queries, density is 0 (coverage handles completeness separately)\n let densityScore = 0;\n \n if (queryTokens.length === 1) {\n // Single-word query: reward repetition\n const totalOccurrences = allWordMatches.length;\n // Cap at reasonable maximum to avoid runaway scores\n densityScore = Math.min(1.0, totalOccurrences / 10);\n }\n // For multi-word queries: densityScore stays 0\n // Coverage is applied as a multiplier at the end instead\n\n // Semantic score (TF-IDF based)\n const semanticScore = calculateSemanticScore(\n phraseWords,\n documentFrequency,\n totalDocuments\n );\n\n // Weighted combination\n const weights = config.weights;\n \n // Calculate weighted components\n const weightedBase = baseScore;\n const weightedOrder = orderScore * weights.order;\n const weightedProximity = proximityScore * weights.proximity;\n const weightedDensity = densityScore * weights.density;\n const weightedSemantic = semanticScore * weights.semantic;\n \n const totalScore = weightedBase + weightedOrder + weightedProximity + weightedDensity + weightedSemantic;\n\n // Calculate max possible score\n // FIX: Use actual max base weight (highest of exact/fuzzy) instead of hardcoded 1.0\n // When tolerance=0 or fuzzy weight=0, only exact matches are possible\n const canHaveFuzzyMatches = config.tolerance > 0 && weights.fuzzy > 0;\n const maxBaseWeight = canHaveFuzzyMatches ? Math.max(weights.exact, weights.fuzzy) : weights.exact;\n // Only include proximity in max if it can actually contribute (avoids penalizing scores when maxGap=0 or single-word)\n const effectiveProximityWeight = (config.maxGap > 0 && weights.proximity > 0 && queryTokens.length > 1) ? weights.proximity : 0;\n const maxPossibleScore = maxBaseWeight + weights.order + effectiveProximityWeight + weights.density + weights.semantic;\n \n // Normalize to 0-1 range\n const normalizedScore = totalScore / maxPossibleScore;\n \n // FIX: Apply coverage as a MULTIPLIER for multi-word queries\n // This ensures incomplete matches (2/3) can never outscore complete matches (3/3)\n const coverageMultiplier = queryTokens.length > 1 ? coverage : 1.0;\n const score = normalizedScore * coverageMultiplier;\n\n // Component contributions to the final normalized score (before coverage multiplier)\n const base = weightedBase / maxPossibleScore;\n const order = weightedOrder / maxPossibleScore;\n const proximity = weightedProximity / maxPossibleScore;\n const density = weightedDensity / maxPossibleScore;\n const semantic = weightedSemantic / maxPossibleScore;\n\n return {\n score,\n breakdown: {\n base,\n order,\n proximity,\n density,\n semantic,\n coverage: coverageMultiplier // Show coverage multiplier in breakdown\n }\n };\n}\n\n/**\n * Check if words are in the same order as query tokens\n * \n * @param phraseWords - Words in the phrase\n * @param queryTokens - Original query tokens\n * @returns True if in order\n */\nfunction isInOrder(phraseWords: WordMatch[], queryTokens: string[]): boolean {\n const tokenOrder = new Map(queryTokens.map((token, index) => [token, index]));\n \n for (let i = 1; i < phraseWords.length; i++) {\n const prevOrder = tokenOrder.get(phraseWords[i - 1].queryToken) ?? -1;\n const currOrder = tokenOrder.get(phraseWords[i].queryToken) ?? -1;\n \n if (currOrder < prevOrder) {\n return false;\n }\n }\n \n return true;\n}\n\n/**\n * Calculate semantic score using TF-IDF\n * \n * @param phraseWords - Words in the phrase\n * @param documentFrequency - Document frequency map\n * @param totalDocuments - Total document count\n * @returns Semantic score (0-1)\n */\nfunction calculateSemanticScore(\n phraseWords: WordMatch[],\n documentFrequency: Map<string, number>,\n totalDocuments: number\n): number {\n // Handle edge case: no documents\n if (totalDocuments === 0) {\n return 0;\n }\n \n let tfidfSum = 0;\n \n for (const word of phraseWords) {\n const df = documentFrequency.get(word.word) || 1;\n const idf = Math.log(totalDocuments / df);\n tfidfSum += idf;\n }\n \n // Normalize by phrase length\n const avgTfidf = tfidfSum / phraseWords.length;\n \n // Normalize to 0-1 range (assuming max IDF of ~10)\n return Math.min(1.0, avgTfidf / 10);\n}\n\n/**\n * Deduplicate overlapping phrases, keeping highest scoring ones\n * \n * @param phrases - Array of phrase matches\n * @returns Deduplicated phrases sorted by score\n */\nfunction deduplicatePhrases(phrases: PhraseMatch[]): PhraseMatch[] {\n if (phrases.length === 0) return [];\n\n // Sort by score descending\n const sorted = phrases.slice().sort((a, b) => b.score - a.score);\n const result: PhraseMatch[] = [];\n const covered = new Set<number>();\n\n for (const phrase of sorted) {\n // Check if this phrase overlaps with already selected phrases\n let overlaps = false;\n for (let pos = phrase.startPosition; pos <= phrase.endPosition; pos++) {\n if (covered.has(pos)) {\n overlaps = true;\n break;\n }\n }\n\n if (!overlaps) {\n result.push(phrase);\n // Mark positions as covered\n for (let pos = phrase.startPosition; pos <= phrase.endPosition; pos++) {\n covered.add(pos);\n }\n }\n }\n\n return result.sort((a, b) => b.score - a.score);\n}\n","/**\n * Fuzzy Phrase Plugin for Orama\n * \n * Advanced fuzzy phrase matching with semantic weighting and synonym expansion.\n * Completely independent from QPS - accesses Orama's radix tree directly.\n */\n\nimport type { AnyOrama, OramaPlugin, Results, TypedDocument } from '@wcs-colab/orama';\nimport type { FuzzyPhraseConfig, PluginState, SynonymMap, DocumentMatch } from './types.js';\nimport { calculateAdaptiveTolerance } from './fuzzy.js';\nimport { \n extractVocabularyFromRadixTree, \n findAllCandidates,\n filterCandidatesByScore \n} from './candidates.js';\nimport { findPhrasesInDocument } from './scoring.js';\n\n/**\n * Default configuration\n */\nconst DEFAULT_CONFIG: Required<FuzzyPhraseConfig> = {\n textProperty: 'normalized_content', // Must match server's field name\n tolerance: 1,\n adaptiveTolerance: true,\n enableSynonyms: false,\n supabase: undefined as any,\n synonymMatchScore: 0.8,\n weights: {\n exact: 1.0,\n fuzzy: 0.8,\n order: 0.3,\n proximity: 0.2,\n density: 0.2,\n semantic: 0.15\n },\n maxGap: 5,\n minScore: 0.1,\n enableFinalScoreMinimum: false,\n finalScoreMinimum: 0.3,\n proximitySpanMultiplier: 5\n};\n\n/**\n * Plugin state storage (keyed by Orama instance)\n */\nconst pluginStates = new WeakMap<AnyOrama, PluginState>();\n\n/**\n * Create the Fuzzy Phrase Plugin\n * \n * @param userConfig - User configuration options\n * @returns Orama plugin instance\n */\nexport function pluginFuzzyPhrase(userConfig: FuzzyPhraseConfig = {}): OramaPlugin {\n // Merge user config with defaults\n const config: Required<FuzzyPhraseConfig> = {\n textProperty: userConfig.textProperty ?? DEFAULT_CONFIG.textProperty,\n tolerance: userConfig.tolerance ?? DEFAULT_CONFIG.tolerance,\n adaptiveTolerance: userConfig.adaptiveTolerance ?? DEFAULT_CONFIG.adaptiveTolerance,\n enableSynonyms: userConfig.enableSynonyms ?? DEFAULT_CONFIG.enableSynonyms,\n supabase: userConfig.supabase || DEFAULT_CONFIG.supabase,\n synonymMatchScore: userConfig.synonymMatchScore ?? DEFAULT_CONFIG.synonymMatchScore,\n weights: {\n exact: userConfig.weights?.exact ?? DEFAULT_CONFIG.weights.exact,\n fuzzy: userConfig.weights?.fuzzy ?? DEFAULT_CONFIG.weights.fuzzy,\n order: userConfig.weights?.order ?? DEFAULT_CONFIG.weights.order,\n proximity: userConfig.weights?.proximity ?? DEFAULT_CONFIG.weights.proximity,\n density: userConfig.weights?.density ?? DEFAULT_CONFIG.weights.density,\n semantic: userConfig.weights?.semantic ?? DEFAULT_CONFIG.weights.semantic\n },\n maxGap: userConfig.maxGap ?? DEFAULT_CONFIG.maxGap,\n minScore: userConfig.minScore ?? DEFAULT_CONFIG.minScore,\n enableFinalScoreMinimum: userConfig.enableFinalScoreMinimum ?? DEFAULT_CONFIG.enableFinalScoreMinimum,\n finalScoreMinimum: userConfig.finalScoreMinimum ?? DEFAULT_CONFIG.finalScoreMinimum,\n proximitySpanMultiplier: userConfig.proximitySpanMultiplier ?? DEFAULT_CONFIG.proximitySpanMultiplier\n };\n\n const plugin: OramaPlugin = {\n name: 'fuzzy-phrase',\n\n /**\n * Initialize plugin after index is created\n */\n afterCreate: async (orama: AnyOrama) => {\n console.log('๐Ÿ”ฎ Initializing Fuzzy Phrase Plugin...');\n\n // Initialize state\n const state: PluginState = {\n synonymMap: {},\n config,\n documentFrequency: new Map(),\n totalDocuments: 0\n };\n\n // Load synonyms from Supabase if enabled\n if (config.enableSynonyms && config.supabase) {\n try {\n console.log('๐Ÿ“– Loading synonyms from Supabase...');\n state.synonymMap = await loadSynonymsFromSupabase(config.supabase);\n console.log(`โœ… Loaded ${Object.keys(state.synonymMap).length} words with synonyms`);\n } catch (error) {\n console.error('โš ๏ธ Failed to load synonyms:', error);\n // Continue without synonyms\n }\n }\n\n // Calculate document frequencies for TF-IDF from document store\n const docs = (orama.data as any)?.docs?.docs;\n if (docs) {\n state.totalDocuments = Object.keys(docs).length;\n state.documentFrequency = calculateDocumentFrequencies(docs, config.textProperty);\n console.log(`๐Ÿ“Š Calculated document frequencies for ${state.totalDocuments} documents`);\n }\n\n // Store state\n pluginStates.set(orama, state);\n console.log('โœ… Fuzzy Phrase Plugin initialized');\n \n // Signal ready - emit a custom event that can be listened to\n // Use setImmediate to ensure this runs after the afterCreate hook completes\n setImmediate(() => {\n if (typeof (globalThis as any).fuzzyPhrasePluginReady === 'function') {\n console.log('๐Ÿ“ก Signaling plugin ready...');\n (globalThis as any).fuzzyPhrasePluginReady();\n } else {\n console.warn('โš ๏ธ fuzzyPhrasePluginReady callback not found');\n }\n });\n }\n };\n\n return plugin;\n}\n\n/**\n * Search with fuzzy phrase matching\n * \n * This function should be called instead of the regular search() function\n * to enable fuzzy phrase matching.\n */\nexport async function searchWithFuzzyPhrase<T extends AnyOrama>(\n orama: T, \n params: { term?: string; properties?: string[]; limit?: number },\n language?: string\n): Promise<Results<TypedDocument<T>>> {\n const startTime = performance.now();\n \n // Get plugin state\n const state = pluginStates.get(orama);\n \n if (!state) {\n console.error('โŒ Plugin state not initialized');\n throw new Error('Fuzzy Phrase Plugin not properly initialized');\n }\n\n const { term, properties } = params;\n \n if (!term || typeof term !== 'string') {\n return { elapsed: { formatted: '0ms', raw: 0 }, hits: [], count: 0 };\n }\n\n // Use specified property or default\n const textProperty = (properties && properties[0]) || state.config.textProperty;\n\n // Tokenize query\n const queryTokens = tokenize(term);\n \n if (queryTokens.length === 0) {\n return { elapsed: { formatted: '0ms', raw: 0 }, hits: [], count: 0 };\n }\n\n // Calculate tolerance (adaptive or fixed)\n const tolerance = state.config.adaptiveTolerance\n ? calculateAdaptiveTolerance(queryTokens, state.config.tolerance)\n : state.config.tolerance;\n\n console.log(`๐Ÿ” Fuzzy phrase search: \"${term}\" (${queryTokens.length} tokens, tolerance: ${tolerance})`);\n\n // Extract vocabulary from radix tree\n let vocabulary: Set<string>;\n \n try {\n // Access radix tree - the actual index data is in orama.data.index, not orama.index\n // orama.index is just the component interface (methods)\n const indexData = (orama as any).data?.index;\n \n if (!indexData) {\n console.error('โŒ No index data found in orama.data.index');\n return { elapsed: { formatted: '0ms', raw: 0 }, hits: [], count: 0 };\n }\n \n console.log('๐Ÿ” DEBUG: Index data keys:', Object.keys(indexData || {}));\n \n // Try different paths to find the radix tree\n let radixNode = null;\n \n // Path 1: QPS-style (orama.data.index.indexes[property].node)\n if (indexData.indexes?.[textProperty]?.node) {\n radixNode = indexData.indexes[textProperty].node;\n console.log('โœ… Found radix via QPS-style path (data.index.indexes)');\n }\n // Path 2: Standard Orama (orama.data.index[property].node)\n else if (indexData[textProperty]?.node) {\n radixNode = indexData[textProperty].node;\n console.log('โœ… Found radix via standard path (data.index[property])');\n }\n \n if (!radixNode) {\n console.error('โŒ Radix tree not found for property:', textProperty);\n console.error(' Available properties in index:', Object.keys(indexData));\n return { elapsed: { formatted: '0ms', raw: 0 }, hits: [], count: 0 };\n }\n\n vocabulary = extractVocabularyFromRadixTree(radixNode);\n console.log(`๐Ÿ“š Extracted ${vocabulary.size} unique words from index`);\n } catch (error) {\n console.error('โŒ Failed to extract vocabulary:', error);\n return { elapsed: { formatted: '0ms', raw: 0 }, hits: [], count: 0 };\n }\n\n // Find candidates for all query tokens\n const candidatesMap = findAllCandidates(\n queryTokens,\n vocabulary,\n tolerance,\n state.config.enableSynonyms ? state.synonymMap : undefined,\n state.config.synonymMatchScore\n );\n\n // Filter by minimum score (skip when tolerance=0 since all matches are exact with score 1.0)\n const filteredCandidates = tolerance === 0\n ? candidatesMap // Skip filtering - all matches are exact\n : filterCandidatesByScore(candidatesMap, state.config.minScore);\n\n console.log(`๐ŸŽฏ Found candidates: ${Array.from(filteredCandidates.values()).reduce((sum, c) => sum + c.length, 0)} total`);\n\n // Search through all documents\n const documentMatches: DocumentMatch[] = [];\n \n console.log('๐Ÿ” DEBUG orama.data structure:', {\n dataKeys: Object.keys((orama as any).data || {}),\n hasDocs: !!((orama as any).data?.docs),\n docsType: (orama as any).data?.docs ? typeof (orama as any).data.docs : 'undefined'\n });\n \n // Try multiple possible document storage locations\n let docs: Record<string, any> = {};\n \n // Access the actual documents - they're nested in orama.data.docs.docs\n if ((orama as any).data?.docs?.docs) {\n docs = (orama as any).data.docs.docs;\n console.log('โœ… Found docs at orama.data.docs.docs');\n }\n // Fallback: orama.data.docs (might be the correct structure in some cases)\n else if ((orama as any).data?.docs && typeof (orama as any).data.docs === 'object') {\n // Check if it has document-like properties (not sharedInternalDocumentStore, etc.)\n const firstKey = Object.keys((orama as any).data.docs)[0];\n if (firstKey && firstKey !== 'sharedInternalDocumentStore' && firstKey !== 'count') {\n docs = (orama as any).data.docs;\n console.log('โœ… Found docs at orama.data.docs (direct)');\n }\n }\n \n if (Object.keys(docs).length === 0) {\n console.log('โŒ Could not find documents - available structure:', {\n hasDataDocs: !!((orama as any).data?.docs),\n dataDocsKeys: (orama as any).data?.docs ? Object.keys((orama as any).data.docs) : 'none',\n hasDataDocsDocs: !!((orama as any).data?.docs?.docs),\n dataDocsDocsCount: (orama as any).data?.docs?.docs ? Object.keys((orama as any).data.docs.docs).length : 0\n });\n }\n \n console.log(`๐Ÿ“„ Searching through ${Object.keys(docs).length} documents`);\n\n for (const [docId, doc] of Object.entries(docs)) {\n const text = doc[textProperty];\n \n if (!text || typeof text !== 'string') {\n continue;\n }\n\n // Tokenize document\n const docTokens = tokenize(text);\n\n // Find phrases in this document\n // Note: state.config.weights is guaranteed to have all properties from default merge\n const phrases = findPhrasesInDocument(\n docTokens,\n filteredCandidates,\n {\n weights: state.config.weights as { exact: number; fuzzy: number; order: number; proximity: number; density: number; semantic: number },\n maxGap: state.config.maxGap,\n proximitySpanMultiplier: state.config.proximitySpanMultiplier,\n tolerance\n },\n state.documentFrequency,\n state.totalDocuments\n );\n\n if (phrases.length > 0) {\n // Calculate overall document score (highest phrase score)\n const docScore = Math.max(...phrases.map(p => p.score));\n\n documentMatches.push({\n id: docId,\n phrases,\n score: docScore,\n document: doc\n });\n }\n }\n\n // Sort by score descending\n documentMatches.sort((a, b) => b.score - a.score);\n\n // Apply final score minimum filter if enabled\n let filteredMatches = documentMatches;\n if (state.config.enableFinalScoreMinimum && state.config.finalScoreMinimum > 0) {\n const threshold = state.config.finalScoreMinimum;\n const beforeCount = filteredMatches.length;\n filteredMatches = filteredMatches.filter(m => m.score >= threshold);\n console.log(`๐ŸŽš๏ธ Final score filter: ${beforeCount} โ†’ ${filteredMatches.length} (threshold: ${threshold})`);\n }\n\n // Apply limit if specified\n const limit = params.limit ?? filteredMatches.length;\n const limitedMatches = filteredMatches.slice(0, limit);\n\n // Convert to Orama results format\n const hits = limitedMatches.map(match => ({\n id: match.id,\n score: match.score,\n document: match.document,\n // Store phrases for highlighting\n _phrases: match.phrases\n })) as any[];\n\n const elapsed = performance.now() - startTime;\n\n console.log(`โœ… Found ${hits.length} results in ${elapsed.toFixed(2)}ms (limit: ${limit})`);\n\n return {\n elapsed: {\n formatted: `${elapsed.toFixed(2)}ms`,\n raw: Math.floor(elapsed * 1000000) // nanoseconds\n },\n hits,\n count: hits.length\n } as any;\n}\n\n/**\n * Load synonyms from Supabase\n */\nasync function loadSynonymsFromSupabase(\n supabaseConfig: { url: string; serviceKey: string }\n): Promise<SynonymMap> {\n try {\n console.log('๐Ÿ” DEBUG: Calling Supabase RPC get_synonym_map...');\n \n // Dynamic import to avoid bundling Supabase client if not needed\n const { createClient } = await import('@supabase/supabase-js');\n \n const supabase = createClient(supabaseConfig.url, supabaseConfig.serviceKey);\n \n // Call the get_synonym_map function\n const { data, error } = await supabase.rpc('get_synonym_map');\n \n console.log('๐Ÿ” DEBUG: Supabase RPC response:', {\n hasError: !!error,\n errorMessage: error?.message,\n hasData: !!data,\n dataType: typeof data,\n dataKeys: data ? Object.keys(data).length : 0\n });\n \n if (error) {\n throw new Error(`Supabase error: ${error.message}`);\n }\n \n const synonymMap = data || {};\n console.log(`๐Ÿ“š Loaded ${Object.keys(synonymMap).length} synonym entries from Supabase`);\n \n return synonymMap;\n } catch (error) {\n console.error('โŒ Failed to load synonyms from Supabase:', error);\n throw error;\n }\n}\n\n/**\n * Calculate document frequencies for TF-IDF\n */\nfunction calculateDocumentFrequencies(\n docs: Record<string, any>,\n textProperty: string\n): Map<string, number> {\n const df = new Map<string, number>();\n\n for (const doc of Object.values(docs)) {\n const text = doc[textProperty];\n \n if (!text || typeof text !== 'string') {\n continue;\n }\n\n // Get unique words in this document\n const words = new Set(tokenize(text));\n\n // Increment document frequency for each unique word\n for (const word of words) {\n df.set(word, (df.get(word) || 0) + 1);\n }\n }\n\n return df;\n}\n\n/**\n * Normalize text using the same rules as server-side\n * \n * CRITICAL: This must match the normalizeText() function in server/index.js exactly\n * PLUS we remove all punctuation to match Orama's French tokenizer behavior\n */\nfunction normalizeText(text: string): string {\n return text\n .toLowerCase()\n .normalize('NFD')\n .replace(/[\\u0300-\\u036f]/g, '') // Remove diacritics\n // Replace French elisions (l', d', etc.) with space to preserve word boundaries\n .replace(/\\b[ldcjmnst][\\u2018\\u2019\\u201A\\u201B\\u2032\\u2035\\u0027\\u0060\\u00B4](?=\\w)/gi, ' ')\n .replace(/[\\u2018\\u2019\\u201A\\u201B\\u2032\\u2035\\u0027\\u0060\\u00B4]/g, '') // Remove remaining apostrophes\n .replace(/[\\u201c\\u201d]/g, '\"') // Normalize curly quotes to straight quotes\n .replace(/[.,;:!?()[\\]{}\\-โ€”โ€“ยซยป\"\"]/g, ' ') // Remove punctuation (replace with space to preserve word boundaries)\n .replace(/\\s+/g, ' ') // Normalize multiple spaces to single space\n .trim();\n}\n\n/**\n * Tokenization matching normalized text behavior\n * \n * Note: Text should already be normalized before indexing, so we normalize again\n * to ensure plugin tokenization matches index tokenization\n */\nfunction tokenize(text: string): string[] {\n // Normalize first (same as indexing), then split by whitespace\n return normalizeText(text)\n .split(/\\s+/)\n .filter(token => token.length > 0);\n}\n\n/**\n * Export types for external use\n */\nexport type {\n FuzzyPhraseConfig,\n WordMatch,\n PhraseMatch,\n DocumentMatch,\n SynonymMap,\n Candidate\n} from './types.js';\n"]}
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@wcs-colab/plugin-fuzzy-phrase",
3
- "version": "3.1.16-custom.newbase.7",
3
+ "version": "3.1.16-custom.newbase.9",
4
4
  "description": "Advanced fuzzy phrase matching plugin for Orama with semantic weighting and synonym expansion",
5
5
  "keywords": [
6
6
  "orama",