sensemaking 0.27.1 → 0.28.0

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 (126) hide show
  1. package/bin/cli.js +16 -7
  2. package/dist/cjs/cli/constants.d.cts +2 -0
  3. package/dist/cjs/cli/constants.d.ts +2 -0
  4. package/dist/cjs/cli/constants.js +21 -0
  5. package/dist/cjs/cli/constants.js.map +1 -0
  6. package/dist/cjs/cli/named.js +2 -1
  7. package/dist/cjs/cli/named.js.map +1 -1
  8. package/dist/cjs/cli/shared.js +5 -4
  9. package/dist/cjs/cli/shared.js.map +1 -1
  10. package/dist/cjs/cli.js +10 -9
  11. package/dist/cjs/cli.js.map +1 -1
  12. package/dist/cjs/embed/query.js.map +1 -1
  13. package/dist/cjs/store/duckdb/engine.d.cts +11 -0
  14. package/dist/cjs/store/duckdb/engine.d.ts +11 -0
  15. package/dist/cjs/store/duckdb/engine.js +193 -0
  16. package/dist/cjs/store/duckdb/engine.js.map +1 -0
  17. package/dist/cjs/store/duckdb/lexical.js +34 -20
  18. package/dist/cjs/store/duckdb/lexical.js.map +1 -1
  19. package/dist/cjs/store/duckdb/native.d.cts +5 -2
  20. package/dist/cjs/store/duckdb/native.d.ts +5 -2
  21. package/dist/cjs/store/duckdb/native.js +157 -13
  22. package/dist/cjs/store/duckdb/native.js.map +1 -1
  23. package/dist/cjs/store/duckdb/open.js +151 -3
  24. package/dist/cjs/store/duckdb/open.js.map +1 -1
  25. package/dist/cjs/store/duckdb/ordered-bm25.d.cts +3 -3
  26. package/dist/cjs/store/duckdb/ordered-bm25.d.ts +3 -3
  27. package/dist/cjs/store/duckdb/ordered-bm25.js +4 -8
  28. package/dist/cjs/store/duckdb/ordered-bm25.js.map +1 -1
  29. package/dist/cjs/store/engine.d.cts +28 -0
  30. package/dist/cjs/store/engine.d.ts +28 -0
  31. package/dist/cjs/store/engine.js +362 -0
  32. package/dist/cjs/store/engine.js.map +1 -0
  33. package/dist/cjs/store/native.d.cts +8 -5
  34. package/dist/cjs/store/native.d.ts +8 -5
  35. package/dist/cjs/store/native.js +78 -73
  36. package/dist/cjs/store/native.js.map +1 -1
  37. package/dist/cjs/store/open.js +100 -40
  38. package/dist/cjs/store/open.js.map +1 -1
  39. package/dist/cjs/store/sqlite/engine.d.cts +12 -0
  40. package/dist/cjs/store/sqlite/engine.d.ts +12 -0
  41. package/dist/cjs/store/sqlite/engine.js +53 -0
  42. package/dist/cjs/store/sqlite/engine.js.map +1 -0
  43. package/dist/cjs/store/sqlite/lexical.js +3 -4
  44. package/dist/cjs/store/sqlite/lexical.js.map +1 -1
  45. package/dist/cjs/store/sqlite/open.js +119 -6
  46. package/dist/cjs/store/sqlite/open.js.map +1 -1
  47. package/dist/cjs/store/turso/engine.d.cts +9 -0
  48. package/dist/cjs/store/turso/engine.d.ts +9 -0
  49. package/dist/cjs/store/turso/engine.js +189 -0
  50. package/dist/cjs/store/turso/engine.js.map +1 -0
  51. package/dist/cjs/store/turso/lexical.js +16 -11
  52. package/dist/cjs/store/turso/lexical.js.map +1 -1
  53. package/dist/cjs/store/turso/native.d.cts +5 -1
  54. package/dist/cjs/store/turso/native.d.ts +5 -1
  55. package/dist/cjs/store/turso/native.js +157 -8
  56. package/dist/cjs/store/turso/native.js.map +1 -1
  57. package/dist/cjs/store/turso/open.js +172 -27
  58. package/dist/cjs/store/turso/open.js.map +1 -1
  59. package/dist/cjs/store/turso/reconcile.d.cts +0 -1
  60. package/dist/cjs/store/turso/reconcile.d.ts +0 -1
  61. package/dist/cjs/store/turso/reconcile.js +22 -23
  62. package/dist/cjs/store/turso/reconcile.js.map +1 -1
  63. package/dist/cjs/store/types.d.cts +2 -0
  64. package/dist/cjs/store/types.d.ts +2 -0
  65. package/dist/cjs/store/types.js.map +1 -1
  66. package/dist/esm/cli/constants.d.ts +2 -0
  67. package/dist/esm/cli/constants.js +2 -0
  68. package/dist/esm/cli/constants.js.map +1 -0
  69. package/dist/esm/cli/named.js +2 -1
  70. package/dist/esm/cli/named.js.map +1 -1
  71. package/dist/esm/cli/shared.js +6 -5
  72. package/dist/esm/cli/shared.js.map +1 -1
  73. package/dist/esm/cli.js +10 -9
  74. package/dist/esm/cli.js.map +1 -1
  75. package/dist/esm/embed/query.js +1 -1
  76. package/dist/esm/embed/query.js.map +1 -1
  77. package/dist/esm/store/duckdb/engine.d.ts +11 -0
  78. package/dist/esm/store/duckdb/engine.js +18 -0
  79. package/dist/esm/store/duckdb/engine.js.map +1 -0
  80. package/dist/esm/store/duckdb/lexical.js +11 -11
  81. package/dist/esm/store/duckdb/lexical.js.map +1 -1
  82. package/dist/esm/store/duckdb/native.d.ts +5 -2
  83. package/dist/esm/store/duckdb/native.js +16 -12
  84. package/dist/esm/store/duckdb/native.js.map +1 -1
  85. package/dist/esm/store/duckdb/open.js +40 -2
  86. package/dist/esm/store/duckdb/open.js.map +1 -1
  87. package/dist/esm/store/duckdb/ordered-bm25.d.ts +3 -3
  88. package/dist/esm/store/duckdb/ordered-bm25.js +4 -9
  89. package/dist/esm/store/duckdb/ordered-bm25.js.map +1 -1
  90. package/dist/esm/store/engine.d.ts +28 -0
  91. package/dist/esm/store/engine.js +85 -0
  92. package/dist/esm/store/engine.js.map +1 -0
  93. package/dist/esm/store/native.d.ts +8 -5
  94. package/dist/esm/store/native.js +59 -57
  95. package/dist/esm/store/native.js.map +1 -1
  96. package/dist/esm/store/open.js +24 -7
  97. package/dist/esm/store/open.js.map +1 -1
  98. package/dist/esm/store/sqlite/engine.d.ts +12 -0
  99. package/dist/esm/store/sqlite/engine.js +16 -0
  100. package/dist/esm/store/sqlite/engine.js.map +1 -0
  101. package/dist/esm/store/sqlite/lexical.js +2 -3
  102. package/dist/esm/store/sqlite/lexical.js.map +1 -1
  103. package/dist/esm/store/sqlite/open.js +38 -7
  104. package/dist/esm/store/sqlite/open.js.map +1 -1
  105. package/dist/esm/store/turso/engine.d.ts +9 -0
  106. package/dist/esm/store/turso/engine.js +14 -0
  107. package/dist/esm/store/turso/engine.js.map +1 -0
  108. package/dist/esm/store/turso/lexical.js +9 -9
  109. package/dist/esm/store/turso/lexical.js.map +1 -1
  110. package/dist/esm/store/turso/native.d.ts +5 -1
  111. package/dist/esm/store/turso/native.js +16 -10
  112. package/dist/esm/store/turso/native.js.map +1 -1
  113. package/dist/esm/store/turso/open.js +44 -3
  114. package/dist/esm/store/turso/open.js.map +1 -1
  115. package/dist/esm/store/turso/reconcile.d.ts +0 -1
  116. package/dist/esm/store/turso/reconcile.js +3 -8
  117. package/dist/esm/store/turso/reconcile.js.map +1 -1
  118. package/dist/esm/store/types.d.ts +2 -0
  119. package/dist/esm/store/types.js.map +1 -1
  120. package/package.json +21 -6
  121. package/skills/sense/SKILL.md +1 -1
  122. package/skills/sense/references/stores/duckdb.md +1 -1
  123. package/skills/sense/references/stores/sqlite.md +1 -1
  124. package/skills/sense/references/stores/turso.md +1 -1
  125. package/skills/sense-setup/references/stores/duckdb.md +1 -1
  126. package/skills/sense-setup/references/stores/sqlite.md +1 -1
@@ -0,0 +1,193 @@
1
+ "use strict";
2
+ Object.defineProperty(exports, "__esModule", {
3
+ value: true
4
+ });
5
+ function _export(target, all) {
6
+ for(var name in all)Object.defineProperty(target, name, {
7
+ enumerable: true,
8
+ get: Object.getOwnPropertyDescriptor(all, name).get
9
+ });
10
+ }
11
+ _export(exports, {
12
+ get DUCKDB_FTS_BUILDERS () {
13
+ return DUCKDB_FTS_BUILDERS;
14
+ },
15
+ get DUCKDB_PATCHES () {
16
+ return DUCKDB_PATCHES;
17
+ },
18
+ get FTS_STEMMER () {
19
+ return FTS_STEMMER;
20
+ },
21
+ get duckdbFts () {
22
+ return duckdbFts;
23
+ }
24
+ });
25
+ var _enginets = require("../engine.js");
26
+ var _nativets = require("./native.js");
27
+ function asyncGeneratorStep(gen, resolve, reject, _next, _throw, key, arg) {
28
+ try {
29
+ var info = gen[key](arg);
30
+ var value = info.value;
31
+ } catch (error) {
32
+ reject(error);
33
+ return;
34
+ }
35
+ if (info.done) resolve(value);
36
+ else Promise.resolve(value).then(_next, _throw);
37
+ }
38
+ function _async_to_generator(fn) {
39
+ return function() {
40
+ var self = this, args = arguments;
41
+ return new Promise(function(resolve, reject) {
42
+ var gen = fn.apply(self, args);
43
+ function _next(value) {
44
+ asyncGeneratorStep(gen, resolve, reject, _next, _throw, "next", value);
45
+ }
46
+ function _throw(err) {
47
+ asyncGeneratorStep(gen, resolve, reject, _next, _throw, "throw", err);
48
+ }
49
+ _next(undefined);
50
+ });
51
+ };
52
+ }
53
+ function _ts_generator(thisArg, body) {
54
+ var f, y, t, _ = {
55
+ label: 0,
56
+ sent: function() {
57
+ if (t[0] & 1) throw t[1];
58
+ return t[1];
59
+ },
60
+ trys: [],
61
+ ops: []
62
+ }, g = Object.create((typeof Iterator === "function" ? Iterator : Object).prototype), d = Object.defineProperty;
63
+ return d(g, "next", {
64
+ value: verb(0)
65
+ }), d(g, "throw", {
66
+ value: verb(1)
67
+ }), d(g, "return", {
68
+ value: verb(2)
69
+ }), typeof Symbol === "function" && d(g, Symbol.iterator, {
70
+ value: function() {
71
+ return this;
72
+ }
73
+ }), g;
74
+ function verb(n) {
75
+ return function(v) {
76
+ return step([
77
+ n,
78
+ v
79
+ ]);
80
+ };
81
+ }
82
+ function step(op) {
83
+ if (f) throw new TypeError("Generator is already executing.");
84
+ while(g && (g = 0, op[0] && (_ = 0)), _)try {
85
+ if (f = 1, y && (t = op[0] & 2 ? y["return"] : op[0] ? y["throw"] || ((t = y["return"]) && t.call(y), 0) : y.next) && !(t = t.call(y, op[1])).done) return t;
86
+ if (y = 0, t) op = [
87
+ op[0] & 2,
88
+ t.value
89
+ ];
90
+ switch(op[0]){
91
+ case 0:
92
+ case 1:
93
+ t = op;
94
+ break;
95
+ case 4:
96
+ _.label++;
97
+ return {
98
+ value: op[1],
99
+ done: false
100
+ };
101
+ case 5:
102
+ _.label++;
103
+ y = op[1];
104
+ op = [
105
+ 0
106
+ ];
107
+ continue;
108
+ case 7:
109
+ op = _.ops.pop();
110
+ _.trys.pop();
111
+ continue;
112
+ default:
113
+ if (!(t = _.trys, t = t.length > 0 && t[t.length - 1]) && (op[0] === 6 || op[0] === 2)) {
114
+ _ = 0;
115
+ continue;
116
+ }
117
+ if (op[0] === 3 && (!t || op[1] > t[0] && op[1] < t[3])) {
118
+ _.label = op[1];
119
+ break;
120
+ }
121
+ if (op[0] === 6 && _.label < t[1]) {
122
+ _.label = t[1];
123
+ t = op;
124
+ break;
125
+ }
126
+ if (t && _.label < t[2]) {
127
+ _.label = t[2];
128
+ _.ops.push(op);
129
+ break;
130
+ }
131
+ if (t[2]) _.ops.pop();
132
+ _.trys.pop();
133
+ continue;
134
+ }
135
+ op = body.call(thisArg, _);
136
+ } catch (e) {
137
+ op = [
138
+ 6,
139
+ e
140
+ ];
141
+ y = 0;
142
+ } finally{
143
+ f = t = 0;
144
+ }
145
+ if (op[0] & 5) throw op[1];
146
+ return {
147
+ value: op[0] ? op[1] : void 0,
148
+ done: true
149
+ };
150
+ }
151
+ }
152
+ var FTS_STEMMER = 'porter';
153
+ var DUCKDB_FTS_BUILDERS = {
154
+ createIndex: function createIndex() {
155
+ return "PRAGMA create_fts_index('content', 'path', 'title', 'summary', 'text', stemmer='".concat(FTS_STEMMER, "', stopwords='none', overwrite=1)");
156
+ },
157
+ queryTerms: function queryTerms() {
158
+ return "\n WITH tokens AS (SELECT unnest(fts_main_content.tokenize(?)) AS raw_term)\n SELECT min(raw_term) AS term FROM tokens\n WHERE raw_term <> '' GROUP BY stem(raw_term, '".concat(FTS_STEMMER, "') ORDER BY term\n ");
159
+ },
160
+ termScore: function termScore(field) {
161
+ return 'SELECT fts_main_content.match_bm25(content."path", ?, fields := \''.concat(field, "', conjunctive := false) AS score");
162
+ },
163
+ conjunctiveGate: function conjunctiveGate() {
164
+ return 'fts_main_content.match_bm25(content."path", ?, conjunctive := true) IS NOT NULL';
165
+ }
166
+ };
167
+ var DUCKDB_PATCHES = [];
168
+ function duckdbFts() {
169
+ return _async_to_generator(function() {
170
+ var _tmp;
171
+ return _ts_generator(this, function(_state) {
172
+ switch(_state.label){
173
+ case 0:
174
+ _tmp = [
175
+ DUCKDB_FTS_BUILDERS,
176
+ DUCKDB_PATCHES
177
+ ];
178
+ return [
179
+ 4,
180
+ (0, _nativets.duckdbNative)()
181
+ ];
182
+ case 1:
183
+ return [
184
+ 2,
185
+ _enginets.selectBuilders.apply(void 0, _tmp.concat([
186
+ _state.sent().version
187
+ ]))
188
+ ];
189
+ }
190
+ });
191
+ })();
192
+ }
193
+ /* CJS INTEROP */ if (exports.__esModule && exports.default) { try { Object.defineProperty(exports.default, '__esModule', { value: true }); for (var key in exports) { exports.default[key] = exports[key]; } } catch (_) {}; module.exports = exports.default; }
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["/Users/kevin/Dev/open-source/ai/sensemaking/src/store/duckdb/engine.ts"],"sourcesContent":["import { type EnginePatch, selectBuilders } from '../engine.ts';\nimport { duckdbNative } from './native.ts';\n\nexport const FTS_STEMMER = 'porter';\n\n// Every statement this store sends to the fts extension; a version that behaves differently gets a DUCKDB_PATCHES entry.\n// match_bm25 tokenizes its query with the generated fts_main_content.tokenize macro, so queryTerms uses that same macro.\nexport interface DuckdbFtsBuilders {\n // stopwords='none': sqlite's porter/unicode61 tokenizer never removes stopwords; the extension's default list would\n // drop common words (e.g. \"and\") from the index but not match_bm25's conjunctive gate, breaking multi-word matches.\n createIndex(): string;\n // One bound parameter, the folded query; returns one original token per stem so native scoring stems once.\n queryTerms(): string;\n // One bound parameter, a single query term, scored within one field.\n termScore(field: string): string;\n // One bound parameter, the space-joined words that must all occur.\n conjunctiveGate(): string;\n}\n\nexport const DUCKDB_FTS_BUILDERS: DuckdbFtsBuilders = {\n createIndex: () => `PRAGMA create_fts_index('content', 'path', 'title', 'summary', 'text', stemmer='${FTS_STEMMER}', stopwords='none', overwrite=1)`,\n queryTerms: () => `\n WITH tokens AS (SELECT unnest(fts_main_content.tokenize(?)) AS raw_term)\n SELECT min(raw_term) AS term FROM tokens\n WHERE raw_term <> '' GROUP BY stem(raw_term, '${FTS_STEMMER}') ORDER BY term\n `,\n termScore: (field) => `SELECT fts_main_content.match_bm25(content.\"path\", ?, fields := '${field}', conjunctive := false) AS score`,\n conjunctiveGate: () => `fts_main_content.match_bm25(content.\"path\", ?, conjunctive := true) IS NOT NULL`,\n};\n\n// Empty: DuckDB 1.4.5, 1.5.0 and 1.5.6 produce identical macros, tokens and scores (probed 2026-10-11).\nexport const DUCKDB_PATCHES: readonly EnginePatch<DuckdbFtsBuilders>[] = [];\n\nexport async function duckdbFts(): Promise<DuckdbFtsBuilders> {\n return selectBuilders(DUCKDB_FTS_BUILDERS, DUCKDB_PATCHES, (await duckdbNative()).version);\n}\n"],"names":["DUCKDB_FTS_BUILDERS","DUCKDB_PATCHES","FTS_STEMMER","duckdbFts","createIndex","queryTerms","termScore","field","conjunctiveGate","duckdbNative","selectBuilders","version"],"mappings":";;;;;;;;;;;QAmBaA;eAAAA;;QAYAC;eAAAA;;QA5BAC;eAAAA;;QA8BSC;eAAAA;;;wBAjC2B;wBACpB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAEtB,IAAMD,cAAc;AAgBpB,IAAMF,sBAAyC;IACpDI,aAAa,SAAbA;eAAmB,AAAC,mFAA8F,OAAZF,aAAY;;IAClHG,YAAY,SAAZA;eAAkB,AAAC,mLAG2C,OAAZH,aAAY;;IAE9DI,WAAW,SAAXA,UAAYC;eAAU,AAAC,qEAAyE,OAANA,OAAM;;IAChGC,iBAAiB,SAAjBA;eAAuB;;AACzB;AAGO,IAAMP,iBAA4D,EAAE;AAEpE,SAAeE;;;;;;;wBACEH;wBAAqBC;;oBAAiB;;wBAAMQ,IAAAA,sBAAY;;;oBAA9E;;wBAAOC,wBAAc;4BAAuC,cAAsBC,OAAO;;;;;IAC3F"}
@@ -31,6 +31,7 @@ var _errorsts = require("../../errors.js");
31
31
  var _segmentts = require("../../text/segment.js");
32
32
  var _sharedts = require("../shared.js");
33
33
  var _transactionts = require("../transaction.js");
34
+ var _enginets = require("./engine.js");
34
35
  var _orderedbm25ts = require("./ordered-bm25.js");
35
36
  function _array_like_to_array(arr, len) {
36
37
  if (len == null || len > arr.length) len = arr.length;
@@ -196,7 +197,7 @@ var FIELDS = [
196
197
  'text'
197
198
  ];
198
199
  // Mirrors sqlite's bm25(content, 10.0, 5.0, 1.0, ...) column-weight intent; DuckDB's match_bm25 has no
199
- // field_weights argument (verified 1.5.5), so each field is scored separately and combined here.
200
+ // field_weights argument (verified 1.4.5 through 1.5.6), so each field is scored separately and combined here.
200
201
  var FIELD_WEIGHT = {
201
202
  title: 10,
202
203
  summary: 5,
@@ -472,16 +473,23 @@ function rebuildFts(conn, state) {
472
473
  4,
473
474
  (0, _transactionts.withTransaction)(conn, function() {
474
475
  return _async_to_generator(function() {
476
+ var _;
475
477
  return _ts_generator(this, function(_state) {
476
478
  switch(_state.label){
477
479
  case 0:
478
- // stopwords='none': sqlite's porter/unicode61 tokenizer never removes stopwords either (verified 1.5.5); the fts extension's
479
- // default 571-word list would otherwise drop common words (e.g. "and") from the index but not match_bm25's conjunctive gate, breaking multi-word matches.
480
+ _ = conn.exec;
480
481
  return [
481
482
  4,
482
- conn.exec("PRAGMA create_fts_index('content', 'path', 'title', 'summary', 'text', stemmer='".concat(_orderedbm25ts.FTS_STEMMER, "', stopwords='none', overwrite=1)"))
483
+ (0, _enginets.duckdbFts)()
483
484
  ];
484
485
  case 1:
486
+ return [
487
+ 4,
488
+ _.apply(conn, [
489
+ _state.sent().createIndex()
490
+ ])
491
+ ];
492
+ case 2:
485
493
  _state.sent();
486
494
  // Same transaction as the rebuild: a crash before commit leaves the mark set (a redundant
487
495
  // rebuild next time), never a cleared mark serving a half-built index as fresh.
@@ -489,7 +497,7 @@ function rebuildFts(conn, state) {
489
497
  4,
490
498
  (0, _sharedts.setMeta)(conn, FTS_STALE_META_KEY, '0')
491
499
  ];
492
- case 2:
500
+ case 3:
493
501
  _state.sent();
494
502
  return [
495
503
  2
@@ -591,7 +599,7 @@ function assertFtsReady(conn) {
591
599
  }
592
600
  // Field-weighted fragments (title 10 / summary 5 / text 1), with params emitted in the exact left-to-right order the assembled SQL needs for positional `?` binding.
593
601
  // score/gate params stay in separate arrays until joined at the end, since the SQL puts every score part before every gate part; sharing one array would shift binds once a query has both words and substrings (PRINCIPLES: no-silent-modes).
594
- function buildScoreAndGate(words, nativeTerms, bareWords, phraseWords, substrings) {
602
+ function buildScoreAndGate(fts, words, nativeTerms, bareWords, phraseWords, substrings) {
595
603
  var scoreParts = [];
596
604
  var gateParts = [];
597
605
  var scoreParams = [];
@@ -602,7 +610,7 @@ function buildScoreAndGate(words, nativeTerms, bareWords, phraseWords, substring
602
610
  for(var _iterator = FIELDS[Symbol.iterator](), _step; !(_iteratorNormalCompletion = (_step = _iterator.next()).done); _iteratorNormalCompletion = true){
603
611
  var field = _step.value;
604
612
  var _scoreParams;
605
- scoreParts.push("".concat(FIELD_WEIGHT[field], ".0 * COALESCE(").concat((0, _orderedbm25ts.orderedBm25Sql)(field, nativeTerms.length), ", 0)"));
613
+ scoreParts.push("".concat(FIELD_WEIGHT[field], ".0 * COALESCE(").concat((0, _orderedbm25ts.orderedBm25Sql)(fts, field, nativeTerms.length), ", 0)"));
606
614
  (_scoreParams = scoreParams).push.apply(_scoreParams, _to_consumable_array(nativeTerms));
607
615
  }
608
616
  } catch (err) {
@@ -620,13 +628,13 @@ function buildScoreAndGate(words, nativeTerms, bareWords, phraseWords, substring
620
628
  }
621
629
  }
622
630
  if (bareWords.length > 0) {
623
- var nativeGate = 'fts_main_content.match_bm25(content."path", ?, conjunctive := true) IS NOT NULL';
631
+ var nativeGate = fts.conjunctiveGate();
624
632
  gateParts.push(nativeGate);
625
633
  gateParams.push(bareWords.map(_segmentts.foldForSearch).join(' '));
626
634
  }
627
635
  if (phraseWords.length > 0) {
628
636
  var _gateParams;
629
- var nativeGate1 = 'fts_main_content.match_bm25(content."path", ?, conjunctive := true) IS NOT NULL';
637
+ var nativeGate1 = fts.conjunctiveGate();
630
638
  var fallback = phraseWords.map(function(_word) {
631
639
  return "(".concat(FIELDS.map(function(field) {
632
640
  return "contains(strip_accents(lower(content.".concat(field, ")), ?)");
@@ -698,7 +706,7 @@ function buildScoreAndGate(words, nativeTerms, bareWords, phraseWords, substring
698
706
  }
699
707
  function queryLexical(conn, terms, opts) {
700
708
  return _async_to_generator(function() {
701
- var _stmt, unsupported, whereJoin, whereCond, scopeCond, limit, _splitTerms, words, bareWords, phraseWords, substrings, phrases, emptyPhrase, nativeTerms, _tmp, _buildScoreAndGate, scoreSql, gateSql, params, selectFields, sqlLimit, sql, stmt, rows, matches;
709
+ var _stmt, unsupported, whereJoin, whereCond, scopeCond, limit, _splitTerms, words, bareWords, phraseWords, substrings, phrases, emptyPhrase, fts, nativeTerms, _tmp, _buildScoreAndGate, scoreSql, gateSql, params, selectFields, sqlLimit, sql, stmt, rows, matches;
702
710
  return _ts_generator(this, function(_state) {
703
711
  switch(_state.label){
704
712
  case 0:
@@ -724,26 +732,32 @@ function queryLexical(conn, terms, opts) {
724
732
  _state.sent();
725
733
  _state.label = 2;
726
734
  case 2:
735
+ return [
736
+ 4,
737
+ (0, _enginets.duckdbFts)()
738
+ ];
739
+ case 3:
740
+ fts = _state.sent();
727
741
  if (!(words.length > 0)) return [
728
742
  3,
729
- 4
743
+ 5
730
744
  ];
731
745
  return [
732
746
  4,
733
- (0, _orderedbm25ts.nativeQueryTerms)(conn, words.map(_segmentts.foldForSearch).join(' '))
747
+ (0, _orderedbm25ts.nativeQueryTerms)(conn, fts, words.map(_segmentts.foldForSearch).join(' '))
734
748
  ];
735
- case 3:
749
+ case 4:
736
750
  _tmp = _state.sent();
737
751
  return [
738
752
  3,
739
- 5
753
+ 6
740
754
  ];
741
- case 4:
742
- _tmp = [];
743
- _state.label = 5;
744
755
  case 5:
756
+ _tmp = [];
757
+ _state.label = 6;
758
+ case 6:
745
759
  nativeTerms = _tmp;
746
- _buildScoreAndGate = buildScoreAndGate(words, nativeTerms, bareWords, phraseWords, substrings), scoreSql = _buildScoreAndGate.scoreSql, gateSql = _buildScoreAndGate.gateSql, params = _buildScoreAndGate.params;
760
+ _buildScoreAndGate = buildScoreAndGate(fts, words, nativeTerms, bareWords, phraseWords, substrings), scoreSql = _buildScoreAndGate.scoreSql, gateSql = _buildScoreAndGate.gateSql, params = _buildScoreAndGate.params;
747
761
  selectFields = phrases.length > 0 ? ', content.title, content.summary, content.text' : '';
748
762
  sqlLimit = phrases.length > 0 ? '' : ' LIMIT ?';
749
763
  sql = "SELECT path, score".concat(phrases.length > 0 ? ', title, summary, text' : '', ' FROM (\n SELECT content."path" AS path, (').concat(scoreSql, ") AS score").concat(selectFields, "\n FROM content\n ").concat(whereJoin, "\n WHERE ").concat(gateSql, "\n ").concat(whereCond, " ").concat(scopeCond, "\n ) sq ORDER BY score DESC, path").concat(sqlLimit);
@@ -751,7 +765,7 @@ function queryLexical(conn, terms, opts) {
751
765
  4,
752
766
  conn.prepare(sql)
753
767
  ];
754
- case 6:
768
+ case 7:
755
769
  stmt = _state.sent();
756
770
  return [
757
771
  4,
@@ -759,7 +773,7 @@ function queryLexical(conn, terms, opts) {
759
773
  limit
760
774
  ])))
761
775
  ];
762
- case 7:
776
+ case 8:
763
777
  rows = _state.sent();
764
778
  matches = phrases.length === 0 ? rows : rows.filter(function(row) {
765
779
  return phrases.every(function(phrase) {
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/open-source/ai/sensemaking/src/store/duckdb/lexical.ts"],"sourcesContent":["// D1: fts BM25 for ranking, contains() scans for exact substring / phrase verification /\n// unspaced scripts, JS snippets for every doc (no snippet()).\nimport { SenseError } from '../../errors.ts';\nimport { foldForSearch, matchesSearchPhrase, type SearchToken, searchTokens } from '../../text/segment.ts';\nimport { getMeta, setMeta } from '../shared.ts';\nimport { withTransaction } from '../transaction.ts';\nimport type { Connection, LexicalHit, LexicalQueryOptions } from '../types.ts';\nimport { FTS_STEMMER, nativeQueryTerms, orderedBm25Sql } from './ordered-bm25.ts';\n\nconst FIELDS = ['title', 'summary', 'text'] as const;\ntype Field = (typeof FIELDS)[number];\n// Mirrors sqlite's bm25(content, 10.0, 5.0, 1.0, ...) column-weight intent; DuckDB's match_bm25 has no\n// field_weights argument (verified 1.5.5), so each field is scored separately and combined here.\nconst FIELD_WEIGHT: Record<Field, number> = { title: 10, summary: 5, text: 1 };\n\n// FTS5 operator syntax (sqlite.org/fts5.html sec. 3) that words/substrings below would otherwise silently\n// treat as literal terms (PRINCIPLES: no-silent-modes). Quoted contents stay ignored while their boundary remains detectable.\nconst FTS5_OPERATORS: Array<{ label: string; re: RegExp }> = [\n { label: 'prefix query', re: /[\\p{L}\\p{N}_]+\\*(?=\\s|$)/u },\n { label: 'boolean operator', re: /(?:^|\\s)(?:AND|OR|NOT)(?=\\s|$)/ },\n { label: 'NEAR operator', re: /(?:^|\\s)NEAR\\b/ },\n { label: 'initial-token operator', re: /(?:^|\\s)\\^\\S+/ },\n { label: 'column filter', re: /(?:^|\\s)[\\p{L}_]\\w*\\s*:/u },\n { label: 'grouping operator', re: /[()]/ },\n];\n\nfunction unsupportedOperator(terms: string): { label: string; token: string } | null {\n const withoutPhrases = terms.replace(/\"[^\"]*\"/g, 'phrase');\n for (const { label, re } of FTS5_OPERATORS) {\n const m = withoutPhrases.match(re);\n if (m) return { label, token: m[0].trim() };\n }\n return null;\n}\n\ninterface FtsIndexState {\n stale: boolean;\n}\n\n// Durable across connections/processes (meta table, shared.ts); `stale` is the in-memory cache of\n// it for this connection's lifetime, so a watcher's repeat queries don't re-read meta every time.\nconst FTS_STALE_META_KEY = 'fts_stale';\n\n// One state per connection, not per store instance: reconcileContent (reconcile.ts) marks it\n// stale from outside createLexicalIndex's closure, whether a one-shot open or a watcher's builder ran it.\nconst ftsState = new WeakMap<Connection, FtsIndexState>();\nasync function stateFor(conn: Connection): Promise<FtsIndexState> {\n let state = ftsState.get(conn);\n if (!state) {\n // Anything but a persisted '0' (missing key, or '1') means a rebuild is owed: a cache that\n // never built an index, or one a prior process marked stale and never got to clear.\n state = { stale: (await getMeta(conn, FTS_STALE_META_KEY)) !== '0' };\n ftsState.set(conn, state);\n }\n return state;\n}\n\nasync function hasFtsArtifact(conn: Connection): Promise<boolean> {\n const stmt = await conn.prepare(`\n SELECT 1 AS ready\n FROM duckdb_functions()\n WHERE schema_name = 'fts_main_content'\n AND function_name = 'match_bm25'\n AND function_type = 'macro'\n LIMIT 1\n `);\n return (await stmt.get()) !== undefined;\n}\n\n// Called whenever this store's reconcileContent changes `content`; must run inside that same\n// transaction (reconcile.ts) so a crash never lands the content write without the stale mark.\nexport async function markContentStale(conn: Connection): Promise<void> {\n ftsState.set(conn, { stale: true });\n await setMeta(conn, FTS_STALE_META_KEY, '1');\n}\n\n// A run whose script marks no word boundaries makes match_bm25's whitespace tokenizer index the whole run as one token,\n// so unspaced terms use contains() and quoted phrases use the shared authored-field verifier.\nfunction splitTerms(terms: string): { words: string[]; bareWords: string[]; phraseWords: string[]; substrings: string[]; phrases: SearchToken[][]; emptyPhrase: boolean } {\n const words: string[] = [];\n const bareWords: string[] = [];\n const phraseWords: string[] = [];\n const substrings: string[] = [];\n const phrases: SearchToken[][] = [];\n let emptyPhrase = false;\n const withoutPhrases = terms.replace(/\"([^\"]*)\"/g, (_m, inner: string) => {\n const phrase = inner.trim();\n const phraseTokens = searchTokens(phrase);\n if (phraseTokens.length === 0) {\n emptyPhrase = true;\n return ' ';\n }\n phrases.push(phraseTokens);\n for (const token of phraseTokens) {\n if (token.unspaced) substrings.push(token.text);\n else {\n words.push(token.text);\n phraseWords.push(token.text);\n }\n }\n return ' ';\n });\n for (const tok of withoutPhrases.split(/\\s+/)) {\n if (tok.length === 0) continue;\n for (const token of searchTokens(tok)) {\n if (token.unspaced) substrings.push(token.text);\n else {\n words.push(token.text);\n bareWords.push(token.text);\n }\n }\n }\n return { words, bareWords, phraseWords, substrings, phrases, emptyPhrase };\n}\n\n// No incremental update (verified 1.5.5: PRAGMA create_fts_index is rebuild-only), so this pays\n// the full rebuild whenever `state.stale`, which meta.fts_stale keeps true across processes until cleared below.\nasync function rebuildFts(conn: Connection, state: FtsIndexState): Promise<void> {\n await conn.exec('INSTALL fts; LOAD fts;');\n if (!state.stale && (await hasFtsArtifact(conn))) return;\n state.stale = true;\n await withTransaction(conn, async () => {\n // stopwords='none': sqlite's porter/unicode61 tokenizer never removes stopwords either (verified 1.5.5); the fts extension's\n // default 571-word list would otherwise drop common words (e.g. \"and\") from the index but not match_bm25's conjunctive gate, breaking multi-word matches.\n await conn.exec(`PRAGMA create_fts_index('content', 'path', 'title', 'summary', 'text', stemmer='${FTS_STEMMER}', stopwords='none', overwrite=1)`);\n // Same transaction as the rebuild: a crash before commit leaves the mark set (a redundant\n // rebuild next time), never a cleared mark serving a half-built index as fresh.\n await setMeta(conn, FTS_STALE_META_KEY, '0');\n });\n state.stale = false;\n}\n\nexport async function prepareFts(conn: Connection): Promise<void> {\n await rebuildFts(conn, await stateFor(conn));\n}\n\nexport async function assertFtsReady(conn: Connection): Promise<void> {\n const state = await stateFor(conn);\n if (state.stale) throw new SenseError('INDEX_NOT_READY', 'the DuckDB lexical index is missing or stale; run `sense build` (or library build(config)) before lexical querying with build disabled');\n try {\n await conn.exec('LOAD fts;');\n if (!(await hasFtsArtifact(conn))) {\n throw new SenseError('INDEX_NOT_READY', 'the DuckDB lexical index artifact is missing; run `sense build` (or library build(config)) before lexical querying with build disabled');\n }\n } catch (err) {\n if (err instanceof SenseError) throw err;\n throw new SenseError('INDEX_NOT_READY', `the DuckDB lexical extension is not prepared (${(err as Error).message}); run \\`sense build\\` (or library build(config))`);\n }\n}\n\n// Field-weighted fragments (title 10 / summary 5 / text 1), with params emitted in the exact left-to-right order the assembled SQL needs for positional `?` binding.\n// score/gate params stay in separate arrays until joined at the end, since the SQL puts every score part before every gate part; sharing one array would shift binds once a query has both words and substrings (PRINCIPLES: no-silent-modes).\nfunction buildScoreAndGate(words: string[], nativeTerms: string[], bareWords: string[], phraseWords: string[], substrings: string[]): { scoreSql: string; gateSql: string; params: unknown[] } {\n const scoreParts: string[] = [];\n const gateParts: string[] = [];\n const scoreParams: unknown[] = [];\n const gateParams: unknown[] = [];\n\n if (words.length > 0) {\n for (const field of FIELDS) {\n scoreParts.push(`${FIELD_WEIGHT[field]}.0 * COALESCE(${orderedBm25Sql(field, nativeTerms.length)}, 0)`);\n scoreParams.push(...nativeTerms);\n }\n if (bareWords.length > 0) {\n const nativeGate = `fts_main_content.match_bm25(content.\"path\", ?, conjunctive := true) IS NOT NULL`;\n gateParts.push(nativeGate);\n gateParams.push(bareWords.map(foldForSearch).join(' '));\n }\n if (phraseWords.length > 0) {\n const nativeGate = `fts_main_content.match_bm25(content.\"path\", ?, conjunctive := true) IS NOT NULL`;\n const fallback = phraseWords.map((_word) => `(${FIELDS.map((field) => `contains(strip_accents(lower(content.${field})), ?)`).join(' OR ')})`).join(' AND ');\n gateParts.push(`(${nativeGate} OR (${fallback}))`);\n gateParams.push(phraseWords.map(foldForSearch).join(' '));\n gateParams.push(...phraseWords.flatMap((word) => FIELDS.map(() => foldForSearch(word))));\n }\n }\n\n for (const raw of substrings) {\n const needle = raw.toLowerCase();\n for (const field of FIELDS) {\n scoreParts.push(`${FIELD_WEIGHT[field]}.0 * CAST(contains(strip_accents(lower(content.${field})), ?) AS INTEGER)`);\n scoreParams.push(foldForSearch(needle));\n }\n gateParts.push(`(${FIELDS.map((field) => `contains(strip_accents(lower(content.${field})), ?)`).join(' OR ')})`);\n gateParams.push(...FIELDS.map(() => foldForSearch(needle)));\n }\n\n return { scoreSql: scoreParts.join(' + '), gateSql: gateParts.join(' AND ') || 'TRUE', params: [...scoreParams, ...gateParams] };\n}\n\n// Ranked query, scoped by the caller-built SQL fragments (same fragments sqlite's queryLexical takes).\n// Phrase rows carry authored fields through the post-filter; the public result still contains paths only.\nexport async function queryLexical(conn: Connection, terms: string, opts: LexicalQueryOptions): Promise<LexicalHit[]> {\n const unsupported = unsupportedOperator(terms);\n if (unsupported !== null) {\n throw new SenseError('STORE_CAPABILITY_MISSING', `store \"duckdb\" does not implement FTS5's ${unsupported.label} (\"${unsupported.token}\") in this build; rephrase \"${terms.trim()}\" without it, or set \"store\" to \"sqlite\" in this tree's config to search it as written`);\n }\n const { whereJoin, whereCond, scopeCond, limit } = opts;\n const { words, bareWords, phraseWords, substrings, phrases, emptyPhrase } = splitTerms(terms);\n if (emptyPhrase || (words.length === 0 && substrings.length === 0)) return [];\n\n if (words.length > 0) await assertFtsReady(conn);\n const nativeTerms = words.length > 0 ? await nativeQueryTerms(conn, words.map(foldForSearch).join(' ')) : [];\n\n const { scoreSql, gateSql, params } = buildScoreAndGate(words, nativeTerms, bareWords, phraseWords, substrings);\n const selectFields = phrases.length > 0 ? ', content.title, content.summary, content.text' : '';\n const sqlLimit = phrases.length > 0 ? '' : ' LIMIT ?';\n const sql = `SELECT path, score${phrases.length > 0 ? ', title, summary, text' : ''} FROM (\n SELECT content.\"path\" AS path, (${scoreSql}) AS score${selectFields}\n FROM content\n ${whereJoin}\n WHERE ${gateSql}\n ${whereCond} ${scopeCond}\n ) sq ORDER BY score DESC, path${sqlLimit}`;\n const stmt = await conn.prepare(sql);\n const rows = (await stmt.all(...params, ...(phrases.length > 0 ? [] : [limit]))) as Array<LexicalHit & Partial<Record<Field, string>>>;\n const matches = phrases.length === 0 ? rows : rows.filter((row) => phrases.every((phrase) => FIELDS.some((field) => matchesSearchPhrase(row[field] ?? '', phrase))));\n return matches.slice(0, limit).map(({ path }) => ({ path }));\n}\n\n// `stale`'s first read for a connection comes from meta.fts_stale (a fresh connection can't know\n// otherwise), and reconcileContent (reconcile.ts) marks it stale again on every content change.\nexport function createLexicalIndex(conn: Connection): { query: (terms: string, opts: LexicalQueryOptions) => Promise<LexicalHit[]>; markStale: () => Promise<void> } {\n return {\n query: async (terms, opts) => queryLexical(conn, terms, opts),\n markStale: () => markContentStale(conn),\n };\n}\n"],"names":["assertFtsReady","createLexicalIndex","markContentStale","prepareFts","queryLexical","FIELDS","FIELD_WEIGHT","title","summary","text","FTS5_OPERATORS","label","re","unsupportedOperator","terms","withoutPhrases","replace","m","match","token","trim","FTS_STALE_META_KEY","ftsState","WeakMap","stateFor","conn","state","get","getMeta","stale","set","hasFtsArtifact","stmt","prepare","undefined","setMeta","splitTerms","words","bareWords","phraseWords","substrings","phrases","emptyPhrase","_m","inner","phrase","phraseTokens","searchTokens","length","push","unspaced","split","tok","rebuildFts","exec","withTransaction","FTS_STEMMER","err","SenseError","message","buildScoreAndGate","nativeTerms","scoreParts","gateParts","scoreParams","gateParams","field","orderedBm25Sql","nativeGate","map","foldForSearch","join","fallback","_word","flatMap","word","raw","needle","toLowerCase","scoreSql","gateSql","params","opts","unsupported","whereJoin","whereCond","scopeCond","limit","selectFields","sqlLimit","sql","rows","matches","nativeQueryTerms","all","filter","row","every","some","matchesSearchPhrase","slice","path","query","markStale"],"mappings":"AAAA,yFAAyF;AACzF,8DAA8D;;;;;;;;;;;;QAuIxCA;eAAAA;;QAsFNC;eAAAA;;QAvJMC;eAAAA;;QA6DAC;eAAAA;;QA4DAC;eAAAA;;;wBA9LK;yBACwD;wBAClD;6BACD;6BAE8B;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAE9D,IAAMC,SAAS;IAAC;IAAS;IAAW;CAAO;AAE3C,uGAAuG;AACvG,iGAAiG;AACjG,IAAMC,eAAsC;IAAEC,OAAO;IAAIC,SAAS;IAAGC,MAAM;AAAE;AAE7E,0GAA0G;AAC1G,8HAA8H;AAC9H,IAAMC,iBAAuD;IAC3D;QAAEC,OAAO;QAAgBC,IAAI;IAA4B;IACzD;QAAED,OAAO;QAAoBC,IAAI;IAAiC;IAClE;QAAED,OAAO;QAAiBC,IAAI;IAAiB;IAC/C;QAAED,OAAO;QAA0BC,IAAI;IAAgB;IACvD;QAAED,OAAO;QAAiBC,IAAI;IAA2B;IACzD;QAAED,OAAO;QAAqBC,IAAI;IAAO;CAC1C;AAED,SAASC,oBAAoBC,KAAa;IACxC,IAAMC,iBAAiBD,MAAME,OAAO,CAAC,YAAY;QAC5C,kCAAA,2BAAA;;QAAL,QAAK,YAAuBN,mCAAvB,SAAA,6BAAA,QAAA,yBAAA,iCAAuC;YAAvC,kBAAA,aAAQC,oBAAAA,OAAOC,iBAAAA;YAClB,IAAMK,IAAIF,eAAeG,KAAK,CAACN;YAC/B,IAAIK,GAAG,OAAO;gBAAEN,OAAAA;gBAAOQ,OAAOF,CAAC,CAAC,EAAE,CAACG,IAAI;YAAG;QAC5C;;QAHK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAIL,OAAO;AACT;AAMA,kGAAkG;AAClG,kGAAkG;AAClG,IAAMC,qBAAqB;AAE3B,6FAA6F;AAC7F,0GAA0G;AAC1G,IAAMC,WAAW,IAAIC;AACrB,SAAeC,SAASC,IAAgB;;YAClCC;;;;oBAAAA,QAAQJ,SAASK,GAAG,CAACF;yBACrB,CAACC,OAAD;;;;;oBAGgB;;wBAAME,IAAAA,iBAAO,EAACH,MAAMJ;;;oBAFtC,2FAA2F;oBAC3F,oFAAoF;oBACpFK,cAAUG,QAAO,AAAC,kBAA6C;oBAC/DP,SAASQ,GAAG,CAACL,MAAMC;;;oBAErB;;wBAAOA;;;;IACT;;AAEA,SAAeK,eAAeN,IAAgB;;YACtCO;;;;oBAAO;;wBAAMP,KAAKQ,OAAO,CAAC;;;oBAA1BD,OAAO;oBAQL;;wBAAMA,KAAKL,GAAG;;;oBAAtB;;wBAAQ,kBAAsBO;;;;IAChC;;AAIO,SAAehC,iBAAiBuB,IAAgB;;;;;oBACrDH,SAASQ,GAAG,CAACL,MAAM;wBAAEI,OAAO;oBAAK;oBACjC;;wBAAMM,IAAAA,iBAAO,EAACV,MAAMJ,oBAAoB;;;oBAAxC;;;;;;IACF;;AAEA,wHAAwH;AACxH,8FAA8F;AAC9F,SAASe,WAAWtB,KAAa;IAC/B,IAAMuB,QAAkB,EAAE;IAC1B,IAAMC,YAAsB,EAAE;IAC9B,IAAMC,cAAwB,EAAE;IAChC,IAAMC,aAAuB,EAAE;IAC/B,IAAMC,UAA2B,EAAE;IACnC,IAAIC,cAAc;IAClB,IAAM3B,iBAAiBD,MAAME,OAAO,CAAC,cAAc,SAAC2B,IAAIC;QACtD,IAAMC,SAASD,MAAMxB,IAAI;QACzB,IAAM0B,eAAeC,IAAAA,uBAAY,EAACF;QAClC,IAAIC,aAAaE,MAAM,KAAK,GAAG;YAC7BN,cAAc;YACd,OAAO;QACT;QACAD,QAAQQ,IAAI,CAACH;YACR,kCAAA,2BAAA;;YAAL,QAAK,YAAeA,iCAAf,SAAA,6BAAA,QAAA,yBAAA,iCAA6B;gBAA7B,IAAM3B,QAAN;gBACH,IAAIA,MAAM+B,QAAQ,EAAEV,WAAWS,IAAI,CAAC9B,MAAMV,IAAI;qBACzC;oBACH4B,MAAMY,IAAI,CAAC9B,MAAMV,IAAI;oBACrB8B,YAAYU,IAAI,CAAC9B,MAAMV,IAAI;gBAC7B;YACF;;YANK;YAAA;;;qBAAA,6BAAA;oBAAA;;;oBAAA;0BAAA;;;;QAOL,OAAO;IACT;QACK,kCAAA,2BAAA;;QAAL,QAAK,YAAaM,eAAeoC,KAAK,CAAC,2BAAlC,SAAA,6BAAA,QAAA,yBAAA,iCAA0C;YAA1C,IAAMC,MAAN;YACH,IAAIA,IAAIJ,MAAM,KAAK,GAAG;gBACjB,mCAAA,4BAAA;;gBAAL,QAAK,aAAeD,IAAAA,uBAAY,EAACK,yBAA5B,UAAA,8BAAA,SAAA,0BAAA,kCAAkC;oBAAlC,IAAMjC,QAAN;oBACH,IAAIA,MAAM+B,QAAQ,EAAEV,WAAWS,IAAI,CAAC9B,MAAMV,IAAI;yBACzC;wBACH4B,MAAMY,IAAI,CAAC9B,MAAMV,IAAI;wBACrB6B,UAAUW,IAAI,CAAC9B,MAAMV,IAAI;oBAC3B;gBACF;;gBANK;gBAAA;;;yBAAA,8BAAA;wBAAA;;;wBAAA;8BAAA;;;;QAOP;;QATK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAUL,OAAO;QAAE4B,OAAAA;QAAOC,WAAAA;QAAWC,aAAAA;QAAaC,YAAAA;QAAYC,SAAAA;QAASC,aAAAA;IAAY;AAC3E;AAEA,gGAAgG;AAChG,iHAAiH;AACjH,SAAeW,WAAW5B,IAAgB,EAAEC,KAAoB;;;;;;oBAC9D;;wBAAMD,KAAK6B,IAAI,CAAC;;;oBAAhB;2BACI,CAAC5B,MAAMG,KAAK;+BAAZ;;;;oBAAiB;;wBAAME,eAAeN;;;2BAArB;;;oBAArB,UAAkD;;;oBAClDC,MAAMG,KAAK,GAAG;oBACd;;wBAAM0B,IAAAA,8BAAe,EAAC9B,MAAM;;;;;4CAC1B,6HAA6H;4CAC7H,0JAA0J;4CAC1J;;gDAAMA,KAAK6B,IAAI,CAAC,AAAC,mFAA8F,OAAZE,0BAAW,EAAC;;;4CAA/G;4CACA,0FAA0F;4CAC1F,gFAAgF;4CAChF;;gDAAMrB,IAAAA,iBAAO,EAACV,MAAMJ,oBAAoB;;;4CAAxC;;;;;;4BACF;;;;oBAPA;oBAQAK,MAAMG,KAAK,GAAG;;;;;;IAChB;;AAEO,SAAe1B,WAAWsB,IAAgB;;;;;;;wBAC9BA;;oBAAM;;wBAAMD,SAASC;;;oBAAtC;;wBAAM4B;4BAAiB;;;;oBAAvB;;;;;;IACF;;AAEO,SAAerD,eAAeyB,IAAgB;;YAC7CC,OAOG+B;;;;oBAPK;;wBAAMjC,SAASC;;;oBAAvBC,QAAQ;oBACd,IAAIA,MAAMG,KAAK,EAAE,MAAM,IAAI6B,oBAAU,CAAC,mBAAmB;;;;;;;;;oBAEvD;;wBAAMjC,KAAK6B,IAAI,CAAC;;;oBAAhB;oBACM;;wBAAMvB,eAAeN;;;oBAA3B,IAAI,CAAE,eAA6B;wBACjC,MAAM,IAAIiC,oBAAU,CAAC,mBAAmB;oBAC1C;;;;;;oBACOD;oBACP,IAAIA,AAAG,YAAHA,KAAeC,oBAAU,GAAE,MAAMD;oBACrC,MAAM,IAAIC,oBAAU,CAAC,mBAAmB,AAAC,iDAAuE,OAAvB,AAACD,IAAcE,OAAO,EAAC;;;;;;;IAEpH;;AAEA,qKAAqK;AACrK,+OAA+O;AAC/O,SAASC,kBAAkBvB,KAAe,EAAEwB,WAAqB,EAAEvB,SAAmB,EAAEC,WAAqB,EAAEC,UAAoB;IACjI,IAAMsB,aAAuB,EAAE;IAC/B,IAAMC,YAAsB,EAAE;IAC9B,IAAMC,cAAyB,EAAE;IACjC,IAAMC,aAAwB,EAAE;IAEhC,IAAI5B,MAAMW,MAAM,GAAG,GAAG;YACf,kCAAA,2BAAA;;YAAL,QAAK,YAAe3C,2BAAf,SAAA,6BAAA,QAAA,yBAAA,iCAAuB;gBAAvB,IAAM6D,QAAN;oBAEHF;gBADAF,WAAWb,IAAI,CAAC,AAAC,GAAsCkB,OAApC7D,YAAY,CAAC4D,MAAM,EAAC,kBAA0D,OAA1CC,IAAAA,6BAAc,EAACD,OAAOL,YAAYb,MAAM,GAAE;gBACjGgB,CAAAA,eAAAA,aAAYf,IAAI,OAAhBe,cAAiB,qBAAGH;YACtB;;YAHK;YAAA;;;qBAAA,6BAAA;oBAAA;;;oBAAA;0BAAA;;;;QAIL,IAAIvB,UAAUU,MAAM,GAAG,GAAG;YACxB,IAAMoB,aAAa;YACnBL,UAAUd,IAAI,CAACmB;YACfH,WAAWhB,IAAI,CAACX,UAAU+B,GAAG,CAACC,wBAAa,EAAEC,IAAI,CAAC;QACpD;QACA,IAAIhC,YAAYS,MAAM,GAAG,GAAG;gBAK1BiB;YAJA,IAAMG,cAAa;YACnB,IAAMI,WAAWjC,YAAY8B,GAAG,CAAC,SAACI;uBAAU,AAAC,IAA6F,OAA1FpE,OAAOgE,GAAG,CAAC,SAACH;2BAAU,AAAC,wCAA6C,OAANA,OAAM;mBAASK,IAAI,CAAC,SAAQ;eAAIA,IAAI,CAAC;YACnJR,UAAUd,IAAI,CAAC,AAAC,IAAqBuB,OAAlBJ,aAAW,SAAgB,OAATI,UAAS;YAC9CP,WAAWhB,IAAI,CAACV,YAAY8B,GAAG,CAACC,wBAAa,EAAEC,IAAI,CAAC;YACpDN,CAAAA,cAAAA,YAAWhB,IAAI,OAAfgB,aAAgB,qBAAG1B,YAAYmC,OAAO,CAAC,SAACC;uBAAStE,OAAOgE,GAAG,CAAC;2BAAMC,IAAAA,wBAAa,EAACK;;;QAClF;IACF;QAEK,mCAAA,4BAAA;;;YAAA,IAAMC,MAAN;gBAOHX;YANA,IAAMY,SAASD,IAAIE,WAAW;gBACzB,kCAAA,2BAAA;;gBAAL,QAAK,YAAezE,2BAAf,SAAA,6BAAA,QAAA,yBAAA,iCAAuB;oBAAvB,IAAM6D,QAAN;oBACHJ,WAAWb,IAAI,CAAC,AAAC,GAAuEiB,OAArE5D,YAAY,CAAC4D,MAAM,EAAC,mDAAuD,OAANA,OAAM;oBAC9FF,YAAYf,IAAI,CAACqB,IAAAA,wBAAa,EAACO;gBACjC;;gBAHK;gBAAA;;;yBAAA,6BAAA;wBAAA;;;wBAAA;8BAAA;;;;YAILd,UAAUd,IAAI,CAAC,AAAC,IAA6F,OAA1F5C,OAAOgE,GAAG,CAAC,SAACH;uBAAU,AAAC,wCAA6C,OAANA,OAAM;eAASK,IAAI,CAAC,SAAQ;YAC7GN,CAAAA,cAAAA,YAAWhB,IAAI,OAAfgB,aAAgB,qBAAG5D,OAAOgE,GAAG,CAAC;uBAAMC,IAAAA,wBAAa,EAACO;;QACpD;QARA,QAAK,aAAarC,+BAAb,UAAA,8BAAA,SAAA,0BAAA;;QAAA;QAAA;;;iBAAA,8BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAUL,OAAO;QAAEuC,UAAUjB,WAAWS,IAAI,CAAC;QAAQS,SAASjB,UAAUQ,IAAI,CAAC,YAAY;QAAQU,QAAQ,AAAC,qBAAGjB,oBAAa,qBAAGC;IAAY;AACjI;AAIO,SAAe7D,aAAaqB,IAAgB,EAAEX,KAAa,EAAEoE,IAAyB;;YAuBvElD,OAtBdmD,aAIEC,WAAWC,WAAWC,WAAWC,OACmCnD,aAApEC,OAAOC,WAAWC,aAAaC,YAAYC,SAASC,aAItDmB,mBAEgCD,oBAA9BmB,UAAUC,SAASC,QACrBO,cACAC,UACAC,KAOA1D,MACA2D,MACAC;;;;oBAvBAT,cAActE,oBAAoBC;oBACxC,IAAIqE,gBAAgB,MAAM;wBACxB,MAAM,IAAIzB,oBAAU,CAAC,4BAA4B,AAAC,6CAAkEyB,OAAvBA,YAAYxE,KAAK,EAAC,OAAqDG,OAAhDqE,YAAYhE,KAAK,EAAC,gCAA2C,OAAbL,MAAMM,IAAI,IAAG;oBACnL;oBACQgE,YAA2CF,KAA3CE,WAAWC,YAAgCH,KAAhCG,WAAWC,YAAqBJ,KAArBI,WAAWC,QAAUL,KAAVK;oBACmCnD,cAAAA,WAAWtB,QAA/EuB,QAAoED,YAApEC,OAAOC,YAA6DF,YAA7DE,WAAWC,cAAkDH,YAAlDG,aAAaC,aAAqCJ,YAArCI,YAAYC,UAAyBL,YAAzBK,SAASC,cAAgBN,YAAhBM;oBAC5D,IAAIA,eAAgBL,MAAMW,MAAM,KAAK,KAAKR,WAAWQ,MAAM,KAAK,GAAI;;;;yBAEhEX,CAAAA,MAAMW,MAAM,GAAG,CAAA,GAAfX;;;;oBAAkB;;wBAAMrC,eAAeyB;;;oBAArB;;;yBACFY,CAAAA,MAAMW,MAAM,GAAG,CAAA,GAAfX;;;;oBAAmB;;wBAAMwD,IAAAA,+BAAgB,EAACpE,MAAMY,MAAMgC,GAAG,CAACC,wBAAa,EAAEC,IAAI,CAAC;;;2BAA3D;;;;;;;;;oBAAjCV;oBAEgCD,qBAAAA,kBAAkBvB,OAAOwB,aAAavB,WAAWC,aAAaC,aAA5FuC,WAA8BnB,mBAA9BmB,UAAUC,UAAoBpB,mBAApBoB,SAASC,SAAWrB,mBAAXqB;oBACrBO,eAAe/C,QAAQO,MAAM,GAAG,IAAI,mDAAmD;oBACvFyC,WAAWhD,QAAQO,MAAM,GAAG,IAAI,KAAK;oBACrC0C,MAAM,AAAC,qBACuBX,OADHtC,QAAQO,MAAM,GAAG,IAAI,2BAA2B,IAAG,iDAC3BwC,OAArBT,UAAS,cAEzCK,OAFqDI,cAAa,4BAG5DR,OADNI,WAAU,gBAEVC,OADML,SAAQ,UACDM,OAAbD,WAAU,KACoBI,OADjBH,WAAU,wCACgB,OAATG;oBACrB;;wBAAMhE,KAAKQ,OAAO,CAACyD;;;oBAA1B1D,OAAO;oBACC;;wBAAMA,CAAAA,QAAAA,MAAK8D,GAAG,OAAR9D,OAAAA,AAAS,qBAAGiD,eAAQ,qBAAIxC,QAAQO,MAAM,GAAG;4BAAUuC;;;;oBAAjEI,OAAQ;oBACRC,UAAUnD,QAAQO,MAAM,KAAK,IAAI2C,OAAOA,KAAKI,MAAM,CAAC,SAACC;+BAAQvD,QAAQwD,KAAK,CAAC,SAACpD;mCAAWxC,OAAO6F,IAAI,CAAC,SAAChC;oCAA8B8B;uCAApBG,IAAAA,8BAAmB,GAACH,aAAAA,GAAG,CAAC9B,MAAM,cAAV8B,wBAAAA,aAAc,IAAInD;;;;oBAC1J;;wBAAO+C,QAAQQ,KAAK,CAAC,GAAGb,OAAOlB,GAAG,CAAC;gCAAGgC,aAAAA;mCAAY;gCAAEA,MAAAA;4BAAK;;;;;IAC3D;;AAIO,SAASpG,mBAAmBwB,IAAgB;IACjD,OAAO;QACL6E,OAAO,SAAPA,MAAcxF,OAAOoE;;;;;wBAAS9E,aAAaqB,MAAMX,OAAOoE;;;;;QACxDqB,WAAW,SAAXA;mBAAiBrG,iBAAiBuB;;IACpC;AACF"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/open-source/ai/sensemaking/src/store/duckdb/lexical.ts"],"sourcesContent":["// D1: fts BM25 for ranking, contains() scans for exact substring / phrase verification /\n// unspaced scripts, JS snippets for every doc (no snippet()).\nimport { SenseError } from '../../errors.ts';\nimport { foldForSearch, matchesSearchPhrase, type SearchToken, searchTokens } from '../../text/segment.ts';\nimport { getMeta, setMeta } from '../shared.ts';\nimport { withTransaction } from '../transaction.ts';\nimport type { Connection, LexicalHit, LexicalQueryOptions } from '../types.ts';\nimport { type DuckdbFtsBuilders, duckdbFts } from './engine.ts';\nimport { nativeQueryTerms, orderedBm25Sql } from './ordered-bm25.ts';\n\nconst FIELDS = ['title', 'summary', 'text'] as const;\ntype Field = (typeof FIELDS)[number];\n// Mirrors sqlite's bm25(content, 10.0, 5.0, 1.0, ...) column-weight intent; DuckDB's match_bm25 has no\n// field_weights argument (verified 1.4.5 through 1.5.6), so each field is scored separately and combined here.\nconst FIELD_WEIGHT: Record<Field, number> = { title: 10, summary: 5, text: 1 };\n\n// FTS5 operator syntax (sqlite.org/fts5.html sec. 3) that words/substrings below would otherwise silently\n// treat as literal terms (PRINCIPLES: no-silent-modes). Quoted contents stay ignored while their boundary remains detectable.\nconst FTS5_OPERATORS: Array<{ label: string; re: RegExp }> = [\n { label: 'prefix query', re: /[\\p{L}\\p{N}_]+\\*(?=\\s|$)/u },\n { label: 'boolean operator', re: /(?:^|\\s)(?:AND|OR|NOT)(?=\\s|$)/ },\n { label: 'NEAR operator', re: /(?:^|\\s)NEAR\\b/ },\n { label: 'initial-token operator', re: /(?:^|\\s)\\^\\S+/ },\n { label: 'column filter', re: /(?:^|\\s)[\\p{L}_]\\w*\\s*:/u },\n { label: 'grouping operator', re: /[()]/ },\n];\n\nfunction unsupportedOperator(terms: string): { label: string; token: string } | null {\n const withoutPhrases = terms.replace(/\"[^\"]*\"/g, 'phrase');\n for (const { label, re } of FTS5_OPERATORS) {\n const m = withoutPhrases.match(re);\n if (m) return { label, token: m[0].trim() };\n }\n return null;\n}\n\ninterface FtsIndexState {\n stale: boolean;\n}\n\n// Durable across connections/processes (meta table, shared.ts); `stale` is the in-memory cache of\n// it for this connection's lifetime, so a watcher's repeat queries don't re-read meta every time.\nconst FTS_STALE_META_KEY = 'fts_stale';\n\n// One state per connection, not per store instance: reconcileContent (reconcile.ts) marks it\n// stale from outside createLexicalIndex's closure, whether a one-shot open or a watcher's builder ran it.\nconst ftsState = new WeakMap<Connection, FtsIndexState>();\nasync function stateFor(conn: Connection): Promise<FtsIndexState> {\n let state = ftsState.get(conn);\n if (!state) {\n // Anything but a persisted '0' (missing key, or '1') means a rebuild is owed: a cache that\n // never built an index, or one a prior process marked stale and never got to clear.\n state = { stale: (await getMeta(conn, FTS_STALE_META_KEY)) !== '0' };\n ftsState.set(conn, state);\n }\n return state;\n}\n\nasync function hasFtsArtifact(conn: Connection): Promise<boolean> {\n const stmt = await conn.prepare(`\n SELECT 1 AS ready\n FROM duckdb_functions()\n WHERE schema_name = 'fts_main_content'\n AND function_name = 'match_bm25'\n AND function_type = 'macro'\n LIMIT 1\n `);\n return (await stmt.get()) !== undefined;\n}\n\n// Called whenever this store's reconcileContent changes `content`; must run inside that same\n// transaction (reconcile.ts) so a crash never lands the content write without the stale mark.\nexport async function markContentStale(conn: Connection): Promise<void> {\n ftsState.set(conn, { stale: true });\n await setMeta(conn, FTS_STALE_META_KEY, '1');\n}\n\n// A run whose script marks no word boundaries makes match_bm25's whitespace tokenizer index the whole run as one token,\n// so unspaced terms use contains() and quoted phrases use the shared authored-field verifier.\nfunction splitTerms(terms: string): { words: string[]; bareWords: string[]; phraseWords: string[]; substrings: string[]; phrases: SearchToken[][]; emptyPhrase: boolean } {\n const words: string[] = [];\n const bareWords: string[] = [];\n const phraseWords: string[] = [];\n const substrings: string[] = [];\n const phrases: SearchToken[][] = [];\n let emptyPhrase = false;\n const withoutPhrases = terms.replace(/\"([^\"]*)\"/g, (_m, inner: string) => {\n const phrase = inner.trim();\n const phraseTokens = searchTokens(phrase);\n if (phraseTokens.length === 0) {\n emptyPhrase = true;\n return ' ';\n }\n phrases.push(phraseTokens);\n for (const token of phraseTokens) {\n if (token.unspaced) substrings.push(token.text);\n else {\n words.push(token.text);\n phraseWords.push(token.text);\n }\n }\n return ' ';\n });\n for (const tok of withoutPhrases.split(/\\s+/)) {\n if (tok.length === 0) continue;\n for (const token of searchTokens(tok)) {\n if (token.unspaced) substrings.push(token.text);\n else {\n words.push(token.text);\n bareWords.push(token.text);\n }\n }\n }\n return { words, bareWords, phraseWords, substrings, phrases, emptyPhrase };\n}\n\n// No incremental update (verified 1.5.5: PRAGMA create_fts_index is rebuild-only), so this pays\n// the full rebuild whenever `state.stale`, which meta.fts_stale keeps true across processes until cleared below.\nasync function rebuildFts(conn: Connection, state: FtsIndexState): Promise<void> {\n await conn.exec('INSTALL fts; LOAD fts;');\n if (!state.stale && (await hasFtsArtifact(conn))) return;\n state.stale = true;\n await withTransaction(conn, async () => {\n await conn.exec((await duckdbFts()).createIndex());\n // Same transaction as the rebuild: a crash before commit leaves the mark set (a redundant\n // rebuild next time), never a cleared mark serving a half-built index as fresh.\n await setMeta(conn, FTS_STALE_META_KEY, '0');\n });\n state.stale = false;\n}\n\nexport async function prepareFts(conn: Connection): Promise<void> {\n await rebuildFts(conn, await stateFor(conn));\n}\n\nexport async function assertFtsReady(conn: Connection): Promise<void> {\n const state = await stateFor(conn);\n if (state.stale) throw new SenseError('INDEX_NOT_READY', 'the DuckDB lexical index is missing or stale; run `sense build` (or library build(config)) before lexical querying with build disabled');\n try {\n await conn.exec('LOAD fts;');\n if (!(await hasFtsArtifact(conn))) {\n throw new SenseError('INDEX_NOT_READY', 'the DuckDB lexical index artifact is missing; run `sense build` (or library build(config)) before lexical querying with build disabled');\n }\n } catch (err) {\n if (err instanceof SenseError) throw err;\n throw new SenseError('INDEX_NOT_READY', `the DuckDB lexical extension is not prepared (${(err as Error).message}); run \\`sense build\\` (or library build(config))`);\n }\n}\n\n// Field-weighted fragments (title 10 / summary 5 / text 1), with params emitted in the exact left-to-right order the assembled SQL needs for positional `?` binding.\n// score/gate params stay in separate arrays until joined at the end, since the SQL puts every score part before every gate part; sharing one array would shift binds once a query has both words and substrings (PRINCIPLES: no-silent-modes).\nfunction buildScoreAndGate(fts: DuckdbFtsBuilders, words: string[], nativeTerms: string[], bareWords: string[], phraseWords: string[], substrings: string[]): { scoreSql: string; gateSql: string; params: unknown[] } {\n const scoreParts: string[] = [];\n const gateParts: string[] = [];\n const scoreParams: unknown[] = [];\n const gateParams: unknown[] = [];\n\n if (words.length > 0) {\n for (const field of FIELDS) {\n scoreParts.push(`${FIELD_WEIGHT[field]}.0 * COALESCE(${orderedBm25Sql(fts, field, nativeTerms.length)}, 0)`);\n scoreParams.push(...nativeTerms);\n }\n if (bareWords.length > 0) {\n const nativeGate = fts.conjunctiveGate();\n gateParts.push(nativeGate);\n gateParams.push(bareWords.map(foldForSearch).join(' '));\n }\n if (phraseWords.length > 0) {\n const nativeGate = fts.conjunctiveGate();\n const fallback = phraseWords.map((_word) => `(${FIELDS.map((field) => `contains(strip_accents(lower(content.${field})), ?)`).join(' OR ')})`).join(' AND ');\n gateParts.push(`(${nativeGate} OR (${fallback}))`);\n gateParams.push(phraseWords.map(foldForSearch).join(' '));\n gateParams.push(...phraseWords.flatMap((word) => FIELDS.map(() => foldForSearch(word))));\n }\n }\n\n for (const raw of substrings) {\n const needle = raw.toLowerCase();\n for (const field of FIELDS) {\n scoreParts.push(`${FIELD_WEIGHT[field]}.0 * CAST(contains(strip_accents(lower(content.${field})), ?) AS INTEGER)`);\n scoreParams.push(foldForSearch(needle));\n }\n gateParts.push(`(${FIELDS.map((field) => `contains(strip_accents(lower(content.${field})), ?)`).join(' OR ')})`);\n gateParams.push(...FIELDS.map(() => foldForSearch(needle)));\n }\n\n return { scoreSql: scoreParts.join(' + '), gateSql: gateParts.join(' AND ') || 'TRUE', params: [...scoreParams, ...gateParams] };\n}\n\n// Ranked query, scoped by the caller-built SQL fragments (same fragments sqlite's queryLexical takes).\n// Phrase rows carry authored fields through the post-filter; the public result still contains paths only.\nexport async function queryLexical(conn: Connection, terms: string, opts: LexicalQueryOptions): Promise<LexicalHit[]> {\n const unsupported = unsupportedOperator(terms);\n if (unsupported !== null) {\n throw new SenseError('STORE_CAPABILITY_MISSING', `store \"duckdb\" does not implement FTS5's ${unsupported.label} (\"${unsupported.token}\") in this build; rephrase \"${terms.trim()}\" without it, or set \"store\" to \"sqlite\" in this tree's config to search it as written`);\n }\n const { whereJoin, whereCond, scopeCond, limit } = opts;\n const { words, bareWords, phraseWords, substrings, phrases, emptyPhrase } = splitTerms(terms);\n if (emptyPhrase || (words.length === 0 && substrings.length === 0)) return [];\n\n if (words.length > 0) await assertFtsReady(conn);\n const fts = await duckdbFts();\n const nativeTerms = words.length > 0 ? await nativeQueryTerms(conn, fts, words.map(foldForSearch).join(' ')) : [];\n\n const { scoreSql, gateSql, params } = buildScoreAndGate(fts, words, nativeTerms, bareWords, phraseWords, substrings);\n const selectFields = phrases.length > 0 ? ', content.title, content.summary, content.text' : '';\n const sqlLimit = phrases.length > 0 ? '' : ' LIMIT ?';\n const sql = `SELECT path, score${phrases.length > 0 ? ', title, summary, text' : ''} FROM (\n SELECT content.\"path\" AS path, (${scoreSql}) AS score${selectFields}\n FROM content\n ${whereJoin}\n WHERE ${gateSql}\n ${whereCond} ${scopeCond}\n ) sq ORDER BY score DESC, path${sqlLimit}`;\n const stmt = await conn.prepare(sql);\n const rows = (await stmt.all(...params, ...(phrases.length > 0 ? [] : [limit]))) as Array<LexicalHit & Partial<Record<Field, string>>>;\n const matches = phrases.length === 0 ? rows : rows.filter((row) => phrases.every((phrase) => FIELDS.some((field) => matchesSearchPhrase(row[field] ?? '', phrase))));\n return matches.slice(0, limit).map(({ path }) => ({ path }));\n}\n\n// `stale`'s first read for a connection comes from meta.fts_stale (a fresh connection can't know\n// otherwise), and reconcileContent (reconcile.ts) marks it stale again on every content change.\nexport function createLexicalIndex(conn: Connection): { query: (terms: string, opts: LexicalQueryOptions) => Promise<LexicalHit[]>; markStale: () => Promise<void> } {\n return {\n query: async (terms, opts) => queryLexical(conn, terms, opts),\n markStale: () => markContentStale(conn),\n };\n}\n"],"names":["assertFtsReady","createLexicalIndex","markContentStale","prepareFts","queryLexical","FIELDS","FIELD_WEIGHT","title","summary","text","FTS5_OPERATORS","label","re","unsupportedOperator","terms","withoutPhrases","replace","m","match","token","trim","FTS_STALE_META_KEY","ftsState","WeakMap","stateFor","conn","state","get","getMeta","stale","set","hasFtsArtifact","stmt","prepare","undefined","setMeta","splitTerms","words","bareWords","phraseWords","substrings","phrases","emptyPhrase","_m","inner","phrase","phraseTokens","searchTokens","length","push","unspaced","split","tok","rebuildFts","exec","withTransaction","duckdbFts","createIndex","err","SenseError","message","buildScoreAndGate","fts","nativeTerms","scoreParts","gateParts","scoreParams","gateParams","field","orderedBm25Sql","nativeGate","conjunctiveGate","map","foldForSearch","join","fallback","_word","flatMap","word","raw","needle","toLowerCase","scoreSql","gateSql","params","opts","unsupported","whereJoin","whereCond","scopeCond","limit","selectFields","sqlLimit","sql","rows","matches","nativeQueryTerms","all","filter","row","every","some","matchesSearchPhrase","slice","path","query","markStale"],"mappings":"AAAA,yFAAyF;AACzF,8DAA8D;;;;;;;;;;;;QAsIxCA;eAAAA;;QAuFNC;eAAAA;;QAtJMC;eAAAA;;QA2DAC;eAAAA;;QA4DAC;eAAAA;;;wBA7LK;yBACwD;wBAClD;6BACD;wBAEkB;6BACD;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAEjD,IAAMC,SAAS;IAAC;IAAS;IAAW;CAAO;AAE3C,uGAAuG;AACvG,+GAA+G;AAC/G,IAAMC,eAAsC;IAAEC,OAAO;IAAIC,SAAS;IAAGC,MAAM;AAAE;AAE7E,0GAA0G;AAC1G,8HAA8H;AAC9H,IAAMC,iBAAuD;IAC3D;QAAEC,OAAO;QAAgBC,IAAI;IAA4B;IACzD;QAAED,OAAO;QAAoBC,IAAI;IAAiC;IAClE;QAAED,OAAO;QAAiBC,IAAI;IAAiB;IAC/C;QAAED,OAAO;QAA0BC,IAAI;IAAgB;IACvD;QAAED,OAAO;QAAiBC,IAAI;IAA2B;IACzD;QAAED,OAAO;QAAqBC,IAAI;IAAO;CAC1C;AAED,SAASC,oBAAoBC,KAAa;IACxC,IAAMC,iBAAiBD,MAAME,OAAO,CAAC,YAAY;QAC5C,kCAAA,2BAAA;;QAAL,QAAK,YAAuBN,mCAAvB,SAAA,6BAAA,QAAA,yBAAA,iCAAuC;YAAvC,kBAAA,aAAQC,oBAAAA,OAAOC,iBAAAA;YAClB,IAAMK,IAAIF,eAAeG,KAAK,CAACN;YAC/B,IAAIK,GAAG,OAAO;gBAAEN,OAAAA;gBAAOQ,OAAOF,CAAC,CAAC,EAAE,CAACG,IAAI;YAAG;QAC5C;;QAHK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAIL,OAAO;AACT;AAMA,kGAAkG;AAClG,kGAAkG;AAClG,IAAMC,qBAAqB;AAE3B,6FAA6F;AAC7F,0GAA0G;AAC1G,IAAMC,WAAW,IAAIC;AACrB,SAAeC,SAASC,IAAgB;;YAClCC;;;;oBAAAA,QAAQJ,SAASK,GAAG,CAACF;yBACrB,CAACC,OAAD;;;;;oBAGgB;;wBAAME,IAAAA,iBAAO,EAACH,MAAMJ;;;oBAFtC,2FAA2F;oBAC3F,oFAAoF;oBACpFK,cAAUG,QAAO,AAAC,kBAA6C;oBAC/DP,SAASQ,GAAG,CAACL,MAAMC;;;oBAErB;;wBAAOA;;;;IACT;;AAEA,SAAeK,eAAeN,IAAgB;;YACtCO;;;;oBAAO;;wBAAMP,KAAKQ,OAAO,CAAC;;;oBAA1BD,OAAO;oBAQL;;wBAAMA,KAAKL,GAAG;;;oBAAtB;;wBAAQ,kBAAsBO;;;;IAChC;;AAIO,SAAehC,iBAAiBuB,IAAgB;;;;;oBACrDH,SAASQ,GAAG,CAACL,MAAM;wBAAEI,OAAO;oBAAK;oBACjC;;wBAAMM,IAAAA,iBAAO,EAACV,MAAMJ,oBAAoB;;;oBAAxC;;;;;;IACF;;AAEA,wHAAwH;AACxH,8FAA8F;AAC9F,SAASe,WAAWtB,KAAa;IAC/B,IAAMuB,QAAkB,EAAE;IAC1B,IAAMC,YAAsB,EAAE;IAC9B,IAAMC,cAAwB,EAAE;IAChC,IAAMC,aAAuB,EAAE;IAC/B,IAAMC,UAA2B,EAAE;IACnC,IAAIC,cAAc;IAClB,IAAM3B,iBAAiBD,MAAME,OAAO,CAAC,cAAc,SAAC2B,IAAIC;QACtD,IAAMC,SAASD,MAAMxB,IAAI;QACzB,IAAM0B,eAAeC,IAAAA,uBAAY,EAACF;QAClC,IAAIC,aAAaE,MAAM,KAAK,GAAG;YAC7BN,cAAc;YACd,OAAO;QACT;QACAD,QAAQQ,IAAI,CAACH;YACR,kCAAA,2BAAA;;YAAL,QAAK,YAAeA,iCAAf,SAAA,6BAAA,QAAA,yBAAA,iCAA6B;gBAA7B,IAAM3B,QAAN;gBACH,IAAIA,MAAM+B,QAAQ,EAAEV,WAAWS,IAAI,CAAC9B,MAAMV,IAAI;qBACzC;oBACH4B,MAAMY,IAAI,CAAC9B,MAAMV,IAAI;oBACrB8B,YAAYU,IAAI,CAAC9B,MAAMV,IAAI;gBAC7B;YACF;;YANK;YAAA;;;qBAAA,6BAAA;oBAAA;;;oBAAA;0BAAA;;;;QAOL,OAAO;IACT;QACK,kCAAA,2BAAA;;QAAL,QAAK,YAAaM,eAAeoC,KAAK,CAAC,2BAAlC,SAAA,6BAAA,QAAA,yBAAA,iCAA0C;YAA1C,IAAMC,MAAN;YACH,IAAIA,IAAIJ,MAAM,KAAK,GAAG;gBACjB,mCAAA,4BAAA;;gBAAL,QAAK,aAAeD,IAAAA,uBAAY,EAACK,yBAA5B,UAAA,8BAAA,SAAA,0BAAA,kCAAkC;oBAAlC,IAAMjC,QAAN;oBACH,IAAIA,MAAM+B,QAAQ,EAAEV,WAAWS,IAAI,CAAC9B,MAAMV,IAAI;yBACzC;wBACH4B,MAAMY,IAAI,CAAC9B,MAAMV,IAAI;wBACrB6B,UAAUW,IAAI,CAAC9B,MAAMV,IAAI;oBAC3B;gBACF;;gBANK;gBAAA;;;yBAAA,8BAAA;wBAAA;;;wBAAA;8BAAA;;;;QAOP;;QATK;QAAA;;;iBAAA,6BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAUL,OAAO;QAAE4B,OAAAA;QAAOC,WAAAA;QAAWC,aAAAA;QAAaC,YAAAA;QAAYC,SAAAA;QAASC,aAAAA;IAAY;AAC3E;AAEA,gGAAgG;AAChG,iHAAiH;AACjH,SAAeW,WAAW5B,IAAgB,EAAEC,KAAoB;;;;;;oBAC9D;;wBAAMD,KAAK6B,IAAI,CAAC;;;oBAAhB;2BACI,CAAC5B,MAAMG,KAAK;+BAAZ;;;;oBAAiB;;wBAAME,eAAeN;;;2BAArB;;;oBAArB,UAAkD;;;oBAClDC,MAAMG,KAAK,GAAG;oBACd;;wBAAM0B,IAAAA,8BAAe,EAAC9B,MAAM;;;;;;gDACpBA,KAAK6B,IAAI;4CAAE;;gDAAME,IAAAA,mBAAS;;;4CAAhC;;gDAAM/B,QAAAA;oDAAW,cAAmBgC,WAAW;;;;4CAA/C;4CACA,0FAA0F;4CAC1F,gFAAgF;4CAChF;;gDAAMtB,IAAAA,iBAAO,EAACV,MAAMJ,oBAAoB;;;4CAAxC;;;;;;4BACF;;;;oBALA;oBAMAK,MAAMG,KAAK,GAAG;;;;;;IAChB;;AAEO,SAAe1B,WAAWsB,IAAgB;;;;;;;wBAC9BA;;oBAAM;;wBAAMD,SAASC;;;oBAAtC;;wBAAM4B;4BAAiB;;;;oBAAvB;;;;;;IACF;;AAEO,SAAerD,eAAeyB,IAAgB;;YAC7CC,OAOGgC;;;;oBAPK;;wBAAMlC,SAASC;;;oBAAvBC,QAAQ;oBACd,IAAIA,MAAMG,KAAK,EAAE,MAAM,IAAI8B,oBAAU,CAAC,mBAAmB;;;;;;;;;oBAEvD;;wBAAMlC,KAAK6B,IAAI,CAAC;;;oBAAhB;oBACM;;wBAAMvB,eAAeN;;;oBAA3B,IAAI,CAAE,eAA6B;wBACjC,MAAM,IAAIkC,oBAAU,CAAC,mBAAmB;oBAC1C;;;;;;oBACOD;oBACP,IAAIA,AAAG,YAAHA,KAAeC,oBAAU,GAAE,MAAMD;oBACrC,MAAM,IAAIC,oBAAU,CAAC,mBAAmB,AAAC,iDAAuE,OAAvB,AAACD,IAAcE,OAAO,EAAC;;;;;;;IAEpH;;AAEA,qKAAqK;AACrK,+OAA+O;AAC/O,SAASC,kBAAkBC,GAAsB,EAAEzB,KAAe,EAAE0B,WAAqB,EAAEzB,SAAmB,EAAEC,WAAqB,EAAEC,UAAoB;IACzJ,IAAMwB,aAAuB,EAAE;IAC/B,IAAMC,YAAsB,EAAE;IAC9B,IAAMC,cAAyB,EAAE;IACjC,IAAMC,aAAwB,EAAE;IAEhC,IAAI9B,MAAMW,MAAM,GAAG,GAAG;YACf,kCAAA,2BAAA;;YAAL,QAAK,YAAe3C,2BAAf,SAAA,6BAAA,QAAA,yBAAA,iCAAuB;gBAAvB,IAAM+D,QAAN;oBAEHF;gBADAF,WAAWf,IAAI,CAAC,AAAC,GAAsCoB,OAApC/D,YAAY,CAAC8D,MAAM,EAAC,kBAA+D,OAA/CC,IAAAA,6BAAc,EAACP,KAAKM,OAAOL,YAAYf,MAAM,GAAE;gBACtGkB,CAAAA,eAAAA,aAAYjB,IAAI,OAAhBiB,cAAiB,qBAAGH;YACtB;;YAHK;YAAA;;;qBAAA,6BAAA;oBAAA;;;oBAAA;0BAAA;;;;QAIL,IAAIzB,UAAUU,MAAM,GAAG,GAAG;YACxB,IAAMsB,aAAaR,IAAIS,eAAe;YACtCN,UAAUhB,IAAI,CAACqB;YACfH,WAAWlB,IAAI,CAACX,UAAUkC,GAAG,CAACC,wBAAa,EAAEC,IAAI,CAAC;QACpD;QACA,IAAInC,YAAYS,MAAM,GAAG,GAAG;gBAK1BmB;YAJA,IAAMG,cAAaR,IAAIS,eAAe;YACtC,IAAMI,WAAWpC,YAAYiC,GAAG,CAAC,SAACI;uBAAU,AAAC,IAA6F,OAA1FvE,OAAOmE,GAAG,CAAC,SAACJ;2BAAU,AAAC,wCAA6C,OAANA,OAAM;mBAASM,IAAI,CAAC,SAAQ;eAAIA,IAAI,CAAC;YACnJT,UAAUhB,IAAI,CAAC,AAAC,IAAqB0B,OAAlBL,aAAW,SAAgB,OAATK,UAAS;YAC9CR,WAAWlB,IAAI,CAACV,YAAYiC,GAAG,CAACC,wBAAa,EAAEC,IAAI,CAAC;YACpDP,CAAAA,cAAAA,YAAWlB,IAAI,OAAfkB,aAAgB,qBAAG5B,YAAYsC,OAAO,CAAC,SAACC;uBAASzE,OAAOmE,GAAG,CAAC;2BAAMC,IAAAA,wBAAa,EAACK;;;QAClF;IACF;QAEK,mCAAA,4BAAA;;;YAAA,IAAMC,MAAN;gBAOHZ;YANA,IAAMa,SAASD,IAAIE,WAAW;gBACzB,kCAAA,2BAAA;;gBAAL,QAAK,YAAe5E,2BAAf,SAAA,6BAAA,QAAA,yBAAA,iCAAuB;oBAAvB,IAAM+D,QAAN;oBACHJ,WAAWf,IAAI,CAAC,AAAC,GAAuEmB,OAArE9D,YAAY,CAAC8D,MAAM,EAAC,mDAAuD,OAANA,OAAM;oBAC9FF,YAAYjB,IAAI,CAACwB,IAAAA,wBAAa,EAACO;gBACjC;;gBAHK;gBAAA;;;yBAAA,6BAAA;wBAAA;;;wBAAA;8BAAA;;;;YAILf,UAAUhB,IAAI,CAAC,AAAC,IAA6F,OAA1F5C,OAAOmE,GAAG,CAAC,SAACJ;uBAAU,AAAC,wCAA6C,OAANA,OAAM;eAASM,IAAI,CAAC,SAAQ;YAC7GP,CAAAA,cAAAA,YAAWlB,IAAI,OAAfkB,aAAgB,qBAAG9D,OAAOmE,GAAG,CAAC;uBAAMC,IAAAA,wBAAa,EAACO;;QACpD;QARA,QAAK,aAAaxC,+BAAb,UAAA,8BAAA,SAAA,0BAAA;;QAAA;QAAA;;;iBAAA,8BAAA;gBAAA;;;gBAAA;sBAAA;;;;IAUL,OAAO;QAAE0C,UAAUlB,WAAWU,IAAI,CAAC;QAAQS,SAASlB,UAAUS,IAAI,CAAC,YAAY;QAAQU,QAAQ,AAAC,qBAAGlB,oBAAa,qBAAGC;IAAY;AACjI;AAIO,SAAe/D,aAAaqB,IAAgB,EAAEX,KAAa,EAAEuE,IAAyB;;YAwBvErD,OAvBdsD,aAIEC,WAAWC,WAAWC,WAAWC,OACmCtD,aAApEC,OAAOC,WAAWC,aAAaC,YAAYC,SAASC,aAItDoB,KACAC,mBAEgCF,oBAA9BqB,UAAUC,SAASC,QACrBO,cACAC,UACAC,KAOA7D,MACA8D,MACAC;;;;oBAxBAT,cAAczE,oBAAoBC;oBACxC,IAAIwE,gBAAgB,MAAM;wBACxB,MAAM,IAAI3B,oBAAU,CAAC,4BAA4B,AAAC,6CAAkE2B,OAAvBA,YAAY3E,KAAK,EAAC,OAAqDG,OAAhDwE,YAAYnE,KAAK,EAAC,gCAA2C,OAAbL,MAAMM,IAAI,IAAG;oBACnL;oBACQmE,YAA2CF,KAA3CE,WAAWC,YAAgCH,KAAhCG,WAAWC,YAAqBJ,KAArBI,WAAWC,QAAUL,KAAVK;oBACmCtD,cAAAA,WAAWtB,QAA/EuB,QAAoED,YAApEC,OAAOC,YAA6DF,YAA7DE,WAAWC,cAAkDH,YAAlDG,aAAaC,aAAqCJ,YAArCI,YAAYC,UAAyBL,YAAzBK,SAASC,cAAgBN,YAAhBM;oBAC5D,IAAIA,eAAgBL,MAAMW,MAAM,KAAK,KAAKR,WAAWQ,MAAM,KAAK,GAAI;;;;yBAEhEX,CAAAA,MAAMW,MAAM,GAAG,CAAA,GAAfX;;;;oBAAkB;;wBAAMrC,eAAeyB;;;oBAArB;;;oBACV;;wBAAM+B,IAAAA,mBAAS;;;oBAArBM,MAAM;yBACQzB,CAAAA,MAAMW,MAAM,GAAG,CAAA,GAAfX;;;;oBAAmB;;wBAAM2D,IAAAA,+BAAgB,EAACvE,MAAMqC,KAAKzB,MAAMmC,GAAG,CAACC,wBAAa,EAAEC,IAAI,CAAC;;;2BAAhE;;;;;;;;;oBAAjCX;oBAEgCF,qBAAAA,kBAAkBC,KAAKzB,OAAO0B,aAAazB,WAAWC,aAAaC,aAAjG0C,WAA8BrB,mBAA9BqB,UAAUC,UAAoBtB,mBAApBsB,SAASC,SAAWvB,mBAAXuB;oBACrBO,eAAelD,QAAQO,MAAM,GAAG,IAAI,mDAAmD;oBACvF4C,WAAWnD,QAAQO,MAAM,GAAG,IAAI,KAAK;oBACrC6C,MAAM,AAAC,qBACuBX,OADHzC,QAAQO,MAAM,GAAG,IAAI,2BAA2B,IAAG,iDAC3B2C,OAArBT,UAAS,cAEzCK,OAFqDI,cAAa,4BAG5DR,OADNI,WAAU,gBAEVC,OADML,SAAQ,UACDM,OAAbD,WAAU,KACoBI,OADjBH,WAAU,wCACgB,OAATG;oBACrB;;wBAAMnE,KAAKQ,OAAO,CAAC4D;;;oBAA1B7D,OAAO;oBACC;;wBAAMA,CAAAA,QAAAA,MAAKiE,GAAG,OAARjE,OAAAA,AAAS,qBAAGoD,eAAQ,qBAAI3C,QAAQO,MAAM,GAAG;4BAAU0C;;;;oBAAjEI,OAAQ;oBACRC,UAAUtD,QAAQO,MAAM,KAAK,IAAI8C,OAAOA,KAAKI,MAAM,CAAC,SAACC;+BAAQ1D,QAAQ2D,KAAK,CAAC,SAACvD;mCAAWxC,OAAOgG,IAAI,CAAC,SAACjC;oCAA8B+B;uCAApBG,IAAAA,8BAAmB,GAACH,aAAAA,GAAG,CAAC/B,MAAM,cAAV+B,wBAAAA,aAAc,IAAItD;;;;oBAC1J;;wBAAOkD,QAAQQ,KAAK,CAAC,GAAGb,OAAOlB,GAAG,CAAC;gCAAGgC,aAAAA;mCAAY;gCAAEA,MAAAA;4BAAK;;;;;IAC3D;;AAIO,SAASvG,mBAAmBwB,IAAgB;IACjD,OAAO;QACLgF,OAAO,SAAPA,MAAc3F,OAAOuE;;;;;wBAASjF,aAAaqB,MAAMX,OAAOuE;;;;;QACxDqB,WAAW,SAAXA;mBAAiBxG,iBAAiBuB;;IACpC;AACF"}
@@ -1,3 +1,6 @@
1
+ import { type NativeModule } from '../native.js';
1
2
  export declare const DUCKDB_PACKAGE = "@duckdb/node-api";
2
- export declare const DUCKDB_INSTALL_SPEC = "@duckdb/node-api@1.5.6-r.1";
3
- export declare function duckdbApi(): Promise<typeof import('@duckdb/node-api')>;
3
+ type DuckdbApi = typeof import('@duckdb/node-api');
4
+ export declare function duckdbNative(): Promise<NativeModule<DuckdbApi>>;
5
+ export declare function duckdbApi(): Promise<DuckdbApi>;
6
+ export {};
@@ -1,3 +1,6 @@
1
+ import { type NativeModule } from '../native.js';
1
2
  export declare const DUCKDB_PACKAGE = "@duckdb/node-api";
2
- export declare const DUCKDB_INSTALL_SPEC = "@duckdb/node-api@1.5.6-r.1";
3
- export declare function duckdbApi(): Promise<typeof import('@duckdb/node-api')>;
3
+ type DuckdbApi = typeof import('@duckdb/node-api');
4
+ export declare function duckdbNative(): Promise<NativeModule<DuckdbApi>>;
5
+ export declare function duckdbApi(): Promise<DuckdbApi>;
6
+ export {};