sensemaking 0.22.1 → 0.23.0

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 (90) hide show
  1. package/README.md +6 -8
  2. package/dist/cjs/chunk/extract.d.cts +3 -3
  3. package/dist/cjs/chunk/extract.d.ts +3 -3
  4. package/dist/cjs/chunk/extract.js +369 -72
  5. package/dist/cjs/chunk/extract.js.map +1 -1
  6. package/dist/cjs/chunk/group.js +2 -2
  7. package/dist/cjs/chunk/group.js.map +1 -1
  8. package/dist/cjs/chunk/parse.js +63 -67
  9. package/dist/cjs/chunk/parse.js.map +1 -1
  10. package/dist/cjs/chunk/parser.d.cts +2 -0
  11. package/dist/cjs/chunk/parser.d.ts +2 -0
  12. package/dist/cjs/chunk/parser.js +40 -0
  13. package/dist/cjs/chunk/parser.js.map +1 -0
  14. package/dist/cjs/chunk/types.d.cts +2 -2
  15. package/dist/cjs/chunk/types.d.ts +2 -2
  16. package/dist/cjs/chunk/version.d.cts +1 -1
  17. package/dist/cjs/chunk/version.d.ts +1 -1
  18. package/dist/cjs/chunk/version.js +1 -1
  19. package/dist/cjs/chunk/version.js.map +1 -1
  20. package/dist/cjs/features/links.js +67 -71
  21. package/dist/cjs/features/links.js.map +1 -1
  22. package/dist/cjs/features/sections.js +1 -1
  23. package/dist/cjs/features/sections.js.map +1 -1
  24. package/dist/cjs/scan/pool.js +1 -1
  25. package/dist/cjs/scan/pool.js.map +1 -1
  26. package/dist/cjs/store/builder.js +2 -3
  27. package/dist/cjs/store/builder.js.map +1 -1
  28. package/dist/cjs/store/duckdb/open.d.cts +1 -1
  29. package/dist/cjs/store/duckdb/open.d.ts +1 -1
  30. package/dist/cjs/store/duckdb/open.js +1 -1
  31. package/dist/cjs/store/duckdb/open.js.map +1 -1
  32. package/dist/cjs/store/sqlite/open.d.cts +1 -1
  33. package/dist/cjs/store/sqlite/open.d.ts +1 -1
  34. package/dist/cjs/store/sqlite/open.js +1 -1
  35. package/dist/cjs/store/sqlite/open.js.map +1 -1
  36. package/dist/cjs/store/turso/connection.d.cts +1 -0
  37. package/dist/cjs/store/turso/connection.d.ts +1 -0
  38. package/dist/cjs/store/turso/connection.js +134 -13
  39. package/dist/cjs/store/turso/connection.js.map +1 -1
  40. package/dist/cjs/store/turso/open.d.cts +1 -1
  41. package/dist/cjs/store/turso/open.d.ts +1 -1
  42. package/dist/cjs/store/turso/open.js +8 -2
  43. package/dist/cjs/store/turso/open.js.map +1 -1
  44. package/dist/cjs/store/turso/store.js +8 -1
  45. package/dist/cjs/store/turso/store.js.map +1 -1
  46. package/dist/cjs/text/strip.js +1 -3
  47. package/dist/cjs/text/strip.js.map +1 -1
  48. package/dist/cjs/workers/parse.js.map +1 -1
  49. package/dist/esm/chunk/extract.d.ts +3 -3
  50. package/dist/esm/chunk/extract.js +276 -63
  51. package/dist/esm/chunk/extract.js.map +1 -1
  52. package/dist/esm/chunk/group.js +2 -2
  53. package/dist/esm/chunk/group.js.map +1 -1
  54. package/dist/esm/chunk/parse.js +65 -64
  55. package/dist/esm/chunk/parse.js.map +1 -1
  56. package/dist/esm/chunk/parser.d.ts +2 -0
  57. package/dist/esm/chunk/parser.js +26 -0
  58. package/dist/esm/chunk/parser.js.map +1 -0
  59. package/dist/esm/chunk/types.d.ts +2 -2
  60. package/dist/esm/chunk/types.js.map +1 -1
  61. package/dist/esm/chunk/version.d.ts +1 -1
  62. package/dist/esm/chunk/version.js +1 -1
  63. package/dist/esm/chunk/version.js.map +1 -1
  64. package/dist/esm/features/links.js +59 -22
  65. package/dist/esm/features/links.js.map +1 -1
  66. package/dist/esm/features/sections.js +1 -1
  67. package/dist/esm/features/sections.js.map +1 -1
  68. package/dist/esm/scan/pool.js +1 -1
  69. package/dist/esm/scan/pool.js.map +1 -1
  70. package/dist/esm/store/builder.js +2 -3
  71. package/dist/esm/store/builder.js.map +1 -1
  72. package/dist/esm/store/duckdb/open.d.ts +1 -1
  73. package/dist/esm/store/duckdb/open.js +1 -1
  74. package/dist/esm/store/duckdb/open.js.map +1 -1
  75. package/dist/esm/store/sqlite/open.d.ts +1 -1
  76. package/dist/esm/store/sqlite/open.js +1 -1
  77. package/dist/esm/store/sqlite/open.js.map +1 -1
  78. package/dist/esm/store/turso/connection.d.ts +1 -0
  79. package/dist/esm/store/turso/connection.js +21 -8
  80. package/dist/esm/store/turso/connection.js.map +1 -1
  81. package/dist/esm/store/turso/open.d.ts +1 -1
  82. package/dist/esm/store/turso/open.js +3 -2
  83. package/dist/esm/store/turso/open.js.map +1 -1
  84. package/dist/esm/store/turso/store.js +2 -0
  85. package/dist/esm/store/turso/store.js.map +1 -1
  86. package/dist/esm/text/strip.js +1 -1
  87. package/dist/esm/text/strip.js.map +1 -1
  88. package/dist/esm/workers/parse.js.map +1 -1
  89. package/package.json +5 -19
  90. package/skills/sense-setup/SKILL.md +1 -1
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/scan/pool.ts"],"sourcesContent":["import { createRequire } from 'node:module';\nimport { dirname, join } from 'node:path';\n// Type-only: erased at build, but keeps depcheck's usage check satisfied for the tier-2\n// `_require` below (see coding-standards' deferral tiers).\nimport type * as TinypoolNS from 'tinypool';\nimport type { Config } from '../config/index.ts';\nimport type { Feature } from '../features/types.ts';\nimport type { ParsedDoc } from '../scan/index.ts';\nimport type { ParseTask, ParseTaskResult, ParseWorkerData } from '../workers/parse.ts';\nimport type { FileStat } from './list.ts';\nimport { reviveError } from './worker-error.ts';\n\n// Tinypool is ESM-only; our floor (>=22.20) has native require(esm), so the tier-2 house deferral reaches it.\nconst _require = typeof require === 'undefined' ? createRequire(import.meta.url) : require;\n\n// Worker MUST load from dist/cjs/: a worker_threads thread is a fresh realm with no inherited\n// TS loader hook, so a source path cannot work. Resolved on first pooled dispatch, not at import.\nlet workerFile: string | undefined;\nfunction resolveWorkerFile(): string {\n if (workerFile) return workerFile;\n const load = createRequire(import.meta.url);\n for (const rel of ['..', '../..', '../../..']) {\n try {\n if ((load(`${rel}/package.json`) as { name?: string }).name === 'sensemaking') {\n workerFile = join(dirname(load.resolve(`${rel}/package.json`)), 'dist', 'cjs', 'workers', 'parse.js');\n return workerFile;\n }\n } catch {}\n }\n throw new Error('cannot locate the sensemaking package root, so the parse worker cannot be found; run npm run build');\n}\n\n// parseMs is worker-side only (see workers/parse.ts); absent on the serial path.\nexport type FileResult = { doc: ParsedDoc; warnings: string[]; parseMs?: number };\n\n// One Tinypool per instance, created lazily on first dispatch and reused by every later call:\n// a builder owns one of these for its whole lifetime instead of paying pool startup per reconcile.\nexport class ParsePool {\n private pool: TinypoolNS.Tinypool | undefined;\n // Tinypools this instance has constructed. One dispatch or a hundred should leave it at 1;\n // it is how a caller, and the specs, observe that a lifetime reuses its pool rather than churning one.\n poolsCreated = 0;\n\n private ensure(workerData: ParseWorkerData, maxWorkers: number): TinypoolNS.Tinypool {\n if (!this.pool) {\n const { Tinypool } = _require('tinypool') as typeof TinypoolNS;\n this.pool = new Tinypool({ filename: resolveWorkerFile(), minThreads: maxWorkers, maxThreads: maxWorkers, workerData });\n this.poolsCreated++;\n }\n return this.pool;\n }\n\n // Never the tree: a worker task carries one FileStat and returns only what parseFile returns --\n // extracted text and feature values, never the mdast tree.\n async run(files: FileStat[], features: Feature[], cfg: Config, onParsed: ((done: number) => void) | undefined, maxWorkers: number): Promise<FileResult[]> {\n // cfg and the feature selection are constant for the dispatch, so they cross once per worker\n // as workerData. Features carry closures and cannot cross at all; only their names do.\n const workerData: ParseWorkerData = { cfg, featureNames: features.map((feature) => feature.name) };\n const pool = this.ensure(workerData, maxWorkers);\n let done = 0;\n // Promise.all over a mapped array is load-bearing: the resolved array keeps `files` order\n // regardless of task completion order, which first-seen column order depends on downstream.\n return Promise.all(\n files.map(async (file): Promise<FileResult> => {\n const result = (await pool.run(file as ParseTask)) as ParseTaskResult;\n if (!result.ok) throw reviveError(result.error);\n onParsed?.(++done);\n return { doc: result.doc, warnings: result.warnings, parseMs: result.parseMs };\n })\n );\n }\n\n // A pool never created costs nothing to destroy.\n async close(): Promise<void> {\n if (!this.pool) return;\n const pool = this.pool;\n this.pool = undefined;\n await pool.destroy();\n }\n}\n"],"names":["createRequire","dirname","join","reviveError","_require","require","url","workerFile","resolveWorkerFile","load","rel","name","resolve","Error","ParsePool","ensure","workerData","maxWorkers","pool","Tinypool","filename","minThreads","maxThreads","poolsCreated","run","files","features","cfg","onParsed","featureNames","map","feature","done","Promise","all","file","result","ok","error","doc","warnings","parseMs","close","undefined","destroy"],"mappings":"AAAA,SAASA,aAAa,QAAQ,cAAc;AAC5C,SAASC,OAAO,EAAEC,IAAI,QAAQ,YAAY;AAS1C,SAASC,WAAW,QAAQ,oBAAoB;AAEhD,8GAA8G;AAC9G,MAAMC,WAAW,OAAOC,YAAY,cAAcL,cAAc,YAAYM,GAAG,IAAID;AAEnF,8FAA8F;AAC9F,kGAAkG;AAClG,IAAIE;AACJ,SAASC;IACP,IAAID,YAAY,OAAOA;IACvB,MAAME,OAAOT,cAAc,YAAYM,GAAG;IAC1C,KAAK,MAAMI,OAAO;QAAC;QAAM;QAAS;KAAW,CAAE;QAC7C,IAAI;YACF,IAAI,AAACD,KAAK,GAAGC,IAAI,aAAa,CAAC,EAAwBC,IAAI,KAAK,eAAe;gBAC7EJ,aAAaL,KAAKD,QAAQQ,KAAKG,OAAO,CAAC,GAAGF,IAAI,aAAa,CAAC,IAAI,QAAQ,OAAO,WAAW;gBAC1F,OAAOH;YACT;QACF,EAAE,OAAM,CAAC;IACX;IACA,MAAM,IAAIM,MAAM;AAClB;AAKA,8FAA8F;AAC9F,mGAAmG;AACnG,OAAO,MAAMC;IAMHC,OAAOC,UAA2B,EAAEC,UAAkB,EAAuB;QACnF,IAAI,CAAC,IAAI,CAACC,IAAI,EAAE;YACd,MAAM,EAAEC,QAAQ,EAAE,GAAGf,SAAS;YAC9B,IAAI,CAACc,IAAI,GAAG,IAAIC,SAAS;gBAAEC,UAAUZ;gBAAqBa,YAAYJ;gBAAYK,YAAYL;gBAAYD;YAAW;YACrH,IAAI,CAACO,YAAY;QACnB;QACA,OAAO,IAAI,CAACL,IAAI;IAClB;IAEA,gGAAgG;IAChG,2DAA2D;IAC3D,MAAMM,IAAIC,KAAiB,EAAEC,QAAmB,EAAEC,GAAW,EAAEC,QAA8C,EAAEX,UAAkB,EAAyB;QACxJ,6FAA6F;QAC7F,uFAAuF;QACvF,MAAMD,aAA8B;YAAEW;YAAKE,cAAcH,SAASI,GAAG,CAAC,CAACC,UAAYA,QAAQpB,IAAI;QAAE;QACjG,MAAMO,OAAO,IAAI,CAACH,MAAM,CAACC,YAAYC;QACrC,IAAIe,OAAO;QACX,0FAA0F;QAC1F,4FAA4F;QAC5F,OAAOC,QAAQC,GAAG,CAChBT,MAAMK,GAAG,CAAC,OAAOK;YACf,MAAMC,SAAU,MAAMlB,KAAKM,GAAG,CAACW;YAC/B,IAAI,CAACC,OAAOC,EAAE,EAAE,MAAMlC,YAAYiC,OAAOE,KAAK;YAC9CV,qBAAAA,+BAAAA,SAAW,EAAEI;YACb,OAAO;gBAAEO,KAAKH,OAAOG,GAAG;gBAAEC,UAAUJ,OAAOI,QAAQ;gBAAEC,SAASL,OAAOK,OAAO;YAAC;QAC/E;IAEJ;IAEA,iDAAiD;IACjD,MAAMC,QAAuB;QAC3B,IAAI,CAAC,IAAI,CAACxB,IAAI,EAAE;QAChB,MAAMA,OAAO,IAAI,CAACA,IAAI;QACtB,IAAI,CAACA,IAAI,GAAGyB;QACZ,MAAMzB,KAAK0B,OAAO;IACpB;;QAvCA,2FAA2F;QAC3F,uGAAuG;aACvGrB,eAAe;;AAsCjB"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/scan/pool.ts"],"sourcesContent":["import { createRequire } from 'node:module';\nimport { dirname, join } from 'node:path';\n// Type-only: erased at build, but keeps depcheck's usage check satisfied for the tier-2\n// `_require` below (see coding-standards' deferral tiers).\nimport type * as TinypoolNS from 'tinypool';\nimport type { Config } from '../config/index.ts';\nimport type { Feature } from '../features/types.ts';\nimport type { ParsedDoc } from '../scan/index.ts';\nimport type { ParseTask, ParseTaskResult, ParseWorkerData } from '../workers/parse.ts';\nimport type { FileStat } from './list.ts';\nimport { reviveError } from './worker-error.ts';\n\n// Tinypool is ESM-only; our floor (>=22.20) has native require(esm), so the tier-2 house deferral reaches it.\nconst _require = typeof require === 'undefined' ? createRequire(import.meta.url) : require;\n\n// Worker MUST load from dist/cjs/: a worker_threads thread is a fresh realm with no inherited\n// TS loader hook, so a source path cannot work. Resolved on first pooled dispatch, not at import.\nlet workerFile: string | undefined;\nfunction resolveWorkerFile(): string {\n if (workerFile) return workerFile;\n const load = createRequire(import.meta.url);\n for (const rel of ['..', '../..', '../../..']) {\n try {\n if ((load(`${rel}/package.json`) as { name?: string }).name === 'sensemaking') {\n workerFile = join(dirname(load.resolve(`${rel}/package.json`)), 'dist', 'cjs', 'workers', 'parse.js');\n return workerFile;\n }\n } catch {}\n }\n throw new Error('cannot locate the sensemaking package root, so the parse worker cannot be found; run npm run build');\n}\n\n// parseMs is worker-side only (see workers/parse.ts); absent on the serial path.\nexport type FileResult = { doc: ParsedDoc; warnings: string[]; parseMs?: number };\n\n// One Tinypool per instance, created lazily on first dispatch and reused by every later call:\n// a builder owns one of these for its whole lifetime instead of paying pool startup per reconcile.\nexport class ParsePool {\n private pool: TinypoolNS.Tinypool | undefined;\n // Tinypools this instance has constructed. One dispatch or a hundred should leave it at 1;\n // it is how a caller, and the specs, observe that a lifetime reuses its pool rather than churning one.\n poolsCreated = 0;\n\n private ensure(workerData: ParseWorkerData, maxWorkers: number): TinypoolNS.Tinypool {\n if (!this.pool) {\n const { Tinypool } = _require('tinypool') as typeof TinypoolNS;\n this.pool = new Tinypool({ filename: resolveWorkerFile(), minThreads: maxWorkers, maxThreads: maxWorkers, workerData });\n this.poolsCreated++;\n }\n return this.pool;\n }\n\n // Never the tree: a worker task carries one FileStat and returns only what parseFile returns --\n // extracted text and feature values, never the token tree.\n async run(files: FileStat[], features: Feature[], cfg: Config, onParsed: ((done: number) => void) | undefined, maxWorkers: number): Promise<FileResult[]> {\n // cfg and the feature selection are constant for the dispatch, so they cross once per worker\n // as workerData. Features carry closures and cannot cross at all; only their names do.\n const workerData: ParseWorkerData = { cfg, featureNames: features.map((feature) => feature.name) };\n const pool = this.ensure(workerData, maxWorkers);\n let done = 0;\n // Promise.all over a mapped array is load-bearing: the resolved array keeps `files` order\n // regardless of task completion order, which first-seen column order depends on downstream.\n return Promise.all(\n files.map(async (file): Promise<FileResult> => {\n const result = (await pool.run(file as ParseTask)) as ParseTaskResult;\n if (!result.ok) throw reviveError(result.error);\n onParsed?.(++done);\n return { doc: result.doc, warnings: result.warnings, parseMs: result.parseMs };\n })\n );\n }\n\n // A pool never created costs nothing to destroy.\n async close(): Promise<void> {\n if (!this.pool) return;\n const pool = this.pool;\n this.pool = undefined;\n await pool.destroy();\n }\n}\n"],"names":["createRequire","dirname","join","reviveError","_require","require","url","workerFile","resolveWorkerFile","load","rel","name","resolve","Error","ParsePool","ensure","workerData","maxWorkers","pool","Tinypool","filename","minThreads","maxThreads","poolsCreated","run","files","features","cfg","onParsed","featureNames","map","feature","done","Promise","all","file","result","ok","error","doc","warnings","parseMs","close","undefined","destroy"],"mappings":"AAAA,SAASA,aAAa,QAAQ,cAAc;AAC5C,SAASC,OAAO,EAAEC,IAAI,QAAQ,YAAY;AAS1C,SAASC,WAAW,QAAQ,oBAAoB;AAEhD,8GAA8G;AAC9G,MAAMC,WAAW,OAAOC,YAAY,cAAcL,cAAc,YAAYM,GAAG,IAAID;AAEnF,8FAA8F;AAC9F,kGAAkG;AAClG,IAAIE;AACJ,SAASC;IACP,IAAID,YAAY,OAAOA;IACvB,MAAME,OAAOT,cAAc,YAAYM,GAAG;IAC1C,KAAK,MAAMI,OAAO;QAAC;QAAM;QAAS;KAAW,CAAE;QAC7C,IAAI;YACF,IAAI,AAACD,KAAK,GAAGC,IAAI,aAAa,CAAC,EAAwBC,IAAI,KAAK,eAAe;gBAC7EJ,aAAaL,KAAKD,QAAQQ,KAAKG,OAAO,CAAC,GAAGF,IAAI,aAAa,CAAC,IAAI,QAAQ,OAAO,WAAW;gBAC1F,OAAOH;YACT;QACF,EAAE,OAAM,CAAC;IACX;IACA,MAAM,IAAIM,MAAM;AAClB;AAKA,8FAA8F;AAC9F,mGAAmG;AACnG,OAAO,MAAMC;IAMHC,OAAOC,UAA2B,EAAEC,UAAkB,EAAuB;QACnF,IAAI,CAAC,IAAI,CAACC,IAAI,EAAE;YACd,MAAM,EAAEC,QAAQ,EAAE,GAAGf,SAAS;YAC9B,IAAI,CAACc,IAAI,GAAG,IAAIC,SAAS;gBAAEC,UAAUZ;gBAAqBa,YAAYJ;gBAAYK,YAAYL;gBAAYD;YAAW;YACrH,IAAI,CAACO,YAAY;QACnB;QACA,OAAO,IAAI,CAACL,IAAI;IAClB;IAEA,gGAAgG;IAChG,2DAA2D;IAC3D,MAAMM,IAAIC,KAAiB,EAAEC,QAAmB,EAAEC,GAAW,EAAEC,QAA8C,EAAEX,UAAkB,EAAyB;QACxJ,6FAA6F;QAC7F,uFAAuF;QACvF,MAAMD,aAA8B;YAAEW;YAAKE,cAAcH,SAASI,GAAG,CAAC,CAACC,UAAYA,QAAQpB,IAAI;QAAE;QACjG,MAAMO,OAAO,IAAI,CAACH,MAAM,CAACC,YAAYC;QACrC,IAAIe,OAAO;QACX,0FAA0F;QAC1F,4FAA4F;QAC5F,OAAOC,QAAQC,GAAG,CAChBT,MAAMK,GAAG,CAAC,OAAOK;YACf,MAAMC,SAAU,MAAMlB,KAAKM,GAAG,CAACW;YAC/B,IAAI,CAACC,OAAOC,EAAE,EAAE,MAAMlC,YAAYiC,OAAOE,KAAK;YAC9CV,qBAAAA,+BAAAA,SAAW,EAAEI;YACb,OAAO;gBAAEO,KAAKH,OAAOG,GAAG;gBAAEC,UAAUJ,OAAOI,QAAQ;gBAAEC,SAASL,OAAOK,OAAO;YAAC;QAC/E;IAEJ;IAEA,iDAAiD;IACjD,MAAMC,QAAuB;QAC3B,IAAI,CAAC,IAAI,CAACxB,IAAI,EAAE;QAChB,MAAMA,OAAO,IAAI,CAACA,IAAI;QACtB,IAAI,CAACA,IAAI,GAAGyB;QACZ,MAAMzB,KAAK0B,OAAO;IACpB;;QAvCA,2FAA2F;QAC3F,uGAAuG;aACvGrB,eAAe;;AAsCjB"}
@@ -76,9 +76,8 @@ async function invalidateFeatureTable(conn, cfg, baseDir, dialect, pool, toggle)
76
76
  extracted: d.extracted[feature.name]
77
77
  }));
78
78
  const paths = docs.map((d)=>d.relPath);
79
- // Every path counts as added, not just touched: the table was just dropped, so there is nothing
80
- // stale to diff against, and links' own afterReconcile takes its cold-build (resolveAll) path
81
- // only when added.length equals files.length.
79
+ // Every path counts as added, not just touched: nothing stale to diff against. Links resolves
80
+ // dst inside store() itself when added.length equals files.length (a cold build).
82
81
  const delta = {
83
82
  files,
84
83
  reparsed: paths,
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/builder.ts"],"sourcesContent":["import type { Config } from '../config/index.ts';\nimport { embed } from '../features/embed.ts';\nimport { FEATURES } from '../features/index.ts';\nimport { rank } from '../features/rank.ts';\nimport type { ReconcileDelta } from '../features/types.ts';\nimport { listFiles } from '../scan/index.ts';\nimport { ParsePool } from '../scan/pool.ts';\nimport { reparseFiles } from '../scan/reparse.ts';\nimport type { EmbedChangeKind } from './embed-scope.ts';\nimport type { FeatureToggle } from './feature-scope.ts';\nimport { NARROW_FEATURE_TABLE } from './feature-scope.ts';\nimport { reconcile } from './reconcile.ts';\nimport { getColumns } from './shared.ts';\nimport type { Stages } from './stages.ts';\nimport { withTransaction } from './transaction.ts';\nimport type { Connection, ReconcileDialect } from './types.ts';\n\n// Owns bringing a store's index current: the write half `Store` (types.ts) deliberately does not\n// carry. A one-shot open calls build() once; a watcher calls it repeatedly on the same instance.\nexport interface Builder {\n // forcedPaths (open.ts's preset-only narrow rebuild) applies to this call alone, not future ones.\n build(forcedPaths?: ReadonlySet<string>): Promise<{ parsed: number; warnings: string[]; stages: Stages }>;\n // Narrow embed invalidation (open.ts, embed-scope.ts's classifyEmbedChange): 'model' nulls every\n // vector/scale in place; 'chunk' rebuilds every embedding row. Neither touches another table.\n invalidate(kind: EmbedChangeKind): Promise<{ parsed: number }>;\n // Narrow feature-toggle invalidation (open.ts, feature-scope.ts's classifyFeatureToggles): a\n // toggle drops (or, turning back on, fully re-derives) that feature's own table; rank instead\n // nulls or recomputes its `_rank` column, since rows written while a feature was off are\n // untrustworthy (reconcile.ts's activeFeatures skips a disabled feature's hooks entirely).\n invalidateFeatures(toggles: FeatureToggle[]): Promise<{ parsed: number }>;\n // Releases the parse worker pool, if this lifetime ever created one; never the connection.\n close(): Promise<void>;\n // Pools this lifetime has constructed, so reuse is observable rather than inferred.\n readonly poolsCreated: number;\n}\n\n// 'model': the provider or model name moved but chunk boundaries did not, so only the vector\n// values are stale. No reparse, and no table but embeddings is touched.\nasync function nullEmbedVectors(conn: Connection, dialect: ReconcileDialect): Promise<{ parsed: number }> {\n await withTransaction(conn, () => conn.exec('UPDATE embeddings SET vector = NULL, scale = NULL'), dialect.beginMode());\n return { parsed: 0 };\n}\n\n// 'chunk': chunkTokens or the chunk version moved, so chunk boundaries genuinely changed. Rebuilds\n// every embedding row through the embed feature's own remove/store, touching no other table.\nasync function rebuildEmbeddings(conn: Connection, cfg: Config, baseDir: string, dialect: ReconcileDialect, pool: ParsePool): Promise<{ parsed: number }> {\n const existingStmt = await conn.prepare('SELECT \"path\" FROM frontmatter');\n const existingPaths = new Set((await existingStmt.all()).map((r) => (r as { path: string }).path));\n // A file added since the last build has no row yet; leaving it for build() to insert avoids a\n // primary-key collision between this INSERT and build()'s own insert for the same new file.\n const files = listFiles(cfg, baseDir).filter((f) => f.embed && existingPaths.has(f.relPath));\n if (files.length === 0) return { parsed: 0 };\n\n const { docs } = await reparseFiles(files, [embed], cfg, new Set(), undefined, { pool });\n const paths = docs.map((d) => d.relPath);\n const delta: ReconcileDelta = { files, reparsed: paths, added: [], vanished: [] };\n await withTransaction(\n conn,\n async () => {\n await embed.remove?.(conn, paths, delta);\n await embed.store?.(\n conn,\n docs.map((d) => ({ path: d.relPath, extracted: d.extracted.embed })),\n delta\n );\n },\n dialect.beginMode()\n );\n return { parsed: docs.length };\n}\n\n// One feature's own table, dropped and (turning on) fully re-derived across every indexed file.\n// Rows left over from before this feature was disabled are untrustworthy by construction (files\n// added, edited, or deleted got no remove/store for it), so this never diffs against them --\n// dropping every row and reparsing the whole tree is the only safe path back to consistent state.\nasync function invalidateFeatureTable(conn: Connection, cfg: Config, baseDir: string, dialect: ReconcileDialect, pool: ParsePool, toggle: FeatureToggle): Promise<{ parsed: number }> {\n const table = NARROW_FEATURE_TABLE[toggle.name];\n const feature = FEATURES.find((f) => f.name === toggle.name);\n if (!feature) throw new Error(`invalidateFeatureTable: no registered feature named \"${toggle.name}\"`);\n\n if (!toggle.turnedOn) {\n await withTransaction(conn, () => conn.exec(`DELETE FROM ${table}`), dialect.beginMode());\n return { parsed: 0 };\n }\n\n const files = listFiles(cfg, baseDir);\n const { docs } = await reparseFiles(files, [feature], cfg, new Set(), undefined, { pool });\n const docsForFeature = docs.map((d) => ({ path: d.relPath, extracted: d.extracted[feature.name] }));\n const paths = docs.map((d) => d.relPath);\n // Every path counts as added, not just touched: the table was just dropped, so there is nothing\n // stale to diff against, and links' own afterReconcile takes its cold-build (resolveAll) path\n // only when added.length equals files.length.\n const delta: ReconcileDelta = { files, reparsed: paths, added: paths, vanished: [] };\n await withTransaction(\n conn,\n async () => {\n await conn.exec(`DELETE FROM ${table}`);\n await feature.store?.(conn, docsForFeature, delta);\n await feature.afterReconcile?.(conn, delta);\n },\n dialect.beginMode()\n );\n return { parsed: docs.length };\n}\n\n// rank has no table of its own: frontmatter._rank is the only state it owns. Guarded on column\n// presence, since links can toggle off while rank has never been enabled for this tree.\nasync function nullRank(conn: Connection, dialect: ReconcileDialect): Promise<void> {\n const columns = await getColumns(conn);\n if (!columns.has('_rank')) return;\n await withTransaction(conn, () => conn.exec('UPDATE frontmatter SET \"_rank\" = NULL'), dialect.beginMode());\n}\n\n// Whole-tree and reparse-free: PageRank is computed from the links table, not from files, so\n// turning rank back on needs no reparseFiles pass at all.\nasync function rerunRank(conn: Connection, dialect: ReconcileDialect): Promise<void> {\n const delta: ReconcileDelta = { files: [], reparsed: [], added: [], vanished: [] };\n await withTransaction(\n conn,\n async () => {\n await rank.afterReconcile?.(conn, delta);\n },\n dialect.beginMode()\n );\n}\n\nasync function invalidateFeatureToggles(conn: Connection, cfg: Config, baseDir: string, dialect: ReconcileDialect, pool: ParsePool, toggles: FeatureToggle[]): Promise<{ parsed: number }> {\n const byName = new Map(toggles.map((toggle) => [toggle.name, toggle]));\n let parsed = 0;\n\n // links first, if present: rank's rerun below (whether from its own toggle or the co-change\n // rankToggle carries) must read the links table after it is back in its new state.\n const linksToggle = byName.get('links');\n if (linksToggle) {\n parsed += (await invalidateFeatureTable(conn, cfg, baseDir, dialect, pool, linksToggle)).parsed;\n // Unconditional, per rank's dependency on links: nulling is a no-op (guarded) when rank was\n // never enabled for this tree, so this is safe even when no feature:rank segment changed.\n if (!linksToggle.turnedOn) await nullRank(conn, dialect);\n }\n\n // rank's effective on/off (featureEnabled's links dependency) only appears as a changed\n // `feature:rank` segment when it actually flips -- whether that's rank's own config or a\n // links toggle taking it along -- so honoring it here is complete on its own.\n const rankToggle = byName.get('rank');\n if (rankToggle) {\n if (rankToggle.turnedOn) await rerunRank(conn, dialect);\n else await nullRank(conn, dialect);\n }\n\n for (const toggle of toggles) {\n if (toggle.name === 'links' || toggle.name === 'rank') continue;\n parsed += (await invalidateFeatureTable(conn, cfg, baseDir, dialect, pool, toggle)).parsed;\n }\n return { parsed };\n}\n\n// The pool is created at most once, lazily, on whichever build() or invalidate() call first\n// needs it, and reused by every later call on this instance.\nexport function createBuilder(conn: Connection, cfg: Config, baseDir: string, dialect: ReconcileDialect): Builder {\n const pool = new ParsePool();\n return {\n build: (forcedPaths) => reconcile(conn, cfg, baseDir, dialect, pool, forcedPaths),\n invalidate: (kind) => (kind === 'model' ? nullEmbedVectors(conn, dialect) : rebuildEmbeddings(conn, cfg, baseDir, dialect, pool)),\n invalidateFeatures: (toggles) => invalidateFeatureToggles(conn, cfg, baseDir, dialect, pool, toggles),\n close: () => pool.close(),\n get poolsCreated() {\n return pool.poolsCreated;\n },\n };\n}\n"],"names":["embed","FEATURES","rank","listFiles","ParsePool","reparseFiles","NARROW_FEATURE_TABLE","reconcile","getColumns","withTransaction","nullEmbedVectors","conn","dialect","exec","beginMode","parsed","rebuildEmbeddings","cfg","baseDir","pool","existingStmt","prepare","existingPaths","Set","all","map","r","path","files","filter","f","has","relPath","length","docs","undefined","paths","d","delta","reparsed","added","vanished","remove","store","extracted","invalidateFeatureTable","toggle","table","name","feature","find","Error","turnedOn","docsForFeature","afterReconcile","nullRank","columns","rerunRank","invalidateFeatureToggles","toggles","byName","Map","linksToggle","get","rankToggle","createBuilder","build","forcedPaths","invalidate","kind","invalidateFeatures","close","poolsCreated"],"mappings":"AACA,SAASA,KAAK,QAAQ,uBAAuB;AAC7C,SAASC,QAAQ,QAAQ,uBAAuB;AAChD,SAASC,IAAI,QAAQ,sBAAsB;AAE3C,SAASC,SAAS,QAAQ,mBAAmB;AAC7C,SAASC,SAAS,QAAQ,kBAAkB;AAC5C,SAASC,YAAY,QAAQ,qBAAqB;AAGlD,SAASC,oBAAoB,QAAQ,qBAAqB;AAC1D,SAASC,SAAS,QAAQ,iBAAiB;AAC3C,SAASC,UAAU,QAAQ,cAAc;AAEzC,SAASC,eAAe,QAAQ,mBAAmB;AAsBnD,6FAA6F;AAC7F,wEAAwE;AACxE,eAAeC,iBAAiBC,IAAgB,EAAEC,OAAyB;IACzE,MAAMH,gBAAgBE,MAAM,IAAMA,KAAKE,IAAI,CAAC,sDAAsDD,QAAQE,SAAS;IACnH,OAAO;QAAEC,QAAQ;IAAE;AACrB;AAEA,mGAAmG;AACnG,6FAA6F;AAC7F,eAAeC,kBAAkBL,IAAgB,EAAEM,GAAW,EAAEC,OAAe,EAAEN,OAAyB,EAAEO,IAAe;IACzH,MAAMC,eAAe,MAAMT,KAAKU,OAAO,CAAC;IACxC,MAAMC,gBAAgB,IAAIC,IAAI,AAAC,CAAA,MAAMH,aAAaI,GAAG,EAAC,EAAGC,GAAG,CAAC,CAACC,IAAM,AAACA,EAAuBC,IAAI;IAChG,8FAA8F;IAC9F,4FAA4F;IAC5F,MAAMC,QAAQzB,UAAUc,KAAKC,SAASW,MAAM,CAAC,CAACC,IAAMA,EAAE9B,KAAK,IAAIsB,cAAcS,GAAG,CAACD,EAAEE,OAAO;IAC1F,IAAIJ,MAAMK,MAAM,KAAK,GAAG,OAAO;QAAElB,QAAQ;IAAE;IAE3C,MAAM,EAAEmB,IAAI,EAAE,GAAG,MAAM7B,aAAauB,OAAO;QAAC5B;KAAM,EAAEiB,KAAK,IAAIM,OAAOY,WAAW;QAAEhB;IAAK;IACtF,MAAMiB,QAAQF,KAAKT,GAAG,CAAC,CAACY,IAAMA,EAAEL,OAAO;IACvC,MAAMM,QAAwB;QAAEV;QAAOW,UAAUH;QAAOI,OAAO,EAAE;QAAEC,UAAU,EAAE;IAAC;IAChF,MAAMhC,gBACJE,MACA;YACQX,eACAA;QADN,QAAMA,gBAAAA,MAAM0C,MAAM,cAAZ1C,oCAAAA,mBAAAA,OAAeW,MAAMyB,OAAOE;QAClC,QAAMtC,eAAAA,MAAM2C,KAAK,cAAX3C,mCAAAA,kBAAAA,OACJW,MACAuB,KAAKT,GAAG,CAAC,CAACY,IAAO,CAAA;gBAAEV,MAAMU,EAAEL,OAAO;gBAAEY,WAAWP,EAAEO,SAAS,CAAC5C,KAAK;YAAC,CAAA,IACjEsC;IAEJ,GACA1B,QAAQE,SAAS;IAEnB,OAAO;QAAEC,QAAQmB,KAAKD,MAAM;IAAC;AAC/B;AAEA,gGAAgG;AAChG,gGAAgG;AAChG,6FAA6F;AAC7F,kGAAkG;AAClG,eAAeY,uBAAuBlC,IAAgB,EAAEM,GAAW,EAAEC,OAAe,EAAEN,OAAyB,EAAEO,IAAe,EAAE2B,MAAqB;IACrJ,MAAMC,QAAQzC,oBAAoB,CAACwC,OAAOE,IAAI,CAAC;IAC/C,MAAMC,UAAUhD,SAASiD,IAAI,CAAC,CAACpB,IAAMA,EAAEkB,IAAI,KAAKF,OAAOE,IAAI;IAC3D,IAAI,CAACC,SAAS,MAAM,IAAIE,MAAM,CAAC,qDAAqD,EAAEL,OAAOE,IAAI,CAAC,CAAC,CAAC;IAEpG,IAAI,CAACF,OAAOM,QAAQ,EAAE;QACpB,MAAM3C,gBAAgBE,MAAM,IAAMA,KAAKE,IAAI,CAAC,CAAC,YAAY,EAAEkC,OAAO,GAAGnC,QAAQE,SAAS;QACtF,OAAO;YAAEC,QAAQ;QAAE;IACrB;IAEA,MAAMa,QAAQzB,UAAUc,KAAKC;IAC7B,MAAM,EAAEgB,IAAI,EAAE,GAAG,MAAM7B,aAAauB,OAAO;QAACqB;KAAQ,EAAEhC,KAAK,IAAIM,OAAOY,WAAW;QAAEhB;IAAK;IACxF,MAAMkC,iBAAiBnB,KAAKT,GAAG,CAAC,CAACY,IAAO,CAAA;YAAEV,MAAMU,EAAEL,OAAO;YAAEY,WAAWP,EAAEO,SAAS,CAACK,QAAQD,IAAI,CAAC;QAAC,CAAA;IAChG,MAAMZ,QAAQF,KAAKT,GAAG,CAAC,CAACY,IAAMA,EAAEL,OAAO;IACvC,gGAAgG;IAChG,8FAA8F;IAC9F,8CAA8C;IAC9C,MAAMM,QAAwB;QAAEV;QAAOW,UAAUH;QAAOI,OAAOJ;QAAOK,UAAU,EAAE;IAAC;IACnF,MAAMhC,gBACJE,MACA;YAEQsC,gBACAA;QAFN,MAAMtC,KAAKE,IAAI,CAAC,CAAC,YAAY,EAAEkC,OAAO;QACtC,QAAME,iBAAAA,QAAQN,KAAK,cAAbM,qCAAAA,oBAAAA,SAAgBtC,MAAM0C,gBAAgBf;QAC5C,QAAMW,0BAAAA,QAAQK,cAAc,cAAtBL,8CAAAA,6BAAAA,SAAyBtC,MAAM2B;IACvC,GACA1B,QAAQE,SAAS;IAEnB,OAAO;QAAEC,QAAQmB,KAAKD,MAAM;IAAC;AAC/B;AAEA,+FAA+F;AAC/F,wFAAwF;AACxF,eAAesB,SAAS5C,IAAgB,EAAEC,OAAyB;IACjE,MAAM4C,UAAU,MAAMhD,WAAWG;IACjC,IAAI,CAAC6C,QAAQzB,GAAG,CAAC,UAAU;IAC3B,MAAMtB,gBAAgBE,MAAM,IAAMA,KAAKE,IAAI,CAAC,0CAA0CD,QAAQE,SAAS;AACzG;AAEA,6FAA6F;AAC7F,0DAA0D;AAC1D,eAAe2C,UAAU9C,IAAgB,EAAEC,OAAyB;IAClE,MAAM0B,QAAwB;QAAEV,OAAO,EAAE;QAAEW,UAAU,EAAE;QAAEC,OAAO,EAAE;QAAEC,UAAU,EAAE;IAAC;IACjF,MAAMhC,gBACJE,MACA;YACQT;QAAN,QAAMA,uBAAAA,KAAKoD,cAAc,cAAnBpD,2CAAAA,0BAAAA,MAAsBS,MAAM2B;IACpC,GACA1B,QAAQE,SAAS;AAErB;AAEA,eAAe4C,yBAAyB/C,IAAgB,EAAEM,GAAW,EAAEC,OAAe,EAAEN,OAAyB,EAAEO,IAAe,EAAEwC,OAAwB;IAC1J,MAAMC,SAAS,IAAIC,IAAIF,QAAQlC,GAAG,CAAC,CAACqB,SAAW;YAACA,OAAOE,IAAI;YAAEF;SAAO;IACpE,IAAI/B,SAAS;IAEb,4FAA4F;IAC5F,mFAAmF;IACnF,MAAM+C,cAAcF,OAAOG,GAAG,CAAC;IAC/B,IAAID,aAAa;QACf/C,UAAU,AAAC,CAAA,MAAM8B,uBAAuBlC,MAAMM,KAAKC,SAASN,SAASO,MAAM2C,YAAW,EAAG/C,MAAM;QAC/F,4FAA4F;QAC5F,0FAA0F;QAC1F,IAAI,CAAC+C,YAAYV,QAAQ,EAAE,MAAMG,SAAS5C,MAAMC;IAClD;IAEA,wFAAwF;IACxF,yFAAyF;IACzF,8EAA8E;IAC9E,MAAMoD,aAAaJ,OAAOG,GAAG,CAAC;IAC9B,IAAIC,YAAY;QACd,IAAIA,WAAWZ,QAAQ,EAAE,MAAMK,UAAU9C,MAAMC;aAC1C,MAAM2C,SAAS5C,MAAMC;IAC5B;IAEA,KAAK,MAAMkC,UAAUa,QAAS;QAC5B,IAAIb,OAAOE,IAAI,KAAK,WAAWF,OAAOE,IAAI,KAAK,QAAQ;QACvDjC,UAAU,AAAC,CAAA,MAAM8B,uBAAuBlC,MAAMM,KAAKC,SAASN,SAASO,MAAM2B,OAAM,EAAG/B,MAAM;IAC5F;IACA,OAAO;QAAEA;IAAO;AAClB;AAEA,4FAA4F;AAC5F,6DAA6D;AAC7D,OAAO,SAASkD,cAActD,IAAgB,EAAEM,GAAW,EAAEC,OAAe,EAAEN,OAAyB;IACrG,MAAMO,OAAO,IAAIf;IACjB,OAAO;QACL8D,OAAO,CAACC,cAAgB5D,UAAUI,MAAMM,KAAKC,SAASN,SAASO,MAAMgD;QACrEC,YAAY,CAACC,OAAUA,SAAS,UAAU3D,iBAAiBC,MAAMC,WAAWI,kBAAkBL,MAAMM,KAAKC,SAASN,SAASO;QAC3HmD,oBAAoB,CAACX,UAAYD,yBAAyB/C,MAAMM,KAAKC,SAASN,SAASO,MAAMwC;QAC7FY,OAAO,IAAMpD,KAAKoD,KAAK;QACvB,IAAIC;YACF,OAAOrD,KAAKqD,YAAY;QAC1B;IACF;AACF"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/builder.ts"],"sourcesContent":["import type { Config } from '../config/index.ts';\nimport { embed } from '../features/embed.ts';\nimport { FEATURES } from '../features/index.ts';\nimport { rank } from '../features/rank.ts';\nimport type { ReconcileDelta } from '../features/types.ts';\nimport { listFiles } from '../scan/index.ts';\nimport { ParsePool } from '../scan/pool.ts';\nimport { reparseFiles } from '../scan/reparse.ts';\nimport type { EmbedChangeKind } from './embed-scope.ts';\nimport type { FeatureToggle } from './feature-scope.ts';\nimport { NARROW_FEATURE_TABLE } from './feature-scope.ts';\nimport { reconcile } from './reconcile.ts';\nimport { getColumns } from './shared.ts';\nimport type { Stages } from './stages.ts';\nimport { withTransaction } from './transaction.ts';\nimport type { Connection, ReconcileDialect } from './types.ts';\n\n// Owns bringing a store's index current: the write half `Store` (types.ts) deliberately does not\n// carry. A one-shot open calls build() once; a watcher calls it repeatedly on the same instance.\nexport interface Builder {\n // forcedPaths (open.ts's preset-only narrow rebuild) applies to this call alone, not future ones.\n build(forcedPaths?: ReadonlySet<string>): Promise<{ parsed: number; warnings: string[]; stages: Stages }>;\n // Narrow embed invalidation (open.ts, embed-scope.ts's classifyEmbedChange): 'model' nulls every\n // vector/scale in place; 'chunk' rebuilds every embedding row. Neither touches another table.\n invalidate(kind: EmbedChangeKind): Promise<{ parsed: number }>;\n // Narrow feature-toggle invalidation (open.ts, feature-scope.ts's classifyFeatureToggles): a\n // toggle drops (or, turning back on, fully re-derives) that feature's own table; rank instead\n // nulls or recomputes its `_rank` column, since rows written while a feature was off are\n // untrustworthy (reconcile.ts's activeFeatures skips a disabled feature's hooks entirely).\n invalidateFeatures(toggles: FeatureToggle[]): Promise<{ parsed: number }>;\n // Releases the parse worker pool, if this lifetime ever created one; never the connection.\n close(): Promise<void>;\n // Pools this lifetime has constructed, so reuse is observable rather than inferred.\n readonly poolsCreated: number;\n}\n\n// 'model': the provider or model name moved but chunk boundaries did not, so only the vector\n// values are stale. No reparse, and no table but embeddings is touched.\nasync function nullEmbedVectors(conn: Connection, dialect: ReconcileDialect): Promise<{ parsed: number }> {\n await withTransaction(conn, () => conn.exec('UPDATE embeddings SET vector = NULL, scale = NULL'), dialect.beginMode());\n return { parsed: 0 };\n}\n\n// 'chunk': chunkTokens or the chunk version moved, so chunk boundaries genuinely changed. Rebuilds\n// every embedding row through the embed feature's own remove/store, touching no other table.\nasync function rebuildEmbeddings(conn: Connection, cfg: Config, baseDir: string, dialect: ReconcileDialect, pool: ParsePool): Promise<{ parsed: number }> {\n const existingStmt = await conn.prepare('SELECT \"path\" FROM frontmatter');\n const existingPaths = new Set((await existingStmt.all()).map((r) => (r as { path: string }).path));\n // A file added since the last build has no row yet; leaving it for build() to insert avoids a\n // primary-key collision between this INSERT and build()'s own insert for the same new file.\n const files = listFiles(cfg, baseDir).filter((f) => f.embed && existingPaths.has(f.relPath));\n if (files.length === 0) return { parsed: 0 };\n\n const { docs } = await reparseFiles(files, [embed], cfg, new Set(), undefined, { pool });\n const paths = docs.map((d) => d.relPath);\n const delta: ReconcileDelta = { files, reparsed: paths, added: [], vanished: [] };\n await withTransaction(\n conn,\n async () => {\n await embed.remove?.(conn, paths, delta);\n await embed.store?.(\n conn,\n docs.map((d) => ({ path: d.relPath, extracted: d.extracted.embed })),\n delta\n );\n },\n dialect.beginMode()\n );\n return { parsed: docs.length };\n}\n\n// One feature's own table, dropped and (turning on) fully re-derived across every indexed file.\n// Rows left over from before this feature was disabled are untrustworthy by construction (files\n// added, edited, or deleted got no remove/store for it), so this never diffs against them --\n// dropping every row and reparsing the whole tree is the only safe path back to consistent state.\nasync function invalidateFeatureTable(conn: Connection, cfg: Config, baseDir: string, dialect: ReconcileDialect, pool: ParsePool, toggle: FeatureToggle): Promise<{ parsed: number }> {\n const table = NARROW_FEATURE_TABLE[toggle.name];\n const feature = FEATURES.find((f) => f.name === toggle.name);\n if (!feature) throw new Error(`invalidateFeatureTable: no registered feature named \"${toggle.name}\"`);\n\n if (!toggle.turnedOn) {\n await withTransaction(conn, () => conn.exec(`DELETE FROM ${table}`), dialect.beginMode());\n return { parsed: 0 };\n }\n\n const files = listFiles(cfg, baseDir);\n const { docs } = await reparseFiles(files, [feature], cfg, new Set(), undefined, { pool });\n const docsForFeature = docs.map((d) => ({ path: d.relPath, extracted: d.extracted[feature.name] }));\n const paths = docs.map((d) => d.relPath);\n // Every path counts as added, not just touched: nothing stale to diff against. Links resolves\n // dst inside store() itself when added.length equals files.length (a cold build).\n const delta: ReconcileDelta = { files, reparsed: paths, added: paths, vanished: [] };\n await withTransaction(\n conn,\n async () => {\n await conn.exec(`DELETE FROM ${table}`);\n await feature.store?.(conn, docsForFeature, delta);\n await feature.afterReconcile?.(conn, delta);\n },\n dialect.beginMode()\n );\n return { parsed: docs.length };\n}\n\n// rank has no table of its own: frontmatter._rank is the only state it owns. Guarded on column\n// presence, since links can toggle off while rank has never been enabled for this tree.\nasync function nullRank(conn: Connection, dialect: ReconcileDialect): Promise<void> {\n const columns = await getColumns(conn);\n if (!columns.has('_rank')) return;\n await withTransaction(conn, () => conn.exec('UPDATE frontmatter SET \"_rank\" = NULL'), dialect.beginMode());\n}\n\n// Whole-tree and reparse-free: PageRank is computed from the links table, not from files, so\n// turning rank back on needs no reparseFiles pass at all.\nasync function rerunRank(conn: Connection, dialect: ReconcileDialect): Promise<void> {\n const delta: ReconcileDelta = { files: [], reparsed: [], added: [], vanished: [] };\n await withTransaction(\n conn,\n async () => {\n await rank.afterReconcile?.(conn, delta);\n },\n dialect.beginMode()\n );\n}\n\nasync function invalidateFeatureToggles(conn: Connection, cfg: Config, baseDir: string, dialect: ReconcileDialect, pool: ParsePool, toggles: FeatureToggle[]): Promise<{ parsed: number }> {\n const byName = new Map(toggles.map((toggle) => [toggle.name, toggle]));\n let parsed = 0;\n\n // links first, if present: rank's rerun below (whether from its own toggle or the co-change\n // rankToggle carries) must read the links table after it is back in its new state.\n const linksToggle = byName.get('links');\n if (linksToggle) {\n parsed += (await invalidateFeatureTable(conn, cfg, baseDir, dialect, pool, linksToggle)).parsed;\n // Unconditional, per rank's dependency on links: nulling is a no-op (guarded) when rank was\n // never enabled for this tree, so this is safe even when no feature:rank segment changed.\n if (!linksToggle.turnedOn) await nullRank(conn, dialect);\n }\n\n // rank's effective on/off (featureEnabled's links dependency) only appears as a changed\n // `feature:rank` segment when it actually flips -- whether that's rank's own config or a\n // links toggle taking it along -- so honoring it here is complete on its own.\n const rankToggle = byName.get('rank');\n if (rankToggle) {\n if (rankToggle.turnedOn) await rerunRank(conn, dialect);\n else await nullRank(conn, dialect);\n }\n\n for (const toggle of toggles) {\n if (toggle.name === 'links' || toggle.name === 'rank') continue;\n parsed += (await invalidateFeatureTable(conn, cfg, baseDir, dialect, pool, toggle)).parsed;\n }\n return { parsed };\n}\n\n// The pool is created at most once, lazily, on whichever build() or invalidate() call first\n// needs it, and reused by every later call on this instance.\nexport function createBuilder(conn: Connection, cfg: Config, baseDir: string, dialect: ReconcileDialect): Builder {\n const pool = new ParsePool();\n return {\n build: (forcedPaths) => reconcile(conn, cfg, baseDir, dialect, pool, forcedPaths),\n invalidate: (kind) => (kind === 'model' ? nullEmbedVectors(conn, dialect) : rebuildEmbeddings(conn, cfg, baseDir, dialect, pool)),\n invalidateFeatures: (toggles) => invalidateFeatureToggles(conn, cfg, baseDir, dialect, pool, toggles),\n close: () => pool.close(),\n get poolsCreated() {\n return pool.poolsCreated;\n },\n };\n}\n"],"names":["embed","FEATURES","rank","listFiles","ParsePool","reparseFiles","NARROW_FEATURE_TABLE","reconcile","getColumns","withTransaction","nullEmbedVectors","conn","dialect","exec","beginMode","parsed","rebuildEmbeddings","cfg","baseDir","pool","existingStmt","prepare","existingPaths","Set","all","map","r","path","files","filter","f","has","relPath","length","docs","undefined","paths","d","delta","reparsed","added","vanished","remove","store","extracted","invalidateFeatureTable","toggle","table","name","feature","find","Error","turnedOn","docsForFeature","afterReconcile","nullRank","columns","rerunRank","invalidateFeatureToggles","toggles","byName","Map","linksToggle","get","rankToggle","createBuilder","build","forcedPaths","invalidate","kind","invalidateFeatures","close","poolsCreated"],"mappings":"AACA,SAASA,KAAK,QAAQ,uBAAuB;AAC7C,SAASC,QAAQ,QAAQ,uBAAuB;AAChD,SAASC,IAAI,QAAQ,sBAAsB;AAE3C,SAASC,SAAS,QAAQ,mBAAmB;AAC7C,SAASC,SAAS,QAAQ,kBAAkB;AAC5C,SAASC,YAAY,QAAQ,qBAAqB;AAGlD,SAASC,oBAAoB,QAAQ,qBAAqB;AAC1D,SAASC,SAAS,QAAQ,iBAAiB;AAC3C,SAASC,UAAU,QAAQ,cAAc;AAEzC,SAASC,eAAe,QAAQ,mBAAmB;AAsBnD,6FAA6F;AAC7F,wEAAwE;AACxE,eAAeC,iBAAiBC,IAAgB,EAAEC,OAAyB;IACzE,MAAMH,gBAAgBE,MAAM,IAAMA,KAAKE,IAAI,CAAC,sDAAsDD,QAAQE,SAAS;IACnH,OAAO;QAAEC,QAAQ;IAAE;AACrB;AAEA,mGAAmG;AACnG,6FAA6F;AAC7F,eAAeC,kBAAkBL,IAAgB,EAAEM,GAAW,EAAEC,OAAe,EAAEN,OAAyB,EAAEO,IAAe;IACzH,MAAMC,eAAe,MAAMT,KAAKU,OAAO,CAAC;IACxC,MAAMC,gBAAgB,IAAIC,IAAI,AAAC,CAAA,MAAMH,aAAaI,GAAG,EAAC,EAAGC,GAAG,CAAC,CAACC,IAAM,AAACA,EAAuBC,IAAI;IAChG,8FAA8F;IAC9F,4FAA4F;IAC5F,MAAMC,QAAQzB,UAAUc,KAAKC,SAASW,MAAM,CAAC,CAACC,IAAMA,EAAE9B,KAAK,IAAIsB,cAAcS,GAAG,CAACD,EAAEE,OAAO;IAC1F,IAAIJ,MAAMK,MAAM,KAAK,GAAG,OAAO;QAAElB,QAAQ;IAAE;IAE3C,MAAM,EAAEmB,IAAI,EAAE,GAAG,MAAM7B,aAAauB,OAAO;QAAC5B;KAAM,EAAEiB,KAAK,IAAIM,OAAOY,WAAW;QAAEhB;IAAK;IACtF,MAAMiB,QAAQF,KAAKT,GAAG,CAAC,CAACY,IAAMA,EAAEL,OAAO;IACvC,MAAMM,QAAwB;QAAEV;QAAOW,UAAUH;QAAOI,OAAO,EAAE;QAAEC,UAAU,EAAE;IAAC;IAChF,MAAMhC,gBACJE,MACA;YACQX,eACAA;QADN,QAAMA,gBAAAA,MAAM0C,MAAM,cAAZ1C,oCAAAA,mBAAAA,OAAeW,MAAMyB,OAAOE;QAClC,QAAMtC,eAAAA,MAAM2C,KAAK,cAAX3C,mCAAAA,kBAAAA,OACJW,MACAuB,KAAKT,GAAG,CAAC,CAACY,IAAO,CAAA;gBAAEV,MAAMU,EAAEL,OAAO;gBAAEY,WAAWP,EAAEO,SAAS,CAAC5C,KAAK;YAAC,CAAA,IACjEsC;IAEJ,GACA1B,QAAQE,SAAS;IAEnB,OAAO;QAAEC,QAAQmB,KAAKD,MAAM;IAAC;AAC/B;AAEA,gGAAgG;AAChG,gGAAgG;AAChG,6FAA6F;AAC7F,kGAAkG;AAClG,eAAeY,uBAAuBlC,IAAgB,EAAEM,GAAW,EAAEC,OAAe,EAAEN,OAAyB,EAAEO,IAAe,EAAE2B,MAAqB;IACrJ,MAAMC,QAAQzC,oBAAoB,CAACwC,OAAOE,IAAI,CAAC;IAC/C,MAAMC,UAAUhD,SAASiD,IAAI,CAAC,CAACpB,IAAMA,EAAEkB,IAAI,KAAKF,OAAOE,IAAI;IAC3D,IAAI,CAACC,SAAS,MAAM,IAAIE,MAAM,CAAC,qDAAqD,EAAEL,OAAOE,IAAI,CAAC,CAAC,CAAC;IAEpG,IAAI,CAACF,OAAOM,QAAQ,EAAE;QACpB,MAAM3C,gBAAgBE,MAAM,IAAMA,KAAKE,IAAI,CAAC,CAAC,YAAY,EAAEkC,OAAO,GAAGnC,QAAQE,SAAS;QACtF,OAAO;YAAEC,QAAQ;QAAE;IACrB;IAEA,MAAMa,QAAQzB,UAAUc,KAAKC;IAC7B,MAAM,EAAEgB,IAAI,EAAE,GAAG,MAAM7B,aAAauB,OAAO;QAACqB;KAAQ,EAAEhC,KAAK,IAAIM,OAAOY,WAAW;QAAEhB;IAAK;IACxF,MAAMkC,iBAAiBnB,KAAKT,GAAG,CAAC,CAACY,IAAO,CAAA;YAAEV,MAAMU,EAAEL,OAAO;YAAEY,WAAWP,EAAEO,SAAS,CAACK,QAAQD,IAAI,CAAC;QAAC,CAAA;IAChG,MAAMZ,QAAQF,KAAKT,GAAG,CAAC,CAACY,IAAMA,EAAEL,OAAO;IACvC,8FAA8F;IAC9F,kFAAkF;IAClF,MAAMM,QAAwB;QAAEV;QAAOW,UAAUH;QAAOI,OAAOJ;QAAOK,UAAU,EAAE;IAAC;IACnF,MAAMhC,gBACJE,MACA;YAEQsC,gBACAA;QAFN,MAAMtC,KAAKE,IAAI,CAAC,CAAC,YAAY,EAAEkC,OAAO;QACtC,QAAME,iBAAAA,QAAQN,KAAK,cAAbM,qCAAAA,oBAAAA,SAAgBtC,MAAM0C,gBAAgBf;QAC5C,QAAMW,0BAAAA,QAAQK,cAAc,cAAtBL,8CAAAA,6BAAAA,SAAyBtC,MAAM2B;IACvC,GACA1B,QAAQE,SAAS;IAEnB,OAAO;QAAEC,QAAQmB,KAAKD,MAAM;IAAC;AAC/B;AAEA,+FAA+F;AAC/F,wFAAwF;AACxF,eAAesB,SAAS5C,IAAgB,EAAEC,OAAyB;IACjE,MAAM4C,UAAU,MAAMhD,WAAWG;IACjC,IAAI,CAAC6C,QAAQzB,GAAG,CAAC,UAAU;IAC3B,MAAMtB,gBAAgBE,MAAM,IAAMA,KAAKE,IAAI,CAAC,0CAA0CD,QAAQE,SAAS;AACzG;AAEA,6FAA6F;AAC7F,0DAA0D;AAC1D,eAAe2C,UAAU9C,IAAgB,EAAEC,OAAyB;IAClE,MAAM0B,QAAwB;QAAEV,OAAO,EAAE;QAAEW,UAAU,EAAE;QAAEC,OAAO,EAAE;QAAEC,UAAU,EAAE;IAAC;IACjF,MAAMhC,gBACJE,MACA;YACQT;QAAN,QAAMA,uBAAAA,KAAKoD,cAAc,cAAnBpD,2CAAAA,0BAAAA,MAAsBS,MAAM2B;IACpC,GACA1B,QAAQE,SAAS;AAErB;AAEA,eAAe4C,yBAAyB/C,IAAgB,EAAEM,GAAW,EAAEC,OAAe,EAAEN,OAAyB,EAAEO,IAAe,EAAEwC,OAAwB;IAC1J,MAAMC,SAAS,IAAIC,IAAIF,QAAQlC,GAAG,CAAC,CAACqB,SAAW;YAACA,OAAOE,IAAI;YAAEF;SAAO;IACpE,IAAI/B,SAAS;IAEb,4FAA4F;IAC5F,mFAAmF;IACnF,MAAM+C,cAAcF,OAAOG,GAAG,CAAC;IAC/B,IAAID,aAAa;QACf/C,UAAU,AAAC,CAAA,MAAM8B,uBAAuBlC,MAAMM,KAAKC,SAASN,SAASO,MAAM2C,YAAW,EAAG/C,MAAM;QAC/F,4FAA4F;QAC5F,0FAA0F;QAC1F,IAAI,CAAC+C,YAAYV,QAAQ,EAAE,MAAMG,SAAS5C,MAAMC;IAClD;IAEA,wFAAwF;IACxF,yFAAyF;IACzF,8EAA8E;IAC9E,MAAMoD,aAAaJ,OAAOG,GAAG,CAAC;IAC9B,IAAIC,YAAY;QACd,IAAIA,WAAWZ,QAAQ,EAAE,MAAMK,UAAU9C,MAAMC;aAC1C,MAAM2C,SAAS5C,MAAMC;IAC5B;IAEA,KAAK,MAAMkC,UAAUa,QAAS;QAC5B,IAAIb,OAAOE,IAAI,KAAK,WAAWF,OAAOE,IAAI,KAAK,QAAQ;QACvDjC,UAAU,AAAC,CAAA,MAAM8B,uBAAuBlC,MAAMM,KAAKC,SAASN,SAASO,MAAM2B,OAAM,EAAG/B,MAAM;IAC5F;IACA,OAAO;QAAEA;IAAO;AAClB;AAEA,4FAA4F;AAC5F,6DAA6D;AAC7D,OAAO,SAASkD,cAActD,IAAgB,EAAEM,GAAW,EAAEC,OAAe,EAAEN,OAAyB;IACrG,MAAMO,OAAO,IAAIf;IACjB,OAAO;QACL8D,OAAO,CAACC,cAAgB5D,UAAUI,MAAMM,KAAKC,SAASN,SAASO,MAAMgD;QACrEC,YAAY,CAACC,OAAUA,SAAS,UAAU3D,iBAAiBC,MAAMC,WAAWI,kBAAkBL,MAAMM,KAAKC,SAASN,SAASO;QAC3HmD,oBAAoB,CAACX,UAAYD,yBAAyB/C,MAAMM,KAAKC,SAASN,SAASO,MAAMwC;QAC7FY,OAAO,IAAMpD,KAAKoD,KAAK;QACvB,IAAIC;YACF,OAAOrD,KAAKqD,YAAY;QAC1B;IACF;AACF"}
@@ -3,7 +3,7 @@ import type { ResolvedConfig } from '../../config/index.js';
3
3
  import type { OpenResult } from '../open.js';
4
4
  import type { OpenDialect } from '../types.js';
5
5
  export declare const DB_FILENAME = "cache.duckdb";
6
- export declare const SCHEMA_VERSION = "3";
6
+ export declare const SCHEMA_VERSION = "4";
7
7
  export type { OpenResult };
8
8
  interface DuckdbHandle {
9
9
  instance: DuckDBInstance;
@@ -12,7 +12,7 @@ import { createStore } from './store.js';
12
12
  export const DB_FILENAME = 'cache.duckdb';
13
13
  // Independent of sqlite's SCHEMA_VERSION -- the two stores' cache shapes evolve separately (VARIANT frontmatter columns vs untyped).
14
14
  // The store name already joins the feature signature, so switching a config's `store` key rebuilds rather than reusing the other engine's cache.
15
- export const SCHEMA_VERSION = '3';
15
+ export const SCHEMA_VERSION = '4';
16
16
  // `content` is a plain table (not FTS-virtual): the fts index is built over it lazily, only when a lexical query runs (lexical.ts), and read directly for contains() verification either way.
17
17
  // No tokenizer resolution: this store always uses the fts extension's default (porter) stemmer.
18
18
  async function ensureSchema(_handle, conn, cfg) {
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/duckdb/open.ts"],"sourcesContent":["import type { DuckDBConnection, DuckDBInstance } from '@duckdb/node-api';\nimport type { Config, ResolvedConfig } from '../../config/index.ts';\nimport { featureSignature } from '../../config/index.ts';\nimport { STORE_DIMS } from '../../embed/types.ts';\nimport { SenseError } from '../../errors.ts';\nimport { activeFeatures, FEATURES } from '../../features/index.ts';\nimport type { OpenResult } from '../open.ts';\nimport { openWithDialect } from '../open.ts';\nimport { getMeta, setMeta } from '../shared.ts';\nimport type { Connection, OpenDialect } from '../types.ts';\nimport { createConnection } from './connection.ts';\nimport { DUCKDB_PACKAGE, duckdbApi } from './native.ts';\nimport { duckdbDialect } from './reconcile.ts';\nimport { registerFunctions } from './sql-functions.ts';\nimport { createStore } from './store.ts';\n\nexport const DB_FILENAME = 'cache.duckdb';\n// Independent of sqlite's SCHEMA_VERSION -- the two stores' cache shapes evolve separately (VARIANT frontmatter columns vs untyped).\n// The store name already joins the feature signature, so switching a config's `store` key rebuilds rather than reusing the other engine's cache.\nexport const SCHEMA_VERSION = '3';\n\nexport type { OpenResult };\n\n// The connection plus this store's own native handles (types.ts's OpenDialect<Handle>): the\n// instance owns the WAL and must outlive the connection borrowed from it.\ninterface DuckdbHandle {\n instance: DuckDBInstance;\n duckdb: DuckDBConnection;\n}\n\n// `content` is a plain table (not FTS-virtual): the fts index is built over it lazily, only when a lexical query runs (lexical.ts), and read directly for contains() verification either way.\n// No tokenizer resolution: this store always uses the fts extension's default (porter) stemmer.\nasync function ensureSchema(_handle: DuckdbHandle, conn: Connection, cfg: Config): Promise<void> {\n await conn.exec(`CREATE TABLE IF NOT EXISTS frontmatter (\"path\" TEXT PRIMARY KEY, \"_mtime\" DOUBLE, \"_ctime\" DOUBLE, \"_size\" INTEGER, \"_parse_error\" TEXT)`);\n await conn.exec(`CREATE TABLE IF NOT EXISTS content (\"path\" TEXT PRIMARY KEY, title TEXT, summary TEXT, text TEXT)`);\n await conn.exec(`CREATE TABLE IF NOT EXISTS preset_files (\"path\" TEXT, preset TEXT, PRIMARY KEY (\"path\", preset))`);\n await conn.exec('CREATE INDEX IF NOT EXISTS preset_files_preset ON preset_files(preset)');\n for (const feature of activeFeatures(cfg)) {\n // Native FLOAT[STORE_DIMS] instead of the embed feature's engine-neutral BLOB+scale DDL (vectors.ts). `scale` is kept, unused,\n // so the feature's shared reconcile-time INSERT/DELETE (features/embed.ts) names a column that exists on both stores.\n if (feature.name === 'embed') {\n await conn.exec(`CREATE TABLE IF NOT EXISTS embeddings (\"path\" TEXT, chunk INTEGER, start_line INTEGER, end_line INTEGER, scale REAL, vector FLOAT[${STORE_DIMS}], PRIMARY KEY (\"path\", chunk))`);\n continue;\n }\n await feature.schema(conn);\n }\n if ((await getMeta(conn, 'schema_version')) === null) await setMeta(conn, 'schema_version', SCHEMA_VERSION);\n if ((await getMeta(conn, 'features')) === null) await setMeta(conn, 'features', featureSignature(cfg, FEATURES));\n}\n\n// Order matters: the connection must be gone before the instance closes the WAL.\nasync function close(handle: DuckdbHandle): Promise<void> {\n handle.duckdb.disconnectSync();\n handle.instance.closeSync();\n}\n\nasync function connect(dbPath: string, _cfg: ResolvedConfig): Promise<{ handle: DuckdbHandle; conn: Connection }> {\n // Dynamic, not a top-level import: sqlite trees must never attempt to resolve this optional peer dependency, so nothing imports\n // it as a value until a duckdb tree opens (types-only imports are erased). Installed on first use if missing, shared with sql-functions.ts and vectors.ts via native.ts's duckdbApi.\n let DuckDBInstance: typeof import('@duckdb/node-api').DuckDBInstance;\n try {\n ({ DuckDBInstance } = await duckdbApi());\n } catch (err) {\n if (err instanceof SenseError) throw err;\n throw new SenseError('STORE_DEPENDENCY_MISSING', `store \"duckdb\" needs the ${DUCKDB_PACKAGE} package (${(err as Error).message})`);\n }\n\n let duckdb: DuckDBConnection;\n let instance: DuckDBInstance | undefined;\n try {\n // On-disk files default to an older storage format for cross-version compatibility, which rejects VARIANT columns (\"VARIANT\n // columns are not supported in storage versions prior to v1.5.0\"); this store's dynamic frontmatter columns need VARIANT (reconcile.ts), so the floor is pinned explicitly.\n instance = await DuckDBInstance.create(dbPath, { storage_compatibility_version: 'v1.5.0' });\n duckdb = await instance.connect();\n } catch (err) {\n // create() may have succeeded before connect() failed: close it, or its WAL stays open.\n instance?.closeSync();\n throw new SenseError('STORE_DEPENDENCY_MISSING', `store \"duckdb\" failed to open ${dbPath}: ${(err as Error).message}`);\n }\n // A throw below would leak the open instance, whose WAL then locks the .duckdb file undeletable on Windows.\n try {\n await registerFunctions(duckdb);\n const conn = createConnection(duckdb);\n return { handle: { instance, duckdb }, conn };\n } catch (err) {\n await close({ instance, duckdb });\n throw err;\n }\n}\n\n// This store's dialect (types.ts's OpenDialect) for the shared orchestration in store/open.ts.\nexport const duckdbOpenDialect: OpenDialect<DuckdbHandle> = {\n filename: DB_FILENAME,\n schemaVersion: SCHEMA_VERSION,\n reconcileDialect: duckdbDialect,\n connect,\n close,\n // duckdb's file lock spans the connection's life, and each platform words the refusal its own\n // way: posix \"Could not set lock on file ... Conflicting lock is held in <exe> (PID n)\", Windows\n // \"Cannot open file ... being used by another process\". Both are the same condition.\n isLocked: (err) => /Could not set lock on file|being used by another process/.test(err.message),\n ensureSchema,\n createStore: (handle, conn) => createStore(handle.instance, handle.duckdb, conn),\n};\n\nexport async function openDuckdb(cfg: ResolvedConfig): Promise<OpenResult> {\n return openWithDialect(cfg, duckdbOpenDialect);\n}\n"],"names":["featureSignature","STORE_DIMS","SenseError","activeFeatures","FEATURES","openWithDialect","getMeta","setMeta","createConnection","DUCKDB_PACKAGE","duckdbApi","duckdbDialect","registerFunctions","createStore","DB_FILENAME","SCHEMA_VERSION","ensureSchema","_handle","conn","cfg","exec","feature","name","schema","close","handle","duckdb","disconnectSync","instance","closeSync","connect","dbPath","_cfg","DuckDBInstance","err","message","create","storage_compatibility_version","duckdbOpenDialect","filename","schemaVersion","reconcileDialect","isLocked","test","openDuckdb"],"mappings":"AAEA,SAASA,gBAAgB,QAAQ,wBAAwB;AACzD,SAASC,UAAU,QAAQ,uBAAuB;AAClD,SAASC,UAAU,QAAQ,kBAAkB;AAC7C,SAASC,cAAc,EAAEC,QAAQ,QAAQ,0BAA0B;AAEnE,SAASC,eAAe,QAAQ,aAAa;AAC7C,SAASC,OAAO,EAAEC,OAAO,QAAQ,eAAe;AAEhD,SAASC,gBAAgB,QAAQ,kBAAkB;AACnD,SAASC,cAAc,EAAEC,SAAS,QAAQ,cAAc;AACxD,SAASC,aAAa,QAAQ,iBAAiB;AAC/C,SAASC,iBAAiB,QAAQ,qBAAqB;AACvD,SAASC,WAAW,QAAQ,aAAa;AAEzC,OAAO,MAAMC,cAAc,eAAe;AAC1C,qIAAqI;AACrI,iJAAiJ;AACjJ,OAAO,MAAMC,iBAAiB,IAAI;AAWlC,8LAA8L;AAC9L,gGAAgG;AAChG,eAAeC,aAAaC,OAAqB,EAAEC,IAAgB,EAAEC,GAAW;IAC9E,MAAMD,KAAKE,IAAI,CAAC,CAAC,wIAAwI,CAAC;IAC1J,MAAMF,KAAKE,IAAI,CAAC,CAAC,iGAAiG,CAAC;IACnH,MAAMF,KAAKE,IAAI,CAAC,CAAC,gGAAgG,CAAC;IAClH,MAAMF,KAAKE,IAAI,CAAC;IAChB,KAAK,MAAMC,WAAWlB,eAAegB,KAAM;QACzC,+HAA+H;QAC/H,sHAAsH;QACtH,IAAIE,QAAQC,IAAI,KAAK,SAAS;YAC5B,MAAMJ,KAAKE,IAAI,CAAC,CAAC,kIAAkI,EAAEnB,WAAW,+BAA+B,CAAC;YAChM;QACF;QACA,MAAMoB,QAAQE,MAAM,CAACL;IACvB;IACA,IAAI,AAAC,MAAMZ,QAAQY,MAAM,sBAAuB,MAAM,MAAMX,QAAQW,MAAM,kBAAkBH;IAC5F,IAAI,AAAC,MAAMT,QAAQY,MAAM,gBAAiB,MAAM,MAAMX,QAAQW,MAAM,YAAYlB,iBAAiBmB,KAAKf;AACxG;AAEA,iFAAiF;AACjF,eAAeoB,MAAMC,MAAoB;IACvCA,OAAOC,MAAM,CAACC,cAAc;IAC5BF,OAAOG,QAAQ,CAACC,SAAS;AAC3B;AAEA,eAAeC,QAAQC,MAAc,EAAEC,IAAoB;IACzD,gIAAgI;IAChI,qLAAqL;IACrL,IAAIC;IACJ,IAAI;QACD,CAAA,EAAEA,cAAc,EAAE,GAAG,MAAMvB,WAAU;IACxC,EAAE,OAAOwB,KAAK;QACZ,IAAIA,eAAehC,YAAY,MAAMgC;QACrC,MAAM,IAAIhC,WAAW,4BAA4B,CAAC,yBAAyB,EAAEO,eAAe,UAAU,EAAE,AAACyB,IAAcC,OAAO,CAAC,CAAC,CAAC;IACnI;IAEA,IAAIT;IACJ,IAAIE;IACJ,IAAI;QACF,4HAA4H;QAC5H,4KAA4K;QAC5KA,WAAW,MAAMK,eAAeG,MAAM,CAACL,QAAQ;YAAEM,+BAA+B;QAAS;QACzFX,SAAS,MAAME,SAASE,OAAO;IACjC,EAAE,OAAOI,KAAK;QACZ,wFAAwF;QACxFN,qBAAAA,+BAAAA,SAAUC,SAAS;QACnB,MAAM,IAAI3B,WAAW,4BAA4B,CAAC,8BAA8B,EAAE6B,OAAO,EAAE,EAAE,AAACG,IAAcC,OAAO,EAAE;IACvH;IACA,4GAA4G;IAC5G,IAAI;QACF,MAAMvB,kBAAkBc;QACxB,MAAMR,OAAOV,iBAAiBkB;QAC9B,OAAO;YAAED,QAAQ;gBAAEG;gBAAUF;YAAO;YAAGR;QAAK;IAC9C,EAAE,OAAOgB,KAAK;QACZ,MAAMV,MAAM;YAAEI;YAAUF;QAAO;QAC/B,MAAMQ;IACR;AACF;AAEA,+FAA+F;AAC/F,OAAO,MAAMI,oBAA+C;IAC1DC,UAAUzB;IACV0B,eAAezB;IACf0B,kBAAkB9B;IAClBmB;IACAN;IACA,8FAA8F;IAC9F,iGAAiG;IACjG,qFAAqF;IACrFkB,UAAU,CAACR,MAAQ,2DAA2DS,IAAI,CAACT,IAAIC,OAAO;IAC9FnB;IACAH,aAAa,CAACY,QAAQP,OAASL,YAAYY,OAAOG,QAAQ,EAAEH,OAAOC,MAAM,EAAER;AAC7E,EAAE;AAEF,OAAO,eAAe0B,WAAWzB,GAAmB;IAClD,OAAOd,gBAAgBc,KAAKmB;AAC9B"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/duckdb/open.ts"],"sourcesContent":["import type { DuckDBConnection, DuckDBInstance } from '@duckdb/node-api';\nimport type { Config, ResolvedConfig } from '../../config/index.ts';\nimport { featureSignature } from '../../config/index.ts';\nimport { STORE_DIMS } from '../../embed/types.ts';\nimport { SenseError } from '../../errors.ts';\nimport { activeFeatures, FEATURES } from '../../features/index.ts';\nimport type { OpenResult } from '../open.ts';\nimport { openWithDialect } from '../open.ts';\nimport { getMeta, setMeta } from '../shared.ts';\nimport type { Connection, OpenDialect } from '../types.ts';\nimport { createConnection } from './connection.ts';\nimport { DUCKDB_PACKAGE, duckdbApi } from './native.ts';\nimport { duckdbDialect } from './reconcile.ts';\nimport { registerFunctions } from './sql-functions.ts';\nimport { createStore } from './store.ts';\n\nexport const DB_FILENAME = 'cache.duckdb';\n// Independent of sqlite's SCHEMA_VERSION -- the two stores' cache shapes evolve separately (VARIANT frontmatter columns vs untyped).\n// The store name already joins the feature signature, so switching a config's `store` key rebuilds rather than reusing the other engine's cache.\nexport const SCHEMA_VERSION = '4';\n\nexport type { OpenResult };\n\n// The connection plus this store's own native handles (types.ts's OpenDialect<Handle>): the\n// instance owns the WAL and must outlive the connection borrowed from it.\ninterface DuckdbHandle {\n instance: DuckDBInstance;\n duckdb: DuckDBConnection;\n}\n\n// `content` is a plain table (not FTS-virtual): the fts index is built over it lazily, only when a lexical query runs (lexical.ts), and read directly for contains() verification either way.\n// No tokenizer resolution: this store always uses the fts extension's default (porter) stemmer.\nasync function ensureSchema(_handle: DuckdbHandle, conn: Connection, cfg: Config): Promise<void> {\n await conn.exec(`CREATE TABLE IF NOT EXISTS frontmatter (\"path\" TEXT PRIMARY KEY, \"_mtime\" DOUBLE, \"_ctime\" DOUBLE, \"_size\" INTEGER, \"_parse_error\" TEXT)`);\n await conn.exec(`CREATE TABLE IF NOT EXISTS content (\"path\" TEXT PRIMARY KEY, title TEXT, summary TEXT, text TEXT)`);\n await conn.exec(`CREATE TABLE IF NOT EXISTS preset_files (\"path\" TEXT, preset TEXT, PRIMARY KEY (\"path\", preset))`);\n await conn.exec('CREATE INDEX IF NOT EXISTS preset_files_preset ON preset_files(preset)');\n for (const feature of activeFeatures(cfg)) {\n // Native FLOAT[STORE_DIMS] instead of the embed feature's engine-neutral BLOB+scale DDL (vectors.ts). `scale` is kept, unused,\n // so the feature's shared reconcile-time INSERT/DELETE (features/embed.ts) names a column that exists on both stores.\n if (feature.name === 'embed') {\n await conn.exec(`CREATE TABLE IF NOT EXISTS embeddings (\"path\" TEXT, chunk INTEGER, start_line INTEGER, end_line INTEGER, scale REAL, vector FLOAT[${STORE_DIMS}], PRIMARY KEY (\"path\", chunk))`);\n continue;\n }\n await feature.schema(conn);\n }\n if ((await getMeta(conn, 'schema_version')) === null) await setMeta(conn, 'schema_version', SCHEMA_VERSION);\n if ((await getMeta(conn, 'features')) === null) await setMeta(conn, 'features', featureSignature(cfg, FEATURES));\n}\n\n// Order matters: the connection must be gone before the instance closes the WAL.\nasync function close(handle: DuckdbHandle): Promise<void> {\n handle.duckdb.disconnectSync();\n handle.instance.closeSync();\n}\n\nasync function connect(dbPath: string, _cfg: ResolvedConfig): Promise<{ handle: DuckdbHandle; conn: Connection }> {\n // Dynamic, not a top-level import: sqlite trees must never attempt to resolve this optional peer dependency, so nothing imports\n // it as a value until a duckdb tree opens (types-only imports are erased). Installed on first use if missing, shared with sql-functions.ts and vectors.ts via native.ts's duckdbApi.\n let DuckDBInstance: typeof import('@duckdb/node-api').DuckDBInstance;\n try {\n ({ DuckDBInstance } = await duckdbApi());\n } catch (err) {\n if (err instanceof SenseError) throw err;\n throw new SenseError('STORE_DEPENDENCY_MISSING', `store \"duckdb\" needs the ${DUCKDB_PACKAGE} package (${(err as Error).message})`);\n }\n\n let duckdb: DuckDBConnection;\n let instance: DuckDBInstance | undefined;\n try {\n // On-disk files default to an older storage format for cross-version compatibility, which rejects VARIANT columns (\"VARIANT\n // columns are not supported in storage versions prior to v1.5.0\"); this store's dynamic frontmatter columns need VARIANT (reconcile.ts), so the floor is pinned explicitly.\n instance = await DuckDBInstance.create(dbPath, { storage_compatibility_version: 'v1.5.0' });\n duckdb = await instance.connect();\n } catch (err) {\n // create() may have succeeded before connect() failed: close it, or its WAL stays open.\n instance?.closeSync();\n throw new SenseError('STORE_DEPENDENCY_MISSING', `store \"duckdb\" failed to open ${dbPath}: ${(err as Error).message}`);\n }\n // A throw below would leak the open instance, whose WAL then locks the .duckdb file undeletable on Windows.\n try {\n await registerFunctions(duckdb);\n const conn = createConnection(duckdb);\n return { handle: { instance, duckdb }, conn };\n } catch (err) {\n await close({ instance, duckdb });\n throw err;\n }\n}\n\n// This store's dialect (types.ts's OpenDialect) for the shared orchestration in store/open.ts.\nexport const duckdbOpenDialect: OpenDialect<DuckdbHandle> = {\n filename: DB_FILENAME,\n schemaVersion: SCHEMA_VERSION,\n reconcileDialect: duckdbDialect,\n connect,\n close,\n // duckdb's file lock spans the connection's life, and each platform words the refusal its own\n // way: posix \"Could not set lock on file ... Conflicting lock is held in <exe> (PID n)\", Windows\n // \"Cannot open file ... being used by another process\". Both are the same condition.\n isLocked: (err) => /Could not set lock on file|being used by another process/.test(err.message),\n ensureSchema,\n createStore: (handle, conn) => createStore(handle.instance, handle.duckdb, conn),\n};\n\nexport async function openDuckdb(cfg: ResolvedConfig): Promise<OpenResult> {\n return openWithDialect(cfg, duckdbOpenDialect);\n}\n"],"names":["featureSignature","STORE_DIMS","SenseError","activeFeatures","FEATURES","openWithDialect","getMeta","setMeta","createConnection","DUCKDB_PACKAGE","duckdbApi","duckdbDialect","registerFunctions","createStore","DB_FILENAME","SCHEMA_VERSION","ensureSchema","_handle","conn","cfg","exec","feature","name","schema","close","handle","duckdb","disconnectSync","instance","closeSync","connect","dbPath","_cfg","DuckDBInstance","err","message","create","storage_compatibility_version","duckdbOpenDialect","filename","schemaVersion","reconcileDialect","isLocked","test","openDuckdb"],"mappings":"AAEA,SAASA,gBAAgB,QAAQ,wBAAwB;AACzD,SAASC,UAAU,QAAQ,uBAAuB;AAClD,SAASC,UAAU,QAAQ,kBAAkB;AAC7C,SAASC,cAAc,EAAEC,QAAQ,QAAQ,0BAA0B;AAEnE,SAASC,eAAe,QAAQ,aAAa;AAC7C,SAASC,OAAO,EAAEC,OAAO,QAAQ,eAAe;AAEhD,SAASC,gBAAgB,QAAQ,kBAAkB;AACnD,SAASC,cAAc,EAAEC,SAAS,QAAQ,cAAc;AACxD,SAASC,aAAa,QAAQ,iBAAiB;AAC/C,SAASC,iBAAiB,QAAQ,qBAAqB;AACvD,SAASC,WAAW,QAAQ,aAAa;AAEzC,OAAO,MAAMC,cAAc,eAAe;AAC1C,qIAAqI;AACrI,iJAAiJ;AACjJ,OAAO,MAAMC,iBAAiB,IAAI;AAWlC,8LAA8L;AAC9L,gGAAgG;AAChG,eAAeC,aAAaC,OAAqB,EAAEC,IAAgB,EAAEC,GAAW;IAC9E,MAAMD,KAAKE,IAAI,CAAC,CAAC,wIAAwI,CAAC;IAC1J,MAAMF,KAAKE,IAAI,CAAC,CAAC,iGAAiG,CAAC;IACnH,MAAMF,KAAKE,IAAI,CAAC,CAAC,gGAAgG,CAAC;IAClH,MAAMF,KAAKE,IAAI,CAAC;IAChB,KAAK,MAAMC,WAAWlB,eAAegB,KAAM;QACzC,+HAA+H;QAC/H,sHAAsH;QACtH,IAAIE,QAAQC,IAAI,KAAK,SAAS;YAC5B,MAAMJ,KAAKE,IAAI,CAAC,CAAC,kIAAkI,EAAEnB,WAAW,+BAA+B,CAAC;YAChM;QACF;QACA,MAAMoB,QAAQE,MAAM,CAACL;IACvB;IACA,IAAI,AAAC,MAAMZ,QAAQY,MAAM,sBAAuB,MAAM,MAAMX,QAAQW,MAAM,kBAAkBH;IAC5F,IAAI,AAAC,MAAMT,QAAQY,MAAM,gBAAiB,MAAM,MAAMX,QAAQW,MAAM,YAAYlB,iBAAiBmB,KAAKf;AACxG;AAEA,iFAAiF;AACjF,eAAeoB,MAAMC,MAAoB;IACvCA,OAAOC,MAAM,CAACC,cAAc;IAC5BF,OAAOG,QAAQ,CAACC,SAAS;AAC3B;AAEA,eAAeC,QAAQC,MAAc,EAAEC,IAAoB;IACzD,gIAAgI;IAChI,qLAAqL;IACrL,IAAIC;IACJ,IAAI;QACD,CAAA,EAAEA,cAAc,EAAE,GAAG,MAAMvB,WAAU;IACxC,EAAE,OAAOwB,KAAK;QACZ,IAAIA,eAAehC,YAAY,MAAMgC;QACrC,MAAM,IAAIhC,WAAW,4BAA4B,CAAC,yBAAyB,EAAEO,eAAe,UAAU,EAAE,AAACyB,IAAcC,OAAO,CAAC,CAAC,CAAC;IACnI;IAEA,IAAIT;IACJ,IAAIE;IACJ,IAAI;QACF,4HAA4H;QAC5H,4KAA4K;QAC5KA,WAAW,MAAMK,eAAeG,MAAM,CAACL,QAAQ;YAAEM,+BAA+B;QAAS;QACzFX,SAAS,MAAME,SAASE,OAAO;IACjC,EAAE,OAAOI,KAAK;QACZ,wFAAwF;QACxFN,qBAAAA,+BAAAA,SAAUC,SAAS;QACnB,MAAM,IAAI3B,WAAW,4BAA4B,CAAC,8BAA8B,EAAE6B,OAAO,EAAE,EAAE,AAACG,IAAcC,OAAO,EAAE;IACvH;IACA,4GAA4G;IAC5G,IAAI;QACF,MAAMvB,kBAAkBc;QACxB,MAAMR,OAAOV,iBAAiBkB;QAC9B,OAAO;YAAED,QAAQ;gBAAEG;gBAAUF;YAAO;YAAGR;QAAK;IAC9C,EAAE,OAAOgB,KAAK;QACZ,MAAMV,MAAM;YAAEI;YAAUF;QAAO;QAC/B,MAAMQ;IACR;AACF;AAEA,+FAA+F;AAC/F,OAAO,MAAMI,oBAA+C;IAC1DC,UAAUzB;IACV0B,eAAezB;IACf0B,kBAAkB9B;IAClBmB;IACAN;IACA,8FAA8F;IAC9F,iGAAiG;IACjG,qFAAqF;IACrFkB,UAAU,CAACR,MAAQ,2DAA2DS,IAAI,CAACT,IAAIC,OAAO;IAC9FnB;IACAH,aAAa,CAACY,QAAQP,OAASL,YAAYY,OAAOG,QAAQ,EAAEH,OAAOC,MAAM,EAAER;AAC7E,EAAE;AAEF,OAAO,eAAe0B,WAAWzB,GAAmB;IAClD,OAAOd,gBAAgBc,KAAKmB;AAC9B"}
@@ -3,7 +3,7 @@ import type { ResolvedConfig } from '../../config/index.js';
3
3
  import type { OpenResult } from '../open.js';
4
4
  import type { OpenDialect } from '../types.js';
5
5
  export declare const DB_FILENAME = "cache.db";
6
- export declare const SCHEMA_VERSION = "19";
6
+ export declare const SCHEMA_VERSION = "20";
7
7
  export type { OpenResult };
8
8
  interface SqliteHandle {
9
9
  db: DatabaseSync;
@@ -13,7 +13,7 @@ import { createStore } from './store.js';
13
13
  export const DB_FILENAME = 'cache.db';
14
14
  // Cache shape version, independent of the config's own `version`. Bumping it rebuilds
15
15
  // existing trees on first query.
16
- export const SCHEMA_VERSION = '19';
16
+ export const SCHEMA_VERSION = '20';
17
17
  // unicode61 splits on spaces, so a language written without them indexes a whole run as one
18
18
  // token and word search finds nothing; the `_seg` sidecars (always populated) cover that case.
19
19
  const TOKENIZE = 'porter unicode61';
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/sqlite/open.ts"],"sourcesContent":["// The Node floor (>=22.20) is explained here and nowhere else: 22.20 is the first release with\n// both FTS5 and row-returning INSERT ... RETURNING. Raise it only for a load-bearing capability.\nimport { DatabaseSync } from 'node:sqlite';\nimport type { Config, ResolvedConfig } from '../../config/index.ts';\nimport { featureSignature } from '../../config/index.ts';\nimport { activeFeatures, FEATURES } from '../../features/index.ts';\nimport type { OpenResult } from '../open.ts';\nimport { openWithDialect } from '../open.ts';\nimport { getMeta, setMeta } from '../shared.ts';\nimport { BEGIN_WRITE, withTransaction } from '../transaction.ts';\nimport type { Connection, OpenDialect } from '../types.ts';\nimport { createConnection } from './connection.ts';\nimport { sqliteDialect } from './reconcile.ts';\nimport { registerFunctions } from './sql-functions.ts';\nimport { createStore } from './store.ts';\n\nexport const DB_FILENAME = 'cache.db';\n// Cache shape version, independent of the config's own `version`. Bumping it rebuilds\n// existing trees on first query.\nexport const SCHEMA_VERSION = '19';\n\nexport type { OpenResult };\n\n// unicode61 splits on spaces, so a language written without them indexes a whole run as one\n// token and word search finds nothing; the `_seg` sidecars (always populated) cover that case.\nconst TOKENIZE = 'porter unicode61';\n\ninterface SqliteHandle {\n db: DatabaseSync;\n}\n\n// The `_seg` sidecars are appended after path, never inserted: bm25() and snippet(content, 2)\n// are documented against the first three columns. Each holds its field's exploded unspaced runs.\nasync function createContentTable(conn: Connection): Promise<void> {\n await conn.exec(`CREATE VIRTUAL TABLE IF NOT EXISTS content USING fts5(title, summary, text, path UNINDEXED, title_seg, summary_seg, text_seg, tokenize = '${TOKENIZE}')`);\n}\n\n// Content is a separate table (not a column on frontmatter) so `SELECT * FROM frontmatter`\n// can't dump file text into context. Features add their own tables after the core ones.\nasync function ensureSchemaTables(conn: Connection, cfg: Config): Promise<void> {\n await conn.exec(`CREATE TABLE IF NOT EXISTS frontmatter (\"path\" TEXT PRIMARY KEY, \"_mtime\" REAL, \"_ctime\" REAL, \"_size\" INTEGER, \"_parse_error\" TEXT)`);\n await createContentTable(conn);\n // Coverage, not ownership: a path can appear under several presets. path leads the PK so the\n // per-doc delete is an index hit -- keyed the other way, cold builds went quadratic.\n await conn.exec(`CREATE TABLE IF NOT EXISTS preset_files (\"path\" TEXT, preset TEXT, PRIMARY KEY (\"path\", preset))`);\n await conn.exec('CREATE INDEX IF NOT EXISTS preset_files_preset ON preset_files(preset)');\n for (const feature of activeFeatures(cfg)) await feature.schema(conn);\n if ((await getMeta(conn, 'schema_version')) === null) await setMeta(conn, 'schema_version', SCHEMA_VERSION);\n if ((await getMeta(conn, 'features')) === null) await setMeta(conn, 'features', featureSignature(cfg, FEATURES));\n}\n\n// Two processes opening the same fresh tree both try to convert it, and the loser gets SQLITE_BUSY\n// with no busy handler behind it. Bounded because a lock held past this is a real problem, not a race.\nfunction setJournalWal(db: DatabaseSync): void {\n const deadline = Date.now() + 5000;\n for (;;) {\n try {\n db.exec('PRAGMA journal_mode = WAL');\n return;\n } catch (err) {\n if (Date.now() >= deadline || !/database is locked|busy/i.test((err as Error).message)) throw err;\n Atomics.wait(new Int32Array(new SharedArrayBuffer(4)), 0, 0, 20);\n }\n }\n}\n\nasync function connect(dbPath: string, _cfg: ResolvedConfig): Promise<{ handle: SqliteHandle; conn: Connection }> {\n const db = new DatabaseSync(dbPath);\n try {\n // Before journal_mode, not after: converting a fresh database to WAL takes a brief exclusive\n // lock, and with no timeout set yet a second process opening the same tree fails in 1ms.\n db.exec('PRAGMA busy_timeout = 30000');\n // busy_timeout does not cover the WAL conversion itself: SQLite does not invoke the busy\n // handler for it, so a concurrent cold open needs its own bounded wait.\n setJournalWal(db);\n registerFunctions(db);\n const conn = createConnection(db);\n return { handle: { db }, conn };\n } catch (err) {\n // A throw here must release this handle, or on Windows the leaked WAL db is undeletable and\n // scratch cleanup fails with EPERM/EBUSY.\n db.close();\n throw err;\n }\n}\n\nasync function close(handle: SqliteHandle): Promise<void> {\n handle.db.close();\n}\n\nasync function ensureSchema(_handle: SqliteHandle, conn: Connection, cfg: Config): Promise<void> {\n // One writer at a time: the feature hooks check a column then add it, so two cold opens racing\n // here both see it missing and the second ALTER fails with a duplicate column.\n await withTransaction(conn, () => ensureSchemaTables(conn, cfg), BEGIN_WRITE);\n}\n\nasync function setDerivedBusyTimeout(handle: SqliteHandle, _conn: Connection, ms: number): Promise<void> {\n handle.db.exec(`PRAGMA busy_timeout = ${ms}`);\n}\n\n// This store's dialect (types.ts's OpenDialect) for the shared orchestration in store/open.ts.\nexport const sqliteOpenDialect: OpenDialect<SqliteHandle> = {\n filename: DB_FILENAME,\n schemaVersion: SCHEMA_VERSION,\n reconcileDialect: sqliteDialect,\n connect,\n close,\n ensureSchema,\n setDerivedBusyTimeout,\n createStore: (handle, conn) => createStore(handle.db, conn),\n};\n\nexport async function openSqlite(cfg: ResolvedConfig): Promise<OpenResult> {\n return openWithDialect(cfg, sqliteOpenDialect);\n}\n"],"names":["DatabaseSync","featureSignature","activeFeatures","FEATURES","openWithDialect","getMeta","setMeta","BEGIN_WRITE","withTransaction","createConnection","sqliteDialect","registerFunctions","createStore","DB_FILENAME","SCHEMA_VERSION","TOKENIZE","createContentTable","conn","exec","ensureSchemaTables","cfg","feature","schema","setJournalWal","db","deadline","Date","now","err","test","message","Atomics","wait","Int32Array","SharedArrayBuffer","connect","dbPath","_cfg","handle","close","ensureSchema","_handle","setDerivedBusyTimeout","_conn","ms","sqliteOpenDialect","filename","schemaVersion","reconcileDialect","openSqlite"],"mappings":"AAAA,+FAA+F;AAC/F,iGAAiG;AACjG,SAASA,YAAY,QAAQ,cAAc;AAE3C,SAASC,gBAAgB,QAAQ,wBAAwB;AACzD,SAASC,cAAc,EAAEC,QAAQ,QAAQ,0BAA0B;AAEnE,SAASC,eAAe,QAAQ,aAAa;AAC7C,SAASC,OAAO,EAAEC,OAAO,QAAQ,eAAe;AAChD,SAASC,WAAW,EAAEC,eAAe,QAAQ,oBAAoB;AAEjE,SAASC,gBAAgB,QAAQ,kBAAkB;AACnD,SAASC,aAAa,QAAQ,iBAAiB;AAC/C,SAASC,iBAAiB,QAAQ,qBAAqB;AACvD,SAASC,WAAW,QAAQ,aAAa;AAEzC,OAAO,MAAMC,cAAc,WAAW;AACtC,sFAAsF;AACtF,iCAAiC;AACjC,OAAO,MAAMC,iBAAiB,KAAK;AAInC,4FAA4F;AAC5F,+FAA+F;AAC/F,MAAMC,WAAW;AAMjB,8FAA8F;AAC9F,iGAAiG;AACjG,eAAeC,mBAAmBC,IAAgB;IAChD,MAAMA,KAAKC,IAAI,CAAC,CAAC,0IAA0I,EAAEH,SAAS,EAAE,CAAC;AAC3K;AAEA,2FAA2F;AAC3F,wFAAwF;AACxF,eAAeI,mBAAmBF,IAAgB,EAAEG,GAAW;IAC7D,MAAMH,KAAKC,IAAI,CAAC,CAAC,oIAAoI,CAAC;IACtJ,MAAMF,mBAAmBC;IACzB,6FAA6F;IAC7F,qFAAqF;IACrF,MAAMA,KAAKC,IAAI,CAAC,CAAC,gGAAgG,CAAC;IAClH,MAAMD,KAAKC,IAAI,CAAC;IAChB,KAAK,MAAMG,WAAWnB,eAAekB,KAAM,MAAMC,QAAQC,MAAM,CAACL;IAChE,IAAI,AAAC,MAAMZ,QAAQY,MAAM,sBAAuB,MAAM,MAAMX,QAAQW,MAAM,kBAAkBH;IAC5F,IAAI,AAAC,MAAMT,QAAQY,MAAM,gBAAiB,MAAM,MAAMX,QAAQW,MAAM,YAAYhB,iBAAiBmB,KAAKjB;AACxG;AAEA,mGAAmG;AACnG,uGAAuG;AACvG,SAASoB,cAAcC,EAAgB;IACrC,MAAMC,WAAWC,KAAKC,GAAG,KAAK;IAC9B,OAAS;QACP,IAAI;YACFH,GAAGN,IAAI,CAAC;YACR;QACF,EAAE,OAAOU,KAAK;YACZ,IAAIF,KAAKC,GAAG,MAAMF,YAAY,CAAC,2BAA2BI,IAAI,CAAC,AAACD,IAAcE,OAAO,GAAG,MAAMF;YAC9FG,QAAQC,IAAI,CAAC,IAAIC,WAAW,IAAIC,kBAAkB,KAAK,GAAG,GAAG;QAC/D;IACF;AACF;AAEA,eAAeC,QAAQC,MAAc,EAAEC,IAAoB;IACzD,MAAMb,KAAK,IAAIxB,aAAaoC;IAC5B,IAAI;QACF,6FAA6F;QAC7F,yFAAyF;QACzFZ,GAAGN,IAAI,CAAC;QACR,yFAAyF;QACzF,wEAAwE;QACxEK,cAAcC;QACdb,kBAAkBa;QAClB,MAAMP,OAAOR,iBAAiBe;QAC9B,OAAO;YAAEc,QAAQ;gBAAEd;YAAG;YAAGP;QAAK;IAChC,EAAE,OAAOW,KAAK;QACZ,4FAA4F;QAC5F,0CAA0C;QAC1CJ,GAAGe,KAAK;QACR,MAAMX;IACR;AACF;AAEA,eAAeW,MAAMD,MAAoB;IACvCA,OAAOd,EAAE,CAACe,KAAK;AACjB;AAEA,eAAeC,aAAaC,OAAqB,EAAExB,IAAgB,EAAEG,GAAW;IAC9E,+FAA+F;IAC/F,+EAA+E;IAC/E,MAAMZ,gBAAgBS,MAAM,IAAME,mBAAmBF,MAAMG,MAAMb;AACnE;AAEA,eAAemC,sBAAsBJ,MAAoB,EAAEK,KAAiB,EAAEC,EAAU;IACtFN,OAAOd,EAAE,CAACN,IAAI,CAAC,CAAC,sBAAsB,EAAE0B,IAAI;AAC9C;AAEA,+FAA+F;AAC/F,OAAO,MAAMC,oBAA+C;IAC1DC,UAAUjC;IACVkC,eAAejC;IACfkC,kBAAkBtC;IAClByB;IACAI;IACAC;IACAE;IACA9B,aAAa,CAAC0B,QAAQrB,OAASL,YAAY0B,OAAOd,EAAE,EAAEP;AACxD,EAAE;AAEF,OAAO,eAAegC,WAAW7B,GAAmB;IAClD,OAAOhB,gBAAgBgB,KAAKyB;AAC9B"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/sqlite/open.ts"],"sourcesContent":["// The Node floor (>=22.20) is explained here and nowhere else: 22.20 is the first release with\n// both FTS5 and row-returning INSERT ... RETURNING. Raise it only for a load-bearing capability.\nimport { DatabaseSync } from 'node:sqlite';\nimport type { Config, ResolvedConfig } from '../../config/index.ts';\nimport { featureSignature } from '../../config/index.ts';\nimport { activeFeatures, FEATURES } from '../../features/index.ts';\nimport type { OpenResult } from '../open.ts';\nimport { openWithDialect } from '../open.ts';\nimport { getMeta, setMeta } from '../shared.ts';\nimport { BEGIN_WRITE, withTransaction } from '../transaction.ts';\nimport type { Connection, OpenDialect } from '../types.ts';\nimport { createConnection } from './connection.ts';\nimport { sqliteDialect } from './reconcile.ts';\nimport { registerFunctions } from './sql-functions.ts';\nimport { createStore } from './store.ts';\n\nexport const DB_FILENAME = 'cache.db';\n// Cache shape version, independent of the config's own `version`. Bumping it rebuilds\n// existing trees on first query.\nexport const SCHEMA_VERSION = '20';\n\nexport type { OpenResult };\n\n// unicode61 splits on spaces, so a language written without them indexes a whole run as one\n// token and word search finds nothing; the `_seg` sidecars (always populated) cover that case.\nconst TOKENIZE = 'porter unicode61';\n\ninterface SqliteHandle {\n db: DatabaseSync;\n}\n\n// The `_seg` sidecars are appended after path, never inserted: bm25() and snippet(content, 2)\n// are documented against the first three columns. Each holds its field's exploded unspaced runs.\nasync function createContentTable(conn: Connection): Promise<void> {\n await conn.exec(`CREATE VIRTUAL TABLE IF NOT EXISTS content USING fts5(title, summary, text, path UNINDEXED, title_seg, summary_seg, text_seg, tokenize = '${TOKENIZE}')`);\n}\n\n// Content is a separate table (not a column on frontmatter) so `SELECT * FROM frontmatter`\n// can't dump file text into context. Features add their own tables after the core ones.\nasync function ensureSchemaTables(conn: Connection, cfg: Config): Promise<void> {\n await conn.exec(`CREATE TABLE IF NOT EXISTS frontmatter (\"path\" TEXT PRIMARY KEY, \"_mtime\" REAL, \"_ctime\" REAL, \"_size\" INTEGER, \"_parse_error\" TEXT)`);\n await createContentTable(conn);\n // Coverage, not ownership: a path can appear under several presets. path leads the PK so the\n // per-doc delete is an index hit -- keyed the other way, cold builds went quadratic.\n await conn.exec(`CREATE TABLE IF NOT EXISTS preset_files (\"path\" TEXT, preset TEXT, PRIMARY KEY (\"path\", preset))`);\n await conn.exec('CREATE INDEX IF NOT EXISTS preset_files_preset ON preset_files(preset)');\n for (const feature of activeFeatures(cfg)) await feature.schema(conn);\n if ((await getMeta(conn, 'schema_version')) === null) await setMeta(conn, 'schema_version', SCHEMA_VERSION);\n if ((await getMeta(conn, 'features')) === null) await setMeta(conn, 'features', featureSignature(cfg, FEATURES));\n}\n\n// Two processes opening the same fresh tree both try to convert it, and the loser gets SQLITE_BUSY\n// with no busy handler behind it. Bounded because a lock held past this is a real problem, not a race.\nfunction setJournalWal(db: DatabaseSync): void {\n const deadline = Date.now() + 5000;\n for (;;) {\n try {\n db.exec('PRAGMA journal_mode = WAL');\n return;\n } catch (err) {\n if (Date.now() >= deadline || !/database is locked|busy/i.test((err as Error).message)) throw err;\n Atomics.wait(new Int32Array(new SharedArrayBuffer(4)), 0, 0, 20);\n }\n }\n}\n\nasync function connect(dbPath: string, _cfg: ResolvedConfig): Promise<{ handle: SqliteHandle; conn: Connection }> {\n const db = new DatabaseSync(dbPath);\n try {\n // Before journal_mode, not after: converting a fresh database to WAL takes a brief exclusive\n // lock, and with no timeout set yet a second process opening the same tree fails in 1ms.\n db.exec('PRAGMA busy_timeout = 30000');\n // busy_timeout does not cover the WAL conversion itself: SQLite does not invoke the busy\n // handler for it, so a concurrent cold open needs its own bounded wait.\n setJournalWal(db);\n registerFunctions(db);\n const conn = createConnection(db);\n return { handle: { db }, conn };\n } catch (err) {\n // A throw here must release this handle, or on Windows the leaked WAL db is undeletable and\n // scratch cleanup fails with EPERM/EBUSY.\n db.close();\n throw err;\n }\n}\n\nasync function close(handle: SqliteHandle): Promise<void> {\n handle.db.close();\n}\n\nasync function ensureSchema(_handle: SqliteHandle, conn: Connection, cfg: Config): Promise<void> {\n // One writer at a time: the feature hooks check a column then add it, so two cold opens racing\n // here both see it missing and the second ALTER fails with a duplicate column.\n await withTransaction(conn, () => ensureSchemaTables(conn, cfg), BEGIN_WRITE);\n}\n\nasync function setDerivedBusyTimeout(handle: SqliteHandle, _conn: Connection, ms: number): Promise<void> {\n handle.db.exec(`PRAGMA busy_timeout = ${ms}`);\n}\n\n// This store's dialect (types.ts's OpenDialect) for the shared orchestration in store/open.ts.\nexport const sqliteOpenDialect: OpenDialect<SqliteHandle> = {\n filename: DB_FILENAME,\n schemaVersion: SCHEMA_VERSION,\n reconcileDialect: sqliteDialect,\n connect,\n close,\n ensureSchema,\n setDerivedBusyTimeout,\n createStore: (handle, conn) => createStore(handle.db, conn),\n};\n\nexport async function openSqlite(cfg: ResolvedConfig): Promise<OpenResult> {\n return openWithDialect(cfg, sqliteOpenDialect);\n}\n"],"names":["DatabaseSync","featureSignature","activeFeatures","FEATURES","openWithDialect","getMeta","setMeta","BEGIN_WRITE","withTransaction","createConnection","sqliteDialect","registerFunctions","createStore","DB_FILENAME","SCHEMA_VERSION","TOKENIZE","createContentTable","conn","exec","ensureSchemaTables","cfg","feature","schema","setJournalWal","db","deadline","Date","now","err","test","message","Atomics","wait","Int32Array","SharedArrayBuffer","connect","dbPath","_cfg","handle","close","ensureSchema","_handle","setDerivedBusyTimeout","_conn","ms","sqliteOpenDialect","filename","schemaVersion","reconcileDialect","openSqlite"],"mappings":"AAAA,+FAA+F;AAC/F,iGAAiG;AACjG,SAASA,YAAY,QAAQ,cAAc;AAE3C,SAASC,gBAAgB,QAAQ,wBAAwB;AACzD,SAASC,cAAc,EAAEC,QAAQ,QAAQ,0BAA0B;AAEnE,SAASC,eAAe,QAAQ,aAAa;AAC7C,SAASC,OAAO,EAAEC,OAAO,QAAQ,eAAe;AAChD,SAASC,WAAW,EAAEC,eAAe,QAAQ,oBAAoB;AAEjE,SAASC,gBAAgB,QAAQ,kBAAkB;AACnD,SAASC,aAAa,QAAQ,iBAAiB;AAC/C,SAASC,iBAAiB,QAAQ,qBAAqB;AACvD,SAASC,WAAW,QAAQ,aAAa;AAEzC,OAAO,MAAMC,cAAc,WAAW;AACtC,sFAAsF;AACtF,iCAAiC;AACjC,OAAO,MAAMC,iBAAiB,KAAK;AAInC,4FAA4F;AAC5F,+FAA+F;AAC/F,MAAMC,WAAW;AAMjB,8FAA8F;AAC9F,iGAAiG;AACjG,eAAeC,mBAAmBC,IAAgB;IAChD,MAAMA,KAAKC,IAAI,CAAC,CAAC,0IAA0I,EAAEH,SAAS,EAAE,CAAC;AAC3K;AAEA,2FAA2F;AAC3F,wFAAwF;AACxF,eAAeI,mBAAmBF,IAAgB,EAAEG,GAAW;IAC7D,MAAMH,KAAKC,IAAI,CAAC,CAAC,oIAAoI,CAAC;IACtJ,MAAMF,mBAAmBC;IACzB,6FAA6F;IAC7F,qFAAqF;IACrF,MAAMA,KAAKC,IAAI,CAAC,CAAC,gGAAgG,CAAC;IAClH,MAAMD,KAAKC,IAAI,CAAC;IAChB,KAAK,MAAMG,WAAWnB,eAAekB,KAAM,MAAMC,QAAQC,MAAM,CAACL;IAChE,IAAI,AAAC,MAAMZ,QAAQY,MAAM,sBAAuB,MAAM,MAAMX,QAAQW,MAAM,kBAAkBH;IAC5F,IAAI,AAAC,MAAMT,QAAQY,MAAM,gBAAiB,MAAM,MAAMX,QAAQW,MAAM,YAAYhB,iBAAiBmB,KAAKjB;AACxG;AAEA,mGAAmG;AACnG,uGAAuG;AACvG,SAASoB,cAAcC,EAAgB;IACrC,MAAMC,WAAWC,KAAKC,GAAG,KAAK;IAC9B,OAAS;QACP,IAAI;YACFH,GAAGN,IAAI,CAAC;YACR;QACF,EAAE,OAAOU,KAAK;YACZ,IAAIF,KAAKC,GAAG,MAAMF,YAAY,CAAC,2BAA2BI,IAAI,CAAC,AAACD,IAAcE,OAAO,GAAG,MAAMF;YAC9FG,QAAQC,IAAI,CAAC,IAAIC,WAAW,IAAIC,kBAAkB,KAAK,GAAG,GAAG;QAC/D;IACF;AACF;AAEA,eAAeC,QAAQC,MAAc,EAAEC,IAAoB;IACzD,MAAMb,KAAK,IAAIxB,aAAaoC;IAC5B,IAAI;QACF,6FAA6F;QAC7F,yFAAyF;QACzFZ,GAAGN,IAAI,CAAC;QACR,yFAAyF;QACzF,wEAAwE;QACxEK,cAAcC;QACdb,kBAAkBa;QAClB,MAAMP,OAAOR,iBAAiBe;QAC9B,OAAO;YAAEc,QAAQ;gBAAEd;YAAG;YAAGP;QAAK;IAChC,EAAE,OAAOW,KAAK;QACZ,4FAA4F;QAC5F,0CAA0C;QAC1CJ,GAAGe,KAAK;QACR,MAAMX;IACR;AACF;AAEA,eAAeW,MAAMD,MAAoB;IACvCA,OAAOd,EAAE,CAACe,KAAK;AACjB;AAEA,eAAeC,aAAaC,OAAqB,EAAExB,IAAgB,EAAEG,GAAW;IAC9E,+FAA+F;IAC/F,+EAA+E;IAC/E,MAAMZ,gBAAgBS,MAAM,IAAME,mBAAmBF,MAAMG,MAAMb;AACnE;AAEA,eAAemC,sBAAsBJ,MAAoB,EAAEK,KAAiB,EAAEC,EAAU;IACtFN,OAAOd,EAAE,CAACN,IAAI,CAAC,CAAC,sBAAsB,EAAE0B,IAAI;AAC9C;AAEA,+FAA+F;AAC/F,OAAO,MAAMC,oBAA+C;IAC1DC,UAAUjC;IACVkC,eAAejC;IACfkC,kBAAkBtC;IAClByB;IACAI;IACAC;IACAE;IACA9B,aAAa,CAAC0B,QAAQrB,OAASL,YAAY0B,OAAOd,EAAE,EAAEP;AACxD,EAAE;AAEF,OAAO,eAAegC,WAAW7B,GAAmB;IAClD,OAAOhB,gBAAgBgB,KAAKyB;AAC9B"}
@@ -1,3 +1,4 @@
1
1
  import type { Database } from '@tursodatabase/database';
2
2
  import type { Connection } from '../types.js';
3
+ export declare function checkpointWal(db: Database): Promise<void>;
3
4
  export declare function createConnection(db: Database): Connection;
@@ -1,6 +1,6 @@
1
1
  import { BEGIN_WRITE, withTransaction } from '../transaction.js';
2
- // Turso's client is async end to end (no synchronous escape hatch like node:sqlite's), so this
3
- // wraps it in the same Connection/Statement shape 1:1, no sync-over-async and no worker bridge.
2
+ // @tursodatabase/database/compat offers a synchronous escape hatch, but it measured only 4-12%
3
+ // faster than this promise client, so this wraps the async client in the same Connection/Statement shape 1:1.
4
4
  let TursoStatementWrapper = class TursoStatementWrapper {
5
5
  async run(...params) {
6
6
  return this.stmt.run(...params);
@@ -24,6 +24,15 @@ let TursoStatementWrapper = class TursoStatementWrapper {
24
24
  this.stmt = stmt;
25
25
  }
26
26
  };
27
+ // This client leaves the WAL behind for the next opener, where node:sqlite checkpoints on the last
28
+ // close, so a tree reconciled over and over grows one without bound. Best-effort: close must not throw.
29
+ export async function checkpointWal(db) {
30
+ try {
31
+ await db.exec('PRAGMA wal_checkpoint(TRUNCATE)');
32
+ } catch (err) {
33
+ console.error(`sense: turso WAL checkpoint failed, the -wal file will keep growing until one succeeds: ${err.message}`);
34
+ }
35
+ }
27
36
  export function createConnection(db) {
28
37
  const conn = {
29
38
  async exec (sql) {
@@ -32,15 +41,19 @@ export function createConnection(db) {
32
41
  async prepare (sql) {
33
42
  return new TursoStatementWrapper(await db.prepare(sql));
34
43
  },
35
- // db.batch() is the engine's own bulk idiom: one crossing, N rows, each its own statement,
36
- // no bind-variable ceiling to chunk around. A literal nested BEGIN hard-errors, so withTransaction's join-not-savepoint helper makes this safe inside reconcile's own transaction too.
44
+ // Prepares once and awaits run() per row: db.batch() re-prepares each statement, which cost as
45
+ // much as preparing per row. A literal nested BEGIN hard-errors, so withTransaction's join-not-savepoint helper makes this safe inside reconcile's own transaction too.
37
46
  async runBatch (sql, paramRows) {
38
47
  if (paramRows.length === 0) return;
39
48
  await withTransaction(conn, async ()=>{
40
- await db.batch(paramRows.map((args)=>({
41
- sql,
42
- args
43
- })));
49
+ const stmt = await db.prepare(sql);
50
+ // Finalized here because nothing else will: open() hands back a connection the caller
51
+ // can hold across many batches, and the db.batch() this replaced finalized its own.
52
+ try {
53
+ for (const row of paramRows)await stmt.run(...row);
54
+ } finally{
55
+ await stmt.close();
56
+ }
44
57
  }, BEGIN_WRITE);
45
58
  }
46
59
  };
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/turso/connection.ts"],"sourcesContent":["import type { Database } from '@tursodatabase/database';\nimport { BEGIN_WRITE, withTransaction } from '../transaction.ts';\nimport type { Connection, RunResult, Statement } from '../types.ts';\n\n// The client's own Statement class isn't re-exported by name from '@tursodatabase/database',\n// so its type is derived structurally from Database.prepare()'s return type instead.\ntype TursoStatement = Awaited<ReturnType<Database['prepare']>>;\n\n// Turso's client is async end to end (no synchronous escape hatch like node:sqlite's), so this\n// wraps it in the same Connection/Statement shape 1:1, no sync-over-async and no worker bridge.\nclass TursoStatementWrapper implements Statement {\n private stmt: TursoStatement;\n\n constructor(stmt: TursoStatement) {\n this.stmt = stmt;\n }\n\n async run(...params: unknown[]): Promise<RunResult> {\n return this.stmt.run(...params);\n }\n\n async get(...params: unknown[]): Promise<unknown> {\n return this.stmt.get(...params);\n }\n\n async all(...params: unknown[]): Promise<unknown[]> {\n return this.stmt.all(...params);\n }\n\n async *iterate(...params: unknown[]): AsyncIterable<unknown> {\n yield* this.stmt.iterate(...params);\n }\n\n columns(): Array<{ name: string }> {\n return this.stmt.columns();\n }\n\n setReadBigInts(enabled: boolean): void {\n this.stmt.safeIntegers(enabled);\n }\n}\n\nexport function createConnection(db: Database): Connection {\n const conn: Connection = {\n async exec(sql: string): Promise<void> {\n await db.exec(sql);\n },\n async prepare(sql: string): Promise<Statement> {\n return new TursoStatementWrapper(await db.prepare(sql));\n },\n // db.batch() is the engine's own bulk idiom: one crossing, N rows, each its own statement,\n // no bind-variable ceiling to chunk around. A literal nested BEGIN hard-errors, so withTransaction's join-not-savepoint helper makes this safe inside reconcile's own transaction too.\n async runBatch(sql: string, paramRows: unknown[][]): Promise<void> {\n if (paramRows.length === 0) return;\n await withTransaction(\n conn,\n async () => {\n await db.batch(paramRows.map((args) => ({ sql, args })));\n },\n BEGIN_WRITE\n );\n },\n };\n return conn;\n}\n"],"names":["BEGIN_WRITE","withTransaction","TursoStatementWrapper","run","params","stmt","get","all","iterate","columns","setReadBigInts","enabled","safeIntegers","createConnection","db","conn","exec","sql","prepare","runBatch","paramRows","length","batch","map","args"],"mappings":"AACA,SAASA,WAAW,EAAEC,eAAe,QAAQ,oBAAoB;AAOjE,+FAA+F;AAC/F,gGAAgG;AAChG,IAAA,AAAMC,wBAAN,MAAMA;IAOJ,MAAMC,IAAI,GAAGC,MAAiB,EAAsB;QAClD,OAAO,IAAI,CAACC,IAAI,CAACF,GAAG,IAAIC;IAC1B;IAEA,MAAME,IAAI,GAAGF,MAAiB,EAAoB;QAChD,OAAO,IAAI,CAACC,IAAI,CAACC,GAAG,IAAIF;IAC1B;IAEA,MAAMG,IAAI,GAAGH,MAAiB,EAAsB;QAClD,OAAO,IAAI,CAACC,IAAI,CAACE,GAAG,IAAIH;IAC1B;IAEA,OAAOI,QAAQ,GAAGJ,MAAiB,EAA0B;QAC3D,OAAO,IAAI,CAACC,IAAI,CAACG,OAAO,IAAIJ;IAC9B;IAEAK,UAAmC;QACjC,OAAO,IAAI,CAACJ,IAAI,CAACI,OAAO;IAC1B;IAEAC,eAAeC,OAAgB,EAAQ;QACrC,IAAI,CAACN,IAAI,CAACO,YAAY,CAACD;IACzB;IA1BA,YAAYN,IAAoB,CAAE;QAChC,IAAI,CAACA,IAAI,GAAGA;IACd;AAyBF;AAEA,OAAO,SAASQ,iBAAiBC,EAAY;IAC3C,MAAMC,OAAmB;QACvB,MAAMC,MAAKC,GAAW;YACpB,MAAMH,GAAGE,IAAI,CAACC;QAChB;QACA,MAAMC,SAAQD,GAAW;YACvB,OAAO,IAAIf,sBAAsB,MAAMY,GAAGI,OAAO,CAACD;QACpD;QACA,2FAA2F;QAC3F,uLAAuL;QACvL,MAAME,UAASF,GAAW,EAAEG,SAAsB;YAChD,IAAIA,UAAUC,MAAM,KAAK,GAAG;YAC5B,MAAMpB,gBACJc,MACA;gBACE,MAAMD,GAAGQ,KAAK,CAACF,UAAUG,GAAG,CAAC,CAACC,OAAU,CAAA;wBAAEP;wBAAKO;oBAAK,CAAA;YACtD,GACAxB;QAEJ;IACF;IACA,OAAOe;AACT"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/turso/connection.ts"],"sourcesContent":["import type { Database } from '@tursodatabase/database';\nimport { BEGIN_WRITE, withTransaction } from '../transaction.ts';\nimport type { Connection, RunResult, Statement } from '../types.ts';\n\n// The client's own Statement class isn't re-exported by name from '@tursodatabase/database',\n// so its type is derived structurally from Database.prepare()'s return type instead.\ntype TursoStatement = Awaited<ReturnType<Database['prepare']>>;\n\n// @tursodatabase/database/compat offers a synchronous escape hatch, but it measured only 4-12%\n// faster than this promise client, so this wraps the async client in the same Connection/Statement shape 1:1.\nclass TursoStatementWrapper implements Statement {\n private stmt: TursoStatement;\n\n constructor(stmt: TursoStatement) {\n this.stmt = stmt;\n }\n\n async run(...params: unknown[]): Promise<RunResult> {\n return this.stmt.run(...params);\n }\n\n async get(...params: unknown[]): Promise<unknown> {\n return this.stmt.get(...params);\n }\n\n async all(...params: unknown[]): Promise<unknown[]> {\n return this.stmt.all(...params);\n }\n\n async *iterate(...params: unknown[]): AsyncIterable<unknown> {\n yield* this.stmt.iterate(...params);\n }\n\n columns(): Array<{ name: string }> {\n return this.stmt.columns();\n }\n\n setReadBigInts(enabled: boolean): void {\n this.stmt.safeIntegers(enabled);\n }\n}\n\n// This client leaves the WAL behind for the next opener, where node:sqlite checkpoints on the last\n// close, so a tree reconciled over and over grows one without bound. Best-effort: close must not throw.\nexport async function checkpointWal(db: Database): Promise<void> {\n try {\n await db.exec('PRAGMA wal_checkpoint(TRUNCATE)');\n } catch (err) {\n console.error(`sense: turso WAL checkpoint failed, the -wal file will keep growing until one succeeds: ${(err as Error).message}`);\n }\n}\n\nexport function createConnection(db: Database): Connection {\n const conn: Connection = {\n async exec(sql: string): Promise<void> {\n await db.exec(sql);\n },\n async prepare(sql: string): Promise<Statement> {\n return new TursoStatementWrapper(await db.prepare(sql));\n },\n // Prepares once and awaits run() per row: db.batch() re-prepares each statement, which cost as\n // much as preparing per row. A literal nested BEGIN hard-errors, so withTransaction's join-not-savepoint helper makes this safe inside reconcile's own transaction too.\n async runBatch(sql: string, paramRows: unknown[][]): Promise<void> {\n if (paramRows.length === 0) return;\n await withTransaction(\n conn,\n async () => {\n const stmt = await db.prepare(sql);\n // Finalized here because nothing else will: open() hands back a connection the caller\n // can hold across many batches, and the db.batch() this replaced finalized its own.\n try {\n for (const row of paramRows) await stmt.run(...row);\n } finally {\n await stmt.close();\n }\n },\n BEGIN_WRITE\n );\n },\n };\n return conn;\n}\n"],"names":["BEGIN_WRITE","withTransaction","TursoStatementWrapper","run","params","stmt","get","all","iterate","columns","setReadBigInts","enabled","safeIntegers","checkpointWal","db","exec","err","console","error","message","createConnection","conn","sql","prepare","runBatch","paramRows","length","row","close"],"mappings":"AACA,SAASA,WAAW,EAAEC,eAAe,QAAQ,oBAAoB;AAOjE,+FAA+F;AAC/F,8GAA8G;AAC9G,IAAA,AAAMC,wBAAN,MAAMA;IAOJ,MAAMC,IAAI,GAAGC,MAAiB,EAAsB;QAClD,OAAO,IAAI,CAACC,IAAI,CAACF,GAAG,IAAIC;IAC1B;IAEA,MAAME,IAAI,GAAGF,MAAiB,EAAoB;QAChD,OAAO,IAAI,CAACC,IAAI,CAACC,GAAG,IAAIF;IAC1B;IAEA,MAAMG,IAAI,GAAGH,MAAiB,EAAsB;QAClD,OAAO,IAAI,CAACC,IAAI,CAACE,GAAG,IAAIH;IAC1B;IAEA,OAAOI,QAAQ,GAAGJ,MAAiB,EAA0B;QAC3D,OAAO,IAAI,CAACC,IAAI,CAACG,OAAO,IAAIJ;IAC9B;IAEAK,UAAmC;QACjC,OAAO,IAAI,CAACJ,IAAI,CAACI,OAAO;IAC1B;IAEAC,eAAeC,OAAgB,EAAQ;QACrC,IAAI,CAACN,IAAI,CAACO,YAAY,CAACD;IACzB;IA1BA,YAAYN,IAAoB,CAAE;QAChC,IAAI,CAACA,IAAI,GAAGA;IACd;AAyBF;AAEA,mGAAmG;AACnG,wGAAwG;AACxG,OAAO,eAAeQ,cAAcC,EAAY;IAC9C,IAAI;QACF,MAAMA,GAAGC,IAAI,CAAC;IAChB,EAAE,OAAOC,KAAK;QACZC,QAAQC,KAAK,CAAC,CAAC,wFAAwF,EAAE,AAACF,IAAcG,OAAO,EAAE;IACnI;AACF;AAEA,OAAO,SAASC,iBAAiBN,EAAY;IAC3C,MAAMO,OAAmB;QACvB,MAAMN,MAAKO,GAAW;YACpB,MAAMR,GAAGC,IAAI,CAACO;QAChB;QACA,MAAMC,SAAQD,GAAW;YACvB,OAAO,IAAIpB,sBAAsB,MAAMY,GAAGS,OAAO,CAACD;QACpD;QACA,+FAA+F;QAC/F,wKAAwK;QACxK,MAAME,UAASF,GAAW,EAAEG,SAAsB;YAChD,IAAIA,UAAUC,MAAM,KAAK,GAAG;YAC5B,MAAMzB,gBACJoB,MACA;gBACE,MAAMhB,OAAO,MAAMS,GAAGS,OAAO,CAACD;gBAC9B,sFAAsF;gBACtF,oFAAoF;gBACpF,IAAI;oBACF,KAAK,MAAMK,OAAOF,UAAW,MAAMpB,KAAKF,GAAG,IAAIwB;gBACjD,SAAU;oBACR,MAAMtB,KAAKuB,KAAK;gBAClB;YACF,GACA5B;QAEJ;IACF;IACA,OAAOqB;AACT"}
@@ -3,7 +3,7 @@ import type { ResolvedConfig } from '../../config/index.js';
3
3
  import type { OpenResult } from '../open.js';
4
4
  import type { OpenDialect } from '../types.js';
5
5
  export declare const DB_FILENAME = "cache.turso.db";
6
- export declare const SCHEMA_VERSION = "4";
6
+ export declare const SCHEMA_VERSION = "5";
7
7
  export type { OpenResult };
8
8
  export declare const tursoOpenDialect: OpenDialect<Database>;
9
9
  export declare function openTurso(cfg: ResolvedConfig): Promise<OpenResult>;
@@ -4,14 +4,14 @@ import { SenseError } from '../../errors.js';
4
4
  import { activeFeatures, FEATURES } from '../../features/index.js';
5
5
  import { openWithDialect } from '../open.js';
6
6
  import { getMeta, setMeta } from '../shared.js';
7
- import { createConnection } from './connection.js';
7
+ import { checkpointWal, createConnection } from './connection.js';
8
8
  import { TURSO_PACKAGE, tursoApi } from './native.js';
9
9
  import { CONTENT_FTS_DDL, tursoDialect } from './reconcile.js';
10
10
  import { createStore } from './store.js';
11
11
  export const DB_FILENAME = 'cache.turso.db';
12
12
  // Independent of sqlite's and duckdb's SCHEMA_VERSION: each store's cache shape evolves
13
13
  // separately. Covers the FTS indexes, the "_ngram" sidecar columns, and embeddings.vector's width.
14
- export const SCHEMA_VERSION = '4';
14
+ export const SCHEMA_VERSION = '5';
15
15
  // This store's Handle (types.ts's OpenDialect<Handle>) is the connected Database itself: no
16
16
  // extra native state to thread, unlike duckdb's separate instance/connection pair.
17
17
  async function ensureSchema(_handle, conn, cfg) {
@@ -33,6 +33,7 @@ async function ensureSchema(_handle, conn, cfg) {
33
33
  if (await getMeta(conn, 'features') === null) await setMeta(conn, 'features', featureSignature(cfg, FEATURES));
34
34
  }
35
35
  async function close(handle) {
36
+ await checkpointWal(handle);
36
37
  await handle.close();
37
38
  }
38
39
  async function setDerivedBusyTimeout(_handle, conn, ms) {
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/turso/open.ts"],"sourcesContent":["import type { Database } from '@tursodatabase/database';\nimport type { Config, ResolvedConfig } from '../../config/index.ts';\nimport { featureSignature } from '../../config/index.ts';\nimport { STORE_DIMS } from '../../embed/types.ts';\nimport { SenseError } from '../../errors.ts';\nimport { activeFeatures, FEATURES } from '../../features/index.ts';\nimport type { OpenResult } from '../open.ts';\nimport { openWithDialect } from '../open.ts';\nimport { getMeta, setMeta } from '../shared.ts';\nimport type { Connection, OpenDialect } from '../types.ts';\nimport { createConnection } from './connection.ts';\nimport { TURSO_PACKAGE, tursoApi } from './native.ts';\nimport { CONTENT_FTS_DDL, tursoDialect } from './reconcile.ts';\nimport { createStore } from './store.ts';\n\nexport const DB_FILENAME = 'cache.turso.db';\n// Independent of sqlite's and duckdb's SCHEMA_VERSION: each store's cache shape evolves\n// separately. Covers the FTS indexes, the \"_ngram\" sidecar columns, and embeddings.vector's width.\nexport const SCHEMA_VERSION = '4';\n\nexport type { OpenResult };\n\n// This store's Handle (types.ts's OpenDialect<Handle>) is the connected Database itself: no\n// extra native state to thread, unlike duckdb's separate instance/connection pair.\nasync function ensureSchema(_handle: Database, conn: Connection, cfg: Config): Promise<void> {\n await conn.exec(`CREATE TABLE IF NOT EXISTS frontmatter (\"path\" TEXT PRIMARY KEY, \"_mtime\" REAL, \"_ctime\" REAL, \"_size\" INTEGER, \"_parse_error\" TEXT)`);\n await conn.exec(`CREATE TABLE IF NOT EXISTS content (\"path\" TEXT PRIMARY KEY, title TEXT, summary TEXT, text TEXT, title_ngram TEXT, summary_ngram TEXT, text_ngram TEXT)`);\n for (const ddl of CONTENT_FTS_DDL) await conn.exec(ddl);\n await conn.exec(`CREATE TABLE IF NOT EXISTS preset_files (\"path\" TEXT, preset TEXT, PRIMARY KEY (\"path\", preset))`);\n await conn.exec('CREATE INDEX IF NOT EXISTS preset_files_preset ON preset_files(preset)');\n for (const feature of activeFeatures(cfg)) {\n // Native F32_BLOB(STORE_DIMS) instead of the embed feature's engine-neutral BLOB DDL. `scale`\n // is kept unused, so the shared reconcile-time INSERT/DELETE names a column both stores have.\n if (feature.name === 'embed') {\n await conn.exec(`CREATE TABLE IF NOT EXISTS embeddings (\"path\" TEXT, chunk INTEGER, start_line INTEGER, end_line INTEGER, scale REAL, vector F32_BLOB(${STORE_DIMS}), PRIMARY KEY (\"path\", chunk))`);\n continue;\n }\n await feature.schema(conn);\n }\n if ((await getMeta(conn, 'schema_version')) === null) await setMeta(conn, 'schema_version', SCHEMA_VERSION);\n if ((await getMeta(conn, 'features')) === null) await setMeta(conn, 'features', featureSignature(cfg, FEATURES));\n}\n\nasync function close(handle: Database): Promise<void> {\n await handle.close();\n}\n\nasync function setDerivedBusyTimeout(_handle: Database, conn: Connection, ms: number): Promise<void> {\n await conn.exec(`PRAGMA busy_timeout = ${ms}`);\n}\n\nasync function connect(dbPath: string, _cfg: ResolvedConfig): Promise<{ handle: Database; conn: Connection }> {\n // Dynamic, not a top-level import: a sqlite or duckdb tree must never attempt to resolve this\n // optional dependency until a turso tree is actually opened. Installed on first use if missing.\n let turso: Awaited<ReturnType<typeof tursoApi>>;\n try {\n turso = await tursoApi();\n } catch (err) {\n if (err instanceof SenseError) throw err;\n throw new SenseError('STORE_DEPENDENCY_MISSING', `store \"turso\" needs the ${TURSO_PACKAGE} package (${(err as Error).message})`);\n }\n\n let db: Database;\n try {\n // Floored at the same 30s sqlite opens with. `timeout` is connect-time only in this client;\n // the derived value is set via runtime PRAGMA below. `index_method` is required for ensureSchema()'s FTS indexes (T1).\n db = await turso.connect(dbPath, { timeout: 30_000, experimental: ['index_method'] });\n } catch (err) {\n throw new SenseError('STORE_DEPENDENCY_MISSING', `store \"turso\" failed to open ${dbPath}: ${(err as Error).message}`);\n }\n return { handle: db, conn: createConnection(db) };\n}\n\n// This store's dialect (types.ts's OpenDialect) for the shared orchestration in store/open.ts.\nexport const tursoOpenDialect: OpenDialect<Database> = {\n filename: DB_FILENAME,\n schemaVersion: SCHEMA_VERSION,\n reconcileDialect: tursoDialect,\n connect,\n close,\n // \"Locking error: Failed locking file ...\", worded per platform: posix \"File is locked by another\n // process\", Windows \"another process has locked a portion of the file (os error 33)\". Distinct from\n // the write-time \"database is locked\" its connect-time `timeout` covers; that one never reaches here.\n isLocked: (err) => /File is locked by another process|locked a portion of the file/.test(err.message),\n ensureSchema,\n setDerivedBusyTimeout,\n createStore: (handle, conn) => createStore(handle, conn),\n};\n\nexport async function openTurso(cfg: ResolvedConfig): Promise<OpenResult> {\n return openWithDialect(cfg, tursoOpenDialect);\n}\n"],"names":["featureSignature","STORE_DIMS","SenseError","activeFeatures","FEATURES","openWithDialect","getMeta","setMeta","createConnection","TURSO_PACKAGE","tursoApi","CONTENT_FTS_DDL","tursoDialect","createStore","DB_FILENAME","SCHEMA_VERSION","ensureSchema","_handle","conn","cfg","exec","ddl","feature","name","schema","close","handle","setDerivedBusyTimeout","ms","connect","dbPath","_cfg","turso","err","message","db","timeout","experimental","tursoOpenDialect","filename","schemaVersion","reconcileDialect","isLocked","test","openTurso"],"mappings":"AAEA,SAASA,gBAAgB,QAAQ,wBAAwB;AACzD,SAASC,UAAU,QAAQ,uBAAuB;AAClD,SAASC,UAAU,QAAQ,kBAAkB;AAC7C,SAASC,cAAc,EAAEC,QAAQ,QAAQ,0BAA0B;AAEnE,SAASC,eAAe,QAAQ,aAAa;AAC7C,SAASC,OAAO,EAAEC,OAAO,QAAQ,eAAe;AAEhD,SAASC,gBAAgB,QAAQ,kBAAkB;AACnD,SAASC,aAAa,EAAEC,QAAQ,QAAQ,cAAc;AACtD,SAASC,eAAe,EAAEC,YAAY,QAAQ,iBAAiB;AAC/D,SAASC,WAAW,QAAQ,aAAa;AAEzC,OAAO,MAAMC,cAAc,iBAAiB;AAC5C,wFAAwF;AACxF,mGAAmG;AACnG,OAAO,MAAMC,iBAAiB,IAAI;AAIlC,4FAA4F;AAC5F,mFAAmF;AACnF,eAAeC,aAAaC,OAAiB,EAAEC,IAAgB,EAAEC,GAAW;IAC1E,MAAMD,KAAKE,IAAI,CAAC,CAAC,oIAAoI,CAAC;IACtJ,MAAMF,KAAKE,IAAI,CAAC,CAAC,wJAAwJ,CAAC;IAC1K,KAAK,MAAMC,OAAOV,gBAAiB,MAAMO,KAAKE,IAAI,CAACC;IACnD,MAAMH,KAAKE,IAAI,CAAC,CAAC,gGAAgG,CAAC;IAClH,MAAMF,KAAKE,IAAI,CAAC;IAChB,KAAK,MAAME,WAAWnB,eAAegB,KAAM;QACzC,8FAA8F;QAC9F,8FAA8F;QAC9F,IAAIG,QAAQC,IAAI,KAAK,SAAS;YAC5B,MAAML,KAAKE,IAAI,CAAC,CAAC,qIAAqI,EAAEnB,WAAW,+BAA+B,CAAC;YACnM;QACF;QACA,MAAMqB,QAAQE,MAAM,CAACN;IACvB;IACA,IAAI,AAAC,MAAMZ,QAAQY,MAAM,sBAAuB,MAAM,MAAMX,QAAQW,MAAM,kBAAkBH;IAC5F,IAAI,AAAC,MAAMT,QAAQY,MAAM,gBAAiB,MAAM,MAAMX,QAAQW,MAAM,YAAYlB,iBAAiBmB,KAAKf;AACxG;AAEA,eAAeqB,MAAMC,MAAgB;IACnC,MAAMA,OAAOD,KAAK;AACpB;AAEA,eAAeE,sBAAsBV,OAAiB,EAAEC,IAAgB,EAAEU,EAAU;IAClF,MAAMV,KAAKE,IAAI,CAAC,CAAC,sBAAsB,EAAEQ,IAAI;AAC/C;AAEA,eAAeC,QAAQC,MAAc,EAAEC,IAAoB;IACzD,8FAA8F;IAC9F,gGAAgG;IAChG,IAAIC;IACJ,IAAI;QACFA,QAAQ,MAAMtB;IAChB,EAAE,OAAOuB,KAAK;QACZ,IAAIA,eAAe/B,YAAY,MAAM+B;QACrC,MAAM,IAAI/B,WAAW,4BAA4B,CAAC,wBAAwB,EAAEO,cAAc,UAAU,EAAE,AAACwB,IAAcC,OAAO,CAAC,CAAC,CAAC;IACjI;IAEA,IAAIC;IACJ,IAAI;QACF,4FAA4F;QAC5F,uHAAuH;QACvHA,KAAK,MAAMH,MAAMH,OAAO,CAACC,QAAQ;YAAEM,SAAS;YAAQC,cAAc;gBAAC;aAAe;QAAC;IACrF,EAAE,OAAOJ,KAAK;QACZ,MAAM,IAAI/B,WAAW,4BAA4B,CAAC,6BAA6B,EAAE4B,OAAO,EAAE,EAAE,AAACG,IAAcC,OAAO,EAAE;IACtH;IACA,OAAO;QAAER,QAAQS;QAAIjB,MAAMV,iBAAiB2B;IAAI;AAClD;AAEA,+FAA+F;AAC/F,OAAO,MAAMG,mBAA0C;IACrDC,UAAUzB;IACV0B,eAAezB;IACf0B,kBAAkB7B;IAClBiB;IACAJ;IACA,kGAAkG;IAClG,oGAAoG;IACpG,sGAAsG;IACtGiB,UAAU,CAACT,MAAQ,iEAAiEU,IAAI,CAACV,IAAIC,OAAO;IACpGlB;IACAW;IACAd,aAAa,CAACa,QAAQR,OAASL,YAAYa,QAAQR;AACrD,EAAE;AAEF,OAAO,eAAe0B,UAAUzB,GAAmB;IACjD,OAAOd,gBAAgBc,KAAKmB;AAC9B"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/turso/open.ts"],"sourcesContent":["import type { Database } from '@tursodatabase/database';\nimport type { Config, ResolvedConfig } from '../../config/index.ts';\nimport { featureSignature } from '../../config/index.ts';\nimport { STORE_DIMS } from '../../embed/types.ts';\nimport { SenseError } from '../../errors.ts';\nimport { activeFeatures, FEATURES } from '../../features/index.ts';\nimport type { OpenResult } from '../open.ts';\nimport { openWithDialect } from '../open.ts';\nimport { getMeta, setMeta } from '../shared.ts';\nimport type { Connection, OpenDialect } from '../types.ts';\nimport { checkpointWal, createConnection } from './connection.ts';\nimport { TURSO_PACKAGE, tursoApi } from './native.ts';\nimport { CONTENT_FTS_DDL, tursoDialect } from './reconcile.ts';\nimport { createStore } from './store.ts';\n\nexport const DB_FILENAME = 'cache.turso.db';\n// Independent of sqlite's and duckdb's SCHEMA_VERSION: each store's cache shape evolves\n// separately. Covers the FTS indexes, the \"_ngram\" sidecar columns, and embeddings.vector's width.\nexport const SCHEMA_VERSION = '5';\n\nexport type { OpenResult };\n\n// This store's Handle (types.ts's OpenDialect<Handle>) is the connected Database itself: no\n// extra native state to thread, unlike duckdb's separate instance/connection pair.\nasync function ensureSchema(_handle: Database, conn: Connection, cfg: Config): Promise<void> {\n await conn.exec(`CREATE TABLE IF NOT EXISTS frontmatter (\"path\" TEXT PRIMARY KEY, \"_mtime\" REAL, \"_ctime\" REAL, \"_size\" INTEGER, \"_parse_error\" TEXT)`);\n await conn.exec(`CREATE TABLE IF NOT EXISTS content (\"path\" TEXT PRIMARY KEY, title TEXT, summary TEXT, text TEXT, title_ngram TEXT, summary_ngram TEXT, text_ngram TEXT)`);\n for (const ddl of CONTENT_FTS_DDL) await conn.exec(ddl);\n await conn.exec(`CREATE TABLE IF NOT EXISTS preset_files (\"path\" TEXT, preset TEXT, PRIMARY KEY (\"path\", preset))`);\n await conn.exec('CREATE INDEX IF NOT EXISTS preset_files_preset ON preset_files(preset)');\n for (const feature of activeFeatures(cfg)) {\n // Native F32_BLOB(STORE_DIMS) instead of the embed feature's engine-neutral BLOB DDL. `scale`\n // is kept unused, so the shared reconcile-time INSERT/DELETE names a column both stores have.\n if (feature.name === 'embed') {\n await conn.exec(`CREATE TABLE IF NOT EXISTS embeddings (\"path\" TEXT, chunk INTEGER, start_line INTEGER, end_line INTEGER, scale REAL, vector F32_BLOB(${STORE_DIMS}), PRIMARY KEY (\"path\", chunk))`);\n continue;\n }\n await feature.schema(conn);\n }\n if ((await getMeta(conn, 'schema_version')) === null) await setMeta(conn, 'schema_version', SCHEMA_VERSION);\n if ((await getMeta(conn, 'features')) === null) await setMeta(conn, 'features', featureSignature(cfg, FEATURES));\n}\n\nasync function close(handle: Database): Promise<void> {\n await checkpointWal(handle);\n await handle.close();\n}\n\nasync function setDerivedBusyTimeout(_handle: Database, conn: Connection, ms: number): Promise<void> {\n await conn.exec(`PRAGMA busy_timeout = ${ms}`);\n}\n\nasync function connect(dbPath: string, _cfg: ResolvedConfig): Promise<{ handle: Database; conn: Connection }> {\n // Dynamic, not a top-level import: a sqlite or duckdb tree must never attempt to resolve this\n // optional dependency until a turso tree is actually opened. Installed on first use if missing.\n let turso: Awaited<ReturnType<typeof tursoApi>>;\n try {\n turso = await tursoApi();\n } catch (err) {\n if (err instanceof SenseError) throw err;\n throw new SenseError('STORE_DEPENDENCY_MISSING', `store \"turso\" needs the ${TURSO_PACKAGE} package (${(err as Error).message})`);\n }\n\n let db: Database;\n try {\n // Floored at the same 30s sqlite opens with. `timeout` is connect-time only in this client;\n // the derived value is set via runtime PRAGMA below. `index_method` is required for ensureSchema()'s FTS indexes (T1).\n db = await turso.connect(dbPath, { timeout: 30_000, experimental: ['index_method'] });\n } catch (err) {\n throw new SenseError('STORE_DEPENDENCY_MISSING', `store \"turso\" failed to open ${dbPath}: ${(err as Error).message}`);\n }\n return { handle: db, conn: createConnection(db) };\n}\n\n// This store's dialect (types.ts's OpenDialect) for the shared orchestration in store/open.ts.\nexport const tursoOpenDialect: OpenDialect<Database> = {\n filename: DB_FILENAME,\n schemaVersion: SCHEMA_VERSION,\n reconcileDialect: tursoDialect,\n connect,\n close,\n // \"Locking error: Failed locking file ...\", worded per platform: posix \"File is locked by another\n // process\", Windows \"another process has locked a portion of the file (os error 33)\". Distinct from\n // the write-time \"database is locked\" its connect-time `timeout` covers; that one never reaches here.\n isLocked: (err) => /File is locked by another process|locked a portion of the file/.test(err.message),\n ensureSchema,\n setDerivedBusyTimeout,\n createStore: (handle, conn) => createStore(handle, conn),\n};\n\nexport async function openTurso(cfg: ResolvedConfig): Promise<OpenResult> {\n return openWithDialect(cfg, tursoOpenDialect);\n}\n"],"names":["featureSignature","STORE_DIMS","SenseError","activeFeatures","FEATURES","openWithDialect","getMeta","setMeta","checkpointWal","createConnection","TURSO_PACKAGE","tursoApi","CONTENT_FTS_DDL","tursoDialect","createStore","DB_FILENAME","SCHEMA_VERSION","ensureSchema","_handle","conn","cfg","exec","ddl","feature","name","schema","close","handle","setDerivedBusyTimeout","ms","connect","dbPath","_cfg","turso","err","message","db","timeout","experimental","tursoOpenDialect","filename","schemaVersion","reconcileDialect","isLocked","test","openTurso"],"mappings":"AAEA,SAASA,gBAAgB,QAAQ,wBAAwB;AACzD,SAASC,UAAU,QAAQ,uBAAuB;AAClD,SAASC,UAAU,QAAQ,kBAAkB;AAC7C,SAASC,cAAc,EAAEC,QAAQ,QAAQ,0BAA0B;AAEnE,SAASC,eAAe,QAAQ,aAAa;AAC7C,SAASC,OAAO,EAAEC,OAAO,QAAQ,eAAe;AAEhD,SAASC,aAAa,EAAEC,gBAAgB,QAAQ,kBAAkB;AAClE,SAASC,aAAa,EAAEC,QAAQ,QAAQ,cAAc;AACtD,SAASC,eAAe,EAAEC,YAAY,QAAQ,iBAAiB;AAC/D,SAASC,WAAW,QAAQ,aAAa;AAEzC,OAAO,MAAMC,cAAc,iBAAiB;AAC5C,wFAAwF;AACxF,mGAAmG;AACnG,OAAO,MAAMC,iBAAiB,IAAI;AAIlC,4FAA4F;AAC5F,mFAAmF;AACnF,eAAeC,aAAaC,OAAiB,EAAEC,IAAgB,EAAEC,GAAW;IAC1E,MAAMD,KAAKE,IAAI,CAAC,CAAC,oIAAoI,CAAC;IACtJ,MAAMF,KAAKE,IAAI,CAAC,CAAC,wJAAwJ,CAAC;IAC1K,KAAK,MAAMC,OAAOV,gBAAiB,MAAMO,KAAKE,IAAI,CAACC;IACnD,MAAMH,KAAKE,IAAI,CAAC,CAAC,gGAAgG,CAAC;IAClH,MAAMF,KAAKE,IAAI,CAAC;IAChB,KAAK,MAAME,WAAWpB,eAAeiB,KAAM;QACzC,8FAA8F;QAC9F,8FAA8F;QAC9F,IAAIG,QAAQC,IAAI,KAAK,SAAS;YAC5B,MAAML,KAAKE,IAAI,CAAC,CAAC,qIAAqI,EAAEpB,WAAW,+BAA+B,CAAC;YACnM;QACF;QACA,MAAMsB,QAAQE,MAAM,CAACN;IACvB;IACA,IAAI,AAAC,MAAMb,QAAQa,MAAM,sBAAuB,MAAM,MAAMZ,QAAQY,MAAM,kBAAkBH;IAC5F,IAAI,AAAC,MAAMV,QAAQa,MAAM,gBAAiB,MAAM,MAAMZ,QAAQY,MAAM,YAAYnB,iBAAiBoB,KAAKhB;AACxG;AAEA,eAAesB,MAAMC,MAAgB;IACnC,MAAMnB,cAAcmB;IACpB,MAAMA,OAAOD,KAAK;AACpB;AAEA,eAAeE,sBAAsBV,OAAiB,EAAEC,IAAgB,EAAEU,EAAU;IAClF,MAAMV,KAAKE,IAAI,CAAC,CAAC,sBAAsB,EAAEQ,IAAI;AAC/C;AAEA,eAAeC,QAAQC,MAAc,EAAEC,IAAoB;IACzD,8FAA8F;IAC9F,gGAAgG;IAChG,IAAIC;IACJ,IAAI;QACFA,QAAQ,MAAMtB;IAChB,EAAE,OAAOuB,KAAK;QACZ,IAAIA,eAAehC,YAAY,MAAMgC;QACrC,MAAM,IAAIhC,WAAW,4BAA4B,CAAC,wBAAwB,EAAEQ,cAAc,UAAU,EAAE,AAACwB,IAAcC,OAAO,CAAC,CAAC,CAAC;IACjI;IAEA,IAAIC;IACJ,IAAI;QACF,4FAA4F;QAC5F,uHAAuH;QACvHA,KAAK,MAAMH,MAAMH,OAAO,CAACC,QAAQ;YAAEM,SAAS;YAAQC,cAAc;gBAAC;aAAe;QAAC;IACrF,EAAE,OAAOJ,KAAK;QACZ,MAAM,IAAIhC,WAAW,4BAA4B,CAAC,6BAA6B,EAAE6B,OAAO,EAAE,EAAE,AAACG,IAAcC,OAAO,EAAE;IACtH;IACA,OAAO;QAAER,QAAQS;QAAIjB,MAAMV,iBAAiB2B;IAAI;AAClD;AAEA,+FAA+F;AAC/F,OAAO,MAAMG,mBAA0C;IACrDC,UAAUzB;IACV0B,eAAezB;IACf0B,kBAAkB7B;IAClBiB;IACAJ;IACA,kGAAkG;IAClG,oGAAoG;IACpG,sGAAsG;IACtGiB,UAAU,CAACT,MAAQ,iEAAiEU,IAAI,CAACV,IAAIC,OAAO;IACpGlB;IACAW;IACAd,aAAa,CAACa,QAAQR,OAASL,YAAYa,QAAQR;AACrD,EAAE;AAEF,OAAO,eAAe0B,UAAUzB,GAAmB;IACjD,OAAOf,gBAAgBe,KAAKmB;AAC9B"}
@@ -2,6 +2,7 @@ import { STORE_DIMS } from '../../embed/types.js';
2
2
  import { getColumns } from '../shared.js';
3
3
  import { withTransaction } from '../transaction.js';
4
4
  import { hasVectorRow, pendingRows } from '../vectors.js';
5
+ import { checkpointWal } from './connection.js';
5
6
  import { fieldStats } from './fieldStats.js';
6
7
  import { queryLexical } from './lexical.js';
7
8
  import { scanCandidates, scanSimilar, writeVectorBatch } from './vectors.js';
@@ -72,6 +73,7 @@ export function createStore(db, conn) {
72
73
  }
73
74
  },
74
75
  async close () {
76
+ await checkpointWal(db);
75
77
  await db.close();
76
78
  }
77
79
  };
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/turso/store.ts"],"sourcesContent":["import type { Database } from '@tursodatabase/database';\nimport { STORE_DIMS } from '../../embed/types.ts';\nimport { getColumns } from '../shared.ts';\nimport { withTransaction } from '../transaction.ts';\nimport type { Capability, Connection, Statement, Store } from '../types.ts';\nimport { hasVectorRow, pendingRows } from '../vectors.ts';\nimport { fieldStats } from './fieldStats.ts';\nimport { queryLexical } from './lexical.ts';\nimport { scanCandidates, scanSimilar, writeVectorBatch } from './vectors.ts';\n\n// No 'snippets': fts_highlight returns the whole column, not a bounded window, so hits use\n// the caller's JS excerpt.\nexport const CAPABILITIES: ReadonlySet<Capability> = new Set(['lexical', 'phrases', 'vectors']);\n\n// Shares one Connection instance (conn) with the builder's own reconcile call so transaction depth\n// (see transaction.ts) is tracked against the same object everywhere.\nexport function createStore(db: Database, conn: Connection): Store {\n return {\n name: 'turso',\n capabilities: CAPABILITIES,\n async exec(sql: string): Promise<void> {\n await conn.exec(sql);\n },\n async prepare(sql: string): Promise<Statement> {\n return conn.prepare(sql);\n },\n async runBatch(sql: string, paramRows: unknown[][]): Promise<void> {\n await conn.runBatch(sql, paramRows);\n },\n async transaction<T>(fn: () => Promise<T>): Promise<T> {\n return withTransaction(conn, fn);\n },\n docs: {\n async columns() {\n return [...(await getColumns(conn))];\n },\n fieldStats: (columns, scopeWhere) => fieldStats(conn, columns, scopeWhere),\n },\n lexical: {\n query: (terms, opts) => queryLexical(conn, terms, opts),\n },\n // The column's fixed DDL width (STORE_DIMS) is what every scan binds against, not the\n // interface's per-call storeDims -- see vectors.ts's padded() for why a shorter vector is still correct against a wider column.\n vectors: {\n pending: () => pendingRows(conn),\n writeVectors: (rows) => writeVectorBatch(conn, STORE_DIMS, rows),\n candidates: (qv, _storeDims, fetch, allowed) => scanCandidates(conn, qv, STORE_DIMS, fetch, allowed),\n similar: (path, opts) => scanSimilar(conn, STORE_DIMS, path, opts),\n hasVector: (path) => hasVectorRow(conn, path),\n },\n async engineStatus() {\n // Read back rather than recomputed: this is what open() actually set (3x the largest\n // recorded reconcile, floored at 30s, capped at 10min). Turso's PRAGMA busy_timeout names its column \"busy_timeout\" (spike-verified), not \"timeout\" like real SQLite.\n const row = (await (await db.prepare('PRAGMA busy_timeout')).get()) as { busy_timeout: number };\n return { busy_timeout: `${row.busy_timeout}ms (derived: 3x the largest reconcile this cache has recorded, floored at 30000ms)` };\n },\n raw: {\n async prepare(sql: string) {\n const stmt = await db.prepare(sql);\n stmt.safeIntegers(true); // int64 past 2^53 arrives as BigInt instead of losing precision\n return {\n columns: () => stmt.columns(),\n // sense sql streams through the client's own async generator.\n iterate: async function* (...params: unknown[]) {\n yield* stmt.iterate(...params);\n },\n };\n },\n },\n async close() {\n await db.close();\n },\n };\n}\n"],"names":["STORE_DIMS","getColumns","withTransaction","hasVectorRow","pendingRows","fieldStats","queryLexical","scanCandidates","scanSimilar","writeVectorBatch","CAPABILITIES","Set","createStore","db","conn","name","capabilities","exec","sql","prepare","runBatch","paramRows","transaction","fn","docs","columns","scopeWhere","lexical","query","terms","opts","vectors","pending","writeVectors","rows","candidates","qv","_storeDims","fetch","allowed","similar","path","hasVector","engineStatus","row","get","busy_timeout","raw","stmt","safeIntegers","iterate","params","close"],"mappings":"AACA,SAASA,UAAU,QAAQ,uBAAuB;AAClD,SAASC,UAAU,QAAQ,eAAe;AAC1C,SAASC,eAAe,QAAQ,oBAAoB;AAEpD,SAASC,YAAY,EAAEC,WAAW,QAAQ,gBAAgB;AAC1D,SAASC,UAAU,QAAQ,kBAAkB;AAC7C,SAASC,YAAY,QAAQ,eAAe;AAC5C,SAASC,cAAc,EAAEC,WAAW,EAAEC,gBAAgB,QAAQ,eAAe;AAE7E,2FAA2F;AAC3F,2BAA2B;AAC3B,OAAO,MAAMC,eAAwC,IAAIC,IAAI;IAAC;IAAW;IAAW;CAAU,EAAE;AAEhG,mGAAmG;AACnG,sEAAsE;AACtE,OAAO,SAASC,YAAYC,EAAY,EAAEC,IAAgB;IACxD,OAAO;QACLC,MAAM;QACNC,cAAcN;QACd,MAAMO,MAAKC,GAAW;YACpB,MAAMJ,KAAKG,IAAI,CAACC;QAClB;QACA,MAAMC,SAAQD,GAAW;YACvB,OAAOJ,KAAKK,OAAO,CAACD;QACtB;QACA,MAAME,UAASF,GAAW,EAAEG,SAAsB;YAChD,MAAMP,KAAKM,QAAQ,CAACF,KAAKG;QAC3B;QACA,MAAMC,aAAeC,EAAoB;YACvC,OAAOrB,gBAAgBY,MAAMS;QAC/B;QACAC,MAAM;YACJ,MAAMC;gBACJ,OAAO;uBAAK,MAAMxB,WAAWa;iBAAO;YACtC;YACAT,YAAY,CAACoB,SAASC,aAAerB,WAAWS,MAAMW,SAASC;QACjE;QACAC,SAAS;YACPC,OAAO,CAACC,OAAOC,OAASxB,aAAaQ,MAAMe,OAAOC;QACpD;QACA,sFAAsF;QACtF,gIAAgI;QAChIC,SAAS;YACPC,SAAS,IAAM5B,YAAYU;YAC3BmB,cAAc,CAACC,OAASzB,iBAAiBK,MAAMd,YAAYkC;YAC3DC,YAAY,CAACC,IAAIC,YAAYC,OAAOC,UAAYhC,eAAeO,MAAMsB,IAAIpC,YAAYsC,OAAOC;YAC5FC,SAAS,CAACC,MAAMX,OAAStB,YAAYM,MAAMd,YAAYyC,MAAMX;YAC7DY,WAAW,CAACD,OAAStC,aAAaW,MAAM2B;QAC1C;QACA,MAAME;YACJ,qFAAqF;YACrF,sKAAsK;YACtK,MAAMC,MAAO,MAAM,AAAC,CAAA,MAAM/B,GAAGM,OAAO,CAAC,sBAAqB,EAAG0B,GAAG;YAChE,OAAO;gBAAEC,cAAc,GAAGF,IAAIE,YAAY,CAAC,kFAAkF,CAAC;YAAC;QACjI;QACAC,KAAK;YACH,MAAM5B,SAAQD,GAAW;gBACvB,MAAM8B,OAAO,MAAMnC,GAAGM,OAAO,CAACD;gBAC9B8B,KAAKC,YAAY,CAAC,OAAO,gEAAgE;gBACzF,OAAO;oBACLxB,SAAS,IAAMuB,KAAKvB,OAAO;oBAC3B,8DAA8D;oBAC9DyB,SAAS,gBAAiB,GAAGC,MAAiB;wBAC5C,OAAOH,KAAKE,OAAO,IAAIC;oBACzB;gBACF;YACF;QACF;QACA,MAAMC;YACJ,MAAMvC,GAAGuC,KAAK;QAChB;IACF;AACF"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/turso/store.ts"],"sourcesContent":["import type { Database } from '@tursodatabase/database';\nimport { STORE_DIMS } from '../../embed/types.ts';\nimport { getColumns } from '../shared.ts';\nimport { withTransaction } from '../transaction.ts';\nimport type { Capability, Connection, Statement, Store } from '../types.ts';\nimport { hasVectorRow, pendingRows } from '../vectors.ts';\nimport { checkpointWal } from './connection.ts';\nimport { fieldStats } from './fieldStats.ts';\nimport { queryLexical } from './lexical.ts';\nimport { scanCandidates, scanSimilar, writeVectorBatch } from './vectors.ts';\n\n// No 'snippets': fts_highlight returns the whole column, not a bounded window, so hits use\n// the caller's JS excerpt.\nexport const CAPABILITIES: ReadonlySet<Capability> = new Set(['lexical', 'phrases', 'vectors']);\n\n// Shares one Connection instance (conn) with the builder's own reconcile call so transaction depth\n// (see transaction.ts) is tracked against the same object everywhere.\nexport function createStore(db: Database, conn: Connection): Store {\n return {\n name: 'turso',\n capabilities: CAPABILITIES,\n async exec(sql: string): Promise<void> {\n await conn.exec(sql);\n },\n async prepare(sql: string): Promise<Statement> {\n return conn.prepare(sql);\n },\n async runBatch(sql: string, paramRows: unknown[][]): Promise<void> {\n await conn.runBatch(sql, paramRows);\n },\n async transaction<T>(fn: () => Promise<T>): Promise<T> {\n return withTransaction(conn, fn);\n },\n docs: {\n async columns() {\n return [...(await getColumns(conn))];\n },\n fieldStats: (columns, scopeWhere) => fieldStats(conn, columns, scopeWhere),\n },\n lexical: {\n query: (terms, opts) => queryLexical(conn, terms, opts),\n },\n // The column's fixed DDL width (STORE_DIMS) is what every scan binds against, not the\n // interface's per-call storeDims -- see vectors.ts's padded() for why a shorter vector is still correct against a wider column.\n vectors: {\n pending: () => pendingRows(conn),\n writeVectors: (rows) => writeVectorBatch(conn, STORE_DIMS, rows),\n candidates: (qv, _storeDims, fetch, allowed) => scanCandidates(conn, qv, STORE_DIMS, fetch, allowed),\n similar: (path, opts) => scanSimilar(conn, STORE_DIMS, path, opts),\n hasVector: (path) => hasVectorRow(conn, path),\n },\n async engineStatus() {\n // Read back rather than recomputed: this is what open() actually set (3x the largest\n // recorded reconcile, floored at 30s, capped at 10min). Turso's PRAGMA busy_timeout names its column \"busy_timeout\" (spike-verified), not \"timeout\" like real SQLite.\n const row = (await (await db.prepare('PRAGMA busy_timeout')).get()) as { busy_timeout: number };\n return { busy_timeout: `${row.busy_timeout}ms (derived: 3x the largest reconcile this cache has recorded, floored at 30000ms)` };\n },\n raw: {\n async prepare(sql: string) {\n const stmt = await db.prepare(sql);\n stmt.safeIntegers(true); // int64 past 2^53 arrives as BigInt instead of losing precision\n return {\n columns: () => stmt.columns(),\n // sense sql streams through the client's own async generator.\n iterate: async function* (...params: unknown[]) {\n yield* stmt.iterate(...params);\n },\n };\n },\n },\n async close() {\n await checkpointWal(db);\n await db.close();\n },\n };\n}\n"],"names":["STORE_DIMS","getColumns","withTransaction","hasVectorRow","pendingRows","checkpointWal","fieldStats","queryLexical","scanCandidates","scanSimilar","writeVectorBatch","CAPABILITIES","Set","createStore","db","conn","name","capabilities","exec","sql","prepare","runBatch","paramRows","transaction","fn","docs","columns","scopeWhere","lexical","query","terms","opts","vectors","pending","writeVectors","rows","candidates","qv","_storeDims","fetch","allowed","similar","path","hasVector","engineStatus","row","get","busy_timeout","raw","stmt","safeIntegers","iterate","params","close"],"mappings":"AACA,SAASA,UAAU,QAAQ,uBAAuB;AAClD,SAASC,UAAU,QAAQ,eAAe;AAC1C,SAASC,eAAe,QAAQ,oBAAoB;AAEpD,SAASC,YAAY,EAAEC,WAAW,QAAQ,gBAAgB;AAC1D,SAASC,aAAa,QAAQ,kBAAkB;AAChD,SAASC,UAAU,QAAQ,kBAAkB;AAC7C,SAASC,YAAY,QAAQ,eAAe;AAC5C,SAASC,cAAc,EAAEC,WAAW,EAAEC,gBAAgB,QAAQ,eAAe;AAE7E,2FAA2F;AAC3F,2BAA2B;AAC3B,OAAO,MAAMC,eAAwC,IAAIC,IAAI;IAAC;IAAW;IAAW;CAAU,EAAE;AAEhG,mGAAmG;AACnG,sEAAsE;AACtE,OAAO,SAASC,YAAYC,EAAY,EAAEC,IAAgB;IACxD,OAAO;QACLC,MAAM;QACNC,cAAcN;QACd,MAAMO,MAAKC,GAAW;YACpB,MAAMJ,KAAKG,IAAI,CAACC;QAClB;QACA,MAAMC,SAAQD,GAAW;YACvB,OAAOJ,KAAKK,OAAO,CAACD;QACtB;QACA,MAAME,UAASF,GAAW,EAAEG,SAAsB;YAChD,MAAMP,KAAKM,QAAQ,CAACF,KAAKG;QAC3B;QACA,MAAMC,aAAeC,EAAoB;YACvC,OAAOtB,gBAAgBa,MAAMS;QAC/B;QACAC,MAAM;YACJ,MAAMC;gBACJ,OAAO;uBAAK,MAAMzB,WAAWc;iBAAO;YACtC;YACAT,YAAY,CAACoB,SAASC,aAAerB,WAAWS,MAAMW,SAASC;QACjE;QACAC,SAAS;YACPC,OAAO,CAACC,OAAOC,OAASxB,aAAaQ,MAAMe,OAAOC;QACpD;QACA,sFAAsF;QACtF,gIAAgI;QAChIC,SAAS;YACPC,SAAS,IAAM7B,YAAYW;YAC3BmB,cAAc,CAACC,OAASzB,iBAAiBK,MAAMf,YAAYmC;YAC3DC,YAAY,CAACC,IAAIC,YAAYC,OAAOC,UAAYhC,eAAeO,MAAMsB,IAAIrC,YAAYuC,OAAOC;YAC5FC,SAAS,CAACC,MAAMX,OAAStB,YAAYM,MAAMf,YAAY0C,MAAMX;YAC7DY,WAAW,CAACD,OAASvC,aAAaY,MAAM2B;QAC1C;QACA,MAAME;YACJ,qFAAqF;YACrF,sKAAsK;YACtK,MAAMC,MAAO,MAAM,AAAC,CAAA,MAAM/B,GAAGM,OAAO,CAAC,sBAAqB,EAAG0B,GAAG;YAChE,OAAO;gBAAEC,cAAc,GAAGF,IAAIE,YAAY,CAAC,kFAAkF,CAAC;YAAC;QACjI;QACAC,KAAK;YACH,MAAM5B,SAAQD,GAAW;gBACvB,MAAM8B,OAAO,MAAMnC,GAAGM,OAAO,CAACD;gBAC9B8B,KAAKC,YAAY,CAAC,OAAO,gEAAgE;gBACzF,OAAO;oBACLxB,SAAS,IAAMuB,KAAKvB,OAAO;oBAC3B,8DAA8D;oBAC9DyB,SAAS,gBAAiB,GAAGC,MAAiB;wBAC5C,OAAOH,KAAKE,OAAO,IAAIC;oBACzB;gBACF;YACF;QACF;QACA,MAAMC;YACJ,MAAMhD,cAAcS;YACpB,MAAMA,GAAGuC,KAAK;QAChB;IACF;AACF"}
@@ -3,7 +3,7 @@ import { normalizeText } from '../scan/frontmatter.js';
3
3
  // Flat prose from blocks a caller already parsed, via the same extractor the chunker uses
4
4
  // (src/chunk) -- words and vectors read one definition of "the prose of this note".
5
5
  export function textFromBlocks(blocks) {
6
- return normalizeText(extractTexts(blocks.map((block)=>block.node)));
6
+ return normalizeText(extractTexts(blocks));
7
7
  }
8
8
  // Thin parse + delegate, for callers with no blocks of their own already.
9
9
  export function stripText(value) {
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/text/strip.ts"],"sourcesContent":["import type { Block } from '../chunk/index.ts';\nimport { extractTexts, parse } from '../chunk/index.ts';\nimport { normalizeText } from '../scan/frontmatter.ts';\n\n// Flat prose from blocks a caller already parsed, via the same extractor the chunker uses\n// (src/chunk) -- words and vectors read one definition of \"the prose of this note\".\nexport function textFromBlocks(blocks: Block[]): string {\n return normalizeText(extractTexts(blocks.map((block) => block.node)));\n}\n\n// Thin parse + delegate, for callers with no blocks of their own already.\nexport function stripText(value: string): string {\n return textFromBlocks(parse(value));\n}\n"],"names":["extractTexts","parse","normalizeText","textFromBlocks","blocks","map","block","node","stripText","value"],"mappings":"AACA,SAASA,YAAY,EAAEC,KAAK,QAAQ,oBAAoB;AACxD,SAASC,aAAa,QAAQ,yBAAyB;AAEvD,0FAA0F;AAC1F,oFAAoF;AACpF,OAAO,SAASC,eAAeC,MAAe;IAC5C,OAAOF,cAAcF,aAAaI,OAAOC,GAAG,CAAC,CAACC,QAAUA,MAAMC,IAAI;AACpE;AAEA,0EAA0E;AAC1E,OAAO,SAASC,UAAUC,KAAa;IACrC,OAAON,eAAeF,MAAMQ;AAC9B"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/text/strip.ts"],"sourcesContent":["import type { Block } from '../chunk/index.ts';\nimport { extractTexts, parse } from '../chunk/index.ts';\nimport { normalizeText } from '../scan/frontmatter.ts';\n\n// Flat prose from blocks a caller already parsed, via the same extractor the chunker uses\n// (src/chunk) -- words and vectors read one definition of \"the prose of this note\".\nexport function textFromBlocks(blocks: Block[]): string {\n return normalizeText(extractTexts(blocks));\n}\n\n// Thin parse + delegate, for callers with no blocks of their own already.\nexport function stripText(value: string): string {\n return textFromBlocks(parse(value));\n}\n"],"names":["extractTexts","parse","normalizeText","textFromBlocks","blocks","stripText","value"],"mappings":"AACA,SAASA,YAAY,EAAEC,KAAK,QAAQ,oBAAoB;AACxD,SAASC,aAAa,QAAQ,yBAAyB;AAEvD,0FAA0F;AAC1F,oFAAoF;AACpF,OAAO,SAASC,eAAeC,MAAe;IAC5C,OAAOF,cAAcF,aAAaI;AACpC;AAEA,0EAA0E;AAC1E,OAAO,SAASC,UAAUC,KAAa;IACrC,OAAOH,eAAeF,MAAMK;AAC9B"}
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/workers/parse.ts"],"sourcesContent":["import Tinypool from 'tinypool';\nimport type { Config, FeatureName } from '../config/index.ts';\nimport { FEATURES } from '../features/index.ts';\nimport type { ParsedDoc } from '../scan/index.ts';\nimport { parseFile } from '../scan/index.ts';\nimport type { FileStat } from '../scan/list.ts';\nimport { featuresForFile } from '../scan/reparse.ts';\nimport type { WorkerErrorPayload } from '../scan/worker-error.ts';\nimport { serializeError } from '../scan/worker-error.ts';\n\n// Constant for the whole dispatch, so it crosses once per worker instead of once per task. A\n// Feature carries closures and cannot cross the thread boundary; its name can, and the registry here resolves it back.\nexport interface ParseWorkerData {\n cfg: Config;\n featureNames: FeatureName[];\n}\n\n// tinypool's task, in and out. The task itself is one FileStat. Result carries only what\n// parseFile already returns -- extracted text and per-feature values, never the mdast tree.\nexport type ParseTask = FileStat;\n\n// parseMs is the worker's own hrtime for parseFile, excluding dispatch and the return clone --\n// what separates it from the pool's dispatch-to-drain `parse` stage.\nexport type ParseTaskResult = { ok: true; doc: ParsedDoc; warnings: string[]; parseMs: number } | { ok: false; error: WorkerErrorPayload };\n\n// Read once per worker, not per task.\nconst { cfg, featureNames } = Tinypool.workerData as ParseWorkerData;\n// Filtering the registry (rather than mapping the names) keeps registry order, which is the\n// order `extracted` keys land in on the serial path.\nconst selected = FEATURES.filter((feature) => featureNames.includes(feature.name));\n\nexport default function parseTask(file: ParseTask): ParseTaskResult {\n try {\n const start = process.hrtime.bigint();\n const { doc, warnings } = parseFile(file, featuresForFile(selected, cfg, file), cfg);\n const parseMs = Number(process.hrtime.bigint() - start) / 1e6;\n return { ok: true, doc, warnings, parseMs };\n } catch (err) {\n return { ok: false, error: serializeError(err) };\n }\n}\n"],"names":["Tinypool","FEATURES","parseFile","featuresForFile","serializeError","cfg","featureNames","workerData","selected","filter","feature","includes","name","parseTask","file","start","process","hrtime","bigint","doc","warnings","parseMs","Number","ok","err","error"],"mappings":"AAAA,OAAOA,cAAc,WAAW;AAEhC,SAASC,QAAQ,QAAQ,uBAAuB;AAEhD,SAASC,SAAS,QAAQ,mBAAmB;AAE7C,SAASC,eAAe,QAAQ,qBAAqB;AAErD,SAASC,cAAc,QAAQ,0BAA0B;AAiBzD,sCAAsC;AACtC,MAAM,EAAEC,GAAG,EAAEC,YAAY,EAAE,GAAGN,SAASO,UAAU;AACjD,4FAA4F;AAC5F,qDAAqD;AACrD,MAAMC,WAAWP,SAASQ,MAAM,CAAC,CAACC,UAAYJ,aAAaK,QAAQ,CAACD,QAAQE,IAAI;AAEhF,eAAe,SAASC,UAAUC,IAAe;IAC/C,IAAI;QACF,MAAMC,QAAQC,QAAQC,MAAM,CAACC,MAAM;QACnC,MAAM,EAAEC,GAAG,EAAEC,QAAQ,EAAE,GAAGlB,UAAUY,MAAMX,gBAAgBK,UAAUH,KAAKS,OAAOT;QAChF,MAAMgB,UAAUC,OAAON,QAAQC,MAAM,CAACC,MAAM,KAAKH,SAAS;QAC1D,OAAO;YAAEQ,IAAI;YAAMJ;YAAKC;YAAUC;QAAQ;IAC5C,EAAE,OAAOG,KAAK;QACZ,OAAO;YAAED,IAAI;YAAOE,OAAOrB,eAAeoB;QAAK;IACjD;AACF"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/workers/parse.ts"],"sourcesContent":["import Tinypool from 'tinypool';\nimport type { Config, FeatureName } from '../config/index.ts';\nimport { FEATURES } from '../features/index.ts';\nimport type { ParsedDoc } from '../scan/index.ts';\nimport { parseFile } from '../scan/index.ts';\nimport type { FileStat } from '../scan/list.ts';\nimport { featuresForFile } from '../scan/reparse.ts';\nimport type { WorkerErrorPayload } from '../scan/worker-error.ts';\nimport { serializeError } from '../scan/worker-error.ts';\n\n// Constant for the whole dispatch, so it crosses once per worker instead of once per task. A\n// Feature carries closures and cannot cross the thread boundary; its name can, and the registry here resolves it back.\nexport interface ParseWorkerData {\n cfg: Config;\n featureNames: FeatureName[];\n}\n\n// tinypool's task, in and out. The task itself is one FileStat. Result carries only what\n// parseFile already returns -- extracted text and per-feature values, never the token tree.\nexport type ParseTask = FileStat;\n\n// parseMs is the worker's own hrtime for parseFile, excluding dispatch and the return clone --\n// what separates it from the pool's dispatch-to-drain `parse` stage.\nexport type ParseTaskResult = { ok: true; doc: ParsedDoc; warnings: string[]; parseMs: number } | { ok: false; error: WorkerErrorPayload };\n\n// Read once per worker, not per task.\nconst { cfg, featureNames } = Tinypool.workerData as ParseWorkerData;\n// Filtering the registry (rather than mapping the names) keeps registry order, which is the\n// order `extracted` keys land in on the serial path.\nconst selected = FEATURES.filter((feature) => featureNames.includes(feature.name));\n\nexport default function parseTask(file: ParseTask): ParseTaskResult {\n try {\n const start = process.hrtime.bigint();\n const { doc, warnings } = parseFile(file, featuresForFile(selected, cfg, file), cfg);\n const parseMs = Number(process.hrtime.bigint() - start) / 1e6;\n return { ok: true, doc, warnings, parseMs };\n } catch (err) {\n return { ok: false, error: serializeError(err) };\n }\n}\n"],"names":["Tinypool","FEATURES","parseFile","featuresForFile","serializeError","cfg","featureNames","workerData","selected","filter","feature","includes","name","parseTask","file","start","process","hrtime","bigint","doc","warnings","parseMs","Number","ok","err","error"],"mappings":"AAAA,OAAOA,cAAc,WAAW;AAEhC,SAASC,QAAQ,QAAQ,uBAAuB;AAEhD,SAASC,SAAS,QAAQ,mBAAmB;AAE7C,SAASC,eAAe,QAAQ,qBAAqB;AAErD,SAASC,cAAc,QAAQ,0BAA0B;AAiBzD,sCAAsC;AACtC,MAAM,EAAEC,GAAG,EAAEC,YAAY,EAAE,GAAGN,SAASO,UAAU;AACjD,4FAA4F;AAC5F,qDAAqD;AACrD,MAAMC,WAAWP,SAASQ,MAAM,CAAC,CAACC,UAAYJ,aAAaK,QAAQ,CAACD,QAAQE,IAAI;AAEhF,eAAe,SAASC,UAAUC,IAAe;IAC/C,IAAI;QACF,MAAMC,QAAQC,QAAQC,MAAM,CAACC,MAAM;QACnC,MAAM,EAAEC,GAAG,EAAEC,QAAQ,EAAE,GAAGlB,UAAUY,MAAMX,gBAAgBK,UAAUH,KAAKS,OAAOT;QAChF,MAAMgB,UAAUC,OAAON,QAAQC,MAAM,CAACC,MAAM,KAAKH,SAAS;QAC1D,OAAO;YAAEQ,IAAI;YAAMJ;YAAKC;YAAUC;QAAQ;IAC5C,EAAE,OAAOG,KAAK;QACZ,OAAO;YAAED,IAAI;YAAOE,OAAOrB,eAAeoB;QAAK;IACjD;AACF"}
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "sensemaking",
3
- "version": "0.22.1",
3
+ "version": "0.23.0",
4
4
  "description": "Query and search your markdown notes with context-aware progressive disclosure: SQL over frontmatter, links, and text, plus semantic search and link-graph ranking. No server, no build step",
5
5
  "keywords": [
6
6
  "markdown",
@@ -84,24 +84,16 @@
84
84
  "@huggingface/tokenizers": "^0.1.3",
85
85
  "franc-min": "^6.2.0",
86
86
  "install-module-linked": "^2.0.0",
87
- "mdast-util-from-markdown": "^2.0.3",
88
- "mdast-util-gfm-autolink-literal": "^2.0.0",
89
- "mdast-util-gfm-footnote": "^2.0.0",
90
- "mdast-util-gfm-strikethrough": "^2.0.0",
91
- "mdast-util-gfm-table": "^2.0.0",
92
- "mdast-util-gfm-task-list-item": "^2.0.0",
93
- "micromark-extension-gfm-autolink-literal": "^2.0.0",
94
- "micromark-extension-gfm-footnote": "^2.0.0",
95
- "micromark-extension-gfm-strikethrough": "^2.0.0",
96
- "micromark-extension-gfm-table": "^2.0.0",
97
- "micromark-extension-gfm-task-list-item": "^2.0.0",
87
+ "markdown-it": "^15.0.1",
88
+ "markdown-it-footnote": "^4.0.0",
89
+ "markdown-it-task-lists": "^2.1.1",
98
90
  "tinypool": "^2.1.2",
99
91
  "yaml": "^2.9.0"
100
92
  },
101
93
  "devDependencies": {
102
94
  "@duckdb/node-api": "*",
103
95
  "@tursodatabase/database": "*",
104
- "@types/mdast": "^4.0.4",
96
+ "@types/markdown-it-footnote": "^3.0.4",
105
97
  "@types/mocha": "*",
106
98
  "@types/node": "*",
107
99
  "cr": "^0.1.0",
@@ -112,11 +104,5 @@
112
104
  },
113
105
  "engines": {
114
106
  "node": ">=22.20"
115
- },
116
- "allowScripts": {
117
- "node-semvers": true,
118
- "node-filename-to-dist-paths": true,
119
- "node-version-use": true,
120
- "thread-sleep-compat": true
121
107
  }
122
108
  }