sensemaking 0.16.0 → 0.17.1

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 (322) hide show
  1. package/README.md +1 -1
  2. package/dist/cjs/chunk/extract.d.cts +2 -0
  3. package/dist/cjs/chunk/extract.d.ts +2 -0
  4. package/dist/cjs/chunk/extract.js +139 -0
  5. package/dist/cjs/chunk/extract.js.map +1 -0
  6. package/dist/cjs/chunk/group.d.cts +4 -0
  7. package/dist/cjs/chunk/group.d.ts +4 -0
  8. package/dist/cjs/chunk/group.js +489 -0
  9. package/dist/cjs/chunk/group.js.map +1 -0
  10. package/dist/cjs/chunk/index.d.cts +8 -0
  11. package/dist/cjs/chunk/index.d.ts +8 -0
  12. package/dist/cjs/chunk/index.js +44 -0
  13. package/dist/cjs/chunk/index.js.map +1 -0
  14. package/dist/cjs/chunk/parse.d.cts +2 -0
  15. package/dist/cjs/chunk/parse.d.ts +2 -0
  16. package/dist/cjs/chunk/parse.js +71 -0
  17. package/dist/cjs/chunk/parse.js.map +1 -0
  18. package/dist/cjs/chunk/types.d.cts +19 -0
  19. package/dist/cjs/chunk/types.d.ts +19 -0
  20. package/dist/cjs/chunk/types.js +5 -0
  21. package/dist/cjs/chunk/types.js.map +1 -0
  22. package/dist/cjs/chunk/version.d.cts +1 -0
  23. package/dist/cjs/chunk/version.d.ts +1 -0
  24. package/dist/cjs/chunk/version.js +14 -0
  25. package/dist/cjs/chunk/version.js.map +1 -0
  26. package/dist/cjs/cli/map.js +5 -5
  27. package/dist/cjs/cli/map.js.map +1 -1
  28. package/dist/cjs/cli/named.js +3 -3
  29. package/dist/cjs/cli/named.js.map +1 -1
  30. package/dist/cjs/cli/path.js +9 -8
  31. package/dist/cjs/cli/path.js.map +1 -1
  32. package/dist/cjs/cli/peek.js +5 -5
  33. package/dist/cjs/cli/peek.js.map +1 -1
  34. package/dist/cjs/cli/related.js +5 -5
  35. package/dist/cjs/cli/related.js.map +1 -1
  36. package/dist/cjs/cli/search.js +5 -5
  37. package/dist/cjs/cli/search.js.map +1 -1
  38. package/dist/cjs/cli/shared.d.cts +1 -1
  39. package/dist/cjs/cli/shared.d.ts +1 -1
  40. package/dist/cjs/cli/shared.js +3 -3
  41. package/dist/cjs/cli/shared.js.map +1 -1
  42. package/dist/cjs/cli/status.js +19 -19
  43. package/dist/cjs/cli/status.js.map +1 -1
  44. package/dist/cjs/commands/map.d.cts +1 -1
  45. package/dist/cjs/commands/map.d.ts +1 -1
  46. package/dist/cjs/commands/map.js.map +1 -1
  47. package/dist/cjs/commands/peek.d.cts +1 -1
  48. package/dist/cjs/commands/peek.d.ts +1 -1
  49. package/dist/cjs/commands/peek.js +13 -6
  50. package/dist/cjs/commands/peek.js.map +1 -1
  51. package/dist/cjs/commands/scope.js +1 -1
  52. package/dist/cjs/commands/scope.js.map +1 -1
  53. package/dist/cjs/commands/search.d.cts +1 -1
  54. package/dist/cjs/commands/search.d.ts +1 -1
  55. package/dist/cjs/commands/search.js +3 -3
  56. package/dist/cjs/commands/search.js.map +1 -1
  57. package/dist/cjs/commands/signals.js +1 -1
  58. package/dist/cjs/commands/signals.js.map +1 -1
  59. package/dist/cjs/config/access.d.cts +3 -1
  60. package/dist/cjs/config/access.d.ts +3 -1
  61. package/dist/cjs/config/access.js +11 -11
  62. package/dist/cjs/config/access.js.map +1 -1
  63. package/dist/cjs/config/index.d.cts +1 -0
  64. package/dist/cjs/config/index.d.ts +1 -0
  65. package/dist/cjs/config/index.js +4 -0
  66. package/dist/cjs/config/index.js.map +1 -1
  67. package/dist/cjs/config/types.d.cts +1 -0
  68. package/dist/cjs/config/types.d.ts +1 -0
  69. package/dist/cjs/config/types.js.map +1 -1
  70. package/dist/cjs/config/validate.d.cts +1 -0
  71. package/dist/cjs/config/validate.d.ts +1 -0
  72. package/dist/cjs/config/validate.js +8 -1
  73. package/dist/cjs/config/validate.js.map +1 -1
  74. package/dist/cjs/db/open.d.cts +1 -1
  75. package/dist/cjs/db/open.d.ts +1 -1
  76. package/dist/cjs/db/open.js +4 -4
  77. package/dist/cjs/db/open.js.map +1 -1
  78. package/dist/cjs/db/reconcile.js +8 -8
  79. package/dist/cjs/db/reconcile.js.map +1 -1
  80. package/dist/cjs/{sql-functions.js → db/sql-functions.js} +1 -1
  81. package/dist/cjs/db/sql-functions.js.map +1 -0
  82. package/dist/cjs/embed/distribution.d.cts +4 -0
  83. package/dist/cjs/embed/distribution.d.ts +4 -0
  84. package/dist/cjs/embed/distribution.js +30 -0
  85. package/dist/cjs/embed/distribution.js.map +1 -0
  86. package/dist/cjs/embed/langfit.d.cts +0 -3
  87. package/dist/cjs/embed/langfit.d.ts +0 -3
  88. package/dist/cjs/embed/langfit.js +53 -96
  89. package/dist/cjs/embed/langfit.js.map +1 -1
  90. package/dist/cjs/embed/languages.js.map +1 -1
  91. package/dist/cjs/embed/query.js +6 -6
  92. package/dist/cjs/embed/query.js.map +1 -1
  93. package/dist/cjs/embed/static.js +8 -35
  94. package/dist/cjs/embed/static.js.map +1 -1
  95. package/dist/cjs/features/embed.d.cts +2 -5
  96. package/dist/cjs/features/embed.d.ts +2 -5
  97. package/dist/cjs/features/embed.js +36 -58
  98. package/dist/cjs/features/embed.js.map +1 -1
  99. package/dist/cjs/{fences.js → features/fences.js} +1 -1
  100. package/dist/cjs/features/fences.js.map +1 -0
  101. package/dist/cjs/features/links.js +1 -1
  102. package/dist/cjs/features/links.js.map +1 -1
  103. package/dist/cjs/features/rank.js +1 -1
  104. package/dist/cjs/features/rank.js.map +1 -1
  105. package/dist/cjs/features/sections.js +4 -3
  106. package/dist/cjs/features/sections.js.map +1 -1
  107. package/dist/cjs/features/tags.js +1 -1
  108. package/dist/cjs/features/tags.js.map +1 -1
  109. package/dist/cjs/features/types.d.cts +4 -2
  110. package/dist/cjs/features/types.d.ts +4 -2
  111. package/dist/cjs/graph/graph.js.map +1 -0
  112. package/dist/cjs/graph/traverse.js.map +1 -0
  113. package/dist/cjs/index.d.cts +4 -4
  114. package/dist/cjs/index.d.ts +4 -4
  115. package/dist/cjs/index.js +4 -1
  116. package/dist/cjs/index.js.map +1 -1
  117. package/dist/cjs/output/column-hint.js.map +1 -0
  118. package/dist/cjs/output/output.js.map +1 -0
  119. package/dist/cjs/output/progress.js.map +1 -0
  120. package/dist/cjs/{search-error.js → output/search-error.js} +1 -1
  121. package/dist/cjs/output/search-error.js.map +1 -0
  122. package/dist/cjs/scan/frontmatter.d.cts +13 -0
  123. package/dist/cjs/scan/frontmatter.d.ts +13 -0
  124. package/dist/cjs/scan/frontmatter.js +229 -0
  125. package/dist/cjs/scan/frontmatter.js.map +1 -0
  126. package/dist/cjs/scan/index.d.cts +25 -0
  127. package/dist/cjs/scan/index.d.ts +25 -0
  128. package/dist/cjs/scan/index.js +107 -0
  129. package/dist/cjs/scan/index.js.map +1 -0
  130. package/dist/cjs/scan/list.d.cts +12 -0
  131. package/dist/cjs/scan/list.d.ts +12 -0
  132. package/dist/cjs/scan/list.js +154 -0
  133. package/dist/cjs/scan/list.js.map +1 -0
  134. package/dist/cjs/text/segment.d.cts +3 -0
  135. package/dist/cjs/text/segment.d.ts +3 -0
  136. package/dist/cjs/{segment.js → text/segment.js} +17 -14
  137. package/dist/cjs/text/segment.js.map +1 -0
  138. package/dist/cjs/text/strip.d.cts +3 -0
  139. package/dist/cjs/text/strip.d.ts +3 -0
  140. package/dist/cjs/text/strip.js +32 -0
  141. package/dist/cjs/text/strip.js.map +1 -0
  142. package/dist/esm/chunk/extract.d.ts +2 -0
  143. package/dist/esm/chunk/extract.js +114 -0
  144. package/dist/esm/chunk/extract.js.map +1 -0
  145. package/dist/esm/chunk/group.d.ts +4 -0
  146. package/dist/esm/chunk/group.js +253 -0
  147. package/dist/esm/chunk/group.js.map +1 -0
  148. package/dist/esm/chunk/index.d.ts +8 -0
  149. package/dist/esm/chunk/index.js +16 -0
  150. package/dist/esm/chunk/index.js.map +1 -0
  151. package/dist/esm/chunk/parse.d.ts +2 -0
  152. package/dist/esm/chunk/parse.js +63 -0
  153. package/dist/esm/chunk/parse.js.map +1 -0
  154. package/dist/esm/chunk/types.d.ts +19 -0
  155. package/dist/esm/chunk/types.js +2 -0
  156. package/dist/esm/chunk/types.js.map +1 -0
  157. package/dist/esm/chunk/version.d.ts +1 -0
  158. package/dist/esm/chunk/version.js +3 -0
  159. package/dist/esm/chunk/version.js.map +1 -0
  160. package/dist/esm/cli/map.js +2 -2
  161. package/dist/esm/cli/map.js.map +1 -1
  162. package/dist/esm/cli/named.js +2 -2
  163. package/dist/esm/cli/named.js.map +1 -1
  164. package/dist/esm/cli/path.js +4 -3
  165. package/dist/esm/cli/path.js.map +1 -1
  166. package/dist/esm/cli/peek.js +2 -2
  167. package/dist/esm/cli/peek.js.map +1 -1
  168. package/dist/esm/cli/related.js +2 -2
  169. package/dist/esm/cli/related.js.map +1 -1
  170. package/dist/esm/cli/search.js +2 -2
  171. package/dist/esm/cli/search.js.map +1 -1
  172. package/dist/esm/cli/shared.d.ts +1 -1
  173. package/dist/esm/cli/shared.js +3 -3
  174. package/dist/esm/cli/shared.js.map +1 -1
  175. package/dist/esm/cli/status.js +3 -3
  176. package/dist/esm/cli/status.js.map +1 -1
  177. package/dist/esm/commands/map.d.ts +1 -1
  178. package/dist/esm/commands/map.js.map +1 -1
  179. package/dist/esm/commands/peek.d.ts +1 -1
  180. package/dist/esm/commands/peek.js +8 -1
  181. package/dist/esm/commands/peek.js.map +1 -1
  182. package/dist/esm/commands/scope.js +1 -1
  183. package/dist/esm/commands/scope.js.map +1 -1
  184. package/dist/esm/commands/search.d.ts +1 -1
  185. package/dist/esm/commands/search.js +3 -3
  186. package/dist/esm/commands/search.js.map +1 -1
  187. package/dist/esm/commands/signals.js +1 -1
  188. package/dist/esm/commands/signals.js.map +1 -1
  189. package/dist/esm/config/access.d.ts +3 -1
  190. package/dist/esm/config/access.js +9 -9
  191. package/dist/esm/config/access.js.map +1 -1
  192. package/dist/esm/config/index.d.ts +1 -0
  193. package/dist/esm/config/index.js +1 -0
  194. package/dist/esm/config/index.js.map +1 -1
  195. package/dist/esm/config/types.d.ts +1 -0
  196. package/dist/esm/config/types.js.map +1 -1
  197. package/dist/esm/config/validate.d.ts +1 -0
  198. package/dist/esm/config/validate.js +8 -2
  199. package/dist/esm/config/validate.js.map +1 -1
  200. package/dist/esm/db/open.d.ts +1 -1
  201. package/dist/esm/db/open.js +5 -5
  202. package/dist/esm/db/open.js.map +1 -1
  203. package/dist/esm/db/reconcile.js +4 -4
  204. package/dist/esm/db/reconcile.js.map +1 -1
  205. package/dist/esm/{sql-functions.js → db/sql-functions.js} +1 -1
  206. package/dist/esm/db/sql-functions.js.map +1 -0
  207. package/dist/esm/embed/distribution.d.ts +4 -0
  208. package/dist/esm/embed/distribution.js +11 -0
  209. package/dist/esm/embed/distribution.js.map +1 -0
  210. package/dist/esm/embed/langfit.d.ts +0 -3
  211. package/dist/esm/embed/langfit.js +8 -8
  212. package/dist/esm/embed/langfit.js.map +1 -1
  213. package/dist/esm/embed/languages.js.map +1 -1
  214. package/dist/esm/embed/query.js +5 -5
  215. package/dist/esm/embed/query.js.map +1 -1
  216. package/dist/esm/embed/static.js +6 -2
  217. package/dist/esm/embed/static.js.map +1 -1
  218. package/dist/esm/features/embed.d.ts +2 -5
  219. package/dist/esm/features/embed.js +34 -32
  220. package/dist/esm/features/embed.js.map +1 -1
  221. package/dist/esm/{fences.js → features/fences.js} +1 -1
  222. package/dist/esm/features/fences.js.map +1 -0
  223. package/dist/esm/features/links.js +1 -1
  224. package/dist/esm/features/links.js.map +1 -1
  225. package/dist/esm/features/rank.js +1 -1
  226. package/dist/esm/features/rank.js.map +1 -1
  227. package/dist/esm/features/sections.js +4 -3
  228. package/dist/esm/features/sections.js.map +1 -1
  229. package/dist/esm/features/tags.js +1 -1
  230. package/dist/esm/features/tags.js.map +1 -1
  231. package/dist/esm/features/types.d.ts +4 -2
  232. package/dist/esm/features/types.js.map +1 -1
  233. package/dist/esm/graph/graph.js.map +1 -0
  234. package/dist/esm/graph/traverse.js.map +1 -0
  235. package/dist/esm/index.d.ts +4 -4
  236. package/dist/esm/index.js +2 -2
  237. package/dist/esm/index.js.map +1 -1
  238. package/dist/esm/output/column-hint.js.map +1 -0
  239. package/dist/esm/output/output.js.map +1 -0
  240. package/dist/esm/output/progress.js.map +1 -0
  241. package/dist/esm/{search-error.js → output/search-error.js} +1 -1
  242. package/dist/esm/output/search-error.js.map +1 -0
  243. package/dist/esm/scan/frontmatter.d.ts +13 -0
  244. package/dist/esm/{scan.js → scan/frontmatter.js} +4 -123
  245. package/dist/esm/scan/frontmatter.js.map +1 -0
  246. package/dist/esm/scan/index.d.ts +25 -0
  247. package/dist/esm/scan/index.js +58 -0
  248. package/dist/esm/scan/index.js.map +1 -0
  249. package/dist/esm/scan/list.d.ts +12 -0
  250. package/dist/esm/scan/list.js +58 -0
  251. package/dist/esm/scan/list.js.map +1 -0
  252. package/dist/esm/text/segment.d.ts +3 -0
  253. package/dist/esm/{segment.js → text/segment.js} +3 -3
  254. package/dist/esm/text/segment.js.map +1 -0
  255. package/dist/esm/text/strip.d.ts +3 -0
  256. package/dist/esm/text/strip.js +12 -0
  257. package/dist/esm/text/strip.js.map +1 -0
  258. package/package.json +15 -2
  259. package/schema.json +8 -1
  260. package/skills/sense/SKILL.md +1 -1
  261. package/skills/sense-setup/SKILL.md +3 -3
  262. package/dist/cjs/column-hint.js.map +0 -1
  263. package/dist/cjs/fences.js.map +0 -1
  264. package/dist/cjs/graph.js.map +0 -1
  265. package/dist/cjs/output.js.map +0 -1
  266. package/dist/cjs/progress.js.map +0 -1
  267. package/dist/cjs/scan.d.cts +0 -35
  268. package/dist/cjs/scan.d.ts +0 -35
  269. package/dist/cjs/scan.js +0 -436
  270. package/dist/cjs/scan.js.map +0 -1
  271. package/dist/cjs/search-error.js.map +0 -1
  272. package/dist/cjs/segment.d.cts +0 -2
  273. package/dist/cjs/segment.d.ts +0 -2
  274. package/dist/cjs/segment.js.map +0 -1
  275. package/dist/cjs/sql-functions.js.map +0 -1
  276. package/dist/cjs/traverse.js.map +0 -1
  277. package/dist/esm/column-hint.js.map +0 -1
  278. package/dist/esm/fences.js.map +0 -1
  279. package/dist/esm/graph.js.map +0 -1
  280. package/dist/esm/output.js.map +0 -1
  281. package/dist/esm/progress.js.map +0 -1
  282. package/dist/esm/scan.d.ts +0 -35
  283. package/dist/esm/scan.js.map +0 -1
  284. package/dist/esm/search-error.js.map +0 -1
  285. package/dist/esm/segment.d.ts +0 -2
  286. package/dist/esm/segment.js.map +0 -1
  287. package/dist/esm/sql-functions.js.map +0 -1
  288. package/dist/esm/traverse.js.map +0 -1
  289. /package/dist/cjs/{sql-functions.d.cts → db/sql-functions.d.cts} +0 -0
  290. /package/dist/cjs/{sql-functions.d.ts → db/sql-functions.d.ts} +0 -0
  291. /package/dist/cjs/{fences.d.cts → features/fences.d.cts} +0 -0
  292. /package/dist/cjs/{fences.d.ts → features/fences.d.ts} +0 -0
  293. /package/dist/cjs/{graph.d.cts → graph/graph.d.cts} +0 -0
  294. /package/dist/cjs/{graph.d.ts → graph/graph.d.ts} +0 -0
  295. /package/dist/cjs/{graph.js → graph/graph.js} +0 -0
  296. /package/dist/cjs/{traverse.d.cts → graph/traverse.d.cts} +0 -0
  297. /package/dist/cjs/{traverse.d.ts → graph/traverse.d.ts} +0 -0
  298. /package/dist/cjs/{traverse.js → graph/traverse.js} +0 -0
  299. /package/dist/cjs/{column-hint.d.cts → output/column-hint.d.cts} +0 -0
  300. /package/dist/cjs/{column-hint.d.ts → output/column-hint.d.ts} +0 -0
  301. /package/dist/cjs/{column-hint.js → output/column-hint.js} +0 -0
  302. /package/dist/cjs/{output.d.cts → output/output.d.cts} +0 -0
  303. /package/dist/cjs/{output.d.ts → output/output.d.ts} +0 -0
  304. /package/dist/cjs/{output.js → output/output.js} +0 -0
  305. /package/dist/cjs/{progress.d.cts → output/progress.d.cts} +0 -0
  306. /package/dist/cjs/{progress.d.ts → output/progress.d.ts} +0 -0
  307. /package/dist/cjs/{progress.js → output/progress.js} +0 -0
  308. /package/dist/cjs/{search-error.d.cts → output/search-error.d.cts} +0 -0
  309. /package/dist/cjs/{search-error.d.ts → output/search-error.d.ts} +0 -0
  310. /package/dist/esm/{sql-functions.d.ts → db/sql-functions.d.ts} +0 -0
  311. /package/dist/esm/{fences.d.ts → features/fences.d.ts} +0 -0
  312. /package/dist/esm/{graph.d.ts → graph/graph.d.ts} +0 -0
  313. /package/dist/esm/{graph.js → graph/graph.js} +0 -0
  314. /package/dist/esm/{traverse.d.ts → graph/traverse.d.ts} +0 -0
  315. /package/dist/esm/{traverse.js → graph/traverse.js} +0 -0
  316. /package/dist/esm/{column-hint.d.ts → output/column-hint.d.ts} +0 -0
  317. /package/dist/esm/{column-hint.js → output/column-hint.js} +0 -0
  318. /package/dist/esm/{output.d.ts → output/output.d.ts} +0 -0
  319. /package/dist/esm/{output.js → output/output.js} +0 -0
  320. /package/dist/esm/{progress.d.ts → output/progress.d.ts} +0 -0
  321. /package/dist/esm/{progress.js → output/progress.js} +0 -0
  322. /package/dist/esm/{search-error.d.ts → output/search-error.d.ts} +0 -0
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/commands/search.ts"],"sourcesContent":["import { readFileSync } from 'node:fs';\nimport { join } from 'node:path';\nimport type { DatabaseSync } from 'node:sqlite';\nimport type { ResolvedConfig } from '../config/index.ts';\nimport { contentTokenize, embedConfig, featureEnabled, resolveSearch } from '../config/index.ts';\nimport { localModelMissing, MODEL_FILENAMES } from '../embed/store.ts';\nimport { SenseError } from '../errors.ts';\nimport type { Row } from '../output.ts';\nimport { searchError } from '../search-error.ts';\nimport { segmentMatch } from '../segment.ts';\nimport { materializeScope, narrowByWhere, rawScope, scopeHasEmbeddings } from './scope.ts';\nimport { linksCandidates, vectorsCandidates, wordsCandidates } from './signals.ts';\n\n// snippet() re-tokenizes each candidate doc, superlinearly: ~10s for one 1MB doc\n// (BENCHMARKING.md). Past this bound, rows get the linear JS excerpt instead.\nconst EXCERPT_WINDOW = 160;\n\n// Bare terms from an FTS5 query string: strips operators/quoting so the oversized-doc excerpt\n// scan matches the same words the query matched on, not FTS5 syntax.\nfunction extractBareTerms(query: string): string[] {\n const cleaned = query\n .replace(/\"/g, ' ')\n .replace(/[()*]/g, ' ')\n .replace(/\\b(AND|OR|NOT|NEAR)\\b(\\/\\d+)?/gi, ' ');\n return cleaned\n .split(/\\s+/)\n .map((tok) => tok.replace(/^[A-Za-z_]\\w*:/, '')) // column filter, e.g. title:term\n .map((tok) => tok.toLowerCase().trim())\n .filter((tok) => tok.length > 0);\n}\n\nfunction findOccurrences(haystackLower: string, terms: string[]): Array<{ start: number; end: number; term: string }> {\n const occ: Array<{ start: number; end: number; term: string }> = [];\n for (const term of terms) {\n let idx = 0;\n for (;;) {\n const found = haystackLower.indexOf(term, idx);\n if (found === -1) break;\n occ.push({ start: found, end: found + term.length, term });\n idx = found + term.length;\n }\n }\n // Longest span first at equal start, so \"tests\" beats its substring \"test\" and the\n // whole word gets highlighted; the emit loop then absorbs the shorter overlap.\n return occ.sort((a, b) => a.start - b.start || b.end - a.end);\n}\n\n// Slides a window over the sorted occurrences, same semantics as FTS5's own best-window\n// pick: most distinct terms, ties broken by most total hits.\nfunction bestWindowStart(occ: Array<{ start: number; end: number; term: string }>, windowSize: number): number {\n let best = occ[0].start;\n let bestDistinct = 0;\n let bestCount = 0;\n for (let i = 0; i < occ.length; i++) {\n const winEnd = occ[i].start + windowSize;\n const seen = new Set<string>();\n let count = 0;\n for (let j = i; j < occ.length && occ[j].start < winEnd; j++) {\n seen.add(occ[j].term);\n count++;\n }\n if (seen.size > bestDistinct || (seen.size === bestDistinct && count > bestCount)) {\n best = occ[i].start;\n bestDistinct = seen.size;\n bestCount = count;\n }\n }\n return best;\n}\n\n// snippet()-shaped excerpt: matched terms wrapped in «», … at cut edges. Linear in doc\n// length, only run for rows actually returned (<= k), unlike snippet()'s per-row cost.\nfunction computeExcerpt(text: string, terms: string[]): { excerpt: string; offset: number } {\n const occ = findOccurrences(text.toLowerCase(), terms);\n if (occ.length === 0) {\n // This is a raw substring scan, not porter-stemmed: a doc matched only through a\n // stemmed variant (query \"negotiate\" vs doc \"negotiating\") finds no occurrence here.\n // Fall back to the doc's start, unmarked, rather than claim a match that isn't there.\n const end = Math.min(text.length, EXCERPT_WINDOW);\n return { excerpt: `${text.slice(0, end).replace(/\\s+/g, ' ').trim()}${end < text.length ? '…' : ''}`, offset: 0 };\n }\n const start = bestWindowStart(occ, EXCERPT_WINDOW);\n const end = Math.min(text.length, start + EXCERPT_WINDOW);\n let out = '';\n let cursor = start;\n for (const o of occ) {\n if (o.start < start || o.end > end) continue;\n // Occurrences of duplicate or substring-overlapping terms (\"test tests\") can overlap;\n // emitting each would duplicate document text. Keep the first, absorb the rest.\n if (o.start < cursor) continue;\n out += `${text.slice(cursor, o.start)}«${text.slice(o.start, o.end)}»`;\n cursor = o.end;\n }\n out += text.slice(cursor, end);\n const prefix = start > 0 ? '…' : '';\n const suffix = end < text.length ? '…' : '';\n return { excerpt: `${prefix}${out.replace(/\\s+/g, ' ')}${suffix}`, offset: start };\n}\n\nfunction lineNumberAt(text: string, offset: number): number {\n let line = 1;\n for (let i = 0; i < offset; i++) if (text.charCodeAt(i) === 10) line++;\n return line;\n}\n\nfunction lineRangeFor(db: DatabaseSync, path: string, line: number): string | null {\n const row = db.prepare('SELECT start_line, end_line FROM sections WHERE \"path\" = ? AND start_line <= ? AND end_line >= ? ORDER BY start_line DESC LIMIT 1').get(path, line, line) as { start_line: number; end_line: number } | undefined;\n return row ? `L${row.start_line}-${row.end_line}` : null;\n}\n\nexport interface SearchOptions {\n k?: number;\n where?: string; // SQL fragment against frontmatter alias `f`, e.g. \"f.status = 'active'\"\n preset?: string; // named preset; unknown name throws listing declared presets, undefined -> \"default\"\n include?: string[]; // ad hoc scope override (repeatable --include); independent of exclude\n exclude?: string[]; // ad hoc scope override (repeatable --exclude); independent of include\n noExclude?: boolean; // --no-exclude: drop the preset's exclude for this command\n}\n\n// The declared or defaulted signals compose via RRF; `via` names which ones produced each row.\n// `opts` arrives already resolved (config.ts:resolveSearch).\nexport async function search(db: DatabaseSync, cfg: ResolvedConfig, terms: string, opts: SearchOptions = {}): Promise<Row[]> {\n const effective = resolveSearch(cfg, opts);\n const { k, signals } = effective;\n\n const allPaths = (db.prepare('SELECT \"path\" FROM frontmatter').all() as Array<{ path: string }>).map((r) => r.path);\n const scopePaths = rawScope(db, cfg, opts, allPaths);\n const scopeActive = scopePaths.size < allPaths.length;\n // The set every candidate pool must be filtered to before truncation: scope narrowed by\n // --where, the same composition scopedPaths() gives the other commands. Runs the same where\n // fragment matchSql does, so a bad column gets the same attributed error either way.\n let allowedPaths: Set<string>;\n try {\n allowedPaths = narrowByWhere(db, scopePaths, effective.where);\n } catch (err) {\n throw searchError(err as Error, terms, effective.where);\n }\n const fetch = Math.max(k * 3, 30);\n\n // A downloadable HF id proceeds -- getProvider fetches it lazily on consent. Only a\n // local path with missing files errors here, since nothing will ever fetch it for itself.\n const wantsVectors = signals.vectors !== undefined;\n if (wantsVectors) {\n const e = embedConfig(cfg); // validate.ts guarantees this is set whenever \"vectors\" is declared\n if (localModelMissing(e)) {\n throw new SenseError('EMBED_MODEL_MISSING', `preset \"${effective.presetName}\" searches with vectors, but the local model path \"${e.model}\" is missing ${MODEL_FILENAMES}; point embed.model at a directory containing them, or drop \"vectors\" from that preset's signals to search without them`);\n }\n }\n const semanticEnabled = wantsVectors && scopeHasEmbeddings(db, cfg, allowedPaths);\n\n // Terms pass to MATCH as written, with one transform: a run of text whose language marks no\n // word boundaries becomes a quoted grapheme phrase (and a title:/summary:/text: qualifier\n // ahead of it retargets its _seg column), matching how it was indexed. Text that already\n // carries boundaries comes through untouched, so this is a no-op for the languages that\n // never needed it. Gated on the same predicate that decided whether the sidecars were\n // populated (contentTokenize(cfg) === undefined), so index and query can't disagree. Invalid\n // syntax still errors rather than being rewritten.\n // --where\n // applies inside the candidate query (a post-filter would drop matches ranked past the pool)\n // and again on the final select, for link-derived rows.\n const scope = effective.where;\n const whereJoin = scope ? `JOIN frontmatter f ON f.\"path\" = content.path` : '';\n const whereCond = scope ? `AND (${scope})` : '';\n // A scope narrower than the whole index must filter the candidate pool before LIMIT, not\n // after -- otherwise scoped notes ranking below the global top-`fetch` never reach the\n // filter. A join against a temp table, not a bound parameter list: real scopes run to\n // thousands of paths, past SQLITE_MAX_VARIABLE_NUMBER on older builds.\n if (scopeActive) materializeScope(db, '_search_scope', scopePaths);\n const scopeCond = scopeActive ? `AND content.path IN (SELECT \"path\" FROM _search_scope)` : '';\n const query = contentTokenize(cfg) === undefined ? segmentMatch(terms) : terms;\n\n const candidates = new Map<string, { score: number; via: string }>();\n let matchRows: Array<{ path: string; hit: string | null }> = [];\n // A query that rewrites to nothing (e.g. only unspaced-script punctuation) has no\n // searchable words: zero lexical rows, and the other signals still compose.\n if (signals.words !== undefined && query.trim() !== '') {\n try {\n matchRows = wordsCandidates(db, candidates, query, whereJoin, whereCond, scopeCond, fetch, signals.words);\n } catch (err) {\n throw searchError(err as Error, terms, scope);\n }\n }\n const hits = new Map(matchRows.map((r) => [r.path, r.hit]));\n // hit === null here means the bound suppressed snippet(), not \"no match\" -- distinguish\n // from via='link' rows (never in matchRows, so absent from this set) below.\n const oversized = new Set(matchRows.filter((r) => r.hit === null).map((r) => r.path));\n\n if (signals.links !== undefined) linksCandidates(db, candidates, matchRows, allPaths, allowedPaths, fetch, signals.links);\n\n let chunkLines = new Map<string, string>();\n let chunkSimilarity = new Map<string, number>();\n if (semanticEnabled) {\n ({ chunkLines, chunkSimilarity } = await vectorsCandidates(db, cfg, candidates, terms, fetch, allowedPaths, signals.vectors as number));\n }\n\n db.exec('DROP TABLE IF EXISTS _search');\n db.exec('CREATE TEMP TABLE _search (\"path\" TEXT PRIMARY KEY, score REAL, via TEXT, hit TEXT, lines TEXT, similarity REAL)');\n const insert = db.prepare('INSERT INTO _search (\"path\", score, via, hit, lines, similarity) VALUES (?, ?, ?, ?, ?, ?)');\n for (const [path, c] of candidates) insert.run(path, c.score, c.via, hits.get(path) ?? null, chunkLines.get(path) ?? null, chunkSimilarity.get(path) ?? null);\n\n // All three candidate paths (match, link, vector) already filtered to scope+where before\n // reaching _search; this reapplies --where anyway since the join to frontmatter is already\n // needed for the path column.\n const where = scope ? `WHERE (${scope})` : '';\n // lines is always present now: semantic rows carry their chunk's range, oversized-doc\n // lexical rows gain one below, everything else stays null. similarity stays semantic-only.\n const similarityCol = semanticEnabled ? ', _search.similarity' : '';\n const rows = db\n .prepare(\n `SELECT f.\"path\" AS path, content.title, content.summary, _search.hit, _search.via, round(_search.score, 4) AS score, _search.lines${similarityCol}\n FROM _search JOIN frontmatter f ON f.\"path\" = _search.\"path\" JOIN content ON content.path = _search.\"path\"\n ${where} ORDER BY _search.score DESC LIMIT ?`\n )\n .all(k) as Row[];\n\n if (oversized.size > 0) {\n const bareTerms = extractBareTerms(terms);\n for (const row of rows) {\n if (row.hit !== null || !oversized.has(row.path as string)) continue;\n let text: string;\n try {\n text = readFileSync(join(cfg.baseDir, row.path as string), 'utf8');\n } catch {\n continue; // vanished since the match; leave hit/lines null rather than throw\n }\n const { excerpt, offset } = computeExcerpt(text, bareTerms);\n row.hit = excerpt;\n if (row.lines == null) row.lines = featureEnabled(cfg, 'sections') ? lineRangeFor(db, row.path as string, lineNumberAt(text, offset)) : null;\n }\n }\n\n return rows;\n}\n"],"names":["readFileSync","join","contentTokenize","embedConfig","featureEnabled","resolveSearch","localModelMissing","MODEL_FILENAMES","SenseError","searchError","segmentMatch","materializeScope","narrowByWhere","rawScope","scopeHasEmbeddings","linksCandidates","vectorsCandidates","wordsCandidates","EXCERPT_WINDOW","extractBareTerms","query","cleaned","replace","split","map","tok","toLowerCase","trim","filter","length","findOccurrences","haystackLower","terms","occ","term","idx","found","indexOf","push","start","end","sort","a","b","bestWindowStart","windowSize","best","bestDistinct","bestCount","i","winEnd","seen","Set","count","j","add","size","computeExcerpt","text","Math","min","excerpt","slice","offset","out","cursor","o","prefix","suffix","lineNumberAt","line","charCodeAt","lineRangeFor","db","path","row","prepare","get","start_line","end_line","search","cfg","opts","hits","chunkLines","chunkSimilarity","effective","k","signals","allPaths","all","r","scopePaths","scopeActive","allowedPaths","where","err","fetch","max","wantsVectors","vectors","undefined","e","presetName","model","semanticEnabled","scope","whereJoin","whereCond","scopeCond","candidates","Map","matchRows","words","hit","oversized","links","exec","insert","c","run","score","via","similarityCol","rows","bareTerms","has","baseDir","lines"],"mappings":"AAAA,SAASA,YAAY,QAAQ,UAAU;AACvC,SAASC,IAAI,QAAQ,YAAY;AAGjC,SAASC,eAAe,EAAEC,WAAW,EAAEC,cAAc,EAAEC,aAAa,QAAQ,qBAAqB;AACjG,SAASC,iBAAiB,EAAEC,eAAe,QAAQ,oBAAoB;AACvE,SAASC,UAAU,QAAQ,eAAe;AAE1C,SAASC,WAAW,QAAQ,qBAAqB;AACjD,SAASC,YAAY,QAAQ,gBAAgB;AAC7C,SAASC,gBAAgB,EAAEC,aAAa,EAAEC,QAAQ,EAAEC,kBAAkB,QAAQ,aAAa;AAC3F,SAASC,eAAe,EAAEC,iBAAiB,EAAEC,eAAe,QAAQ,eAAe;AAEnF,iFAAiF;AACjF,8EAA8E;AAC9E,MAAMC,iBAAiB;AAEvB,8FAA8F;AAC9F,qEAAqE;AACrE,SAASC,iBAAiBC,KAAa;IACrC,MAAMC,UAAUD,MACbE,OAAO,CAAC,MAAM,KACdA,OAAO,CAAC,UAAU,KAClBA,OAAO,CAAC,mCAAmC;IAC9C,OAAOD,QACJE,KAAK,CAAC,OACNC,GAAG,CAAC,CAACC,MAAQA,IAAIH,OAAO,CAAC,kBAAkB,KAAK,iCAAiC;KACjFE,GAAG,CAAC,CAACC,MAAQA,IAAIC,WAAW,GAAGC,IAAI,IACnCC,MAAM,CAAC,CAACH,MAAQA,IAAII,MAAM,GAAG;AAClC;AAEA,SAASC,gBAAgBC,aAAqB,EAAEC,KAAe;IAC7D,MAAMC,MAA2D,EAAE;IACnE,KAAK,MAAMC,QAAQF,MAAO;QACxB,IAAIG,MAAM;QACV,OAAS;YACP,MAAMC,QAAQL,cAAcM,OAAO,CAACH,MAAMC;YAC1C,IAAIC,UAAU,CAAC,GAAG;YAClBH,IAAIK,IAAI,CAAC;gBAAEC,OAAOH;gBAAOI,KAAKJ,QAAQF,KAAKL,MAAM;gBAAEK;YAAK;YACxDC,MAAMC,QAAQF,KAAKL,MAAM;QAC3B;IACF;IACA,mFAAmF;IACnF,+EAA+E;IAC/E,OAAOI,IAAIQ,IAAI,CAAC,CAACC,GAAGC,IAAMD,EAAEH,KAAK,GAAGI,EAAEJ,KAAK,IAAII,EAAEH,GAAG,GAAGE,EAAEF,GAAG;AAC9D;AAEA,wFAAwF;AACxF,6DAA6D;AAC7D,SAASI,gBAAgBX,GAAwD,EAAEY,UAAkB;IACnG,IAAIC,OAAOb,GAAG,CAAC,EAAE,CAACM,KAAK;IACvB,IAAIQ,eAAe;IACnB,IAAIC,YAAY;IAChB,IAAK,IAAIC,IAAI,GAAGA,IAAIhB,IAAIJ,MAAM,EAAEoB,IAAK;QACnC,MAAMC,SAASjB,GAAG,CAACgB,EAAE,CAACV,KAAK,GAAGM;QAC9B,MAAMM,OAAO,IAAIC;QACjB,IAAIC,QAAQ;QACZ,IAAK,IAAIC,IAAIL,GAAGK,IAAIrB,IAAIJ,MAAM,IAAII,GAAG,CAACqB,EAAE,CAACf,KAAK,GAAGW,QAAQI,IAAK;YAC5DH,KAAKI,GAAG,CAACtB,GAAG,CAACqB,EAAE,CAACpB,IAAI;YACpBmB;QACF;QACA,IAAIF,KAAKK,IAAI,GAAGT,gBAAiBI,KAAKK,IAAI,KAAKT,gBAAgBM,QAAQL,WAAY;YACjFF,OAAOb,GAAG,CAACgB,EAAE,CAACV,KAAK;YACnBQ,eAAeI,KAAKK,IAAI;YACxBR,YAAYK;QACd;IACF;IACA,OAAOP;AACT;AAEA,uFAAuF;AACvF,uFAAuF;AACvF,SAASW,eAAeC,IAAY,EAAE1B,KAAe;IACnD,MAAMC,MAAMH,gBAAgB4B,KAAKhC,WAAW,IAAIM;IAChD,IAAIC,IAAIJ,MAAM,KAAK,GAAG;QACpB,iFAAiF;QACjF,qFAAqF;QACrF,sFAAsF;QACtF,MAAMW,MAAMmB,KAAKC,GAAG,CAACF,KAAK7B,MAAM,EAAEX;QAClC,OAAO;YAAE2C,SAAS,GAAGH,KAAKI,KAAK,CAAC,GAAGtB,KAAKlB,OAAO,CAAC,QAAQ,KAAKK,IAAI,KAAKa,MAAMkB,KAAK7B,MAAM,GAAG,MAAM,IAAI;YAAEkC,QAAQ;QAAE;IAClH;IACA,MAAMxB,QAAQK,gBAAgBX,KAAKf;IACnC,MAAMsB,MAAMmB,KAAKC,GAAG,CAACF,KAAK7B,MAAM,EAAEU,QAAQrB;IAC1C,IAAI8C,MAAM;IACV,IAAIC,SAAS1B;IACb,KAAK,MAAM2B,KAAKjC,IAAK;QACnB,IAAIiC,EAAE3B,KAAK,GAAGA,SAAS2B,EAAE1B,GAAG,GAAGA,KAAK;QACpC,sFAAsF;QACtF,gFAAgF;QAChF,IAAI0B,EAAE3B,KAAK,GAAG0B,QAAQ;QACtBD,OAAO,GAAGN,KAAKI,KAAK,CAACG,QAAQC,EAAE3B,KAAK,EAAE,CAAC,EAAEmB,KAAKI,KAAK,CAACI,EAAE3B,KAAK,EAAE2B,EAAE1B,GAAG,EAAE,CAAC,CAAC;QACtEyB,SAASC,EAAE1B,GAAG;IAChB;IACAwB,OAAON,KAAKI,KAAK,CAACG,QAAQzB;IAC1B,MAAM2B,SAAS5B,QAAQ,IAAI,MAAM;IACjC,MAAM6B,SAAS5B,MAAMkB,KAAK7B,MAAM,GAAG,MAAM;IACzC,OAAO;QAAEgC,SAAS,GAAGM,SAASH,IAAI1C,OAAO,CAAC,QAAQ,OAAO8C,QAAQ;QAAEL,QAAQxB;IAAM;AACnF;AAEA,SAAS8B,aAAaX,IAAY,EAAEK,MAAc;IAChD,IAAIO,OAAO;IACX,IAAK,IAAIrB,IAAI,GAAGA,IAAIc,QAAQd,IAAK,IAAIS,KAAKa,UAAU,CAACtB,OAAO,IAAIqB;IAChE,OAAOA;AACT;AAEA,SAASE,aAAaC,EAAgB,EAAEC,IAAY,EAAEJ,IAAY;IAChE,MAAMK,MAAMF,GAAGG,OAAO,CAAC,qIAAqIC,GAAG,CAACH,MAAMJ,MAAMA;IAC5K,OAAOK,MAAM,CAAC,CAAC,EAAEA,IAAIG,UAAU,CAAC,CAAC,EAAEH,IAAII,QAAQ,EAAE,GAAG;AACtD;AAWA,+FAA+F;AAC/F,6DAA6D;AAC7D,OAAO,eAAeC,OAAOP,EAAgB,EAAEQ,GAAmB,EAAEjD,KAAa,EAAEkD,OAAsB,CAAC,CAAC;QA6EpCC,WAAwBC,iBAA8BC;IA5E3H,MAAMC,YAAYjF,cAAc4E,KAAKC;IACrC,MAAM,EAAEK,CAAC,EAAEC,OAAO,EAAE,GAAGF;IAEvB,MAAMG,WAAW,AAAChB,GAAGG,OAAO,CAAC,kCAAkCc,GAAG,GAA+BlE,GAAG,CAAC,CAACmE,IAAMA,EAAEjB,IAAI;IAClH,MAAMkB,aAAa/E,SAAS4D,IAAIQ,KAAKC,MAAMO;IAC3C,MAAMI,cAAcD,WAAWpC,IAAI,GAAGiC,SAAS5D,MAAM;IACrD,wFAAwF;IACxF,4FAA4F;IAC5F,qFAAqF;IACrF,IAAIiE;IACJ,IAAI;QACFA,eAAelF,cAAc6D,IAAImB,YAAYN,UAAUS,KAAK;IAC9D,EAAE,OAAOC,KAAK;QACZ,MAAMvF,YAAYuF,KAAchE,OAAOsD,UAAUS,KAAK;IACxD;IACA,MAAME,QAAQtC,KAAKuC,GAAG,CAACX,IAAI,GAAG;IAE9B,oFAAoF;IACpF,0FAA0F;IAC1F,MAAMY,eAAeX,QAAQY,OAAO,KAAKC;IACzC,IAAIF,cAAc;QAChB,MAAMG,IAAInG,YAAY8E,MAAM,oEAAoE;QAChG,IAAI3E,kBAAkBgG,IAAI;YACxB,MAAM,IAAI9F,WAAW,uBAAuB,CAAC,QAAQ,EAAE8E,UAAUiB,UAAU,CAAC,mDAAmD,EAAED,EAAEE,KAAK,CAAC,aAAa,EAAEjG,gBAAgB,uHAAuH,CAAC;QAClS;IACF;IACA,MAAMkG,kBAAkBN,gBAAgBrF,mBAAmB2D,IAAIQ,KAAKa;IAEpE,4FAA4F;IAC5F,0FAA0F;IAC1F,yFAAyF;IACzF,wFAAwF;IACxF,sFAAsF;IACtF,6FAA6F;IAC7F,mDAAmD;IACnD,UAAU;IACV,6FAA6F;IAC7F,wDAAwD;IACxD,MAAMY,QAAQpB,UAAUS,KAAK;IAC7B,MAAMY,YAAYD,QAAQ,CAAC,6CAA6C,CAAC,GAAG;IAC5E,MAAME,YAAYF,QAAQ,CAAC,KAAK,EAAEA,MAAM,CAAC,CAAC,GAAG;IAC7C,yFAAyF;IACzF,uFAAuF;IACvF,sFAAsF;IACtF,uEAAuE;IACvE,IAAIb,aAAalF,iBAAiB8D,IAAI,iBAAiBmB;IACvD,MAAMiB,YAAYhB,cAAc,CAAC,sDAAsD,CAAC,GAAG;IAC3F,MAAMzE,QAAQlB,gBAAgB+E,SAASoB,YAAY3F,aAAasB,SAASA;IAEzE,MAAM8E,aAAa,IAAIC;IACvB,IAAIC,YAAyD,EAAE;IAC/D,kFAAkF;IAClF,4EAA4E;IAC5E,IAAIxB,QAAQyB,KAAK,KAAKZ,aAAajF,MAAMO,IAAI,OAAO,IAAI;QACtD,IAAI;YACFqF,YAAY/F,gBAAgBwD,IAAIqC,YAAY1F,OAAOuF,WAAWC,WAAWC,WAAWZ,OAAOT,QAAQyB,KAAK;QAC1G,EAAE,OAAOjB,KAAK;YACZ,MAAMvF,YAAYuF,KAAchE,OAAO0E;QACzC;IACF;IACA,MAAMvB,OAAO,IAAI4B,IAAIC,UAAUxF,GAAG,CAAC,CAACmE,IAAM;YAACA,EAAEjB,IAAI;YAAEiB,EAAEuB,GAAG;SAAC;IACzD,wFAAwF;IACxF,4EAA4E;IAC5E,MAAMC,YAAY,IAAI/D,IAAI4D,UAAUpF,MAAM,CAAC,CAAC+D,IAAMA,EAAEuB,GAAG,KAAK,MAAM1F,GAAG,CAAC,CAACmE,IAAMA,EAAEjB,IAAI;IAEnF,IAAIc,QAAQ4B,KAAK,KAAKf,WAAWtF,gBAAgB0D,IAAIqC,YAAYE,WAAWvB,UAAUK,cAAcG,OAAOT,QAAQ4B,KAAK;IAExH,IAAIhC,aAAa,IAAI2B;IACrB,IAAI1B,kBAAkB,IAAI0B;IAC1B,IAAIN,iBAAiB;QAClB,CAAA,EAAErB,UAAU,EAAEC,eAAe,EAAE,GAAG,MAAMrE,kBAAkByD,IAAIQ,KAAK6B,YAAY9E,OAAOiE,OAAOH,cAAcN,QAAQY,OAAO,CAAU;IACvI;IAEA3B,GAAG4C,IAAI,CAAC;IACR5C,GAAG4C,IAAI,CAAC;IACR,MAAMC,SAAS7C,GAAGG,OAAO,CAAC;IAC1B,KAAK,MAAM,CAACF,MAAM6C,EAAE,IAAIT,WAAYQ,OAAOE,GAAG,CAAC9C,MAAM6C,EAAEE,KAAK,EAAEF,EAAEG,GAAG,GAAEvC,YAAAA,KAAKN,GAAG,CAACH,mBAATS,uBAAAA,YAAkB,OAAMC,kBAAAA,WAAWP,GAAG,CAACH,mBAAfU,6BAAAA,kBAAwB,OAAMC,uBAAAA,gBAAgBR,GAAG,CAACH,mBAApBW,kCAAAA,uBAA6B;IAExJ,yFAAyF;IACzF,2FAA2F;IAC3F,8BAA8B;IAC9B,MAAMU,QAAQW,QAAQ,CAAC,OAAO,EAAEA,MAAM,CAAC,CAAC,GAAG;IAC3C,sFAAsF;IACtF,2FAA2F;IAC3F,MAAMiB,gBAAgBlB,kBAAkB,yBAAyB;IACjE,MAAMmB,OAAOnD,GACVG,OAAO,CACN,CAAC,kIAAkI,EAAE+C,cAAc;;OAElJ,EAAE5B,MAAM,oCAAoC,CAAC,EAE/CL,GAAG,CAACH;IAEP,IAAI4B,UAAU3D,IAAI,GAAG,GAAG;QACtB,MAAMqE,YAAY1G,iBAAiBa;QACnC,KAAK,MAAM2C,OAAOiD,KAAM;YACtB,IAAIjD,IAAIuC,GAAG,KAAK,QAAQ,CAACC,UAAUW,GAAG,CAACnD,IAAID,IAAI,GAAa;YAC5D,IAAIhB;YACJ,IAAI;gBACFA,OAAO1D,aAAaC,KAAKgF,IAAI8C,OAAO,EAAEpD,IAAID,IAAI,GAAa;YAC7D,EAAE,OAAM;gBACN,UAAU,mEAAmE;YAC/E;YACA,MAAM,EAAEb,OAAO,EAAEE,MAAM,EAAE,GAAGN,eAAeC,MAAMmE;YACjDlD,IAAIuC,GAAG,GAAGrD;YACV,IAAIc,IAAIqD,KAAK,IAAI,MAAMrD,IAAIqD,KAAK,GAAG5H,eAAe6E,KAAK,cAAcT,aAAaC,IAAIE,IAAID,IAAI,EAAYL,aAAaX,MAAMK,WAAW;QAC1I;IACF;IAEA,OAAO6D;AACT"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/commands/search.ts"],"sourcesContent":["import { readFileSync } from 'node:fs';\nimport { join } from 'node:path';\nimport type { DatabaseSync } from 'node:sqlite';\nimport type { ResolvedConfig } from '../config/index.ts';\nimport { contentTokenize, embedConfig, featureEnabled, resolveSearch } from '../config/index.ts';\nimport { localModelMissing, MODEL_FILENAMES } from '../embed/store.ts';\nimport { SenseError } from '../errors.ts';\nimport type { Row } from '../output/output.ts';\nimport { searchError } from '../output/search-error.ts';\nimport { segmentMatch } from '../text/segment.ts';\nimport { materializeScope, narrowByWhere, rawScope, scopeHasEmbeddings } from './scope.ts';\nimport { linksCandidates, vectorsCandidates, wordsCandidates } from './signals.ts';\n\n// snippet() re-tokenizes each candidate doc, superlinearly: ~10s for one 1MB doc\n// (benchmark/reports/2026-08-23-hub-release-battery.md). Past this bound, rows get the JS excerpt.\nconst EXCERPT_WINDOW = 160;\n\n// Bare terms from an FTS5 query string: strips operators/quoting so the oversized-doc excerpt\n// scan matches the same words the query matched on, not FTS5 syntax.\nfunction extractBareTerms(query: string): string[] {\n const cleaned = query\n .replace(/\"/g, ' ')\n .replace(/[()*]/g, ' ')\n .replace(/\\b(AND|OR|NOT|NEAR)\\b(\\/\\d+)?/gi, ' ');\n return cleaned\n .split(/\\s+/)\n .map((tok) => tok.replace(/^[A-Za-z_]\\w*:/, '')) // column filter, e.g. title:term\n .map((tok) => tok.toLowerCase().trim())\n .filter((tok) => tok.length > 0);\n}\n\nfunction findOccurrences(haystackLower: string, terms: string[]): Array<{ start: number; end: number; term: string }> {\n const occ: Array<{ start: number; end: number; term: string }> = [];\n for (const term of terms) {\n let idx = 0;\n for (;;) {\n const found = haystackLower.indexOf(term, idx);\n if (found === -1) break;\n occ.push({ start: found, end: found + term.length, term });\n idx = found + term.length;\n }\n }\n // Longest span first at equal start, so \"tests\" beats its substring \"test\" and the\n // whole word gets highlighted; the emit loop then absorbs the shorter overlap.\n return occ.sort((a, b) => a.start - b.start || b.end - a.end);\n}\n\n// Slides a window over the sorted occurrences, same semantics as FTS5's own best-window\n// pick: most distinct terms, ties broken by most total hits.\nfunction bestWindowStart(occ: Array<{ start: number; end: number; term: string }>, windowSize: number): number {\n let best = occ[0].start;\n let bestDistinct = 0;\n let bestCount = 0;\n for (let i = 0; i < occ.length; i++) {\n const winEnd = occ[i].start + windowSize;\n const seen = new Set<string>();\n let count = 0;\n for (let j = i; j < occ.length && occ[j].start < winEnd; j++) {\n seen.add(occ[j].term);\n count++;\n }\n if (seen.size > bestDistinct || (seen.size === bestDistinct && count > bestCount)) {\n best = occ[i].start;\n bestDistinct = seen.size;\n bestCount = count;\n }\n }\n return best;\n}\n\n// snippet()-shaped excerpt: matched terms wrapped in «», … at cut edges. Linear in doc\n// length, only run for rows actually returned (<= k), unlike snippet()'s per-row cost.\nfunction computeExcerpt(text: string, terms: string[]): { excerpt: string; offset: number } {\n const occ = findOccurrences(text.toLowerCase(), terms);\n if (occ.length === 0) {\n // This is a raw substring scan, not porter-stemmed: a doc matched only through a\n // stemmed variant (query \"negotiate\" vs doc \"negotiating\") finds no occurrence here.\n // Fall back to the doc's start, unmarked, rather than claim a match that isn't there.\n const end = Math.min(text.length, EXCERPT_WINDOW);\n return { excerpt: `${text.slice(0, end).replace(/\\s+/g, ' ').trim()}${end < text.length ? '…' : ''}`, offset: 0 };\n }\n const start = bestWindowStart(occ, EXCERPT_WINDOW);\n const end = Math.min(text.length, start + EXCERPT_WINDOW);\n let out = '';\n let cursor = start;\n for (const o of occ) {\n if (o.start < start || o.end > end) continue;\n // Occurrences of duplicate or substring-overlapping terms (\"test tests\") can overlap;\n // emitting each would duplicate document text. Keep the first, absorb the rest.\n if (o.start < cursor) continue;\n out += `${text.slice(cursor, o.start)}«${text.slice(o.start, o.end)}»`;\n cursor = o.end;\n }\n out += text.slice(cursor, end);\n const prefix = start > 0 ? '…' : '';\n const suffix = end < text.length ? '…' : '';\n return { excerpt: `${prefix}${out.replace(/\\s+/g, ' ')}${suffix}`, offset: start };\n}\n\nfunction lineNumberAt(text: string, offset: number): number {\n let line = 1;\n for (let i = 0; i < offset; i++) if (text.charCodeAt(i) === 10) line++;\n return line;\n}\n\nfunction lineRangeFor(db: DatabaseSync, path: string, line: number): string | null {\n const row = db.prepare('SELECT start_line, end_line FROM sections WHERE \"path\" = ? AND start_line <= ? AND end_line >= ? ORDER BY start_line DESC LIMIT 1').get(path, line, line) as { start_line: number; end_line: number } | undefined;\n return row ? `L${row.start_line}-${row.end_line}` : null;\n}\n\nexport interface SearchOptions {\n k?: number;\n where?: string; // SQL fragment against frontmatter alias `f`, e.g. \"f.status = 'active'\"\n preset?: string; // named preset; unknown name throws listing declared presets, undefined -> \"default\"\n include?: string[]; // ad hoc scope override (repeatable --include); independent of exclude\n exclude?: string[]; // ad hoc scope override (repeatable --exclude); independent of include\n noExclude?: boolean; // --no-exclude: drop the preset's exclude for this command\n}\n\n// The declared or defaulted signals compose via RRF; `via` names which ones produced each row.\n// `opts` arrives already resolved (config.ts:resolveSearch).\nexport async function search(db: DatabaseSync, cfg: ResolvedConfig, terms: string, opts: SearchOptions = {}): Promise<Row[]> {\n const effective = resolveSearch(cfg, opts);\n const { k, signals } = effective;\n\n const allPaths = (db.prepare('SELECT \"path\" FROM frontmatter').all() as Array<{ path: string }>).map((r) => r.path);\n const scopePaths = rawScope(db, cfg, opts, allPaths);\n const scopeActive = scopePaths.size < allPaths.length;\n // The set every candidate pool must be filtered to before truncation: scope narrowed by\n // --where, the same composition scopedPaths() gives the other commands. Runs the same where\n // fragment matchSql does, so a bad column gets the same attributed error either way.\n let allowedPaths: Set<string>;\n try {\n allowedPaths = narrowByWhere(db, scopePaths, effective.where);\n } catch (err) {\n throw searchError(err as Error, terms, effective.where);\n }\n const fetch = Math.max(k * 3, 30);\n\n // A downloadable HF id proceeds -- getProvider fetches it lazily on consent. Only a\n // local path with missing files errors here, since nothing will ever fetch it for itself.\n const wantsVectors = signals.vectors !== undefined;\n if (wantsVectors) {\n const e = embedConfig(cfg); // validate.ts guarantees this is set whenever \"vectors\" is declared\n if (localModelMissing(e)) {\n throw new SenseError('EMBED_MODEL_MISSING', `preset \"${effective.presetName}\" searches with vectors, but the local model path \"${e.model}\" is missing ${MODEL_FILENAMES}; point embed.model at a directory containing them, or drop \"vectors\" from that preset's signals to search without them`);\n }\n }\n const semanticEnabled = wantsVectors && scopeHasEmbeddings(db, cfg, allowedPaths);\n\n // Terms pass to MATCH as written, with one transform: a run of text whose language marks no\n // word boundaries becomes a quoted grapheme phrase (and a title:/summary:/text: qualifier\n // ahead of it retargets its _seg column), matching how it was indexed. Text that already\n // carries boundaries comes through untouched, so this is a no-op for the languages that\n // never needed it. Gated on the same predicate that decided whether the sidecars were\n // populated (contentTokenize(cfg) === undefined), so index and query can't disagree. Invalid\n // syntax still errors rather than being rewritten.\n // --where\n // applies inside the candidate query (a post-filter would drop matches ranked past the pool)\n // and again on the final select, for link-derived rows.\n const scope = effective.where;\n const whereJoin = scope ? `JOIN frontmatter f ON f.\"path\" = content.path` : '';\n const whereCond = scope ? `AND (${scope})` : '';\n // A scope narrower than the whole index must filter the candidate pool before LIMIT, not\n // after -- otherwise scoped notes ranking below the global top-`fetch` never reach the\n // filter. A join against a temp table, not a bound parameter list: real scopes run to\n // thousands of paths, past SQLITE_MAX_VARIABLE_NUMBER on older builds.\n if (scopeActive) materializeScope(db, '_search_scope', scopePaths);\n const scopeCond = scopeActive ? `AND content.path IN (SELECT \"path\" FROM _search_scope)` : '';\n const query = contentTokenize(cfg) === undefined ? segmentMatch(terms) : terms;\n\n const candidates = new Map<string, { score: number; via: string }>();\n let matchRows: Array<{ path: string; hit: string | null }> = [];\n // A query that rewrites to nothing (e.g. only unspaced-script punctuation) has no\n // searchable words: zero lexical rows, and the other signals still compose.\n if (signals.words !== undefined && query.trim() !== '') {\n try {\n matchRows = wordsCandidates(db, candidates, query, whereJoin, whereCond, scopeCond, fetch, signals.words);\n } catch (err) {\n throw searchError(err as Error, terms, scope);\n }\n }\n const hits = new Map(matchRows.map((r) => [r.path, r.hit]));\n // hit === null here means the bound suppressed snippet(), not \"no match\" -- distinguish\n // from via='link' rows (never in matchRows, so absent from this set) below.\n const oversized = new Set(matchRows.filter((r) => r.hit === null).map((r) => r.path));\n\n if (signals.links !== undefined) linksCandidates(db, candidates, matchRows, allPaths, allowedPaths, fetch, signals.links);\n\n let chunkLines = new Map<string, string>();\n let chunkSimilarity = new Map<string, number>();\n if (semanticEnabled) {\n ({ chunkLines, chunkSimilarity } = await vectorsCandidates(db, cfg, candidates, terms, fetch, allowedPaths, signals.vectors as number));\n }\n\n db.exec('DROP TABLE IF EXISTS _search');\n db.exec('CREATE TEMP TABLE _search (\"path\" TEXT PRIMARY KEY, score REAL, via TEXT, hit TEXT, lines TEXT, similarity REAL)');\n const insert = db.prepare('INSERT INTO _search (\"path\", score, via, hit, lines, similarity) VALUES (?, ?, ?, ?, ?, ?)');\n for (const [path, c] of candidates) insert.run(path, c.score, c.via, hits.get(path) ?? null, chunkLines.get(path) ?? null, chunkSimilarity.get(path) ?? null);\n\n // All three candidate paths (match, link, vector) already filtered to scope+where before\n // reaching _search; this reapplies --where anyway since the join to frontmatter is already\n // needed for the path column.\n const where = scope ? `WHERE (${scope})` : '';\n // lines is always present now: semantic rows carry their chunk's range, oversized-doc\n // lexical rows gain one below, everything else stays null. similarity stays semantic-only.\n const similarityCol = semanticEnabled ? ', _search.similarity' : '';\n const rows = db\n .prepare(\n `SELECT f.\"path\" AS path, content.title, content.summary, _search.hit, _search.via, round(_search.score, 4) AS score, _search.lines${similarityCol}\n FROM _search JOIN frontmatter f ON f.\"path\" = _search.\"path\" JOIN content ON content.path = _search.\"path\"\n ${where} ORDER BY _search.score DESC LIMIT ?`\n )\n .all(k) as Row[];\n\n if (oversized.size > 0) {\n const bareTerms = extractBareTerms(terms);\n for (const row of rows) {\n if (row.hit !== null || !oversized.has(row.path as string)) continue;\n let text: string;\n try {\n text = readFileSync(join(cfg.baseDir, row.path as string), 'utf8');\n } catch {\n continue; // vanished since the match; leave hit/lines null rather than throw\n }\n const { excerpt, offset } = computeExcerpt(text, bareTerms);\n row.hit = excerpt;\n if (row.lines == null) row.lines = featureEnabled(cfg, 'sections') ? lineRangeFor(db, row.path as string, lineNumberAt(text, offset)) : null;\n }\n }\n\n return rows;\n}\n"],"names":["readFileSync","join","contentTokenize","embedConfig","featureEnabled","resolveSearch","localModelMissing","MODEL_FILENAMES","SenseError","searchError","segmentMatch","materializeScope","narrowByWhere","rawScope","scopeHasEmbeddings","linksCandidates","vectorsCandidates","wordsCandidates","EXCERPT_WINDOW","extractBareTerms","query","cleaned","replace","split","map","tok","toLowerCase","trim","filter","length","findOccurrences","haystackLower","terms","occ","term","idx","found","indexOf","push","start","end","sort","a","b","bestWindowStart","windowSize","best","bestDistinct","bestCount","i","winEnd","seen","Set","count","j","add","size","computeExcerpt","text","Math","min","excerpt","slice","offset","out","cursor","o","prefix","suffix","lineNumberAt","line","charCodeAt","lineRangeFor","db","path","row","prepare","get","start_line","end_line","search","cfg","opts","hits","chunkLines","chunkSimilarity","effective","k","signals","allPaths","all","r","scopePaths","scopeActive","allowedPaths","where","err","fetch","max","wantsVectors","vectors","undefined","e","presetName","model","semanticEnabled","scope","whereJoin","whereCond","scopeCond","candidates","Map","matchRows","words","hit","oversized","links","exec","insert","c","run","score","via","similarityCol","rows","bareTerms","has","baseDir","lines"],"mappings":"AAAA,SAASA,YAAY,QAAQ,UAAU;AACvC,SAASC,IAAI,QAAQ,YAAY;AAGjC,SAASC,eAAe,EAAEC,WAAW,EAAEC,cAAc,EAAEC,aAAa,QAAQ,qBAAqB;AACjG,SAASC,iBAAiB,EAAEC,eAAe,QAAQ,oBAAoB;AACvE,SAASC,UAAU,QAAQ,eAAe;AAE1C,SAASC,WAAW,QAAQ,4BAA4B;AACxD,SAASC,YAAY,QAAQ,qBAAqB;AAClD,SAASC,gBAAgB,EAAEC,aAAa,EAAEC,QAAQ,EAAEC,kBAAkB,QAAQ,aAAa;AAC3F,SAASC,eAAe,EAAEC,iBAAiB,EAAEC,eAAe,QAAQ,eAAe;AAEnF,iFAAiF;AACjF,mGAAmG;AACnG,MAAMC,iBAAiB;AAEvB,8FAA8F;AAC9F,qEAAqE;AACrE,SAASC,iBAAiBC,KAAa;IACrC,MAAMC,UAAUD,MACbE,OAAO,CAAC,MAAM,KACdA,OAAO,CAAC,UAAU,KAClBA,OAAO,CAAC,mCAAmC;IAC9C,OAAOD,QACJE,KAAK,CAAC,OACNC,GAAG,CAAC,CAACC,MAAQA,IAAIH,OAAO,CAAC,kBAAkB,KAAK,iCAAiC;KACjFE,GAAG,CAAC,CAACC,MAAQA,IAAIC,WAAW,GAAGC,IAAI,IACnCC,MAAM,CAAC,CAACH,MAAQA,IAAII,MAAM,GAAG;AAClC;AAEA,SAASC,gBAAgBC,aAAqB,EAAEC,KAAe;IAC7D,MAAMC,MAA2D,EAAE;IACnE,KAAK,MAAMC,QAAQF,MAAO;QACxB,IAAIG,MAAM;QACV,OAAS;YACP,MAAMC,QAAQL,cAAcM,OAAO,CAACH,MAAMC;YAC1C,IAAIC,UAAU,CAAC,GAAG;YAClBH,IAAIK,IAAI,CAAC;gBAAEC,OAAOH;gBAAOI,KAAKJ,QAAQF,KAAKL,MAAM;gBAAEK;YAAK;YACxDC,MAAMC,QAAQF,KAAKL,MAAM;QAC3B;IACF;IACA,mFAAmF;IACnF,+EAA+E;IAC/E,OAAOI,IAAIQ,IAAI,CAAC,CAACC,GAAGC,IAAMD,EAAEH,KAAK,GAAGI,EAAEJ,KAAK,IAAII,EAAEH,GAAG,GAAGE,EAAEF,GAAG;AAC9D;AAEA,wFAAwF;AACxF,6DAA6D;AAC7D,SAASI,gBAAgBX,GAAwD,EAAEY,UAAkB;IACnG,IAAIC,OAAOb,GAAG,CAAC,EAAE,CAACM,KAAK;IACvB,IAAIQ,eAAe;IACnB,IAAIC,YAAY;IAChB,IAAK,IAAIC,IAAI,GAAGA,IAAIhB,IAAIJ,MAAM,EAAEoB,IAAK;QACnC,MAAMC,SAASjB,GAAG,CAACgB,EAAE,CAACV,KAAK,GAAGM;QAC9B,MAAMM,OAAO,IAAIC;QACjB,IAAIC,QAAQ;QACZ,IAAK,IAAIC,IAAIL,GAAGK,IAAIrB,IAAIJ,MAAM,IAAII,GAAG,CAACqB,EAAE,CAACf,KAAK,GAAGW,QAAQI,IAAK;YAC5DH,KAAKI,GAAG,CAACtB,GAAG,CAACqB,EAAE,CAACpB,IAAI;YACpBmB;QACF;QACA,IAAIF,KAAKK,IAAI,GAAGT,gBAAiBI,KAAKK,IAAI,KAAKT,gBAAgBM,QAAQL,WAAY;YACjFF,OAAOb,GAAG,CAACgB,EAAE,CAACV,KAAK;YACnBQ,eAAeI,KAAKK,IAAI;YACxBR,YAAYK;QACd;IACF;IACA,OAAOP;AACT;AAEA,uFAAuF;AACvF,uFAAuF;AACvF,SAASW,eAAeC,IAAY,EAAE1B,KAAe;IACnD,MAAMC,MAAMH,gBAAgB4B,KAAKhC,WAAW,IAAIM;IAChD,IAAIC,IAAIJ,MAAM,KAAK,GAAG;QACpB,iFAAiF;QACjF,qFAAqF;QACrF,sFAAsF;QACtF,MAAMW,MAAMmB,KAAKC,GAAG,CAACF,KAAK7B,MAAM,EAAEX;QAClC,OAAO;YAAE2C,SAAS,GAAGH,KAAKI,KAAK,CAAC,GAAGtB,KAAKlB,OAAO,CAAC,QAAQ,KAAKK,IAAI,KAAKa,MAAMkB,KAAK7B,MAAM,GAAG,MAAM,IAAI;YAAEkC,QAAQ;QAAE;IAClH;IACA,MAAMxB,QAAQK,gBAAgBX,KAAKf;IACnC,MAAMsB,MAAMmB,KAAKC,GAAG,CAACF,KAAK7B,MAAM,EAAEU,QAAQrB;IAC1C,IAAI8C,MAAM;IACV,IAAIC,SAAS1B;IACb,KAAK,MAAM2B,KAAKjC,IAAK;QACnB,IAAIiC,EAAE3B,KAAK,GAAGA,SAAS2B,EAAE1B,GAAG,GAAGA,KAAK;QACpC,sFAAsF;QACtF,gFAAgF;QAChF,IAAI0B,EAAE3B,KAAK,GAAG0B,QAAQ;QACtBD,OAAO,GAAGN,KAAKI,KAAK,CAACG,QAAQC,EAAE3B,KAAK,EAAE,CAAC,EAAEmB,KAAKI,KAAK,CAACI,EAAE3B,KAAK,EAAE2B,EAAE1B,GAAG,EAAE,CAAC,CAAC;QACtEyB,SAASC,EAAE1B,GAAG;IAChB;IACAwB,OAAON,KAAKI,KAAK,CAACG,QAAQzB;IAC1B,MAAM2B,SAAS5B,QAAQ,IAAI,MAAM;IACjC,MAAM6B,SAAS5B,MAAMkB,KAAK7B,MAAM,GAAG,MAAM;IACzC,OAAO;QAAEgC,SAAS,GAAGM,SAASH,IAAI1C,OAAO,CAAC,QAAQ,OAAO8C,QAAQ;QAAEL,QAAQxB;IAAM;AACnF;AAEA,SAAS8B,aAAaX,IAAY,EAAEK,MAAc;IAChD,IAAIO,OAAO;IACX,IAAK,IAAIrB,IAAI,GAAGA,IAAIc,QAAQd,IAAK,IAAIS,KAAKa,UAAU,CAACtB,OAAO,IAAIqB;IAChE,OAAOA;AACT;AAEA,SAASE,aAAaC,EAAgB,EAAEC,IAAY,EAAEJ,IAAY;IAChE,MAAMK,MAAMF,GAAGG,OAAO,CAAC,qIAAqIC,GAAG,CAACH,MAAMJ,MAAMA;IAC5K,OAAOK,MAAM,CAAC,CAAC,EAAEA,IAAIG,UAAU,CAAC,CAAC,EAAEH,IAAII,QAAQ,EAAE,GAAG;AACtD;AAWA,+FAA+F;AAC/F,6DAA6D;AAC7D,OAAO,eAAeC,OAAOP,EAAgB,EAAEQ,GAAmB,EAAEjD,KAAa,EAAEkD,OAAsB,CAAC,CAAC;QA6EpCC,WAAwBC,iBAA8BC;IA5E3H,MAAMC,YAAYjF,cAAc4E,KAAKC;IACrC,MAAM,EAAEK,CAAC,EAAEC,OAAO,EAAE,GAAGF;IAEvB,MAAMG,WAAW,AAAChB,GAAGG,OAAO,CAAC,kCAAkCc,GAAG,GAA+BlE,GAAG,CAAC,CAACmE,IAAMA,EAAEjB,IAAI;IAClH,MAAMkB,aAAa/E,SAAS4D,IAAIQ,KAAKC,MAAMO;IAC3C,MAAMI,cAAcD,WAAWpC,IAAI,GAAGiC,SAAS5D,MAAM;IACrD,wFAAwF;IACxF,4FAA4F;IAC5F,qFAAqF;IACrF,IAAIiE;IACJ,IAAI;QACFA,eAAelF,cAAc6D,IAAImB,YAAYN,UAAUS,KAAK;IAC9D,EAAE,OAAOC,KAAK;QACZ,MAAMvF,YAAYuF,KAAchE,OAAOsD,UAAUS,KAAK;IACxD;IACA,MAAME,QAAQtC,KAAKuC,GAAG,CAACX,IAAI,GAAG;IAE9B,oFAAoF;IACpF,0FAA0F;IAC1F,MAAMY,eAAeX,QAAQY,OAAO,KAAKC;IACzC,IAAIF,cAAc;QAChB,MAAMG,IAAInG,YAAY8E,MAAM,oEAAoE;QAChG,IAAI3E,kBAAkBgG,IAAI;YACxB,MAAM,IAAI9F,WAAW,uBAAuB,CAAC,QAAQ,EAAE8E,UAAUiB,UAAU,CAAC,mDAAmD,EAAED,EAAEE,KAAK,CAAC,aAAa,EAAEjG,gBAAgB,uHAAuH,CAAC;QAClS;IACF;IACA,MAAMkG,kBAAkBN,gBAAgBrF,mBAAmB2D,IAAIQ,KAAKa;IAEpE,4FAA4F;IAC5F,0FAA0F;IAC1F,yFAAyF;IACzF,wFAAwF;IACxF,sFAAsF;IACtF,6FAA6F;IAC7F,mDAAmD;IACnD,UAAU;IACV,6FAA6F;IAC7F,wDAAwD;IACxD,MAAMY,QAAQpB,UAAUS,KAAK;IAC7B,MAAMY,YAAYD,QAAQ,CAAC,6CAA6C,CAAC,GAAG;IAC5E,MAAME,YAAYF,QAAQ,CAAC,KAAK,EAAEA,MAAM,CAAC,CAAC,GAAG;IAC7C,yFAAyF;IACzF,uFAAuF;IACvF,sFAAsF;IACtF,uEAAuE;IACvE,IAAIb,aAAalF,iBAAiB8D,IAAI,iBAAiBmB;IACvD,MAAMiB,YAAYhB,cAAc,CAAC,sDAAsD,CAAC,GAAG;IAC3F,MAAMzE,QAAQlB,gBAAgB+E,SAASoB,YAAY3F,aAAasB,SAASA;IAEzE,MAAM8E,aAAa,IAAIC;IACvB,IAAIC,YAAyD,EAAE;IAC/D,kFAAkF;IAClF,4EAA4E;IAC5E,IAAIxB,QAAQyB,KAAK,KAAKZ,aAAajF,MAAMO,IAAI,OAAO,IAAI;QACtD,IAAI;YACFqF,YAAY/F,gBAAgBwD,IAAIqC,YAAY1F,OAAOuF,WAAWC,WAAWC,WAAWZ,OAAOT,QAAQyB,KAAK;QAC1G,EAAE,OAAOjB,KAAK;YACZ,MAAMvF,YAAYuF,KAAchE,OAAO0E;QACzC;IACF;IACA,MAAMvB,OAAO,IAAI4B,IAAIC,UAAUxF,GAAG,CAAC,CAACmE,IAAM;YAACA,EAAEjB,IAAI;YAAEiB,EAAEuB,GAAG;SAAC;IACzD,wFAAwF;IACxF,4EAA4E;IAC5E,MAAMC,YAAY,IAAI/D,IAAI4D,UAAUpF,MAAM,CAAC,CAAC+D,IAAMA,EAAEuB,GAAG,KAAK,MAAM1F,GAAG,CAAC,CAACmE,IAAMA,EAAEjB,IAAI;IAEnF,IAAIc,QAAQ4B,KAAK,KAAKf,WAAWtF,gBAAgB0D,IAAIqC,YAAYE,WAAWvB,UAAUK,cAAcG,OAAOT,QAAQ4B,KAAK;IAExH,IAAIhC,aAAa,IAAI2B;IACrB,IAAI1B,kBAAkB,IAAI0B;IAC1B,IAAIN,iBAAiB;QAClB,CAAA,EAAErB,UAAU,EAAEC,eAAe,EAAE,GAAG,MAAMrE,kBAAkByD,IAAIQ,KAAK6B,YAAY9E,OAAOiE,OAAOH,cAAcN,QAAQY,OAAO,CAAU;IACvI;IAEA3B,GAAG4C,IAAI,CAAC;IACR5C,GAAG4C,IAAI,CAAC;IACR,MAAMC,SAAS7C,GAAGG,OAAO,CAAC;IAC1B,KAAK,MAAM,CAACF,MAAM6C,EAAE,IAAIT,WAAYQ,OAAOE,GAAG,CAAC9C,MAAM6C,EAAEE,KAAK,EAAEF,EAAEG,GAAG,GAAEvC,YAAAA,KAAKN,GAAG,CAACH,mBAATS,uBAAAA,YAAkB,OAAMC,kBAAAA,WAAWP,GAAG,CAACH,mBAAfU,6BAAAA,kBAAwB,OAAMC,uBAAAA,gBAAgBR,GAAG,CAACH,mBAApBW,kCAAAA,uBAA6B;IAExJ,yFAAyF;IACzF,2FAA2F;IAC3F,8BAA8B;IAC9B,MAAMU,QAAQW,QAAQ,CAAC,OAAO,EAAEA,MAAM,CAAC,CAAC,GAAG;IAC3C,sFAAsF;IACtF,2FAA2F;IAC3F,MAAMiB,gBAAgBlB,kBAAkB,yBAAyB;IACjE,MAAMmB,OAAOnD,GACVG,OAAO,CACN,CAAC,kIAAkI,EAAE+C,cAAc;;OAElJ,EAAE5B,MAAM,oCAAoC,CAAC,EAE/CL,GAAG,CAACH;IAEP,IAAI4B,UAAU3D,IAAI,GAAG,GAAG;QACtB,MAAMqE,YAAY1G,iBAAiBa;QACnC,KAAK,MAAM2C,OAAOiD,KAAM;YACtB,IAAIjD,IAAIuC,GAAG,KAAK,QAAQ,CAACC,UAAUW,GAAG,CAACnD,IAAID,IAAI,GAAa;YAC5D,IAAIhB;YACJ,IAAI;gBACFA,OAAO1D,aAAaC,KAAKgF,IAAI8C,OAAO,EAAEpD,IAAID,IAAI,GAAa;YAC7D,EAAE,OAAM;gBACN,UAAU,mEAAmE;YAC/E;YACA,MAAM,EAAEb,OAAO,EAAEE,MAAM,EAAE,GAAGN,eAAeC,MAAMmE;YACjDlD,IAAIuC,GAAG,GAAGrD;YACV,IAAIc,IAAIqD,KAAK,IAAI,MAAMrD,IAAIqD,KAAK,GAAG5H,eAAe6E,KAAK,cAAcT,aAAaC,IAAIE,IAAID,IAAI,EAAYL,aAAaX,MAAMK,WAAW;QAC1I;IACF;IAEA,OAAO6D;AACT"}
@@ -1,6 +1,6 @@
1
1
  import { semanticCandidates } from '../embed/query.js';
2
2
  import { linkEdges } from '../features/index.js';
3
- import { personalizedRank } from '../graph.js';
3
+ import { personalizedRank } from '../graph/graph.js';
4
4
  // RRF composes whatever the preset declares; the constant lives here since it belongs to
5
5
  // the fusion these three functions share, not to search() itself.
6
6
  const RRF_K = 60;
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/commands/signals.ts"],"sourcesContent":["import type { DatabaseSync } from 'node:sqlite';\nimport type { ResolvedConfig } from '../config/index.ts';\nimport { semanticCandidates } from '../embed/query.ts';\nimport { linkEdges } from '../features/index.ts';\nimport { personalizedRank } from '../graph.ts';\n\n// RRF composes whatever the preset declares; the constant lives here since it belongs to\n// the fusion these three functions share, not to search() itself.\nconst RRF_K = 60;\n\nexport type Candidates = Map<string, { score: number; via: string }>;\n\n// BM25 match rows, folded into `candidates` by reciprocal rank scaled by the preset's declared\n// weight for this signal (default 1, bit-identical to the unweighted score). Every other signal\n// is seeded from its output, so search() runs this one first when \"words\" is declared.\nexport function wordsCandidates(db: DatabaseSync, candidates: Candidates, query: string, whereJoin: string, whereCond: string, scopeCond: string, fetch: number, weight: number): Array<{ path: string; hit: string | null }> {\n const WEIGHTED_BM25 = 'bm25(content, 10.0, 5.0, 1.0, 0, 10.0, 5.0, 1.0)';\n const SNIPPET_BOUND = 16_384;\n const matchSql = `SELECT content.path AS path, CASE WHEN length(content.text) <= ${SNIPPET_BOUND} THEN snippet(content, 2, '«', '»', '…', 10) ELSE NULL END AS hit FROM content ${whereJoin} WHERE content MATCH ? ${whereCond} ${scopeCond} ORDER BY ${WEIGHTED_BM25} LIMIT ${fetch}`;\n const matchRows = db.prepare(matchSql).all(query) as Array<{ path: string; hit: string | null }>;\n matchRows.forEach((r, i) => {\n candidates.set(r.path, { score: weight / (RRF_K + i), via: 'match' });\n });\n return matchRows;\n}\n\n// Personalized-PageRank expansion from the word-match seeds, folded into `candidates` at this\n// signal's weight. A no-op when there is nothing to seed from (matchRows empty), same as before\n// this was a signal.\nexport function linksCandidates(db: DatabaseSync, candidates: Candidates, matchRows: Array<{ path: string; hit: string | null }>, allPaths: string[], allowedPaths: Set<string>, fetch: number, weight: number): void {\n if (matchRows.length === 0) return;\n const edges = linkEdges(db);\n if (edges.length === 0) return;\n // Gates the label only, not the score: restart mass ranks every seed without an incident\n // edge, but dropping it from the score reweights fusion (FEVER hit@10 0.997 -> 0.907).\n const linked = new Set(edges.flat());\n const seeds = new Map(matchRows.map((r, i) => [r.path, 1 / (i + 1)]));\n const ranked = [...personalizedRank(allPaths, edges, seeds)]\n .filter(([path, score]) => score > 1e-9 && allowedPaths.has(path))\n .sort((a, b) => b[1] - a[1])\n .slice(0, fetch);\n ranked.forEach(([path], i) => {\n const existing = candidates.get(path);\n if (existing) {\n existing.score += weight / (RRF_K + i);\n if (linked.has(path)) existing.via = 'match+link';\n } else {\n candidates.set(path, { score: weight / (RRF_K + i), via: 'link' });\n }\n });\n}\n\n// Vector expansion at the swept flat-region constants (pool = fetch), folded into `candidates`\n// at this signal's weight. Each row also carries its best chunk's line range and similarity,\n// for the caller.\nexport async function vectorsCandidates(db: DatabaseSync, cfg: ResolvedConfig, candidates: Candidates, terms: string, fetch: number, allowedPaths: Set<string>, weight: number): Promise<{ chunkLines: Map<string, string>; chunkSimilarity: Map<string, number> }> {\n const chunkLines = new Map<string, string>();\n const chunkSimilarity = new Map<string, number>();\n const vec = await semanticCandidates(db, cfg, terms, fetch, allowedPaths);\n vec.forEach(({ path, lines, similarity }, i) => {\n chunkLines.set(path, lines);\n chunkSimilarity.set(path, similarity);\n const existing = candidates.get(path);\n if (existing) {\n existing.score += weight / (RRF_K + i);\n existing.via = `${existing.via}+vector`;\n } else {\n candidates.set(path, { score: weight / (RRF_K + i), via: 'vector' });\n }\n });\n return { chunkLines, chunkSimilarity };\n}\n"],"names":["semanticCandidates","linkEdges","personalizedRank","RRF_K","wordsCandidates","db","candidates","query","whereJoin","whereCond","scopeCond","fetch","weight","WEIGHTED_BM25","SNIPPET_BOUND","matchSql","matchRows","prepare","all","forEach","r","i","set","path","score","via","linksCandidates","allPaths","allowedPaths","length","edges","linked","Set","flat","seeds","Map","map","ranked","filter","has","sort","a","b","slice","existing","get","vectorsCandidates","cfg","terms","chunkLines","chunkSimilarity","vec","lines","similarity"],"mappings":"AAEA,SAASA,kBAAkB,QAAQ,oBAAoB;AACvD,SAASC,SAAS,QAAQ,uBAAuB;AACjD,SAASC,gBAAgB,QAAQ,cAAc;AAE/C,yFAAyF;AACzF,kEAAkE;AAClE,MAAMC,QAAQ;AAId,+FAA+F;AAC/F,gGAAgG;AAChG,uFAAuF;AACvF,OAAO,SAASC,gBAAgBC,EAAgB,EAAEC,UAAsB,EAAEC,KAAa,EAAEC,SAAiB,EAAEC,SAAiB,EAAEC,SAAiB,EAAEC,KAAa,EAAEC,MAAc;IAC7K,MAAMC,gBAAgB;IACtB,MAAMC,gBAAgB;IACtB,MAAMC,WAAW,CAAC,+DAA+D,EAAED,cAAc,+EAA+E,EAAEN,UAAU,uBAAuB,EAAEC,UAAU,CAAC,EAAEC,UAAU,UAAU,EAAEG,cAAc,OAAO,EAAEF,OAAO;IACtR,MAAMK,YAAYX,GAAGY,OAAO,CAACF,UAAUG,GAAG,CAACX;IAC3CS,UAAUG,OAAO,CAAC,CAACC,GAAGC;QACpBf,WAAWgB,GAAG,CAACF,EAAEG,IAAI,EAAE;YAAEC,OAAOZ,SAAUT,CAAAA,QAAQkB,CAAAA;YAAII,KAAK;QAAQ;IACrE;IACA,OAAOT;AACT;AAEA,8FAA8F;AAC9F,gGAAgG;AAChG,qBAAqB;AACrB,OAAO,SAASU,gBAAgBrB,EAAgB,EAAEC,UAAsB,EAAEU,SAAsD,EAAEW,QAAkB,EAAEC,YAAyB,EAAEjB,KAAa,EAAEC,MAAc;IAC5M,IAAII,UAAUa,MAAM,KAAK,GAAG;IAC5B,MAAMC,QAAQ7B,UAAUI;IACxB,IAAIyB,MAAMD,MAAM,KAAK,GAAG;IACxB,yFAAyF;IACzF,uFAAuF;IACvF,MAAME,SAAS,IAAIC,IAAIF,MAAMG,IAAI;IACjC,MAAMC,QAAQ,IAAIC,IAAInB,UAAUoB,GAAG,CAAC,CAAChB,GAAGC,IAAM;YAACD,EAAEG,IAAI;YAAE,IAAKF,CAAAA,IAAI,CAAA;SAAG;IACnE,MAAMgB,SAAS;WAAInC,iBAAiByB,UAAUG,OAAOI;KAAO,CACzDI,MAAM,CAAC,CAAC,CAACf,MAAMC,MAAM,GAAKA,QAAQ,QAAQI,aAAaW,GAAG,CAAChB,OAC3DiB,IAAI,CAAC,CAACC,GAAGC,IAAMA,CAAC,CAAC,EAAE,GAAGD,CAAC,CAAC,EAAE,EAC1BE,KAAK,CAAC,GAAGhC;IACZ0B,OAAOlB,OAAO,CAAC,CAAC,CAACI,KAAK,EAAEF;QACtB,MAAMuB,WAAWtC,WAAWuC,GAAG,CAACtB;QAChC,IAAIqB,UAAU;YACZA,SAASpB,KAAK,IAAIZ,SAAUT,CAAAA,QAAQkB,CAAAA;YACpC,IAAIU,OAAOQ,GAAG,CAAChB,OAAOqB,SAASnB,GAAG,GAAG;QACvC,OAAO;YACLnB,WAAWgB,GAAG,CAACC,MAAM;gBAAEC,OAAOZ,SAAUT,CAAAA,QAAQkB,CAAAA;gBAAII,KAAK;YAAO;QAClE;IACF;AACF;AAEA,+FAA+F;AAC/F,6FAA6F;AAC7F,kBAAkB;AAClB,OAAO,eAAeqB,kBAAkBzC,EAAgB,EAAE0C,GAAmB,EAAEzC,UAAsB,EAAE0C,KAAa,EAAErC,KAAa,EAAEiB,YAAyB,EAAEhB,MAAc;IAC5K,MAAMqC,aAAa,IAAId;IACvB,MAAMe,kBAAkB,IAAIf;IAC5B,MAAMgB,MAAM,MAAMnD,mBAAmBK,IAAI0C,KAAKC,OAAOrC,OAAOiB;IAC5DuB,IAAIhC,OAAO,CAAC,CAAC,EAAEI,IAAI,EAAE6B,KAAK,EAAEC,UAAU,EAAE,EAAEhC;QACxC4B,WAAW3B,GAAG,CAACC,MAAM6B;QACrBF,gBAAgB5B,GAAG,CAACC,MAAM8B;QAC1B,MAAMT,WAAWtC,WAAWuC,GAAG,CAACtB;QAChC,IAAIqB,UAAU;YACZA,SAASpB,KAAK,IAAIZ,SAAUT,CAAAA,QAAQkB,CAAAA;YACpCuB,SAASnB,GAAG,GAAG,GAAGmB,SAASnB,GAAG,CAAC,OAAO,CAAC;QACzC,OAAO;YACLnB,WAAWgB,GAAG,CAACC,MAAM;gBAAEC,OAAOZ,SAAUT,CAAAA,QAAQkB,CAAAA;gBAAII,KAAK;YAAS;QACpE;IACF;IACA,OAAO;QAAEwB;QAAYC;IAAgB;AACvC"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/commands/signals.ts"],"sourcesContent":["import type { DatabaseSync } from 'node:sqlite';\nimport type { ResolvedConfig } from '../config/index.ts';\nimport { semanticCandidates } from '../embed/query.ts';\nimport { linkEdges } from '../features/index.ts';\nimport { personalizedRank } from '../graph/graph.ts';\n\n// RRF composes whatever the preset declares; the constant lives here since it belongs to\n// the fusion these three functions share, not to search() itself.\nconst RRF_K = 60;\n\nexport type Candidates = Map<string, { score: number; via: string }>;\n\n// BM25 match rows, folded into `candidates` by reciprocal rank scaled by the preset's declared\n// weight for this signal (default 1, bit-identical to the unweighted score). Every other signal\n// is seeded from its output, so search() runs this one first when \"words\" is declared.\nexport function wordsCandidates(db: DatabaseSync, candidates: Candidates, query: string, whereJoin: string, whereCond: string, scopeCond: string, fetch: number, weight: number): Array<{ path: string; hit: string | null }> {\n const WEIGHTED_BM25 = 'bm25(content, 10.0, 5.0, 1.0, 0, 10.0, 5.0, 1.0)';\n const SNIPPET_BOUND = 16_384;\n const matchSql = `SELECT content.path AS path, CASE WHEN length(content.text) <= ${SNIPPET_BOUND} THEN snippet(content, 2, '«', '»', '…', 10) ELSE NULL END AS hit FROM content ${whereJoin} WHERE content MATCH ? ${whereCond} ${scopeCond} ORDER BY ${WEIGHTED_BM25} LIMIT ${fetch}`;\n const matchRows = db.prepare(matchSql).all(query) as Array<{ path: string; hit: string | null }>;\n matchRows.forEach((r, i) => {\n candidates.set(r.path, { score: weight / (RRF_K + i), via: 'match' });\n });\n return matchRows;\n}\n\n// Personalized-PageRank expansion from the word-match seeds, folded into `candidates` at this\n// signal's weight. A no-op when there is nothing to seed from (matchRows empty), same as before\n// this was a signal.\nexport function linksCandidates(db: DatabaseSync, candidates: Candidates, matchRows: Array<{ path: string; hit: string | null }>, allPaths: string[], allowedPaths: Set<string>, fetch: number, weight: number): void {\n if (matchRows.length === 0) return;\n const edges = linkEdges(db);\n if (edges.length === 0) return;\n // Gates the label only, not the score: restart mass ranks every seed without an incident\n // edge, but dropping it from the score reweights fusion (FEVER hit@10 0.997 -> 0.907).\n const linked = new Set(edges.flat());\n const seeds = new Map(matchRows.map((r, i) => [r.path, 1 / (i + 1)]));\n const ranked = [...personalizedRank(allPaths, edges, seeds)]\n .filter(([path, score]) => score > 1e-9 && allowedPaths.has(path))\n .sort((a, b) => b[1] - a[1])\n .slice(0, fetch);\n ranked.forEach(([path], i) => {\n const existing = candidates.get(path);\n if (existing) {\n existing.score += weight / (RRF_K + i);\n if (linked.has(path)) existing.via = 'match+link';\n } else {\n candidates.set(path, { score: weight / (RRF_K + i), via: 'link' });\n }\n });\n}\n\n// Vector expansion at the swept flat-region constants (pool = fetch), folded into `candidates`\n// at this signal's weight. Each row also carries its best chunk's line range and similarity,\n// for the caller.\nexport async function vectorsCandidates(db: DatabaseSync, cfg: ResolvedConfig, candidates: Candidates, terms: string, fetch: number, allowedPaths: Set<string>, weight: number): Promise<{ chunkLines: Map<string, string>; chunkSimilarity: Map<string, number> }> {\n const chunkLines = new Map<string, string>();\n const chunkSimilarity = new Map<string, number>();\n const vec = await semanticCandidates(db, cfg, terms, fetch, allowedPaths);\n vec.forEach(({ path, lines, similarity }, i) => {\n chunkLines.set(path, lines);\n chunkSimilarity.set(path, similarity);\n const existing = candidates.get(path);\n if (existing) {\n existing.score += weight / (RRF_K + i);\n existing.via = `${existing.via}+vector`;\n } else {\n candidates.set(path, { score: weight / (RRF_K + i), via: 'vector' });\n }\n });\n return { chunkLines, chunkSimilarity };\n}\n"],"names":["semanticCandidates","linkEdges","personalizedRank","RRF_K","wordsCandidates","db","candidates","query","whereJoin","whereCond","scopeCond","fetch","weight","WEIGHTED_BM25","SNIPPET_BOUND","matchSql","matchRows","prepare","all","forEach","r","i","set","path","score","via","linksCandidates","allPaths","allowedPaths","length","edges","linked","Set","flat","seeds","Map","map","ranked","filter","has","sort","a","b","slice","existing","get","vectorsCandidates","cfg","terms","chunkLines","chunkSimilarity","vec","lines","similarity"],"mappings":"AAEA,SAASA,kBAAkB,QAAQ,oBAAoB;AACvD,SAASC,SAAS,QAAQ,uBAAuB;AACjD,SAASC,gBAAgB,QAAQ,oBAAoB;AAErD,yFAAyF;AACzF,kEAAkE;AAClE,MAAMC,QAAQ;AAId,+FAA+F;AAC/F,gGAAgG;AAChG,uFAAuF;AACvF,OAAO,SAASC,gBAAgBC,EAAgB,EAAEC,UAAsB,EAAEC,KAAa,EAAEC,SAAiB,EAAEC,SAAiB,EAAEC,SAAiB,EAAEC,KAAa,EAAEC,MAAc;IAC7K,MAAMC,gBAAgB;IACtB,MAAMC,gBAAgB;IACtB,MAAMC,WAAW,CAAC,+DAA+D,EAAED,cAAc,+EAA+E,EAAEN,UAAU,uBAAuB,EAAEC,UAAU,CAAC,EAAEC,UAAU,UAAU,EAAEG,cAAc,OAAO,EAAEF,OAAO;IACtR,MAAMK,YAAYX,GAAGY,OAAO,CAACF,UAAUG,GAAG,CAACX;IAC3CS,UAAUG,OAAO,CAAC,CAACC,GAAGC;QACpBf,WAAWgB,GAAG,CAACF,EAAEG,IAAI,EAAE;YAAEC,OAAOZ,SAAUT,CAAAA,QAAQkB,CAAAA;YAAII,KAAK;QAAQ;IACrE;IACA,OAAOT;AACT;AAEA,8FAA8F;AAC9F,gGAAgG;AAChG,qBAAqB;AACrB,OAAO,SAASU,gBAAgBrB,EAAgB,EAAEC,UAAsB,EAAEU,SAAsD,EAAEW,QAAkB,EAAEC,YAAyB,EAAEjB,KAAa,EAAEC,MAAc;IAC5M,IAAII,UAAUa,MAAM,KAAK,GAAG;IAC5B,MAAMC,QAAQ7B,UAAUI;IACxB,IAAIyB,MAAMD,MAAM,KAAK,GAAG;IACxB,yFAAyF;IACzF,uFAAuF;IACvF,MAAME,SAAS,IAAIC,IAAIF,MAAMG,IAAI;IACjC,MAAMC,QAAQ,IAAIC,IAAInB,UAAUoB,GAAG,CAAC,CAAChB,GAAGC,IAAM;YAACD,EAAEG,IAAI;YAAE,IAAKF,CAAAA,IAAI,CAAA;SAAG;IACnE,MAAMgB,SAAS;WAAInC,iBAAiByB,UAAUG,OAAOI;KAAO,CACzDI,MAAM,CAAC,CAAC,CAACf,MAAMC,MAAM,GAAKA,QAAQ,QAAQI,aAAaW,GAAG,CAAChB,OAC3DiB,IAAI,CAAC,CAACC,GAAGC,IAAMA,CAAC,CAAC,EAAE,GAAGD,CAAC,CAAC,EAAE,EAC1BE,KAAK,CAAC,GAAGhC;IACZ0B,OAAOlB,OAAO,CAAC,CAAC,CAACI,KAAK,EAAEF;QACtB,MAAMuB,WAAWtC,WAAWuC,GAAG,CAACtB;QAChC,IAAIqB,UAAU;YACZA,SAASpB,KAAK,IAAIZ,SAAUT,CAAAA,QAAQkB,CAAAA;YACpC,IAAIU,OAAOQ,GAAG,CAAChB,OAAOqB,SAASnB,GAAG,GAAG;QACvC,OAAO;YACLnB,WAAWgB,GAAG,CAACC,MAAM;gBAAEC,OAAOZ,SAAUT,CAAAA,QAAQkB,CAAAA;gBAAII,KAAK;YAAO;QAClE;IACF;AACF;AAEA,+FAA+F;AAC/F,6FAA6F;AAC7F,kBAAkB;AAClB,OAAO,eAAeqB,kBAAkBzC,EAAgB,EAAE0C,GAAmB,EAAEzC,UAAsB,EAAE0C,KAAa,EAAErC,KAAa,EAAEiB,YAAyB,EAAEhB,MAAc;IAC5K,MAAMqC,aAAa,IAAId;IACvB,MAAMe,kBAAkB,IAAIf;IAC5B,MAAMgB,MAAM,MAAMnD,mBAAmBK,IAAI0C,KAAKC,OAAOrC,OAAOiB;IAC5DuB,IAAIhC,OAAO,CAAC,CAAC,EAAEI,IAAI,EAAE6B,KAAK,EAAEC,UAAU,EAAE,EAAEhC;QACxC4B,WAAW3B,GAAG,CAACC,MAAM6B;QACrBF,gBAAgB5B,GAAG,CAACC,MAAM8B;QAC1B,MAAMT,WAAWtC,WAAWuC,GAAG,CAACtB;QAChC,IAAIqB,UAAU;YACZA,SAASpB,KAAK,IAAIZ,SAAUT,CAAAA,QAAQkB,CAAAA;YACpCuB,SAASnB,GAAG,GAAG,GAAGmB,SAASnB,GAAG,CAAC,OAAO,CAAC;QACzC,OAAO;YACLnB,WAAWgB,GAAG,CAACC,MAAM;gBAAEC,OAAOZ,SAAUT,CAAAA,QAAQkB,CAAAA;gBAAII,KAAK;YAAS;QACpE;IACF;IACA,OAAO;QAAEwB;QAAYC;IAAgB;AACvC"}
@@ -1,3 +1,4 @@
1
+ import type { Feature } from '../features/types.js';
1
2
  import type { SignalName, SignalWeights } from './signals.js';
2
3
  import { type Config, type FeatureName } from './types.js';
3
4
  export declare function presetNames(cfg: Config): string[];
@@ -17,6 +18,7 @@ export declare function embedConfig(cfg: Config): {
17
18
  url?: string;
18
19
  key?: string;
19
20
  languages?: string[];
21
+ chunkTokens?: number;
20
22
  } | null;
21
23
  export declare function contentTokenize(cfg: Config): string | undefined;
22
- export declare function featureSignature(cfg: Config): string;
24
+ export declare function featureSignature(cfg: Config, features: Feature[]): string;
@@ -1,4 +1,3 @@
1
- import { modelIdentity } from '../embed/identity.js';
2
1
  import { FEATURE_NAMES } from './types.js';
3
2
  export function presetNames(cfg) {
4
3
  return Object.keys(cfg.presets);
@@ -64,7 +63,8 @@ export function embedConfig(cfg) {
64
63
  provider: (_e_provider = e.provider) !== null && _e_provider !== void 0 ? _e_provider : 'static',
65
64
  url: e.url,
66
65
  key: e.key,
67
- languages: e.languages
66
+ languages: e.languages,
67
+ chunkTokens: e.chunkTokens
68
68
  };
69
69
  }
70
70
  // The configured FTS5 tokenizer, or undefined for the built-in default. Undefined rather than
@@ -76,13 +76,13 @@ export function contentTokenize(cfg) {
76
76
  return tokenize === undefined || tokenize.trim() === '' ? undefined : tokenize.trim();
77
77
  }
78
78
  // Cache key over everything indexing derives from, so any change to those inputs rebuilds.
79
- export function featureSignature(cfg) {
79
+ // Segment order is fixed (features list, then `features` in registry order, then tokenize, then presets) since reordering alone would fire a spurious rebuild; tokenize/presets stay here because no Feature module owns either.
80
+ export function featureSignature(cfg, features) {
80
81
  const globalPart = enabledFeatures(cfg).filter((name)=>name !== 'embed').join(',');
81
- const e = embedConfig(cfg);
82
- // Weight identity from identity.ts, no network: a static model's resolved sha or local
83
- // size+mtime, appended once known so changed weights re-embed.
84
- const identity = e && e.provider === 'static' ? modelIdentity(e.model) : undefined;
85
- const embedPart = e ? `embed:${e.provider}:${e.model}${identity !== undefined ? `@${identity}` : ''}` : 'embed:off';
82
+ const featureParts = features.map((feature)=>{
83
+ var _feature_signature;
84
+ return (_feature_signature = feature.signature) === null || _feature_signature === void 0 ? void 0 : _feature_signature.call(feature, cfg);
85
+ }).filter((part)=>part !== undefined);
86
86
  // One keyed segment per preset so a rebuild notice can name exactly which preset moved.
87
87
  const presetsPart = [
88
88
  ...presetNames(cfg)
@@ -100,7 +100,7 @@ export function featureSignature(cfg) {
100
100
  const tokenize = contentTokenize(cfg);
101
101
  const parts = [
102
102
  `features:${globalPart}`,
103
- embedPart
103
+ ...featureParts
104
104
  ];
105
105
  if (tokenize !== undefined) parts.push(`tokenize:${tokenize}`);
106
106
  parts.push(presetsPart);
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/config/access.ts"],"sourcesContent":["import { modelIdentity } from '../embed/identity.ts';\nimport type { SignalName, SignalWeights } from './signals.ts';\nimport { type Config, type EmbedConfig, FEATURE_NAMES, type FeatureName } from './types.ts';\n\nexport function presetNames(cfg: Config): string[] {\n return Object.keys(cfg.presets);\n}\n\n// Whether the tree names a model at all.\nexport function embedEnabled(cfg: Config): boolean {\n return typeof cfg.embed?.model === 'string' && cfg.embed.model.length > 0;\n}\n\n// A preset's effective signal weights: declared exhaustively when `signals` is present;\n// otherwise every signal whose prerequisite holds, each at weight 1 -- words always, links when\n// the feature is on, vectors when the tree names a model. Single source search() and status/map\n// read.\nexport function presetSignals(cfg: Config, name: string): SignalWeights {\n const declared = cfg.presets[name]?.signals;\n if (declared) return declared;\n const signals: SignalWeights = { words: 1 };\n if (cfg.features?.links !== false) signals.links = 1;\n if (embedEnabled(cfg)) signals.vectors = 1;\n return signals;\n}\n\nexport function presetHasSignal(cfg: Config, name: string, signal: SignalName): boolean {\n return presetSignals(cfg, name)[signal] !== undefined;\n}\n\n// Whether embedding needs to run: a model is named AND some preset's effective signals include vectors.\nexport function anyPresetEmbeds(cfg: Config): boolean {\n return presetNames(cfg).some((name) => presetHasSignal(cfg, name, 'vectors'));\n}\n\n// Opt-out features (default on): absent block or key means enabled. `rank` additionally\n// requires `links`. `embed` is derived, not a features-block member: on iff some preset's\n// effective signals include vectors.\nexport function featureEnabled(cfg: Config, name: FeatureName): boolean {\n if (name === 'embed') return anyPresetEmbeds(cfg);\n const enabled = cfg.features?.[name] !== false;\n if (name === 'rank') return enabled && featureEnabled(cfg, 'links');\n return enabled;\n}\n\nexport function enabledFeatures(cfg: Config): FeatureName[] {\n return FEATURE_NAMES.filter((name) => featureEnabled(cfg, name));\n}\n\n// Feature states, so \"off\" and \"empty\" stay distinguishable in output. `embed` is excluded:\n// it is not a features-block toggle, and `status` reports it on its own line.\nexport function featureStates(cfg: Config): { on: FeatureName[]; off: FeatureName[] } {\n const reported = FEATURE_NAMES.filter((name) => name !== 'embed');\n return {\n on: reported.filter((name) => featureEnabled(cfg, name)),\n off: reported.filter((name) => !featureEnabled(cfg, name)),\n };\n}\n\n// Resolved embed provider settings, or null when the config names no model. There is no\n// fallback to DEFAULT_EMBED_MODEL: that constant is a template `init` and the v4 migration\n// write into the file, never an implicit runtime default.\nexport function embedConfig(cfg: Config): { model: string; provider: 'static' | 'openai' | 'cohere'; url?: string; key?: string; languages?: string[] } | null {\n if (!embedEnabled(cfg)) return null;\n const e = cfg.embed as EmbedConfig;\n return { model: e.model as string, provider: e.provider ?? 'static', url: e.url, key: e.key, languages: e.languages };\n}\n\n// The configured FTS5 tokenizer, or undefined for the built-in default. Undefined rather than\n// the default string so featureSignature can leave the segment off entirely for trees that\n// never set it, which is what keeps existing indexes from rebuilding on upgrade.\nexport function contentTokenize(cfg: Config): string | undefined {\n const tokenize = cfg.content?.tokenize;\n return tokenize === undefined || tokenize.trim() === '' ? undefined : tokenize.trim();\n}\n\n// Cache key over everything indexing derives from, so any change to those inputs rebuilds.\nexport function featureSignature(cfg: Config): string {\n const globalPart = enabledFeatures(cfg)\n .filter((name) => name !== 'embed')\n .join(',');\n const e = embedConfig(cfg);\n // Weight identity from identity.ts, no network: a static model's resolved sha or local\n // size+mtime, appended once known so changed weights re-embed.\n const identity = e && e.provider === 'static' ? modelIdentity(e.model) : undefined;\n const embedPart = e ? `embed:${e.provider}:${e.model}${identity !== undefined ? `@${identity}` : ''}` : 'embed:off';\n // One keyed segment per preset so a rebuild notice can name exactly which preset moved.\n const presetsPart = [...presetNames(cfg)]\n .sort()\n .map((name) => {\n const p = cfg.presets[name];\n const include = [...p.include].sort().join('+');\n const exclude = [...(p.exclude ?? [])].sort().join('+');\n return `preset:${name}:${include}:${exclude}:${presetHasSignal(cfg, name, 'vectors') ? 'on' : 'off'}`;\n })\n .join('|');\n const tokenize = contentTokenize(cfg);\n const parts = [`features:${globalPart}`, embedPart];\n if (tokenize !== undefined) parts.push(`tokenize:${tokenize}`);\n parts.push(presetsPart);\n return parts.join('|');\n}\n"],"names":["modelIdentity","FEATURE_NAMES","presetNames","cfg","Object","keys","presets","embedEnabled","embed","model","length","presetSignals","name","declared","signals","words","features","links","vectors","presetHasSignal","signal","undefined","anyPresetEmbeds","some","featureEnabled","enabled","enabledFeatures","filter","featureStates","reported","on","off","embedConfig","e","provider","url","key","languages","contentTokenize","tokenize","content","trim","featureSignature","globalPart","join","identity","embedPart","presetsPart","sort","map","p","include","exclude","parts","push"],"mappings":"AAAA,SAASA,aAAa,QAAQ,uBAAuB;AAErD,SAAwCC,aAAa,QAA0B,aAAa;AAE5F,OAAO,SAASC,YAAYC,GAAW;IACrC,OAAOC,OAAOC,IAAI,CAACF,IAAIG,OAAO;AAChC;AAEA,yCAAyC;AACzC,OAAO,SAASC,aAAaJ,GAAW;QACxBA;IAAd,OAAO,SAAOA,aAAAA,IAAIK,KAAK,cAATL,iCAAAA,WAAWM,KAAK,MAAK,YAAYN,IAAIK,KAAK,CAACC,KAAK,CAACC,MAAM,GAAG;AAC1E;AAEA,wFAAwF;AACxF,gGAAgG;AAChG,gGAAgG;AAChG,QAAQ;AACR,OAAO,SAASC,cAAcR,GAAW,EAAES,IAAY;QACpCT,mBAGbA;IAHJ,MAAMU,YAAWV,oBAAAA,IAAIG,OAAO,CAACM,KAAK,cAAjBT,wCAAAA,kBAAmBW,OAAO;IAC3C,IAAID,UAAU,OAAOA;IACrB,MAAMC,UAAyB;QAAEC,OAAO;IAAE;IAC1C,IAAIZ,EAAAA,gBAAAA,IAAIa,QAAQ,cAAZb,oCAAAA,cAAcc,KAAK,MAAK,OAAOH,QAAQG,KAAK,GAAG;IACnD,IAAIV,aAAaJ,MAAMW,QAAQI,OAAO,GAAG;IACzC,OAAOJ;AACT;AAEA,OAAO,SAASK,gBAAgBhB,GAAW,EAAES,IAAY,EAAEQ,MAAkB;IAC3E,OAAOT,cAAcR,KAAKS,KAAK,CAACQ,OAAO,KAAKC;AAC9C;AAEA,wGAAwG;AACxG,OAAO,SAASC,gBAAgBnB,GAAW;IACzC,OAAOD,YAAYC,KAAKoB,IAAI,CAAC,CAACX,OAASO,gBAAgBhB,KAAKS,MAAM;AACpE;AAEA,wFAAwF;AACxF,0FAA0F;AAC1F,qCAAqC;AACrC,OAAO,SAASY,eAAerB,GAAW,EAAES,IAAiB;QAE3CT;IADhB,IAAIS,SAAS,SAAS,OAAOU,gBAAgBnB;IAC7C,MAAMsB,UAAUtB,EAAAA,gBAAAA,IAAIa,QAAQ,cAAZb,oCAAAA,aAAc,CAACS,KAAK,MAAK;IACzC,IAAIA,SAAS,QAAQ,OAAOa,WAAWD,eAAerB,KAAK;IAC3D,OAAOsB;AACT;AAEA,OAAO,SAASC,gBAAgBvB,GAAW;IACzC,OAAOF,cAAc0B,MAAM,CAAC,CAACf,OAASY,eAAerB,KAAKS;AAC5D;AAEA,4FAA4F;AAC5F,8EAA8E;AAC9E,OAAO,SAASgB,cAAczB,GAAW;IACvC,MAAM0B,WAAW5B,cAAc0B,MAAM,CAAC,CAACf,OAASA,SAAS;IACzD,OAAO;QACLkB,IAAID,SAASF,MAAM,CAAC,CAACf,OAASY,eAAerB,KAAKS;QAClDmB,KAAKF,SAASF,MAAM,CAAC,CAACf,OAAS,CAACY,eAAerB,KAAKS;IACtD;AACF;AAEA,wFAAwF;AACxF,2FAA2F;AAC3F,0DAA0D;AAC1D,OAAO,SAASoB,YAAY7B,GAAW;QAGQ8B;IAF7C,IAAI,CAAC1B,aAAaJ,MAAM,OAAO;IAC/B,MAAM8B,IAAI9B,IAAIK,KAAK;IACnB,OAAO;QAAEC,OAAOwB,EAAExB,KAAK;QAAYyB,QAAQ,GAAED,cAAAA,EAAEC,QAAQ,cAAVD,yBAAAA,cAAc;QAAUE,KAAKF,EAAEE,GAAG;QAAEC,KAAKH,EAAEG,GAAG;QAAEC,WAAWJ,EAAEI,SAAS;IAAC;AACtH;AAEA,8FAA8F;AAC9F,2FAA2F;AAC3F,iFAAiF;AACjF,OAAO,SAASC,gBAAgBnC,GAAW;QACxBA;IAAjB,MAAMoC,YAAWpC,eAAAA,IAAIqC,OAAO,cAAXrC,mCAAAA,aAAaoC,QAAQ;IACtC,OAAOA,aAAalB,aAAakB,SAASE,IAAI,OAAO,KAAKpB,YAAYkB,SAASE,IAAI;AACrF;AAEA,2FAA2F;AAC3F,OAAO,SAASC,iBAAiBvC,GAAW;IAC1C,MAAMwC,aAAajB,gBAAgBvB,KAChCwB,MAAM,CAAC,CAACf,OAASA,SAAS,SAC1BgC,IAAI,CAAC;IACR,MAAMX,IAAID,YAAY7B;IACtB,uFAAuF;IACvF,+DAA+D;IAC/D,MAAM0C,WAAWZ,KAAKA,EAAEC,QAAQ,KAAK,WAAWlC,cAAciC,EAAExB,KAAK,IAAIY;IACzE,MAAMyB,YAAYb,IAAI,CAAC,MAAM,EAAEA,EAAEC,QAAQ,CAAC,CAAC,EAAED,EAAExB,KAAK,GAAGoC,aAAaxB,YAAY,CAAC,CAAC,EAAEwB,UAAU,GAAG,IAAI,GAAG;IACxG,wFAAwF;IACxF,MAAME,cAAc;WAAI7C,YAAYC;KAAK,CACtC6C,IAAI,GACJC,GAAG,CAAC,CAACrC;YAGiBsC;QAFrB,MAAMA,IAAI/C,IAAIG,OAAO,CAACM,KAAK;QAC3B,MAAMuC,UAAU;eAAID,EAAEC,OAAO;SAAC,CAACH,IAAI,GAAGJ,IAAI,CAAC;QAC3C,MAAMQ,UAAU;gBAAKF,aAAAA,EAAEE,OAAO,cAATF,wBAAAA,aAAa,EAAE;SAAE,CAACF,IAAI,GAAGJ,IAAI,CAAC;QACnD,OAAO,CAAC,OAAO,EAAEhC,KAAK,CAAC,EAAEuC,QAAQ,CAAC,EAAEC,QAAQ,CAAC,EAAEjC,gBAAgBhB,KAAKS,MAAM,aAAa,OAAO,OAAO;IACvG,GACCgC,IAAI,CAAC;IACR,MAAML,WAAWD,gBAAgBnC;IACjC,MAAMkD,QAAQ;QAAC,CAAC,SAAS,EAAEV,YAAY;QAAEG;KAAU;IACnD,IAAIP,aAAalB,WAAWgC,MAAMC,IAAI,CAAC,CAAC,SAAS,EAAEf,UAAU;IAC7Dc,MAAMC,IAAI,CAACP;IACX,OAAOM,MAAMT,IAAI,CAAC;AACpB"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/config/access.ts"],"sourcesContent":["import type { Feature } from '../features/types.ts';\nimport type { SignalName, SignalWeights } from './signals.ts';\nimport { type Config, type EmbedConfig, FEATURE_NAMES, type FeatureName } from './types.ts';\n\nexport function presetNames(cfg: Config): string[] {\n return Object.keys(cfg.presets);\n}\n\n// Whether the tree names a model at all.\nexport function embedEnabled(cfg: Config): boolean {\n return typeof cfg.embed?.model === 'string' && cfg.embed.model.length > 0;\n}\n\n// A preset's effective signal weights: declared exhaustively when `signals` is present;\n// otherwise every signal whose prerequisite holds, each at weight 1 -- words always, links when\n// the feature is on, vectors when the tree names a model. Single source search() and status/map\n// read.\nexport function presetSignals(cfg: Config, name: string): SignalWeights {\n const declared = cfg.presets[name]?.signals;\n if (declared) return declared;\n const signals: SignalWeights = { words: 1 };\n if (cfg.features?.links !== false) signals.links = 1;\n if (embedEnabled(cfg)) signals.vectors = 1;\n return signals;\n}\n\nexport function presetHasSignal(cfg: Config, name: string, signal: SignalName): boolean {\n return presetSignals(cfg, name)[signal] !== undefined;\n}\n\n// Whether embedding needs to run: a model is named AND some preset's effective signals include vectors.\nexport function anyPresetEmbeds(cfg: Config): boolean {\n return presetNames(cfg).some((name) => presetHasSignal(cfg, name, 'vectors'));\n}\n\n// Opt-out features (default on): absent block or key means enabled. `rank` additionally\n// requires `links`. `embed` is derived, not a features-block member: on iff some preset's\n// effective signals include vectors.\nexport function featureEnabled(cfg: Config, name: FeatureName): boolean {\n if (name === 'embed') return anyPresetEmbeds(cfg);\n const enabled = cfg.features?.[name] !== false;\n if (name === 'rank') return enabled && featureEnabled(cfg, 'links');\n return enabled;\n}\n\nexport function enabledFeatures(cfg: Config): FeatureName[] {\n return FEATURE_NAMES.filter((name) => featureEnabled(cfg, name));\n}\n\n// Feature states, so \"off\" and \"empty\" stay distinguishable in output. `embed` is excluded:\n// it is not a features-block toggle, and `status` reports it on its own line.\nexport function featureStates(cfg: Config): { on: FeatureName[]; off: FeatureName[] } {\n const reported = FEATURE_NAMES.filter((name) => name !== 'embed');\n return {\n on: reported.filter((name) => featureEnabled(cfg, name)),\n off: reported.filter((name) => !featureEnabled(cfg, name)),\n };\n}\n\n// Resolved embed provider settings, or null when the config names no model. There is no\n// fallback to DEFAULT_EMBED_MODEL: that constant is a template `init` and the v4 migration\n// write into the file, never an implicit runtime default.\nexport function embedConfig(cfg: Config): { model: string; provider: 'static' | 'openai' | 'cohere'; url?: string; key?: string; languages?: string[]; chunkTokens?: number } | null {\n if (!embedEnabled(cfg)) return null;\n const e = cfg.embed as EmbedConfig;\n return { model: e.model as string, provider: e.provider ?? 'static', url: e.url, key: e.key, languages: e.languages, chunkTokens: e.chunkTokens };\n}\n\n// The configured FTS5 tokenizer, or undefined for the built-in default. Undefined rather than\n// the default string so featureSignature can leave the segment off entirely for trees that\n// never set it, which is what keeps existing indexes from rebuilding on upgrade.\nexport function contentTokenize(cfg: Config): string | undefined {\n const tokenize = cfg.content?.tokenize;\n return tokenize === undefined || tokenize.trim() === '' ? undefined : tokenize.trim();\n}\n\n// Cache key over everything indexing derives from, so any change to those inputs rebuilds.\n// Segment order is fixed (features list, then `features` in registry order, then tokenize, then presets) since reordering alone would fire a spurious rebuild; tokenize/presets stay here because no Feature module owns either.\nexport function featureSignature(cfg: Config, features: Feature[]): string {\n const globalPart = enabledFeatures(cfg)\n .filter((name) => name !== 'embed')\n .join(',');\n const featureParts = features.map((feature) => feature.signature?.(cfg)).filter((part): part is string => part !== undefined);\n // One keyed segment per preset so a rebuild notice can name exactly which preset moved.\n const presetsPart = [...presetNames(cfg)]\n .sort()\n .map((name) => {\n const p = cfg.presets[name];\n const include = [...p.include].sort().join('+');\n const exclude = [...(p.exclude ?? [])].sort().join('+');\n return `preset:${name}:${include}:${exclude}:${presetHasSignal(cfg, name, 'vectors') ? 'on' : 'off'}`;\n })\n .join('|');\n const tokenize = contentTokenize(cfg);\n const parts = [`features:${globalPart}`, ...featureParts];\n if (tokenize !== undefined) parts.push(`tokenize:${tokenize}`);\n parts.push(presetsPart);\n return parts.join('|');\n}\n"],"names":["FEATURE_NAMES","presetNames","cfg","Object","keys","presets","embedEnabled","embed","model","length","presetSignals","name","declared","signals","words","features","links","vectors","presetHasSignal","signal","undefined","anyPresetEmbeds","some","featureEnabled","enabled","enabledFeatures","filter","featureStates","reported","on","off","embedConfig","e","provider","url","key","languages","chunkTokens","contentTokenize","tokenize","content","trim","featureSignature","globalPart","join","featureParts","map","feature","signature","part","presetsPart","sort","p","include","exclude","parts","push"],"mappings":"AAEA,SAAwCA,aAAa,QAA0B,aAAa;AAE5F,OAAO,SAASC,YAAYC,GAAW;IACrC,OAAOC,OAAOC,IAAI,CAACF,IAAIG,OAAO;AAChC;AAEA,yCAAyC;AACzC,OAAO,SAASC,aAAaJ,GAAW;QACxBA;IAAd,OAAO,SAAOA,aAAAA,IAAIK,KAAK,cAATL,iCAAAA,WAAWM,KAAK,MAAK,YAAYN,IAAIK,KAAK,CAACC,KAAK,CAACC,MAAM,GAAG;AAC1E;AAEA,wFAAwF;AACxF,gGAAgG;AAChG,gGAAgG;AAChG,QAAQ;AACR,OAAO,SAASC,cAAcR,GAAW,EAAES,IAAY;QACpCT,mBAGbA;IAHJ,MAAMU,YAAWV,oBAAAA,IAAIG,OAAO,CAACM,KAAK,cAAjBT,wCAAAA,kBAAmBW,OAAO;IAC3C,IAAID,UAAU,OAAOA;IACrB,MAAMC,UAAyB;QAAEC,OAAO;IAAE;IAC1C,IAAIZ,EAAAA,gBAAAA,IAAIa,QAAQ,cAAZb,oCAAAA,cAAcc,KAAK,MAAK,OAAOH,QAAQG,KAAK,GAAG;IACnD,IAAIV,aAAaJ,MAAMW,QAAQI,OAAO,GAAG;IACzC,OAAOJ;AACT;AAEA,OAAO,SAASK,gBAAgBhB,GAAW,EAAES,IAAY,EAAEQ,MAAkB;IAC3E,OAAOT,cAAcR,KAAKS,KAAK,CAACQ,OAAO,KAAKC;AAC9C;AAEA,wGAAwG;AACxG,OAAO,SAASC,gBAAgBnB,GAAW;IACzC,OAAOD,YAAYC,KAAKoB,IAAI,CAAC,CAACX,OAASO,gBAAgBhB,KAAKS,MAAM;AACpE;AAEA,wFAAwF;AACxF,0FAA0F;AAC1F,qCAAqC;AACrC,OAAO,SAASY,eAAerB,GAAW,EAAES,IAAiB;QAE3CT;IADhB,IAAIS,SAAS,SAAS,OAAOU,gBAAgBnB;IAC7C,MAAMsB,UAAUtB,EAAAA,gBAAAA,IAAIa,QAAQ,cAAZb,oCAAAA,aAAc,CAACS,KAAK,MAAK;IACzC,IAAIA,SAAS,QAAQ,OAAOa,WAAWD,eAAerB,KAAK;IAC3D,OAAOsB;AACT;AAEA,OAAO,SAASC,gBAAgBvB,GAAW;IACzC,OAAOF,cAAc0B,MAAM,CAAC,CAACf,OAASY,eAAerB,KAAKS;AAC5D;AAEA,4FAA4F;AAC5F,8EAA8E;AAC9E,OAAO,SAASgB,cAAczB,GAAW;IACvC,MAAM0B,WAAW5B,cAAc0B,MAAM,CAAC,CAACf,OAASA,SAAS;IACzD,OAAO;QACLkB,IAAID,SAASF,MAAM,CAAC,CAACf,OAASY,eAAerB,KAAKS;QAClDmB,KAAKF,SAASF,MAAM,CAAC,CAACf,OAAS,CAACY,eAAerB,KAAKS;IACtD;AACF;AAEA,wFAAwF;AACxF,2FAA2F;AAC3F,0DAA0D;AAC1D,OAAO,SAASoB,YAAY7B,GAAW;QAGQ8B;IAF7C,IAAI,CAAC1B,aAAaJ,MAAM,OAAO;IAC/B,MAAM8B,IAAI9B,IAAIK,KAAK;IACnB,OAAO;QAAEC,OAAOwB,EAAExB,KAAK;QAAYyB,QAAQ,GAAED,cAAAA,EAAEC,QAAQ,cAAVD,yBAAAA,cAAc;QAAUE,KAAKF,EAAEE,GAAG;QAAEC,KAAKH,EAAEG,GAAG;QAAEC,WAAWJ,EAAEI,SAAS;QAAEC,aAAaL,EAAEK,WAAW;IAAC;AAClJ;AAEA,8FAA8F;AAC9F,2FAA2F;AAC3F,iFAAiF;AACjF,OAAO,SAASC,gBAAgBpC,GAAW;QACxBA;IAAjB,MAAMqC,YAAWrC,eAAAA,IAAIsC,OAAO,cAAXtC,mCAAAA,aAAaqC,QAAQ;IACtC,OAAOA,aAAanB,aAAamB,SAASE,IAAI,OAAO,KAAKrB,YAAYmB,SAASE,IAAI;AACrF;AAEA,2FAA2F;AAC3F,iOAAiO;AACjO,OAAO,SAASC,iBAAiBxC,GAAW,EAAEa,QAAmB;IAC/D,MAAM4B,aAAalB,gBAAgBvB,KAChCwB,MAAM,CAAC,CAACf,OAASA,SAAS,SAC1BiC,IAAI,CAAC;IACR,MAAMC,eAAe9B,SAAS+B,GAAG,CAAC,CAACC;YAAYA;gBAAAA,qBAAAA,QAAQC,SAAS,cAAjBD,yCAAAA,wBAAAA,SAAoB7C;OAAMwB,MAAM,CAAC,CAACuB,OAAyBA,SAAS7B;IACnH,wFAAwF;IACxF,MAAM8B,cAAc;WAAIjD,YAAYC;KAAK,CACtCiD,IAAI,GACJL,GAAG,CAAC,CAACnC;YAGiByC;QAFrB,MAAMA,IAAIlD,IAAIG,OAAO,CAACM,KAAK;QAC3B,MAAM0C,UAAU;eAAID,EAAEC,OAAO;SAAC,CAACF,IAAI,GAAGP,IAAI,CAAC;QAC3C,MAAMU,UAAU;gBAAKF,aAAAA,EAAEE,OAAO,cAATF,wBAAAA,aAAa,EAAE;SAAE,CAACD,IAAI,GAAGP,IAAI,CAAC;QACnD,OAAO,CAAC,OAAO,EAAEjC,KAAK,CAAC,EAAE0C,QAAQ,CAAC,EAAEC,QAAQ,CAAC,EAAEpC,gBAAgBhB,KAAKS,MAAM,aAAa,OAAO,OAAO;IACvG,GACCiC,IAAI,CAAC;IACR,MAAML,WAAWD,gBAAgBpC;IACjC,MAAMqD,QAAQ;QAAC,CAAC,SAAS,EAAEZ,YAAY;WAAKE;KAAa;IACzD,IAAIN,aAAanB,WAAWmC,MAAMC,IAAI,CAAC,CAAC,SAAS,EAAEjB,UAAU;IAC7DgB,MAAMC,IAAI,CAACN;IACX,OAAOK,MAAMX,IAAI,CAAC;AACpB"}
@@ -6,3 +6,4 @@ export type { SignalName, SignalWeights } from './signals.js';
6
6
  export { SIGNAL_NAMES, SIGNAL_PREREQUISITES } from './signals.js';
7
7
  export type { Config, EffectiveSearch, EmbedConfig, FeatureName, Preset, ResolvedConfig, SavedQuery, SavedSearch, SearchOverrides } from './types.js';
8
8
  export { CONFIG_FILENAME, DEFAULT_EMBED_MODEL, STATE_DIR, SUPPORTED_CONFIG_VERSION } from './types.js';
9
+ export { KNOWN_EMBED_KEYS } from './validate.js';
@@ -3,3 +3,4 @@ export { findConfigPath, initConfig, loadConfig, migrateConfig } from './load.js
3
3
  export { resolvePreset, resolveSearch } from './resolve.js';
4
4
  export { SIGNAL_NAMES, SIGNAL_PREREQUISITES } from './signals.js';
5
5
  export { CONFIG_FILENAME, DEFAULT_EMBED_MODEL, STATE_DIR, SUPPORTED_CONFIG_VERSION } from './types.js';
6
+ export { KNOWN_EMBED_KEYS } from './validate.js';
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/config/index.ts"],"sourcesContent":["export { anyPresetEmbeds, contentTokenize, embedConfig, embedEnabled, enabledFeatures, featureEnabled, featureSignature, featureStates, presetHasSignal, presetNames, presetSignals } from './access.ts';\nexport type { InitOverrides } from './load.ts';\nexport { findConfigPath, initConfig, loadConfig, migrateConfig } from './load.ts';\nexport { resolvePreset, resolveSearch } from './resolve.ts';\nexport type { SignalName, SignalWeights } from './signals.ts';\nexport { SIGNAL_NAMES, SIGNAL_PREREQUISITES } from './signals.ts';\nexport type { Config, EffectiveSearch, EmbedConfig, FeatureName, Preset, ResolvedConfig, SavedQuery, SavedSearch, SearchOverrides } from './types.ts';\nexport { CONFIG_FILENAME, DEFAULT_EMBED_MODEL, STATE_DIR, SUPPORTED_CONFIG_VERSION } from './types.ts';\n"],"names":["anyPresetEmbeds","contentTokenize","embedConfig","embedEnabled","enabledFeatures","featureEnabled","featureSignature","featureStates","presetHasSignal","presetNames","presetSignals","findConfigPath","initConfig","loadConfig","migrateConfig","resolvePreset","resolveSearch","SIGNAL_NAMES","SIGNAL_PREREQUISITES","CONFIG_FILENAME","DEFAULT_EMBED_MODEL","STATE_DIR","SUPPORTED_CONFIG_VERSION"],"mappings":"AAAA,SAASA,eAAe,EAAEC,eAAe,EAAEC,WAAW,EAAEC,YAAY,EAAEC,eAAe,EAAEC,cAAc,EAAEC,gBAAgB,EAAEC,aAAa,EAAEC,eAAe,EAAEC,WAAW,EAAEC,aAAa,QAAQ,cAAc;AAEzM,SAASC,cAAc,EAAEC,UAAU,EAAEC,UAAU,EAAEC,aAAa,QAAQ,YAAY;AAClF,SAASC,aAAa,EAAEC,aAAa,QAAQ,eAAe;AAE5D,SAASC,YAAY,EAAEC,oBAAoB,QAAQ,eAAe;AAElE,SAASC,eAAe,EAAEC,mBAAmB,EAAEC,SAAS,EAAEC,wBAAwB,QAAQ,aAAa"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/config/index.ts"],"sourcesContent":["export { anyPresetEmbeds, contentTokenize, embedConfig, embedEnabled, enabledFeatures, featureEnabled, featureSignature, featureStates, presetHasSignal, presetNames, presetSignals } from './access.ts';\nexport type { InitOverrides } from './load.ts';\nexport { findConfigPath, initConfig, loadConfig, migrateConfig } from './load.ts';\nexport { resolvePreset, resolveSearch } from './resolve.ts';\nexport type { SignalName, SignalWeights } from './signals.ts';\nexport { SIGNAL_NAMES, SIGNAL_PREREQUISITES } from './signals.ts';\nexport type { Config, EffectiveSearch, EmbedConfig, FeatureName, Preset, ResolvedConfig, SavedQuery, SavedSearch, SearchOverrides } from './types.ts';\nexport { CONFIG_FILENAME, DEFAULT_EMBED_MODEL, STATE_DIR, SUPPORTED_CONFIG_VERSION } from './types.ts';\nexport { KNOWN_EMBED_KEYS } from './validate.ts';\n"],"names":["anyPresetEmbeds","contentTokenize","embedConfig","embedEnabled","enabledFeatures","featureEnabled","featureSignature","featureStates","presetHasSignal","presetNames","presetSignals","findConfigPath","initConfig","loadConfig","migrateConfig","resolvePreset","resolveSearch","SIGNAL_NAMES","SIGNAL_PREREQUISITES","CONFIG_FILENAME","DEFAULT_EMBED_MODEL","STATE_DIR","SUPPORTED_CONFIG_VERSION","KNOWN_EMBED_KEYS"],"mappings":"AAAA,SAASA,eAAe,EAAEC,eAAe,EAAEC,WAAW,EAAEC,YAAY,EAAEC,eAAe,EAAEC,cAAc,EAAEC,gBAAgB,EAAEC,aAAa,EAAEC,eAAe,EAAEC,WAAW,EAAEC,aAAa,QAAQ,cAAc;AAEzM,SAASC,cAAc,EAAEC,UAAU,EAAEC,UAAU,EAAEC,aAAa,QAAQ,YAAY;AAClF,SAASC,aAAa,EAAEC,aAAa,QAAQ,eAAe;AAE5D,SAASC,YAAY,EAAEC,oBAAoB,QAAQ,eAAe;AAElE,SAASC,eAAe,EAAEC,mBAAmB,EAAEC,SAAS,EAAEC,wBAAwB,QAAQ,aAAa;AACvG,SAASC,gBAAgB,QAAQ,gBAAgB"}
@@ -10,6 +10,7 @@ export interface EmbedConfig {
10
10
  url?: string;
11
11
  key?: string;
12
12
  languages?: string[];
13
+ chunkTokens?: number;
13
14
  }
14
15
  export declare const DEFAULT_EMBED_MODEL = "minishlab/potion-retrieval-32M";
15
16
  export interface Preset {
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/config/types.ts"],"sourcesContent":["import type { SignalWeights } from './signals.ts';\n\nexport const CONFIG_FILENAME = 'sense.config.json';\nexport const STATE_DIR = '.sense';\n\n// Highest sense.config.json `version` this build understands. Older versions auto-migrate on load.\nexport const SUPPORTED_CONFIG_VERSION = 5;\n\n// Each feature owns its tables, parse-time extraction, and reconcile step; commands degrade when one is off.\n// links/sections/tags/rank are opt-out toggles in the top-level `features` block; embed is not a\n// member of that block -- it is on exactly when an `embed` block names a model.\nexport const FEATURE_NAMES = ['links', 'sections', 'tags', 'rank', 'embed'] as const;\nexport type FeatureName = (typeof FEATURE_NAMES)[number];\n\n// `embed` names the model and gives the tree vectors at all; a preset's `signals` says which\n// engines that scope's searches use.\nexport interface EmbedConfig {\n model?: string;\n provider?: 'static' | 'openai' | 'cohere';\n url?: string; // openai provider: OpenAI-compatible base URL, e.g. http://localhost:11434/v1\n key?: string; // openai/cohere provider: name of the env var holding the bearer token, if any\n languages?: string[]; // owner-declared model languages; enables the language-fit check where no card exists\n}\n\nexport const DEFAULT_EMBED_MODEL = 'minishlab/potion-retrieval-32M';\n\n// A named, self-contained file-selection scope. include/exclude are globby patterns resolved\n// relative to the config file. No inheritance between presets -- every field a preset wants\n// it declares itself. Presets may overlap freely; they are views, not partitions.\nexport interface Preset {\n include: string[];\n exclude?: string[];\n k?: number; // result count for `search` scoped to this preset; default 10\n // Exhaustive when present: every signal this preset's searches compose, each mapped to\n // its RRF weight (a finite number > 0; presence is enablement). Absent means every signal\n // whose prerequisite holds, each at weight 1 -- words always, links when the feature is on,\n // vectors when the tree names a model. A layer searched for exact wording (ingested sources,\n // archives, generated output) declares {\"words\":1,\"links\":1} and costs no embedding at all.\n signals?: SignalWeights;\n where?: string; // standing SQL condition against frontmatter alias `f`\n}\n\n// A saved `search` invocation: `sense <name>` runs like\n// `sense search <search> [--preset] [--include] [--exclude] [--where] [--k]`.\nexport interface SavedSearch {\n search: string;\n preset?: string;\n include?: string[];\n exclude?: string[];\n where?: string;\n k?: number;\n}\n\n// An entry names the verb it runs, one to one with the commands: { sql } runs like\n// `sense sql`, { search } like `sense search`.\nexport type SavedQuery = { sql: string } | SavedSearch;\n\nexport interface Config {\n // Editor-only pointer to schema.json; never read by sense.\n $schema?: string;\n version?: number;\n // File selection, index-time, and per-preset search defaults. A file is indexed iff any\n // preset's include/exclude covers it (union). `default` is used when a command names no preset.\n presets: Record<string, Preset>;\n // Global feature defaults; absent block or key means enabled. embed is not a member here --\n // see the `embed` block.\n features?: { links?: boolean; sections?: boolean; tags?: boolean; rank?: boolean };\n // Names the embedding model. Present means every indexed file gets vectors; absent means none do.\n embed?: EmbedConfig;\n // Settings for the `content` FTS5 table. `tokenize` decides which languages the tree can be\n // word-searched in at all; changing it rebuilds the index.\n content?: { tokenize?: string };\n queries: Record<string, SavedQuery>;\n}\n\nexport interface ResolvedConfig extends Config {\n baseDir: string;\n configPath: string | null;\n // Keys the file declares that this build does not read; cli reports them.\n unknownKeys?: string[];\n // Set when loadConfig auto-migrated the file on disk; cli reports it.\n migratedFrom?: number;\n}\n\nexport interface SearchOverrides {\n preset?: string;\n k?: number;\n where?: string;\n include?: string[];\n exclude?: string[];\n noExclude?: boolean; // drop the preset's exclude for this command; --exclude still applies\n}\n\nexport interface EffectiveSearch {\n presetName: string;\n k: number;\n where?: string;\n include: string[];\n exclude?: string[];\n signals: SignalWeights; // this preset's effective signal weights\n}\n"],"names":["CONFIG_FILENAME","STATE_DIR","SUPPORTED_CONFIG_VERSION","FEATURE_NAMES","DEFAULT_EMBED_MODEL"],"mappings":"AAEA,OAAO,MAAMA,kBAAkB,oBAAoB;AACnD,OAAO,MAAMC,YAAY,SAAS;AAElC,mGAAmG;AACnG,OAAO,MAAMC,2BAA2B,EAAE;AAE1C,6GAA6G;AAC7G,iGAAiG;AACjG,gFAAgF;AAChF,OAAO,MAAMC,gBAAgB;IAAC;IAAS;IAAY;IAAQ;IAAQ;CAAQ,CAAU;AAarF,OAAO,MAAMC,sBAAsB,iCAAiC"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/config/types.ts"],"sourcesContent":["import type { SignalWeights } from './signals.ts';\n\nexport const CONFIG_FILENAME = 'sense.config.json';\nexport const STATE_DIR = '.sense';\n\n// Highest sense.config.json `version` this build understands. Older versions auto-migrate on load.\nexport const SUPPORTED_CONFIG_VERSION = 5;\n\n// Each feature owns its tables, parse-time extraction, and reconcile step; commands degrade when one is off.\n// links/sections/tags/rank are opt-out toggles in the top-level `features` block; embed is not a\n// member of that block -- it is on exactly when an `embed` block names a model.\nexport const FEATURE_NAMES = ['links', 'sections', 'tags', 'rank', 'embed'] as const;\nexport type FeatureName = (typeof FEATURE_NAMES)[number];\n\n// `embed` names the model and gives the tree vectors at all; a preset's `signals` says which\n// engines that scope's searches use.\nexport interface EmbedConfig {\n model?: string;\n provider?: 'static' | 'openai' | 'cohere';\n url?: string; // openai provider: OpenAI-compatible base URL, e.g. http://localhost:11434/v1\n key?: string; // openai/cohere provider: name of the env var holding the bearer token, if any\n languages?: string[]; // owner-declared model languages; enables the language-fit check where no card exists\n chunkTokens?: number; // chunk size ceiling in estimated tokens for small-context models; default 500\n}\n\nexport const DEFAULT_EMBED_MODEL = 'minishlab/potion-retrieval-32M';\n\n// A named, self-contained file-selection scope. include/exclude are globby patterns resolved\n// relative to the config file. No inheritance between presets -- every field a preset wants\n// it declares itself. Presets may overlap freely; they are views, not partitions.\nexport interface Preset {\n include: string[];\n exclude?: string[];\n k?: number; // result count for `search` scoped to this preset; default 10\n // Exhaustive when present: every signal this preset's searches compose, each mapped to\n // its RRF weight (a finite number > 0; presence is enablement). Absent means every signal\n // whose prerequisite holds, each at weight 1 -- words always, links when the feature is on,\n // vectors when the tree names a model. A layer searched for exact wording (ingested sources,\n // archives, generated output) declares {\"words\":1,\"links\":1} and costs no embedding at all.\n signals?: SignalWeights;\n where?: string; // standing SQL condition against frontmatter alias `f`\n}\n\n// A saved `search` invocation: `sense <name>` runs like\n// `sense search <search> [--preset] [--include] [--exclude] [--where] [--k]`.\nexport interface SavedSearch {\n search: string;\n preset?: string;\n include?: string[];\n exclude?: string[];\n where?: string;\n k?: number;\n}\n\n// An entry names the verb it runs, one to one with the commands: { sql } runs like\n// `sense sql`, { search } like `sense search`.\nexport type SavedQuery = { sql: string } | SavedSearch;\n\nexport interface Config {\n // Editor-only pointer to schema.json; never read by sense.\n $schema?: string;\n version?: number;\n // File selection, index-time, and per-preset search defaults. A file is indexed iff any\n // preset's include/exclude covers it (union). `default` is used when a command names no preset.\n presets: Record<string, Preset>;\n // Global feature defaults; absent block or key means enabled. embed is not a member here --\n // see the `embed` block.\n features?: { links?: boolean; sections?: boolean; tags?: boolean; rank?: boolean };\n // Names the embedding model. Present means every indexed file gets vectors; absent means none do.\n embed?: EmbedConfig;\n // Settings for the `content` FTS5 table. `tokenize` decides which languages the tree can be\n // word-searched in at all; changing it rebuilds the index.\n content?: { tokenize?: string };\n queries: Record<string, SavedQuery>;\n}\n\nexport interface ResolvedConfig extends Config {\n baseDir: string;\n configPath: string | null;\n // Keys the file declares that this build does not read; cli reports them.\n unknownKeys?: string[];\n // Set when loadConfig auto-migrated the file on disk; cli reports it.\n migratedFrom?: number;\n}\n\nexport interface SearchOverrides {\n preset?: string;\n k?: number;\n where?: string;\n include?: string[];\n exclude?: string[];\n noExclude?: boolean; // drop the preset's exclude for this command; --exclude still applies\n}\n\nexport interface EffectiveSearch {\n presetName: string;\n k: number;\n where?: string;\n include: string[];\n exclude?: string[];\n signals: SignalWeights; // this preset's effective signal weights\n}\n"],"names":["CONFIG_FILENAME","STATE_DIR","SUPPORTED_CONFIG_VERSION","FEATURE_NAMES","DEFAULT_EMBED_MODEL"],"mappings":"AAEA,OAAO,MAAMA,kBAAkB,oBAAoB;AACnD,OAAO,MAAMC,YAAY,SAAS;AAElC,mGAAmG;AACnG,OAAO,MAAMC,2BAA2B,EAAE;AAE1C,6GAA6G;AAC7G,iGAAiG;AACjG,gFAAgF;AAChF,OAAO,MAAMC,gBAAgB;IAAC;IAAS;IAAY;IAAQ;IAAQ;CAAQ,CAAU;AAcrF,OAAO,MAAMC,sBAAsB,iCAAiC"}
@@ -1,4 +1,5 @@
1
1
  import type { Config } from './types.js';
2
+ export declare const KNOWN_EMBED_KEYS: Set<string>;
2
3
  export declare function unknownConfigKeys(cfg: Record<string, unknown>): string[];
3
4
  export declare function validateLegacyScan(parsed: unknown, configPath: string): void;
4
5
  export declare function validateConfig(parsed: unknown, configPath: string): Config;
@@ -24,12 +24,15 @@ const KNOWN_FEATURE_KEYS = new Set([
24
24
  'tags',
25
25
  'rank'
26
26
  ]);
27
- const KNOWN_EMBED_KEYS = new Set([
27
+ // Exported so a docs test can assert every key here has a schema.json property -- one source
28
+ // of truth for what the embed block accepts, checked against the other for drift.
29
+ export const KNOWN_EMBED_KEYS = new Set([
28
30
  'model',
29
31
  'provider',
30
32
  'url',
31
33
  'key',
32
- 'languages'
34
+ 'languages',
35
+ 'chunkTokens'
33
36
  ]);
34
37
  const KNOWN_CONTENT_KEYS = new Set([
35
38
  'tokenize'
@@ -121,6 +124,9 @@ function validateEmbedBlock(value, configPath) {
121
124
  if (embed.languages !== undefined && (!Array.isArray(embed.languages) || embed.languages.length === 0 || embed.languages.some((l)=>typeof l !== 'string' || l === ''))) {
122
125
  throw new SenseError('CONFIG_INVALID', `${configPath}: embed.languages must be a non-empty array of language tags (e.g. ["en", "zh"])`);
123
126
  }
127
+ if (embed.chunkTokens !== undefined && (typeof embed.chunkTokens !== 'number' || !Number.isInteger(embed.chunkTokens) || embed.chunkTokens <= 0)) {
128
+ throw new SenseError('CONFIG_INVALID', `${configPath}: embed.chunkTokens must be a positive integer`);
129
+ }
124
130
  }
125
131
  function isSignalName(value) {
126
132
  return typeof value === 'string' && SIGNAL_NAMES.includes(value);
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/config/validate.ts"],"sourcesContent":["import { SenseError } from '../errors.ts';\nimport { SIGNAL_NAMES, SIGNAL_PREREQUISITES, type SignalName } from './signals.ts';\nimport type { Config } from './types.ts';\n\n// Shape check for hand-edited files, so a typo names itself instead of surfacing as a\n// TypeError. Unknown top-level keys warn (forward compat); unknown keys inside a block error.\nconst KNOWN_KEYS = new Set(['$schema', 'version', 'presets', 'features', 'embed', 'content', 'queries']);\nconst KNOWN_PRESET_KEYS = new Set(['include', 'exclude', 'k', 'signals', 'where']);\nconst KNOWN_FEATURE_KEYS = new Set(['links', 'sections', 'tags', 'rank']);\nconst KNOWN_EMBED_KEYS = new Set(['model', 'provider', 'url', 'key', 'languages']);\nconst KNOWN_CONTENT_KEYS = new Set(['tokenize']);\nconst SAVED_SEARCH_KEYS = new Set(['search', 'preset', 'include', 'exclude', 'where', 'k']);\n\nexport function unknownConfigKeys(cfg: Record<string, unknown>): string[] {\n return Object.keys(cfg).filter((k) => !KNOWN_KEYS.has(k));\n}\n\nfunction isNonEmptyStringArray(value: unknown): value is string[] {\n return Array.isArray(value) && value.length > 0 && value.every((g) => typeof g === 'string' && g.length > 0);\n}\n\n// Pre-v3 shape check, just enough for migrateConfig to run safely (it reads scan.include\n// directly). The full v3 shape is checked by validateConfig once migration has produced it.\nexport function validateLegacyScan(parsed: unknown, configPath: string): void {\n const cfg = typeof parsed === 'object' && parsed !== null && !Array.isArray(parsed) ? (parsed as Record<string, unknown>) : null;\n const scan = cfg?.scan as { include?: unknown } | undefined;\n if (!cfg || !scan || !isNonEmptyStringArray(scan.include)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: scan.include must be a non-empty array of glob strings`);\n }\n}\n\n// Only the shape is checked here. Whether the linked SQLite accepts the tokenizer is settled\n// by probing it in db.ts, so this never has to carry a table of which version added what.\nfunction validateContentBlock(value: unknown, configPath: string): void {\n if (typeof value !== 'object' || value === null || Array.isArray(value)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: content must be an object of { tokenize?: string }`);\n }\n const block = value as Record<string, unknown>;\n const unknown = Object.keys(block).filter((k) => !KNOWN_CONTENT_KEYS.has(k));\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: content has unknown key(s) ${unknown.join(', ')}; content takes tokenize`);\n }\n if (block.tokenize !== undefined && (typeof block.tokenize !== 'string' || block.tokenize.trim() === '')) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: content.tokenize must be a non-empty string, e.g. \"trigram\" or \"unicode61 tokenchars '-_'\"`);\n }\n}\n\nfunction validateFeaturesBlock(value: unknown, configPath: string): void {\n if (typeof value !== 'object' || value === null || Array.isArray(value)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: features must be an object of name -> boolean`);\n }\n const block = value as Record<string, unknown>;\n const unknown = Object.keys(block).filter((k) => !KNOWN_FEATURE_KEYS.has(k));\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: features has unknown key(s) ${unknown.join(', ')}; embed is not a features key -- see the top-level \"embed\" block`);\n }\n for (const [name, v] of Object.entries(block)) {\n if (typeof v !== 'boolean') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: features.${name} must be a boolean`);\n }\n }\n}\n\nfunction validateEmbedBlock(value: unknown, configPath: string): void {\n if (typeof value !== 'object' || value === null || Array.isArray(value)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed must be an object of { model?, provider?: \"static\"|\"openai\"|\"cohere\", url?, key? }`);\n }\n const embed = value as Record<string, unknown>;\n const unknown = Object.keys(embed).filter((k) => !KNOWN_EMBED_KEYS.has(k));\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed has unknown key(s) ${unknown.join(', ')}`);\n }\n if (embed.model !== undefined && typeof embed.model !== 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.model must be a string`);\n }\n // The block exists to name a model; provider settings without one would read as \"vectors\n // configured\" while embedding stays off, which is the ambiguity v4 removed.\n if (embed.model === undefined || embed.model === '') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.model is required when the \"embed\" block is present (it is what turns vectors on); remove the block to index without vectors`);\n }\n if (embed.provider !== undefined && embed.provider !== 'static' && embed.provider !== 'openai' && embed.provider !== 'cohere') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.provider must be \"static\", \"openai\", or \"cohere\"`);\n }\n if (embed.url !== undefined && typeof embed.url !== 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.url must be a string`);\n }\n if (embed.key !== undefined && typeof embed.key !== 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.key must be a string`);\n }\n // openai has no default host, so a missing url would otherwise only surface at the first\n // vector search; a config error belongs at config time.\n if (embed.provider === 'openai' && (embed.url === undefined || embed.url === '')) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.provider \"openai\" requires embed.url (e.g. \"http://localhost:11434/v1\" for Ollama)`);\n }\n if (embed.languages !== undefined && (!Array.isArray(embed.languages) || embed.languages.length === 0 || embed.languages.some((l) => typeof l !== 'string' || l === ''))) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.languages must be a non-empty array of language tags (e.g. [\"en\", \"zh\"])`);\n }\n}\n\nfunction isSignalName(value: unknown): value is SignalName {\n return typeof value === 'string' && (SIGNAL_NAMES as readonly string[]).includes(value);\n}\n\nfunction isSignalWeight(value: unknown): value is number {\n return typeof value === 'number' && Number.isFinite(value) && value > 0;\n}\n\n// Shape (an object of signal name -> weight) plus the one cross-signal prerequisite that\n// doesn't need the top-level embed block: links is seeded by word-match rows, so declaring it\n// without words is always empty.\nfunction validatePresetSignals(name: string, value: unknown, configPath: string): void {\n if (typeof value !== 'object' || value === null || Array.isArray(value) || Object.keys(value).length === 0) {\n throw new SenseError(\n 'CONFIG_INVALID',\n `${configPath}: presets.${name}.signals must be a non-empty object of signal name -> weight (a finite number > 0), naming ${SIGNAL_NAMES.join(', ')} -- a preset that searches nothing is a mistake; name at least \"words\", or omit \"signals\" to use every signal whose prerequisites hold at weight 1`\n );\n }\n const entries = Object.entries(value as Record<string, unknown>);\n const unknown = entries.filter(([k]) => !isSignalName(k)).map(([k]) => k);\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.signals names unknown signal(s) ${unknown.join(', ')}; valid signals are ${SIGNAL_NAMES.join(', ')}`);\n }\n for (const [signal, weight] of entries) {\n if (!isSignalWeight(weight)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.signals.${signal} must be a finite number > 0`);\n }\n }\n const names = new Set(entries.map(([k]) => k));\n for (const [signal, requires] of Object.entries(SIGNAL_PREREQUISITES) as Array<[SignalName, SignalName]>) {\n if (names.has(signal) && !names.has(requires)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.signals has \"${signal}\" without \"${requires}\"; add \"${requires}\" to presets.${name}.signals, or drop \"${signal}\"`);\n }\n }\n}\n\nfunction validatePreset(name: string, value: unknown, configPath: string): void {\n if (typeof value !== 'object' || value === null || Array.isArray(value)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name} must be an object`);\n }\n const preset = value as Record<string, unknown>;\n const unknown = Object.keys(preset).filter((k) => !KNOWN_PRESET_KEYS.has(k));\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name} has unknown key(s) ${unknown.join(', ')}; a preset takes include, exclude, k, signals, where`);\n }\n if (!isNonEmptyStringArray(preset.include)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.include must be a non-empty array of glob strings`);\n }\n if (preset.exclude !== undefined && !isNonEmptyStringArray(preset.exclude)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.exclude must be a non-empty array of glob strings`);\n }\n if (preset.signals !== undefined) validatePresetSignals(name, preset.signals, configPath);\n if (preset.k !== undefined && (typeof preset.k !== 'number' || !Number.isInteger(preset.k) || preset.k <= 0)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.k must be a positive integer`);\n }\n if (preset.where !== undefined && typeof preset.where !== 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.where must be a SQL condition string`);\n }\n}\n\n// A queries.<name> entry: { sql } or a saved search { search, preset?, include?, exclude?, where?, k? }.\nfunction validateSavedQuery(name: string, value: unknown, configPath: string): void {\n // A bare string used to mean SQL. It now fails rather than being inferred: an entry says\n // which of the two verbs it runs, the same choice the CLI makes explicit.\n if (typeof value === 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} must say which verb it runs: { \"sql\": ${JSON.stringify(value.length > 40 ? `${value.slice(0, 37)}...` : value)} } to run it as SQL, or { \"search\": \"...\" } for a ranked search`);\n }\n if (typeof value !== 'object' || value === null || Array.isArray(value)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} must be { sql } or { search, preset?, include?, exclude?, where?, k? }`);\n }\n const entry = value as Record<string, unknown>;\n if ('sql' in entry) {\n const unknown = Object.keys(entry).filter((k) => k !== 'sql');\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} has unknown key(s) ${unknown.join(', ')}; a { sql } query takes only sql`);\n }\n if (typeof entry.sql !== 'string' || entry.sql.trim() === '') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.sql must be a non-empty string`);\n }\n return;\n }\n if ('search' in entry) {\n const unknown = Object.keys(entry).filter((k) => !SAVED_SEARCH_KEYS.has(k));\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} has unknown key(s) ${unknown.join(', ')}; a saved search takes search, preset, include, exclude, where, k`);\n }\n // A saved entry saves a question; a scope without a question is just flags.\n if (typeof entry.search !== 'string' || entry.search.trim() === '') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.search must be non-empty text`);\n }\n if (entry.preset !== undefined && (typeof entry.preset !== 'string' || entry.preset.length === 0)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.preset must be a preset name`);\n }\n if (entry.include !== undefined && !isNonEmptyStringArray(entry.include)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.include must be a non-empty array of glob strings`);\n }\n if (entry.exclude !== undefined && !isNonEmptyStringArray(entry.exclude)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.exclude must be a non-empty array of glob strings`);\n }\n if (entry.where !== undefined && typeof entry.where !== 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.where must be a SQL condition string`);\n }\n if (entry.k !== undefined && (typeof entry.k !== 'number' || !Number.isInteger(entry.k) || entry.k <= 0)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.k must be a positive integer`);\n }\n return;\n }\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} must be { sql } or { search, preset?, include?, exclude?, where?, k? }`);\n}\n\nexport function validateConfig(parsed: unknown, configPath: string): Config {\n if (typeof parsed !== 'object' || parsed === null || Array.isArray(parsed)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: config must be a JSON object`);\n }\n const cfg = parsed as Record<string, unknown>;\n\n // `checks` is rejected by name, not warned: silence would hide that assertions are gone.\n if (cfg.checks !== undefined) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: checks was removed in v3 -- sense check no longer asserts on saved queries; a returned row set is the reader's judgment`);\n }\n\n const presets = cfg.presets as Record<string, unknown> | undefined;\n if (!presets || typeof presets !== 'object' || Array.isArray(presets) || Object.keys(presets).length === 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets must be a non-empty object of preset name -> { include, exclude?, k?, signals?, where? }`);\n }\n for (const [name, value] of Object.entries(presets)) {\n validatePreset(name, value, configPath);\n }\n if (presets.default === undefined) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets must include a \"default\" preset`);\n }\n\n if (cfg.queries === undefined) cfg.queries = {};\n if (typeof cfg.queries !== 'object' || cfg.queries === null || Array.isArray(cfg.queries)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries must be an object of name -> { sql } or { search }`);\n }\n const queries = cfg.queries as Record<string, unknown>;\n for (const [name, value] of Object.entries(queries)) {\n validateSavedQuery(name, value, configPath);\n }\n\n if (cfg.features !== undefined) {\n validateFeaturesBlock(cfg.features, configPath);\n }\n if (cfg.embed !== undefined) {\n validateEmbedBlock(cfg.embed, configPath);\n }\n if (cfg.content !== undefined) {\n validateContentBlock(cfg.content, configPath);\n }\n\n // vectors' prerequisite is a named model, checked here once embed is known valid (or absent).\n if (cfg.embed === undefined) {\n for (const [name, value] of Object.entries(presets)) {\n const signals = (value as Record<string, unknown>).signals as Record<string, unknown> | undefined;\n if (signals && 'vectors' in signals) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.signals includes \"vectors\", but no \"embed\" block names a model; add one, or drop \"vectors\" from presets.${name}.signals`);\n }\n }\n }\n\n return cfg as unknown as Config;\n}\n"],"names":["SenseError","SIGNAL_NAMES","SIGNAL_PREREQUISITES","KNOWN_KEYS","Set","KNOWN_PRESET_KEYS","KNOWN_FEATURE_KEYS","KNOWN_EMBED_KEYS","KNOWN_CONTENT_KEYS","SAVED_SEARCH_KEYS","unknownConfigKeys","cfg","Object","keys","filter","k","has","isNonEmptyStringArray","value","Array","isArray","length","every","g","validateLegacyScan","parsed","configPath","scan","include","validateContentBlock","block","unknown","join","tokenize","undefined","trim","validateFeaturesBlock","name","v","entries","validateEmbedBlock","embed","model","provider","url","key","languages","some","l","isSignalName","includes","isSignalWeight","Number","isFinite","validatePresetSignals","map","signal","weight","names","requires","validatePreset","preset","exclude","signals","isInteger","where","validateSavedQuery","JSON","stringify","slice","entry","sql","search","validateConfig","checks","presets","default","queries","features","content"],"mappings":"AAAA,SAASA,UAAU,QAAQ,eAAe;AAC1C,SAASC,YAAY,EAAEC,oBAAoB,QAAyB,eAAe;AAGnF,sFAAsF;AACtF,8FAA8F;AAC9F,MAAMC,aAAa,IAAIC,IAAI;IAAC;IAAW;IAAW;IAAW;IAAY;IAAS;IAAW;CAAU;AACvG,MAAMC,oBAAoB,IAAID,IAAI;IAAC;IAAW;IAAW;IAAK;IAAW;CAAQ;AACjF,MAAME,qBAAqB,IAAIF,IAAI;IAAC;IAAS;IAAY;IAAQ;CAAO;AACxE,MAAMG,mBAAmB,IAAIH,IAAI;IAAC;IAAS;IAAY;IAAO;IAAO;CAAY;AACjF,MAAMI,qBAAqB,IAAIJ,IAAI;IAAC;CAAW;AAC/C,MAAMK,oBAAoB,IAAIL,IAAI;IAAC;IAAU;IAAU;IAAW;IAAW;IAAS;CAAI;AAE1F,OAAO,SAASM,kBAAkBC,GAA4B;IAC5D,OAAOC,OAAOC,IAAI,CAACF,KAAKG,MAAM,CAAC,CAACC,IAAM,CAACZ,WAAWa,GAAG,CAACD;AACxD;AAEA,SAASE,sBAAsBC,KAAc;IAC3C,OAAOC,MAAMC,OAAO,CAACF,UAAUA,MAAMG,MAAM,GAAG,KAAKH,MAAMI,KAAK,CAAC,CAACC,IAAM,OAAOA,MAAM,YAAYA,EAAEF,MAAM,GAAG;AAC5G;AAEA,yFAAyF;AACzF,4FAA4F;AAC5F,OAAO,SAASG,mBAAmBC,MAAe,EAAEC,UAAkB;IACpE,MAAMf,MAAM,OAAOc,WAAW,YAAYA,WAAW,QAAQ,CAACN,MAAMC,OAAO,CAACK,UAAWA,SAAqC;IAC5H,MAAME,OAAOhB,gBAAAA,0BAAAA,IAAKgB,IAAI;IACtB,IAAI,CAAChB,OAAO,CAACgB,QAAQ,CAACV,sBAAsBU,KAAKC,OAAO,GAAG;QACzD,MAAM,IAAI5B,WAAW,kBAAkB,GAAG0B,WAAW,wDAAwD,CAAC;IAChH;AACF;AAEA,6FAA6F;AAC7F,0FAA0F;AAC1F,SAASG,qBAAqBX,KAAc,EAAEQ,UAAkB;IAC9D,IAAI,OAAOR,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,QAAQ;QACvE,MAAM,IAAIlB,WAAW,kBAAkB,GAAG0B,WAAW,oDAAoD,CAAC;IAC5G;IACA,MAAMI,QAAQZ;IACd,MAAMa,UAAUnB,OAAOC,IAAI,CAACiB,OAAOhB,MAAM,CAAC,CAACC,IAAM,CAACP,mBAAmBQ,GAAG,CAACD;IACzE,IAAIgB,QAAQV,MAAM,GAAG,GAAG;QACtB,MAAM,IAAIrB,WAAW,kBAAkB,GAAG0B,WAAW,6BAA6B,EAAEK,QAAQC,IAAI,CAAC,MAAM,wBAAwB,CAAC;IAClI;IACA,IAAIF,MAAMG,QAAQ,KAAKC,aAAc,CAAA,OAAOJ,MAAMG,QAAQ,KAAK,YAAYH,MAAMG,QAAQ,CAACE,IAAI,OAAO,EAAC,GAAI;QACxG,MAAM,IAAInC,WAAW,kBAAkB,GAAG0B,WAAW,4FAA4F,CAAC;IACpJ;AACF;AAEA,SAASU,sBAAsBlB,KAAc,EAAEQ,UAAkB;IAC/D,IAAI,OAAOR,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,QAAQ;QACvE,MAAM,IAAIlB,WAAW,kBAAkB,GAAG0B,WAAW,+CAA+C,CAAC;IACvG;IACA,MAAMI,QAAQZ;IACd,MAAMa,UAAUnB,OAAOC,IAAI,CAACiB,OAAOhB,MAAM,CAAC,CAACC,IAAM,CAACT,mBAAmBU,GAAG,CAACD;IACzE,IAAIgB,QAAQV,MAAM,GAAG,GAAG;QACtB,MAAM,IAAIrB,WAAW,kBAAkB,GAAG0B,WAAW,8BAA8B,EAAEK,QAAQC,IAAI,CAAC,MAAM,gEAAgE,CAAC;IAC3K;IACA,KAAK,MAAM,CAACK,MAAMC,EAAE,IAAI1B,OAAO2B,OAAO,CAACT,OAAQ;QAC7C,IAAI,OAAOQ,MAAM,WAAW;YAC1B,MAAM,IAAItC,WAAW,kBAAkB,GAAG0B,WAAW,WAAW,EAAEW,KAAK,kBAAkB,CAAC;QAC5F;IACF;AACF;AAEA,SAASG,mBAAmBtB,KAAc,EAAEQ,UAAkB;IAC5D,IAAI,OAAOR,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,QAAQ;QACvE,MAAM,IAAIlB,WAAW,kBAAkB,GAAG0B,WAAW,0FAA0F,CAAC;IAClJ;IACA,MAAMe,QAAQvB;IACd,MAAMa,UAAUnB,OAAOC,IAAI,CAAC4B,OAAO3B,MAAM,CAAC,CAACC,IAAM,CAACR,iBAAiBS,GAAG,CAACD;IACvE,IAAIgB,QAAQV,MAAM,GAAG,GAAG;QACtB,MAAM,IAAIrB,WAAW,kBAAkB,GAAG0B,WAAW,2BAA2B,EAAEK,QAAQC,IAAI,CAAC,OAAO;IACxG;IACA,IAAIS,MAAMC,KAAK,KAAKR,aAAa,OAAOO,MAAMC,KAAK,KAAK,UAAU;QAChE,MAAM,IAAI1C,WAAW,kBAAkB,GAAG0B,WAAW,8BAA8B,CAAC;IACtF;IACA,yFAAyF;IACzF,4EAA4E;IAC5E,IAAIe,MAAMC,KAAK,KAAKR,aAAaO,MAAMC,KAAK,KAAK,IAAI;QACnD,MAAM,IAAI1C,WAAW,kBAAkB,GAAG0B,WAAW,oIAAoI,CAAC;IAC5L;IACA,IAAIe,MAAME,QAAQ,KAAKT,aAAaO,MAAME,QAAQ,KAAK,YAAYF,MAAME,QAAQ,KAAK,YAAYF,MAAME,QAAQ,KAAK,UAAU;QAC7H,MAAM,IAAI3C,WAAW,kBAAkB,GAAG0B,WAAW,wDAAwD,CAAC;IAChH;IACA,IAAIe,MAAMG,GAAG,KAAKV,aAAa,OAAOO,MAAMG,GAAG,KAAK,UAAU;QAC5D,MAAM,IAAI5C,WAAW,kBAAkB,GAAG0B,WAAW,4BAA4B,CAAC;IACpF;IACA,IAAIe,MAAMI,GAAG,KAAKX,aAAa,OAAOO,MAAMI,GAAG,KAAK,UAAU;QAC5D,MAAM,IAAI7C,WAAW,kBAAkB,GAAG0B,WAAW,4BAA4B,CAAC;IACpF;IACA,yFAAyF;IACzF,wDAAwD;IACxD,IAAIe,MAAME,QAAQ,KAAK,YAAaF,CAAAA,MAAMG,GAAG,KAAKV,aAAaO,MAAMG,GAAG,KAAK,EAAC,GAAI;QAChF,MAAM,IAAI5C,WAAW,kBAAkB,GAAG0B,WAAW,0FAA0F,CAAC;IAClJ;IACA,IAAIe,MAAMK,SAAS,KAAKZ,aAAc,CAAA,CAACf,MAAMC,OAAO,CAACqB,MAAMK,SAAS,KAAKL,MAAMK,SAAS,CAACzB,MAAM,KAAK,KAAKoB,MAAMK,SAAS,CAACC,IAAI,CAAC,CAACC,IAAM,OAAOA,MAAM,YAAYA,MAAM,GAAE,GAAI;QACxK,MAAM,IAAIhD,WAAW,kBAAkB,GAAG0B,WAAW,gFAAgF,CAAC;IACxI;AACF;AAEA,SAASuB,aAAa/B,KAAc;IAClC,OAAO,OAAOA,UAAU,YAAY,AAACjB,aAAmCiD,QAAQ,CAAChC;AACnF;AAEA,SAASiC,eAAejC,KAAc;IACpC,OAAO,OAAOA,UAAU,YAAYkC,OAAOC,QAAQ,CAACnC,UAAUA,QAAQ;AACxE;AAEA,yFAAyF;AACzF,8FAA8F;AAC9F,iCAAiC;AACjC,SAASoC,sBAAsBjB,IAAY,EAAEnB,KAAc,EAAEQ,UAAkB;IAC7E,IAAI,OAAOR,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,UAAUN,OAAOC,IAAI,CAACK,OAAOG,MAAM,KAAK,GAAG;QAC1G,MAAM,IAAIrB,WACR,kBACA,GAAG0B,WAAW,UAAU,EAAEW,KAAK,2FAA2F,EAAEpC,aAAa+B,IAAI,CAAC,MAAM,kJAAkJ,CAAC;IAE3S;IACA,MAAMO,UAAU3B,OAAO2B,OAAO,CAACrB;IAC/B,MAAMa,UAAUQ,QAAQzB,MAAM,CAAC,CAAC,CAACC,EAAE,GAAK,CAACkC,aAAalC,IAAIwC,GAAG,CAAC,CAAC,CAACxC,EAAE,GAAKA;IACvE,IAAIgB,QAAQV,MAAM,GAAG,GAAG;QACtB,MAAM,IAAIrB,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,iCAAiC,EAAEN,QAAQC,IAAI,CAAC,MAAM,oBAAoB,EAAE/B,aAAa+B,IAAI,CAAC,OAAO;IAC7K;IACA,KAAK,MAAM,CAACwB,QAAQC,OAAO,IAAIlB,QAAS;QACtC,IAAI,CAACY,eAAeM,SAAS;YAC3B,MAAM,IAAIzD,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,SAAS,EAAEmB,OAAO,4BAA4B,CAAC;QACvH;IACF;IACA,MAAME,QAAQ,IAAItD,IAAImC,QAAQgB,GAAG,CAAC,CAAC,CAACxC,EAAE,GAAKA;IAC3C,KAAK,MAAM,CAACyC,QAAQG,SAAS,IAAI/C,OAAO2B,OAAO,CAACrC,sBAA0D;QACxG,IAAIwD,MAAM1C,GAAG,CAACwC,WAAW,CAACE,MAAM1C,GAAG,CAAC2C,WAAW;YAC7C,MAAM,IAAI3D,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,cAAc,EAAEmB,OAAO,WAAW,EAAEG,SAAS,QAAQ,EAAEA,SAAS,aAAa,EAAEtB,KAAK,mBAAmB,EAAEmB,OAAO,CAAC,CAAC;QAC1L;IACF;AACF;AAEA,SAASI,eAAevB,IAAY,EAAEnB,KAAc,EAAEQ,UAAkB;IACtE,IAAI,OAAOR,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,QAAQ;QACvE,MAAM,IAAIlB,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,kBAAkB,CAAC;IAC3F;IACA,MAAMwB,SAAS3C;IACf,MAAMa,UAAUnB,OAAOC,IAAI,CAACgD,QAAQ/C,MAAM,CAAC,CAACC,IAAM,CAACV,kBAAkBW,GAAG,CAACD;IACzE,IAAIgB,QAAQV,MAAM,GAAG,GAAG;QACtB,MAAM,IAAIrB,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,oBAAoB,EAAEN,QAAQC,IAAI,CAAC,MAAM,oDAAoD,CAAC;IACtK;IACA,IAAI,CAACf,sBAAsB4C,OAAOjC,OAAO,GAAG;QAC1C,MAAM,IAAI5B,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,kDAAkD,CAAC;IAC3H;IACA,IAAIwB,OAAOC,OAAO,KAAK5B,aAAa,CAACjB,sBAAsB4C,OAAOC,OAAO,GAAG;QAC1E,MAAM,IAAI9D,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,kDAAkD,CAAC;IAC3H;IACA,IAAIwB,OAAOE,OAAO,KAAK7B,WAAWoB,sBAAsBjB,MAAMwB,OAAOE,OAAO,EAAErC;IAC9E,IAAImC,OAAO9C,CAAC,KAAKmB,aAAc,CAAA,OAAO2B,OAAO9C,CAAC,KAAK,YAAY,CAACqC,OAAOY,SAAS,CAACH,OAAO9C,CAAC,KAAK8C,OAAO9C,CAAC,IAAI,CAAA,GAAI;QAC5G,MAAM,IAAIf,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,6BAA6B,CAAC;IACtG;IACA,IAAIwB,OAAOI,KAAK,KAAK/B,aAAa,OAAO2B,OAAOI,KAAK,KAAK,UAAU;QAClE,MAAM,IAAIjE,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,qCAAqC,CAAC;IAC9G;AACF;AAEA,yGAAyG;AACzG,SAAS6B,mBAAmB7B,IAAY,EAAEnB,KAAc,EAAEQ,UAAkB;IAC1E,yFAAyF;IACzF,0EAA0E;IAC1E,IAAI,OAAOR,UAAU,UAAU;QAC7B,MAAM,IAAIlB,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,uCAAuC,EAAE8B,KAAKC,SAAS,CAAClD,MAAMG,MAAM,GAAG,KAAK,GAAGH,MAAMmD,KAAK,CAAC,GAAG,IAAI,GAAG,CAAC,GAAGnD,OAAO,+DAA+D,CAAC;IACxP;IACA,IAAI,OAAOA,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,QAAQ;QACvE,MAAM,IAAIlB,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,uEAAuE,CAAC;IAChJ;IACA,MAAMiC,QAAQpD;IACd,IAAI,SAASoD,OAAO;QAClB,MAAMvC,UAAUnB,OAAOC,IAAI,CAACyD,OAAOxD,MAAM,CAAC,CAACC,IAAMA,MAAM;QACvD,IAAIgB,QAAQV,MAAM,GAAG,GAAG;YACtB,MAAM,IAAIrB,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,oBAAoB,EAAEN,QAAQC,IAAI,CAAC,MAAM,gCAAgC,CAAC;QAClJ;QACA,IAAI,OAAOsC,MAAMC,GAAG,KAAK,YAAYD,MAAMC,GAAG,CAACpC,IAAI,OAAO,IAAI;YAC5D,MAAM,IAAInC,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,+BAA+B,CAAC;QACxG;QACA;IACF;IACA,IAAI,YAAYiC,OAAO;QACrB,MAAMvC,UAAUnB,OAAOC,IAAI,CAACyD,OAAOxD,MAAM,CAAC,CAACC,IAAM,CAACN,kBAAkBO,GAAG,CAACD;QACxE,IAAIgB,QAAQV,MAAM,GAAG,GAAG;YACtB,MAAM,IAAIrB,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,oBAAoB,EAAEN,QAAQC,IAAI,CAAC,MAAM,iEAAiE,CAAC;QACnL;QACA,4EAA4E;QAC5E,IAAI,OAAOsC,MAAME,MAAM,KAAK,YAAYF,MAAME,MAAM,CAACrC,IAAI,OAAO,IAAI;YAClE,MAAM,IAAInC,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,8BAA8B,CAAC;QACvG;QACA,IAAIiC,MAAMT,MAAM,KAAK3B,aAAc,CAAA,OAAOoC,MAAMT,MAAM,KAAK,YAAYS,MAAMT,MAAM,CAACxC,MAAM,KAAK,CAAA,GAAI;YACjG,MAAM,IAAIrB,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,6BAA6B,CAAC;QACtG;QACA,IAAIiC,MAAM1C,OAAO,KAAKM,aAAa,CAACjB,sBAAsBqD,MAAM1C,OAAO,GAAG;YACxE,MAAM,IAAI5B,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,kDAAkD,CAAC;QAC3H;QACA,IAAIiC,MAAMR,OAAO,KAAK5B,aAAa,CAACjB,sBAAsBqD,MAAMR,OAAO,GAAG;YACxE,MAAM,IAAI9D,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,kDAAkD,CAAC;QAC3H;QACA,IAAIiC,MAAML,KAAK,KAAK/B,aAAa,OAAOoC,MAAML,KAAK,KAAK,UAAU;YAChE,MAAM,IAAIjE,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,qCAAqC,CAAC;QAC9G;QACA,IAAIiC,MAAMvD,CAAC,KAAKmB,aAAc,CAAA,OAAOoC,MAAMvD,CAAC,KAAK,YAAY,CAACqC,OAAOY,SAAS,CAACM,MAAMvD,CAAC,KAAKuD,MAAMvD,CAAC,IAAI,CAAA,GAAI;YACxG,MAAM,IAAIf,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,6BAA6B,CAAC;QACtG;QACA;IACF;IACA,MAAM,IAAIrC,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,uEAAuE,CAAC;AAChJ;AAEA,OAAO,SAASoC,eAAehD,MAAe,EAAEC,UAAkB;IAChE,IAAI,OAAOD,WAAW,YAAYA,WAAW,QAAQN,MAAMC,OAAO,CAACK,SAAS;QAC1E,MAAM,IAAIzB,WAAW,kBAAkB,GAAG0B,WAAW,8BAA8B,CAAC;IACtF;IACA,MAAMf,MAAMc;IAEZ,yFAAyF;IACzF,IAAId,IAAI+D,MAAM,KAAKxC,WAAW;QAC5B,MAAM,IAAIlC,WAAW,kBAAkB,GAAG0B,WAAW,yHAAyH,CAAC;IACjL;IAEA,MAAMiD,UAAUhE,IAAIgE,OAAO;IAC3B,IAAI,CAACA,WAAW,OAAOA,YAAY,YAAYxD,MAAMC,OAAO,CAACuD,YAAY/D,OAAOC,IAAI,CAAC8D,SAAStD,MAAM,KAAK,GAAG;QAC1G,MAAM,IAAIrB,WAAW,kBAAkB,GAAG0B,WAAW,kGAAkG,CAAC;IAC1J;IACA,KAAK,MAAM,CAACW,MAAMnB,MAAM,IAAIN,OAAO2B,OAAO,CAACoC,SAAU;QACnDf,eAAevB,MAAMnB,OAAOQ;IAC9B;IACA,IAAIiD,QAAQC,OAAO,KAAK1C,WAAW;QACjC,MAAM,IAAIlC,WAAW,kBAAkB,GAAG0B,WAAW,yCAAyC,CAAC;IACjG;IAEA,IAAIf,IAAIkE,OAAO,KAAK3C,WAAWvB,IAAIkE,OAAO,GAAG,CAAC;IAC9C,IAAI,OAAOlE,IAAIkE,OAAO,KAAK,YAAYlE,IAAIkE,OAAO,KAAK,QAAQ1D,MAAMC,OAAO,CAACT,IAAIkE,OAAO,GAAG;QACzF,MAAM,IAAI7E,WAAW,kBAAkB,GAAG0B,WAAW,4DAA4D,CAAC;IACpH;IACA,MAAMmD,UAAUlE,IAAIkE,OAAO;IAC3B,KAAK,MAAM,CAACxC,MAAMnB,MAAM,IAAIN,OAAO2B,OAAO,CAACsC,SAAU;QACnDX,mBAAmB7B,MAAMnB,OAAOQ;IAClC;IAEA,IAAIf,IAAImE,QAAQ,KAAK5C,WAAW;QAC9BE,sBAAsBzB,IAAImE,QAAQ,EAAEpD;IACtC;IACA,IAAIf,IAAI8B,KAAK,KAAKP,WAAW;QAC3BM,mBAAmB7B,IAAI8B,KAAK,EAAEf;IAChC;IACA,IAAIf,IAAIoE,OAAO,KAAK7C,WAAW;QAC7BL,qBAAqBlB,IAAIoE,OAAO,EAAErD;IACpC;IAEA,8FAA8F;IAC9F,IAAIf,IAAI8B,KAAK,KAAKP,WAAW;QAC3B,KAAK,MAAM,CAACG,MAAMnB,MAAM,IAAIN,OAAO2B,OAAO,CAACoC,SAAU;YACnD,MAAMZ,UAAU,AAAC7C,MAAkC6C,OAAO;YAC1D,IAAIA,WAAW,aAAaA,SAAS;gBACnC,MAAM,IAAI/D,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,yGAAyG,EAAEA,KAAK,QAAQ,CAAC;YACjM;QACF;IACF;IAEA,OAAO1B;AACT"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/config/validate.ts"],"sourcesContent":["import { SenseError } from '../errors.ts';\nimport { SIGNAL_NAMES, SIGNAL_PREREQUISITES, type SignalName } from './signals.ts';\nimport type { Config } from './types.ts';\n\n// Shape check for hand-edited files, so a typo names itself instead of surfacing as a\n// TypeError. Unknown top-level keys warn (forward compat); unknown keys inside a block error.\nconst KNOWN_KEYS = new Set(['$schema', 'version', 'presets', 'features', 'embed', 'content', 'queries']);\nconst KNOWN_PRESET_KEYS = new Set(['include', 'exclude', 'k', 'signals', 'where']);\nconst KNOWN_FEATURE_KEYS = new Set(['links', 'sections', 'tags', 'rank']);\n// Exported so a docs test can assert every key here has a schema.json property -- one source\n// of truth for what the embed block accepts, checked against the other for drift.\nexport const KNOWN_EMBED_KEYS = new Set(['model', 'provider', 'url', 'key', 'languages', 'chunkTokens']);\nconst KNOWN_CONTENT_KEYS = new Set(['tokenize']);\nconst SAVED_SEARCH_KEYS = new Set(['search', 'preset', 'include', 'exclude', 'where', 'k']);\n\nexport function unknownConfigKeys(cfg: Record<string, unknown>): string[] {\n return Object.keys(cfg).filter((k) => !KNOWN_KEYS.has(k));\n}\n\nfunction isNonEmptyStringArray(value: unknown): value is string[] {\n return Array.isArray(value) && value.length > 0 && value.every((g) => typeof g === 'string' && g.length > 0);\n}\n\n// Pre-v3 shape check, just enough for migrateConfig to run safely (it reads scan.include\n// directly). The full v3 shape is checked by validateConfig once migration has produced it.\nexport function validateLegacyScan(parsed: unknown, configPath: string): void {\n const cfg = typeof parsed === 'object' && parsed !== null && !Array.isArray(parsed) ? (parsed as Record<string, unknown>) : null;\n const scan = cfg?.scan as { include?: unknown } | undefined;\n if (!cfg || !scan || !isNonEmptyStringArray(scan.include)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: scan.include must be a non-empty array of glob strings`);\n }\n}\n\n// Only the shape is checked here. Whether the linked SQLite accepts the tokenizer is settled\n// by probing it in db.ts, so this never has to carry a table of which version added what.\nfunction validateContentBlock(value: unknown, configPath: string): void {\n if (typeof value !== 'object' || value === null || Array.isArray(value)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: content must be an object of { tokenize?: string }`);\n }\n const block = value as Record<string, unknown>;\n const unknown = Object.keys(block).filter((k) => !KNOWN_CONTENT_KEYS.has(k));\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: content has unknown key(s) ${unknown.join(', ')}; content takes tokenize`);\n }\n if (block.tokenize !== undefined && (typeof block.tokenize !== 'string' || block.tokenize.trim() === '')) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: content.tokenize must be a non-empty string, e.g. \"trigram\" or \"unicode61 tokenchars '-_'\"`);\n }\n}\n\nfunction validateFeaturesBlock(value: unknown, configPath: string): void {\n if (typeof value !== 'object' || value === null || Array.isArray(value)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: features must be an object of name -> boolean`);\n }\n const block = value as Record<string, unknown>;\n const unknown = Object.keys(block).filter((k) => !KNOWN_FEATURE_KEYS.has(k));\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: features has unknown key(s) ${unknown.join(', ')}; embed is not a features key -- see the top-level \"embed\" block`);\n }\n for (const [name, v] of Object.entries(block)) {\n if (typeof v !== 'boolean') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: features.${name} must be a boolean`);\n }\n }\n}\n\nfunction validateEmbedBlock(value: unknown, configPath: string): void {\n if (typeof value !== 'object' || value === null || Array.isArray(value)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed must be an object of { model?, provider?: \"static\"|\"openai\"|\"cohere\", url?, key? }`);\n }\n const embed = value as Record<string, unknown>;\n const unknown = Object.keys(embed).filter((k) => !KNOWN_EMBED_KEYS.has(k));\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed has unknown key(s) ${unknown.join(', ')}`);\n }\n if (embed.model !== undefined && typeof embed.model !== 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.model must be a string`);\n }\n // The block exists to name a model; provider settings without one would read as \"vectors\n // configured\" while embedding stays off, which is the ambiguity v4 removed.\n if (embed.model === undefined || embed.model === '') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.model is required when the \"embed\" block is present (it is what turns vectors on); remove the block to index without vectors`);\n }\n if (embed.provider !== undefined && embed.provider !== 'static' && embed.provider !== 'openai' && embed.provider !== 'cohere') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.provider must be \"static\", \"openai\", or \"cohere\"`);\n }\n if (embed.url !== undefined && typeof embed.url !== 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.url must be a string`);\n }\n if (embed.key !== undefined && typeof embed.key !== 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.key must be a string`);\n }\n // openai has no default host, so a missing url would otherwise only surface at the first\n // vector search; a config error belongs at config time.\n if (embed.provider === 'openai' && (embed.url === undefined || embed.url === '')) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.provider \"openai\" requires embed.url (e.g. \"http://localhost:11434/v1\" for Ollama)`);\n }\n if (embed.languages !== undefined && (!Array.isArray(embed.languages) || embed.languages.length === 0 || embed.languages.some((l) => typeof l !== 'string' || l === ''))) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.languages must be a non-empty array of language tags (e.g. [\"en\", \"zh\"])`);\n }\n if (embed.chunkTokens !== undefined && (typeof embed.chunkTokens !== 'number' || !Number.isInteger(embed.chunkTokens) || embed.chunkTokens <= 0)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.chunkTokens must be a positive integer`);\n }\n}\n\nfunction isSignalName(value: unknown): value is SignalName {\n return typeof value === 'string' && (SIGNAL_NAMES as readonly string[]).includes(value);\n}\n\nfunction isSignalWeight(value: unknown): value is number {\n return typeof value === 'number' && Number.isFinite(value) && value > 0;\n}\n\n// Shape (an object of signal name -> weight) plus the one cross-signal prerequisite that\n// doesn't need the top-level embed block: links is seeded by word-match rows, so declaring it\n// without words is always empty.\nfunction validatePresetSignals(name: string, value: unknown, configPath: string): void {\n if (typeof value !== 'object' || value === null || Array.isArray(value) || Object.keys(value).length === 0) {\n throw new SenseError(\n 'CONFIG_INVALID',\n `${configPath}: presets.${name}.signals must be a non-empty object of signal name -> weight (a finite number > 0), naming ${SIGNAL_NAMES.join(', ')} -- a preset that searches nothing is a mistake; name at least \"words\", or omit \"signals\" to use every signal whose prerequisites hold at weight 1`\n );\n }\n const entries = Object.entries(value as Record<string, unknown>);\n const unknown = entries.filter(([k]) => !isSignalName(k)).map(([k]) => k);\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.signals names unknown signal(s) ${unknown.join(', ')}; valid signals are ${SIGNAL_NAMES.join(', ')}`);\n }\n for (const [signal, weight] of entries) {\n if (!isSignalWeight(weight)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.signals.${signal} must be a finite number > 0`);\n }\n }\n const names = new Set(entries.map(([k]) => k));\n for (const [signal, requires] of Object.entries(SIGNAL_PREREQUISITES) as Array<[SignalName, SignalName]>) {\n if (names.has(signal) && !names.has(requires)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.signals has \"${signal}\" without \"${requires}\"; add \"${requires}\" to presets.${name}.signals, or drop \"${signal}\"`);\n }\n }\n}\n\nfunction validatePreset(name: string, value: unknown, configPath: string): void {\n if (typeof value !== 'object' || value === null || Array.isArray(value)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name} must be an object`);\n }\n const preset = value as Record<string, unknown>;\n const unknown = Object.keys(preset).filter((k) => !KNOWN_PRESET_KEYS.has(k));\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name} has unknown key(s) ${unknown.join(', ')}; a preset takes include, exclude, k, signals, where`);\n }\n if (!isNonEmptyStringArray(preset.include)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.include must be a non-empty array of glob strings`);\n }\n if (preset.exclude !== undefined && !isNonEmptyStringArray(preset.exclude)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.exclude must be a non-empty array of glob strings`);\n }\n if (preset.signals !== undefined) validatePresetSignals(name, preset.signals, configPath);\n if (preset.k !== undefined && (typeof preset.k !== 'number' || !Number.isInteger(preset.k) || preset.k <= 0)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.k must be a positive integer`);\n }\n if (preset.where !== undefined && typeof preset.where !== 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.where must be a SQL condition string`);\n }\n}\n\n// A queries.<name> entry: { sql } or a saved search { search, preset?, include?, exclude?, where?, k? }.\nfunction validateSavedQuery(name: string, value: unknown, configPath: string): void {\n // A bare string used to mean SQL. It now fails rather than being inferred: an entry says\n // which of the two verbs it runs, the same choice the CLI makes explicit.\n if (typeof value === 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} must say which verb it runs: { \"sql\": ${JSON.stringify(value.length > 40 ? `${value.slice(0, 37)}...` : value)} } to run it as SQL, or { \"search\": \"...\" } for a ranked search`);\n }\n if (typeof value !== 'object' || value === null || Array.isArray(value)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} must be { sql } or { search, preset?, include?, exclude?, where?, k? }`);\n }\n const entry = value as Record<string, unknown>;\n if ('sql' in entry) {\n const unknown = Object.keys(entry).filter((k) => k !== 'sql');\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} has unknown key(s) ${unknown.join(', ')}; a { sql } query takes only sql`);\n }\n if (typeof entry.sql !== 'string' || entry.sql.trim() === '') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.sql must be a non-empty string`);\n }\n return;\n }\n if ('search' in entry) {\n const unknown = Object.keys(entry).filter((k) => !SAVED_SEARCH_KEYS.has(k));\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} has unknown key(s) ${unknown.join(', ')}; a saved search takes search, preset, include, exclude, where, k`);\n }\n // A saved entry saves a question; a scope without a question is just flags.\n if (typeof entry.search !== 'string' || entry.search.trim() === '') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.search must be non-empty text`);\n }\n if (entry.preset !== undefined && (typeof entry.preset !== 'string' || entry.preset.length === 0)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.preset must be a preset name`);\n }\n if (entry.include !== undefined && !isNonEmptyStringArray(entry.include)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.include must be a non-empty array of glob strings`);\n }\n if (entry.exclude !== undefined && !isNonEmptyStringArray(entry.exclude)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.exclude must be a non-empty array of glob strings`);\n }\n if (entry.where !== undefined && typeof entry.where !== 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.where must be a SQL condition string`);\n }\n if (entry.k !== undefined && (typeof entry.k !== 'number' || !Number.isInteger(entry.k) || entry.k <= 0)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.k must be a positive integer`);\n }\n return;\n }\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} must be { sql } or { search, preset?, include?, exclude?, where?, k? }`);\n}\n\nexport function validateConfig(parsed: unknown, configPath: string): Config {\n if (typeof parsed !== 'object' || parsed === null || Array.isArray(parsed)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: config must be a JSON object`);\n }\n const cfg = parsed as Record<string, unknown>;\n\n // `checks` is rejected by name, not warned: silence would hide that assertions are gone.\n if (cfg.checks !== undefined) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: checks was removed in v3 -- sense check no longer asserts on saved queries; a returned row set is the reader's judgment`);\n }\n\n const presets = cfg.presets as Record<string, unknown> | undefined;\n if (!presets || typeof presets !== 'object' || Array.isArray(presets) || Object.keys(presets).length === 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets must be a non-empty object of preset name -> { include, exclude?, k?, signals?, where? }`);\n }\n for (const [name, value] of Object.entries(presets)) {\n validatePreset(name, value, configPath);\n }\n if (presets.default === undefined) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets must include a \"default\" preset`);\n }\n\n if (cfg.queries === undefined) cfg.queries = {};\n if (typeof cfg.queries !== 'object' || cfg.queries === null || Array.isArray(cfg.queries)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries must be an object of name -> { sql } or { search }`);\n }\n const queries = cfg.queries as Record<string, unknown>;\n for (const [name, value] of Object.entries(queries)) {\n validateSavedQuery(name, value, configPath);\n }\n\n if (cfg.features !== undefined) {\n validateFeaturesBlock(cfg.features, configPath);\n }\n if (cfg.embed !== undefined) {\n validateEmbedBlock(cfg.embed, configPath);\n }\n if (cfg.content !== undefined) {\n validateContentBlock(cfg.content, configPath);\n }\n\n // vectors' prerequisite is a named model, checked here once embed is known valid (or absent).\n if (cfg.embed === undefined) {\n for (const [name, value] of Object.entries(presets)) {\n const signals = (value as Record<string, unknown>).signals as Record<string, unknown> | undefined;\n if (signals && 'vectors' in signals) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.signals includes \"vectors\", but no \"embed\" block names a model; add one, or drop \"vectors\" from presets.${name}.signals`);\n }\n }\n }\n\n return cfg as unknown as Config;\n}\n"],"names":["SenseError","SIGNAL_NAMES","SIGNAL_PREREQUISITES","KNOWN_KEYS","Set","KNOWN_PRESET_KEYS","KNOWN_FEATURE_KEYS","KNOWN_EMBED_KEYS","KNOWN_CONTENT_KEYS","SAVED_SEARCH_KEYS","unknownConfigKeys","cfg","Object","keys","filter","k","has","isNonEmptyStringArray","value","Array","isArray","length","every","g","validateLegacyScan","parsed","configPath","scan","include","validateContentBlock","block","unknown","join","tokenize","undefined","trim","validateFeaturesBlock","name","v","entries","validateEmbedBlock","embed","model","provider","url","key","languages","some","l","chunkTokens","Number","isInteger","isSignalName","includes","isSignalWeight","isFinite","validatePresetSignals","map","signal","weight","names","requires","validatePreset","preset","exclude","signals","where","validateSavedQuery","JSON","stringify","slice","entry","sql","search","validateConfig","checks","presets","default","queries","features","content"],"mappings":"AAAA,SAASA,UAAU,QAAQ,eAAe;AAC1C,SAASC,YAAY,EAAEC,oBAAoB,QAAyB,eAAe;AAGnF,sFAAsF;AACtF,8FAA8F;AAC9F,MAAMC,aAAa,IAAIC,IAAI;IAAC;IAAW;IAAW;IAAW;IAAY;IAAS;IAAW;CAAU;AACvG,MAAMC,oBAAoB,IAAID,IAAI;IAAC;IAAW;IAAW;IAAK;IAAW;CAAQ;AACjF,MAAME,qBAAqB,IAAIF,IAAI;IAAC;IAAS;IAAY;IAAQ;CAAO;AACxE,6FAA6F;AAC7F,kFAAkF;AAClF,OAAO,MAAMG,mBAAmB,IAAIH,IAAI;IAAC;IAAS;IAAY;IAAO;IAAO;IAAa;CAAc,EAAE;AACzG,MAAMI,qBAAqB,IAAIJ,IAAI;IAAC;CAAW;AAC/C,MAAMK,oBAAoB,IAAIL,IAAI;IAAC;IAAU;IAAU;IAAW;IAAW;IAAS;CAAI;AAE1F,OAAO,SAASM,kBAAkBC,GAA4B;IAC5D,OAAOC,OAAOC,IAAI,CAACF,KAAKG,MAAM,CAAC,CAACC,IAAM,CAACZ,WAAWa,GAAG,CAACD;AACxD;AAEA,SAASE,sBAAsBC,KAAc;IAC3C,OAAOC,MAAMC,OAAO,CAACF,UAAUA,MAAMG,MAAM,GAAG,KAAKH,MAAMI,KAAK,CAAC,CAACC,IAAM,OAAOA,MAAM,YAAYA,EAAEF,MAAM,GAAG;AAC5G;AAEA,yFAAyF;AACzF,4FAA4F;AAC5F,OAAO,SAASG,mBAAmBC,MAAe,EAAEC,UAAkB;IACpE,MAAMf,MAAM,OAAOc,WAAW,YAAYA,WAAW,QAAQ,CAACN,MAAMC,OAAO,CAACK,UAAWA,SAAqC;IAC5H,MAAME,OAAOhB,gBAAAA,0BAAAA,IAAKgB,IAAI;IACtB,IAAI,CAAChB,OAAO,CAACgB,QAAQ,CAACV,sBAAsBU,KAAKC,OAAO,GAAG;QACzD,MAAM,IAAI5B,WAAW,kBAAkB,GAAG0B,WAAW,wDAAwD,CAAC;IAChH;AACF;AAEA,6FAA6F;AAC7F,0FAA0F;AAC1F,SAASG,qBAAqBX,KAAc,EAAEQ,UAAkB;IAC9D,IAAI,OAAOR,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,QAAQ;QACvE,MAAM,IAAIlB,WAAW,kBAAkB,GAAG0B,WAAW,oDAAoD,CAAC;IAC5G;IACA,MAAMI,QAAQZ;IACd,MAAMa,UAAUnB,OAAOC,IAAI,CAACiB,OAAOhB,MAAM,CAAC,CAACC,IAAM,CAACP,mBAAmBQ,GAAG,CAACD;IACzE,IAAIgB,QAAQV,MAAM,GAAG,GAAG;QACtB,MAAM,IAAIrB,WAAW,kBAAkB,GAAG0B,WAAW,6BAA6B,EAAEK,QAAQC,IAAI,CAAC,MAAM,wBAAwB,CAAC;IAClI;IACA,IAAIF,MAAMG,QAAQ,KAAKC,aAAc,CAAA,OAAOJ,MAAMG,QAAQ,KAAK,YAAYH,MAAMG,QAAQ,CAACE,IAAI,OAAO,EAAC,GAAI;QACxG,MAAM,IAAInC,WAAW,kBAAkB,GAAG0B,WAAW,4FAA4F,CAAC;IACpJ;AACF;AAEA,SAASU,sBAAsBlB,KAAc,EAAEQ,UAAkB;IAC/D,IAAI,OAAOR,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,QAAQ;QACvE,MAAM,IAAIlB,WAAW,kBAAkB,GAAG0B,WAAW,+CAA+C,CAAC;IACvG;IACA,MAAMI,QAAQZ;IACd,MAAMa,UAAUnB,OAAOC,IAAI,CAACiB,OAAOhB,MAAM,CAAC,CAACC,IAAM,CAACT,mBAAmBU,GAAG,CAACD;IACzE,IAAIgB,QAAQV,MAAM,GAAG,GAAG;QACtB,MAAM,IAAIrB,WAAW,kBAAkB,GAAG0B,WAAW,8BAA8B,EAAEK,QAAQC,IAAI,CAAC,MAAM,gEAAgE,CAAC;IAC3K;IACA,KAAK,MAAM,CAACK,MAAMC,EAAE,IAAI1B,OAAO2B,OAAO,CAACT,OAAQ;QAC7C,IAAI,OAAOQ,MAAM,WAAW;YAC1B,MAAM,IAAItC,WAAW,kBAAkB,GAAG0B,WAAW,WAAW,EAAEW,KAAK,kBAAkB,CAAC;QAC5F;IACF;AACF;AAEA,SAASG,mBAAmBtB,KAAc,EAAEQ,UAAkB;IAC5D,IAAI,OAAOR,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,QAAQ;QACvE,MAAM,IAAIlB,WAAW,kBAAkB,GAAG0B,WAAW,0FAA0F,CAAC;IAClJ;IACA,MAAMe,QAAQvB;IACd,MAAMa,UAAUnB,OAAOC,IAAI,CAAC4B,OAAO3B,MAAM,CAAC,CAACC,IAAM,CAACR,iBAAiBS,GAAG,CAACD;IACvE,IAAIgB,QAAQV,MAAM,GAAG,GAAG;QACtB,MAAM,IAAIrB,WAAW,kBAAkB,GAAG0B,WAAW,2BAA2B,EAAEK,QAAQC,IAAI,CAAC,OAAO;IACxG;IACA,IAAIS,MAAMC,KAAK,KAAKR,aAAa,OAAOO,MAAMC,KAAK,KAAK,UAAU;QAChE,MAAM,IAAI1C,WAAW,kBAAkB,GAAG0B,WAAW,8BAA8B,CAAC;IACtF;IACA,yFAAyF;IACzF,4EAA4E;IAC5E,IAAIe,MAAMC,KAAK,KAAKR,aAAaO,MAAMC,KAAK,KAAK,IAAI;QACnD,MAAM,IAAI1C,WAAW,kBAAkB,GAAG0B,WAAW,oIAAoI,CAAC;IAC5L;IACA,IAAIe,MAAME,QAAQ,KAAKT,aAAaO,MAAME,QAAQ,KAAK,YAAYF,MAAME,QAAQ,KAAK,YAAYF,MAAME,QAAQ,KAAK,UAAU;QAC7H,MAAM,IAAI3C,WAAW,kBAAkB,GAAG0B,WAAW,wDAAwD,CAAC;IAChH;IACA,IAAIe,MAAMG,GAAG,KAAKV,aAAa,OAAOO,MAAMG,GAAG,KAAK,UAAU;QAC5D,MAAM,IAAI5C,WAAW,kBAAkB,GAAG0B,WAAW,4BAA4B,CAAC;IACpF;IACA,IAAIe,MAAMI,GAAG,KAAKX,aAAa,OAAOO,MAAMI,GAAG,KAAK,UAAU;QAC5D,MAAM,IAAI7C,WAAW,kBAAkB,GAAG0B,WAAW,4BAA4B,CAAC;IACpF;IACA,yFAAyF;IACzF,wDAAwD;IACxD,IAAIe,MAAME,QAAQ,KAAK,YAAaF,CAAAA,MAAMG,GAAG,KAAKV,aAAaO,MAAMG,GAAG,KAAK,EAAC,GAAI;QAChF,MAAM,IAAI5C,WAAW,kBAAkB,GAAG0B,WAAW,0FAA0F,CAAC;IAClJ;IACA,IAAIe,MAAMK,SAAS,KAAKZ,aAAc,CAAA,CAACf,MAAMC,OAAO,CAACqB,MAAMK,SAAS,KAAKL,MAAMK,SAAS,CAACzB,MAAM,KAAK,KAAKoB,MAAMK,SAAS,CAACC,IAAI,CAAC,CAACC,IAAM,OAAOA,MAAM,YAAYA,MAAM,GAAE,GAAI;QACxK,MAAM,IAAIhD,WAAW,kBAAkB,GAAG0B,WAAW,gFAAgF,CAAC;IACxI;IACA,IAAIe,MAAMQ,WAAW,KAAKf,aAAc,CAAA,OAAOO,MAAMQ,WAAW,KAAK,YAAY,CAACC,OAAOC,SAAS,CAACV,MAAMQ,WAAW,KAAKR,MAAMQ,WAAW,IAAI,CAAA,GAAI;QAChJ,MAAM,IAAIjD,WAAW,kBAAkB,GAAG0B,WAAW,8CAA8C,CAAC;IACtG;AACF;AAEA,SAAS0B,aAAalC,KAAc;IAClC,OAAO,OAAOA,UAAU,YAAY,AAACjB,aAAmCoD,QAAQ,CAACnC;AACnF;AAEA,SAASoC,eAAepC,KAAc;IACpC,OAAO,OAAOA,UAAU,YAAYgC,OAAOK,QAAQ,CAACrC,UAAUA,QAAQ;AACxE;AAEA,yFAAyF;AACzF,8FAA8F;AAC9F,iCAAiC;AACjC,SAASsC,sBAAsBnB,IAAY,EAAEnB,KAAc,EAAEQ,UAAkB;IAC7E,IAAI,OAAOR,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,UAAUN,OAAOC,IAAI,CAACK,OAAOG,MAAM,KAAK,GAAG;QAC1G,MAAM,IAAIrB,WACR,kBACA,GAAG0B,WAAW,UAAU,EAAEW,KAAK,2FAA2F,EAAEpC,aAAa+B,IAAI,CAAC,MAAM,kJAAkJ,CAAC;IAE3S;IACA,MAAMO,UAAU3B,OAAO2B,OAAO,CAACrB;IAC/B,MAAMa,UAAUQ,QAAQzB,MAAM,CAAC,CAAC,CAACC,EAAE,GAAK,CAACqC,aAAarC,IAAI0C,GAAG,CAAC,CAAC,CAAC1C,EAAE,GAAKA;IACvE,IAAIgB,QAAQV,MAAM,GAAG,GAAG;QACtB,MAAM,IAAIrB,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,iCAAiC,EAAEN,QAAQC,IAAI,CAAC,MAAM,oBAAoB,EAAE/B,aAAa+B,IAAI,CAAC,OAAO;IAC7K;IACA,KAAK,MAAM,CAAC0B,QAAQC,OAAO,IAAIpB,QAAS;QACtC,IAAI,CAACe,eAAeK,SAAS;YAC3B,MAAM,IAAI3D,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,SAAS,EAAEqB,OAAO,4BAA4B,CAAC;QACvH;IACF;IACA,MAAME,QAAQ,IAAIxD,IAAImC,QAAQkB,GAAG,CAAC,CAAC,CAAC1C,EAAE,GAAKA;IAC3C,KAAK,MAAM,CAAC2C,QAAQG,SAAS,IAAIjD,OAAO2B,OAAO,CAACrC,sBAA0D;QACxG,IAAI0D,MAAM5C,GAAG,CAAC0C,WAAW,CAACE,MAAM5C,GAAG,CAAC6C,WAAW;YAC7C,MAAM,IAAI7D,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,cAAc,EAAEqB,OAAO,WAAW,EAAEG,SAAS,QAAQ,EAAEA,SAAS,aAAa,EAAExB,KAAK,mBAAmB,EAAEqB,OAAO,CAAC,CAAC;QAC1L;IACF;AACF;AAEA,SAASI,eAAezB,IAAY,EAAEnB,KAAc,EAAEQ,UAAkB;IACtE,IAAI,OAAOR,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,QAAQ;QACvE,MAAM,IAAIlB,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,kBAAkB,CAAC;IAC3F;IACA,MAAM0B,SAAS7C;IACf,MAAMa,UAAUnB,OAAOC,IAAI,CAACkD,QAAQjD,MAAM,CAAC,CAACC,IAAM,CAACV,kBAAkBW,GAAG,CAACD;IACzE,IAAIgB,QAAQV,MAAM,GAAG,GAAG;QACtB,MAAM,IAAIrB,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,oBAAoB,EAAEN,QAAQC,IAAI,CAAC,MAAM,oDAAoD,CAAC;IACtK;IACA,IAAI,CAACf,sBAAsB8C,OAAOnC,OAAO,GAAG;QAC1C,MAAM,IAAI5B,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,kDAAkD,CAAC;IAC3H;IACA,IAAI0B,OAAOC,OAAO,KAAK9B,aAAa,CAACjB,sBAAsB8C,OAAOC,OAAO,GAAG;QAC1E,MAAM,IAAIhE,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,kDAAkD,CAAC;IAC3H;IACA,IAAI0B,OAAOE,OAAO,KAAK/B,WAAWsB,sBAAsBnB,MAAM0B,OAAOE,OAAO,EAAEvC;IAC9E,IAAIqC,OAAOhD,CAAC,KAAKmB,aAAc,CAAA,OAAO6B,OAAOhD,CAAC,KAAK,YAAY,CAACmC,OAAOC,SAAS,CAACY,OAAOhD,CAAC,KAAKgD,OAAOhD,CAAC,IAAI,CAAA,GAAI;QAC5G,MAAM,IAAIf,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,6BAA6B,CAAC;IACtG;IACA,IAAI0B,OAAOG,KAAK,KAAKhC,aAAa,OAAO6B,OAAOG,KAAK,KAAK,UAAU;QAClE,MAAM,IAAIlE,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,qCAAqC,CAAC;IAC9G;AACF;AAEA,yGAAyG;AACzG,SAAS8B,mBAAmB9B,IAAY,EAAEnB,KAAc,EAAEQ,UAAkB;IAC1E,yFAAyF;IACzF,0EAA0E;IAC1E,IAAI,OAAOR,UAAU,UAAU;QAC7B,MAAM,IAAIlB,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,uCAAuC,EAAE+B,KAAKC,SAAS,CAACnD,MAAMG,MAAM,GAAG,KAAK,GAAGH,MAAMoD,KAAK,CAAC,GAAG,IAAI,GAAG,CAAC,GAAGpD,OAAO,+DAA+D,CAAC;IACxP;IACA,IAAI,OAAOA,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,QAAQ;QACvE,MAAM,IAAIlB,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,uEAAuE,CAAC;IAChJ;IACA,MAAMkC,QAAQrD;IACd,IAAI,SAASqD,OAAO;QAClB,MAAMxC,UAAUnB,OAAOC,IAAI,CAAC0D,OAAOzD,MAAM,CAAC,CAACC,IAAMA,MAAM;QACvD,IAAIgB,QAAQV,MAAM,GAAG,GAAG;YACtB,MAAM,IAAIrB,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,oBAAoB,EAAEN,QAAQC,IAAI,CAAC,MAAM,gCAAgC,CAAC;QAClJ;QACA,IAAI,OAAOuC,MAAMC,GAAG,KAAK,YAAYD,MAAMC,GAAG,CAACrC,IAAI,OAAO,IAAI;YAC5D,MAAM,IAAInC,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,+BAA+B,CAAC;QACxG;QACA;IACF;IACA,IAAI,YAAYkC,OAAO;QACrB,MAAMxC,UAAUnB,OAAOC,IAAI,CAAC0D,OAAOzD,MAAM,CAAC,CAACC,IAAM,CAACN,kBAAkBO,GAAG,CAACD;QACxE,IAAIgB,QAAQV,MAAM,GAAG,GAAG;YACtB,MAAM,IAAIrB,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,oBAAoB,EAAEN,QAAQC,IAAI,CAAC,MAAM,iEAAiE,CAAC;QACnL;QACA,4EAA4E;QAC5E,IAAI,OAAOuC,MAAME,MAAM,KAAK,YAAYF,MAAME,MAAM,CAACtC,IAAI,OAAO,IAAI;YAClE,MAAM,IAAInC,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,8BAA8B,CAAC;QACvG;QACA,IAAIkC,MAAMR,MAAM,KAAK7B,aAAc,CAAA,OAAOqC,MAAMR,MAAM,KAAK,YAAYQ,MAAMR,MAAM,CAAC1C,MAAM,KAAK,CAAA,GAAI;YACjG,MAAM,IAAIrB,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,6BAA6B,CAAC;QACtG;QACA,IAAIkC,MAAM3C,OAAO,KAAKM,aAAa,CAACjB,sBAAsBsD,MAAM3C,OAAO,GAAG;YACxE,MAAM,IAAI5B,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,kDAAkD,CAAC;QAC3H;QACA,IAAIkC,MAAMP,OAAO,KAAK9B,aAAa,CAACjB,sBAAsBsD,MAAMP,OAAO,GAAG;YACxE,MAAM,IAAIhE,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,kDAAkD,CAAC;QAC3H;QACA,IAAIkC,MAAML,KAAK,KAAKhC,aAAa,OAAOqC,MAAML,KAAK,KAAK,UAAU;YAChE,MAAM,IAAIlE,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,qCAAqC,CAAC;QAC9G;QACA,IAAIkC,MAAMxD,CAAC,KAAKmB,aAAc,CAAA,OAAOqC,MAAMxD,CAAC,KAAK,YAAY,CAACmC,OAAOC,SAAS,CAACoB,MAAMxD,CAAC,KAAKwD,MAAMxD,CAAC,IAAI,CAAA,GAAI;YACxG,MAAM,IAAIf,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,6BAA6B,CAAC;QACtG;QACA;IACF;IACA,MAAM,IAAIrC,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,uEAAuE,CAAC;AAChJ;AAEA,OAAO,SAASqC,eAAejD,MAAe,EAAEC,UAAkB;IAChE,IAAI,OAAOD,WAAW,YAAYA,WAAW,QAAQN,MAAMC,OAAO,CAACK,SAAS;QAC1E,MAAM,IAAIzB,WAAW,kBAAkB,GAAG0B,WAAW,8BAA8B,CAAC;IACtF;IACA,MAAMf,MAAMc;IAEZ,yFAAyF;IACzF,IAAId,IAAIgE,MAAM,KAAKzC,WAAW;QAC5B,MAAM,IAAIlC,WAAW,kBAAkB,GAAG0B,WAAW,yHAAyH,CAAC;IACjL;IAEA,MAAMkD,UAAUjE,IAAIiE,OAAO;IAC3B,IAAI,CAACA,WAAW,OAAOA,YAAY,YAAYzD,MAAMC,OAAO,CAACwD,YAAYhE,OAAOC,IAAI,CAAC+D,SAASvD,MAAM,KAAK,GAAG;QAC1G,MAAM,IAAIrB,WAAW,kBAAkB,GAAG0B,WAAW,kGAAkG,CAAC;IAC1J;IACA,KAAK,MAAM,CAACW,MAAMnB,MAAM,IAAIN,OAAO2B,OAAO,CAACqC,SAAU;QACnDd,eAAezB,MAAMnB,OAAOQ;IAC9B;IACA,IAAIkD,QAAQC,OAAO,KAAK3C,WAAW;QACjC,MAAM,IAAIlC,WAAW,kBAAkB,GAAG0B,WAAW,yCAAyC,CAAC;IACjG;IAEA,IAAIf,IAAImE,OAAO,KAAK5C,WAAWvB,IAAImE,OAAO,GAAG,CAAC;IAC9C,IAAI,OAAOnE,IAAImE,OAAO,KAAK,YAAYnE,IAAImE,OAAO,KAAK,QAAQ3D,MAAMC,OAAO,CAACT,IAAImE,OAAO,GAAG;QACzF,MAAM,IAAI9E,WAAW,kBAAkB,GAAG0B,WAAW,4DAA4D,CAAC;IACpH;IACA,MAAMoD,UAAUnE,IAAImE,OAAO;IAC3B,KAAK,MAAM,CAACzC,MAAMnB,MAAM,IAAIN,OAAO2B,OAAO,CAACuC,SAAU;QACnDX,mBAAmB9B,MAAMnB,OAAOQ;IAClC;IAEA,IAAIf,IAAIoE,QAAQ,KAAK7C,WAAW;QAC9BE,sBAAsBzB,IAAIoE,QAAQ,EAAErD;IACtC;IACA,IAAIf,IAAI8B,KAAK,KAAKP,WAAW;QAC3BM,mBAAmB7B,IAAI8B,KAAK,EAAEf;IAChC;IACA,IAAIf,IAAIqE,OAAO,KAAK9C,WAAW;QAC7BL,qBAAqBlB,IAAIqE,OAAO,EAAEtD;IACpC;IAEA,8FAA8F;IAC9F,IAAIf,IAAI8B,KAAK,KAAKP,WAAW;QAC3B,KAAK,MAAM,CAACG,MAAMnB,MAAM,IAAIN,OAAO2B,OAAO,CAACqC,SAAU;YACnD,MAAMX,UAAU,AAAC/C,MAAkC+C,OAAO;YAC1D,IAAIA,WAAW,aAAaA,SAAS;gBACnC,MAAM,IAAIjE,WAAW,kBAAkB,GAAG0B,WAAW,UAAU,EAAEW,KAAK,yGAAyG,EAAEA,KAAK,QAAQ,CAAC;YACjM;QACF;IACF;IAEA,OAAO1B;AACT"}
@@ -1,7 +1,7 @@
1
1
  import { DatabaseSync } from 'node:sqlite';
2
2
  import type { ResolvedConfig } from '../config/index.js';
3
3
  export declare const DB_FILENAME = "cache.db";
4
- export declare const SCHEMA_VERSION = "17";
4
+ export declare const SCHEMA_VERSION = "18";
5
5
  export interface OpenResult {
6
6
  db: DatabaseSync;
7
7
  cfg: ResolvedConfig;
@@ -5,14 +5,14 @@ import { join } from 'node:path';
5
5
  import { DatabaseSync } from 'node:sqlite';
6
6
  import { contentTokenize, featureSignature, STATE_DIR } from '../config/index.js';
7
7
  import { SenseError } from '../errors.js';
8
- import { activeFeatures } from '../features/index.js';
9
- import { registerFunctions } from '../sql-functions.js';
8
+ import { activeFeatures, FEATURES } from '../features/index.js';
10
9
  import { changedSignatureKeys, embedIdentityAdopted, rebuildContentTable, reconcile, signatureDiff } from './reconcile.js';
11
10
  import { getMeta, setMeta } from './shared.js';
11
+ import { registerFunctions } from './sql-functions.js';
12
12
  export const DB_FILENAME = 'cache.db';
13
13
  // Cache shape version, independent of the config's own `version`. Bumping it rebuilds
14
14
  // existing trees on first query.
15
- export const SCHEMA_VERSION = '17';
15
+ export const SCHEMA_VERSION = '18';
16
16
  // Stemming is English-only, but the segmentation underneath it is what decides coverage:
17
17
  // unicode61 splits on spaces, so a language written without them (Chinese, Japanese, Thai)
18
18
  // indexes a whole run as one token and word search finds nothing. `content.tokenize` is how
@@ -61,7 +61,7 @@ function ensureSchema(db, cfg, tokenize) {
61
61
  db.exec('CREATE INDEX IF NOT EXISTS preset_files_preset ON preset_files(preset)');
62
62
  for (const feature of activeFeatures(cfg))feature.schema(db);
63
63
  if (getMeta(db, 'schema_version') === null) setMeta(db, 'schema_version', SCHEMA_VERSION);
64
- if (getMeta(db, 'features') === null) setMeta(db, 'features', featureSignature(cfg));
64
+ if (getMeta(db, 'features') === null) setMeta(db, 'features', featureSignature(cfg, FEATURES));
65
65
  }
66
66
  export function docCount(db) {
67
67
  const row = db.prepare('SELECT COUNT(*) AS n FROM frontmatter').get();
@@ -88,7 +88,7 @@ export function open(cfg) {
88
88
  // old cache can't be patched incrementally -- rebuild instead (cheap: nothing expensive lives here).
89
89
  const version = getMeta(db, 'schema_version');
90
90
  const features = getMeta(db, 'features');
91
- const wantFeatures = featureSignature(cfg);
91
+ const wantFeatures = featureSignature(cfg, FEATURES);
92
92
  let tokenizeOnlyRebuild = false;
93
93
  if (version !== null && version !== SCHEMA_VERSION || features !== null && features !== wantFeatures) {
94
94
  // Indexing derives from presets, so a config edit rebuilding the cache must say so and