sensemaking 0.27.0 → 0.28.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (364) hide show
  1. package/bin/cli.js +16 -7
  2. package/dist/cjs/chunk/extract.js.map +1 -1
  3. package/dist/cjs/chunk/group.js.map +1 -1
  4. package/dist/cjs/chunk/index.js.map +1 -1
  5. package/dist/cjs/chunk/parse.js.map +1 -1
  6. package/dist/cjs/chunk/parser.js.map +1 -1
  7. package/dist/cjs/chunk/tokens.js.map +1 -1
  8. package/dist/cjs/chunk/version.js.map +1 -1
  9. package/dist/cjs/cli/build.js.map +1 -1
  10. package/dist/cjs/cli/constants.d.cts +2 -0
  11. package/dist/cjs/cli/constants.d.ts +2 -0
  12. package/dist/cjs/cli/constants.js +21 -0
  13. package/dist/cjs/cli/constants.js.map +1 -0
  14. package/dist/cjs/cli/download.js.map +1 -1
  15. package/dist/cjs/cli/index.js.map +1 -1
  16. package/dist/cjs/cli/init.js.map +1 -1
  17. package/dist/cjs/cli/map.js.map +1 -1
  18. package/dist/cjs/cli/named.js +2 -1
  19. package/dist/cjs/cli/named.js.map +1 -1
  20. package/dist/cjs/cli/path.js.map +1 -1
  21. package/dist/cjs/cli/peek.js.map +1 -1
  22. package/dist/cjs/cli/related.js.map +1 -1
  23. package/dist/cjs/cli/search.js.map +1 -1
  24. package/dist/cjs/cli/shared.js +5 -4
  25. package/dist/cjs/cli/shared.js.map +1 -1
  26. package/dist/cjs/cli/sql.js.map +1 -1
  27. package/dist/cjs/cli/status.js.map +1 -1
  28. package/dist/cjs/cli/watch.js.map +1 -1
  29. package/dist/cjs/cli.js +10 -9
  30. package/dist/cjs/cli.js.map +1 -1
  31. package/dist/cjs/commands/index.js.map +1 -1
  32. package/dist/cjs/commands/map.js.map +1 -1
  33. package/dist/cjs/commands/path.js.map +1 -1
  34. package/dist/cjs/commands/peek.js.map +1 -1
  35. package/dist/cjs/commands/related.js.map +1 -1
  36. package/dist/cjs/commands/scope.js.map +1 -1
  37. package/dist/cjs/commands/search.js.map +1 -1
  38. package/dist/cjs/commands/signals.js.map +1 -1
  39. package/dist/cjs/commands/snapshot.js.map +1 -1
  40. package/dist/cjs/commands/status.js.map +1 -1
  41. package/dist/cjs/config/access.js.map +1 -1
  42. package/dist/cjs/config/index.js.map +1 -1
  43. package/dist/cjs/config/load.js.map +1 -1
  44. package/dist/cjs/config/resolve.js.map +1 -1
  45. package/dist/cjs/config/signals.js.map +1 -1
  46. package/dist/cjs/config/types.js.map +1 -1
  47. package/dist/cjs/config/validate.js.map +1 -1
  48. package/dist/cjs/embed/cohere.js.map +1 -1
  49. package/dist/cjs/embed/distribution.js.map +1 -1
  50. package/dist/cjs/embed/endpoint.js.map +1 -1
  51. package/dist/cjs/embed/handoff.js.map +1 -1
  52. package/dist/cjs/embed/http.js.map +1 -1
  53. package/dist/cjs/embed/identity.js.map +1 -1
  54. package/dist/cjs/embed/langfit.js.map +1 -1
  55. package/dist/cjs/embed/languages.js.map +1 -1
  56. package/dist/cjs/embed/openai.js.map +1 -1
  57. package/dist/cjs/embed/query.js.map +1 -1
  58. package/dist/cjs/embed/registry.js.map +1 -1
  59. package/dist/cjs/embed/static.js.map +1 -1
  60. package/dist/cjs/embed/store.js.map +1 -1
  61. package/dist/cjs/embed/types.js.map +1 -1
  62. package/dist/cjs/errors.js.map +1 -1
  63. package/dist/cjs/features/embed.js.map +1 -1
  64. package/dist/cjs/features/fences.js.map +1 -1
  65. package/dist/cjs/features/index.js.map +1 -1
  66. package/dist/cjs/features/links.js.map +1 -1
  67. package/dist/cjs/features/rank.js.map +1 -1
  68. package/dist/cjs/features/sections.js.map +1 -1
  69. package/dist/cjs/features/tags.js.map +1 -1
  70. package/dist/cjs/graph/graph.js.map +1 -1
  71. package/dist/cjs/graph/traverse.js.map +1 -1
  72. package/dist/cjs/index.js.map +1 -1
  73. package/dist/cjs/lib/atomic-write.js.map +1 -1
  74. package/dist/cjs/lib/guarded-tick.js.map +1 -1
  75. package/dist/cjs/lib/serial-query.js.map +1 -1
  76. package/dist/cjs/lib/watch-notification.js.map +1 -1
  77. package/dist/cjs/lib/worker-file.js.map +1 -1
  78. package/dist/cjs/output/column-hint.js.map +1 -1
  79. package/dist/cjs/output/output.js.map +1 -1
  80. package/dist/cjs/output/progress.js.map +1 -1
  81. package/dist/cjs/output/search-error.js.map +1 -1
  82. package/dist/cjs/scan/frontmatter.js.map +1 -1
  83. package/dist/cjs/scan/index.js.map +1 -1
  84. package/dist/cjs/scan/list.js.map +1 -1
  85. package/dist/cjs/scan/pool.js.map +1 -1
  86. package/dist/cjs/scan/reparse.js.map +1 -1
  87. package/dist/cjs/scan/worker-error.js.map +1 -1
  88. package/dist/cjs/store/batch.js.map +1 -1
  89. package/dist/cjs/store/builder.js.map +1 -1
  90. package/dist/cjs/store/cache.js.map +1 -1
  91. package/dist/cjs/store/duckdb/batch.js.map +1 -1
  92. package/dist/cjs/store/duckdb/connection.js.map +1 -1
  93. package/dist/cjs/store/duckdb/engine.d.cts +11 -0
  94. package/dist/cjs/store/duckdb/engine.d.ts +11 -0
  95. package/dist/cjs/store/duckdb/engine.js +193 -0
  96. package/dist/cjs/store/duckdb/engine.js.map +1 -0
  97. package/dist/cjs/store/duckdb/fieldStats.js.map +1 -1
  98. package/dist/cjs/store/duckdb/lexical.d.cts +1 -6
  99. package/dist/cjs/store/duckdb/lexical.d.ts +1 -6
  100. package/dist/cjs/store/duckdb/lexical.js +48 -45
  101. package/dist/cjs/store/duckdb/lexical.js.map +1 -1
  102. package/dist/cjs/store/duckdb/native.d.cts +5 -3
  103. package/dist/cjs/store/duckdb/native.d.ts +5 -3
  104. package/dist/cjs/store/duckdb/native.js +157 -18
  105. package/dist/cjs/store/duckdb/native.js.map +1 -1
  106. package/dist/cjs/store/duckdb/open.js +151 -3
  107. package/dist/cjs/store/duckdb/open.js.map +1 -1
  108. package/dist/cjs/store/duckdb/ordered-bm25.d.cts +3 -16
  109. package/dist/cjs/store/duckdb/ordered-bm25.d.ts +3 -16
  110. package/dist/cjs/store/duckdb/ordered-bm25.js +24 -80
  111. package/dist/cjs/store/duckdb/ordered-bm25.js.map +1 -1
  112. package/dist/cjs/store/duckdb/reconcile.js.map +1 -1
  113. package/dist/cjs/store/duckdb/sql-functions.js.map +1 -1
  114. package/dist/cjs/store/duckdb/store.js.map +1 -1
  115. package/dist/cjs/store/duckdb/vectors.js.map +1 -1
  116. package/dist/cjs/store/embed-scope.js.map +1 -1
  117. package/dist/cjs/store/engine.d.cts +28 -0
  118. package/dist/cjs/store/engine.d.ts +28 -0
  119. package/dist/cjs/store/engine.js +362 -0
  120. package/dist/cjs/store/engine.js.map +1 -0
  121. package/dist/cjs/store/feature-scope.js.map +1 -1
  122. package/dist/cjs/store/index.js.map +1 -1
  123. package/dist/cjs/store/lock-wait.js.map +1 -1
  124. package/dist/cjs/store/native.d.cts +8 -5
  125. package/dist/cjs/store/native.d.ts +8 -5
  126. package/dist/cjs/store/native.js +78 -73
  127. package/dist/cjs/store/native.js.map +1 -1
  128. package/dist/cjs/store/open.js +100 -40
  129. package/dist/cjs/store/open.js.map +1 -1
  130. package/dist/cjs/store/preset-scope.js.map +1 -1
  131. package/dist/cjs/store/readiness.js.map +1 -1
  132. package/dist/cjs/store/reconcile.js.map +1 -1
  133. package/dist/cjs/store/shared.js.map +1 -1
  134. package/dist/cjs/store/signature.js.map +1 -1
  135. package/dist/cjs/store/sql-functions.js.map +1 -1
  136. package/dist/cjs/store/sqlite/connection.js.map +1 -1
  137. package/dist/cjs/store/sqlite/engine.d.cts +12 -0
  138. package/dist/cjs/store/sqlite/engine.d.ts +12 -0
  139. package/dist/cjs/store/sqlite/engine.js +53 -0
  140. package/dist/cjs/store/sqlite/engine.js.map +1 -0
  141. package/dist/cjs/store/sqlite/fieldStats.js.map +1 -1
  142. package/dist/cjs/store/sqlite/lexical.js +3 -4
  143. package/dist/cjs/store/sqlite/lexical.js.map +1 -1
  144. package/dist/cjs/store/sqlite/open.js +119 -6
  145. package/dist/cjs/store/sqlite/open.js.map +1 -1
  146. package/dist/cjs/store/sqlite/reconcile.js.map +1 -1
  147. package/dist/cjs/store/sqlite/sql-functions.js.map +1 -1
  148. package/dist/cjs/store/sqlite/store.js.map +1 -1
  149. package/dist/cjs/store/sqlite/vectors.js.map +1 -1
  150. package/dist/cjs/store/stages.js.map +1 -1
  151. package/dist/cjs/store/transaction.js.map +1 -1
  152. package/dist/cjs/store/turso/connection.js.map +1 -1
  153. package/dist/cjs/store/turso/engine.d.cts +9 -0
  154. package/dist/cjs/store/turso/engine.d.ts +9 -0
  155. package/dist/cjs/store/turso/engine.js +189 -0
  156. package/dist/cjs/store/turso/engine.js.map +1 -0
  157. package/dist/cjs/store/turso/fieldStats.js.map +1 -1
  158. package/dist/cjs/store/turso/lexical-text.js.map +1 -1
  159. package/dist/cjs/store/turso/lexical.js +16 -11
  160. package/dist/cjs/store/turso/lexical.js.map +1 -1
  161. package/dist/cjs/store/turso/native.d.cts +5 -1
  162. package/dist/cjs/store/turso/native.d.ts +5 -1
  163. package/dist/cjs/store/turso/native.js +157 -8
  164. package/dist/cjs/store/turso/native.js.map +1 -1
  165. package/dist/cjs/store/turso/open.js +172 -27
  166. package/dist/cjs/store/turso/open.js.map +1 -1
  167. package/dist/cjs/store/turso/reconcile.d.cts +0 -1
  168. package/dist/cjs/store/turso/reconcile.d.ts +0 -1
  169. package/dist/cjs/store/turso/reconcile.js +22 -23
  170. package/dist/cjs/store/turso/reconcile.js.map +1 -1
  171. package/dist/cjs/store/turso/sql-functions.js.map +1 -1
  172. package/dist/cjs/store/turso/store.js.map +1 -1
  173. package/dist/cjs/store/turso/vectors.js.map +1 -1
  174. package/dist/cjs/store/types.d.cts +2 -0
  175. package/dist/cjs/store/types.d.ts +2 -0
  176. package/dist/cjs/store/types.js.map +1 -1
  177. package/dist/cjs/store/vectors.js.map +1 -1
  178. package/dist/cjs/text/segment.js.map +1 -1
  179. package/dist/cjs/text/strip.js.map +1 -1
  180. package/dist/cjs/watch-claim.js.map +1 -1
  181. package/dist/cjs/watch.js.map +1 -1
  182. package/dist/cjs/workers/parse.js.map +1 -1
  183. package/dist/cjs/workers/watch-heartbeat.js.map +1 -1
  184. package/dist/esm/chunk/extract.js.map +1 -1
  185. package/dist/esm/chunk/group.js.map +1 -1
  186. package/dist/esm/chunk/index.js.map +1 -1
  187. package/dist/esm/chunk/parse.js.map +1 -1
  188. package/dist/esm/chunk/parser.js.map +1 -1
  189. package/dist/esm/chunk/tokens.js.map +1 -1
  190. package/dist/esm/chunk/types.js.map +1 -1
  191. package/dist/esm/chunk/version.js.map +1 -1
  192. package/dist/esm/cli/build.js.map +1 -1
  193. package/dist/esm/cli/constants.d.ts +2 -0
  194. package/dist/esm/cli/constants.js +2 -0
  195. package/dist/esm/cli/constants.js.map +1 -0
  196. package/dist/esm/cli/download.js.map +1 -1
  197. package/dist/esm/cli/index.js.map +1 -1
  198. package/dist/esm/cli/init.js.map +1 -1
  199. package/dist/esm/cli/map.js.map +1 -1
  200. package/dist/esm/cli/named.js +2 -1
  201. package/dist/esm/cli/named.js.map +1 -1
  202. package/dist/esm/cli/path.js.map +1 -1
  203. package/dist/esm/cli/peek.js.map +1 -1
  204. package/dist/esm/cli/related.js.map +1 -1
  205. package/dist/esm/cli/search.js.map +1 -1
  206. package/dist/esm/cli/shared.js +6 -5
  207. package/dist/esm/cli/shared.js.map +1 -1
  208. package/dist/esm/cli/sql.js.map +1 -1
  209. package/dist/esm/cli/status.js.map +1 -1
  210. package/dist/esm/cli/types.js.map +1 -1
  211. package/dist/esm/cli/watch.js.map +1 -1
  212. package/dist/esm/cli.js +10 -9
  213. package/dist/esm/cli.js.map +1 -1
  214. package/dist/esm/commands/index.js.map +1 -1
  215. package/dist/esm/commands/map.js.map +1 -1
  216. package/dist/esm/commands/path.js.map +1 -1
  217. package/dist/esm/commands/peek.js.map +1 -1
  218. package/dist/esm/commands/related.js.map +1 -1
  219. package/dist/esm/commands/scope.js.map +1 -1
  220. package/dist/esm/commands/search.js.map +1 -1
  221. package/dist/esm/commands/signals.js.map +1 -1
  222. package/dist/esm/commands/snapshot.js.map +1 -1
  223. package/dist/esm/commands/status.js.map +1 -1
  224. package/dist/esm/config/access.js.map +1 -1
  225. package/dist/esm/config/index.js.map +1 -1
  226. package/dist/esm/config/load.js.map +1 -1
  227. package/dist/esm/config/resolve.js.map +1 -1
  228. package/dist/esm/config/signals.js.map +1 -1
  229. package/dist/esm/config/types.js.map +1 -1
  230. package/dist/esm/config/validate.js.map +1 -1
  231. package/dist/esm/embed/cohere.js.map +1 -1
  232. package/dist/esm/embed/distribution.js.map +1 -1
  233. package/dist/esm/embed/endpoint.js.map +1 -1
  234. package/dist/esm/embed/handoff.js.map +1 -1
  235. package/dist/esm/embed/http.js.map +1 -1
  236. package/dist/esm/embed/identity.js.map +1 -1
  237. package/dist/esm/embed/langfit.js.map +1 -1
  238. package/dist/esm/embed/languages.js.map +1 -1
  239. package/dist/esm/embed/openai.js.map +1 -1
  240. package/dist/esm/embed/query.js +1 -1
  241. package/dist/esm/embed/query.js.map +1 -1
  242. package/dist/esm/embed/registry.js.map +1 -1
  243. package/dist/esm/embed/static.js.map +1 -1
  244. package/dist/esm/embed/store.js.map +1 -1
  245. package/dist/esm/embed/types.js.map +1 -1
  246. package/dist/esm/errors.js.map +1 -1
  247. package/dist/esm/features/embed.js.map +1 -1
  248. package/dist/esm/features/fences.js.map +1 -1
  249. package/dist/esm/features/index.js.map +1 -1
  250. package/dist/esm/features/links.js.map +1 -1
  251. package/dist/esm/features/rank.js.map +1 -1
  252. package/dist/esm/features/sections.js.map +1 -1
  253. package/dist/esm/features/tags.js.map +1 -1
  254. package/dist/esm/features/types.js.map +1 -1
  255. package/dist/esm/graph/graph.js.map +1 -1
  256. package/dist/esm/graph/traverse.js.map +1 -1
  257. package/dist/esm/index.js.map +1 -1
  258. package/dist/esm/lib/atomic-write.js.map +1 -1
  259. package/dist/esm/lib/guarded-tick.js.map +1 -1
  260. package/dist/esm/lib/serial-query.js.map +1 -1
  261. package/dist/esm/lib/watch-notification.js.map +1 -1
  262. package/dist/esm/lib/worker-file.js.map +1 -1
  263. package/dist/esm/output/column-hint.js.map +1 -1
  264. package/dist/esm/output/output.js.map +1 -1
  265. package/dist/esm/output/progress.js.map +1 -1
  266. package/dist/esm/output/search-error.js.map +1 -1
  267. package/dist/esm/scan/frontmatter.js.map +1 -1
  268. package/dist/esm/scan/index.js.map +1 -1
  269. package/dist/esm/scan/list.js.map +1 -1
  270. package/dist/esm/scan/pool.js.map +1 -1
  271. package/dist/esm/scan/reparse.js.map +1 -1
  272. package/dist/esm/scan/worker-error.js.map +1 -1
  273. package/dist/esm/store/batch.js.map +1 -1
  274. package/dist/esm/store/builder.js.map +1 -1
  275. package/dist/esm/store/cache.js.map +1 -1
  276. package/dist/esm/store/duckdb/batch.js.map +1 -1
  277. package/dist/esm/store/duckdb/connection.js.map +1 -1
  278. package/dist/esm/store/duckdb/engine.d.ts +11 -0
  279. package/dist/esm/store/duckdb/engine.js +18 -0
  280. package/dist/esm/store/duckdb/engine.js.map +1 -0
  281. package/dist/esm/store/duckdb/fieldStats.js.map +1 -1
  282. package/dist/esm/store/duckdb/lexical.d.ts +1 -6
  283. package/dist/esm/store/duckdb/lexical.js +16 -20
  284. package/dist/esm/store/duckdb/lexical.js.map +1 -1
  285. package/dist/esm/store/duckdb/native.d.ts +5 -3
  286. package/dist/esm/store/duckdb/native.js +16 -14
  287. package/dist/esm/store/duckdb/native.js.map +1 -1
  288. package/dist/esm/store/duckdb/open.js +40 -2
  289. package/dist/esm/store/duckdb/open.js.map +1 -1
  290. package/dist/esm/store/duckdb/ordered-bm25.d.ts +3 -16
  291. package/dist/esm/store/duckdb/ordered-bm25.js +14 -55
  292. package/dist/esm/store/duckdb/ordered-bm25.js.map +1 -1
  293. package/dist/esm/store/duckdb/reconcile.js.map +1 -1
  294. package/dist/esm/store/duckdb/sql-functions.js.map +1 -1
  295. package/dist/esm/store/duckdb/store.js.map +1 -1
  296. package/dist/esm/store/duckdb/vectors.js.map +1 -1
  297. package/dist/esm/store/embed-scope.js.map +1 -1
  298. package/dist/esm/store/engine.d.ts +28 -0
  299. package/dist/esm/store/engine.js +85 -0
  300. package/dist/esm/store/engine.js.map +1 -0
  301. package/dist/esm/store/feature-scope.js.map +1 -1
  302. package/dist/esm/store/index.js.map +1 -1
  303. package/dist/esm/store/lock-wait.js.map +1 -1
  304. package/dist/esm/store/native.d.ts +8 -5
  305. package/dist/esm/store/native.js +59 -57
  306. package/dist/esm/store/native.js.map +1 -1
  307. package/dist/esm/store/open.js +24 -7
  308. package/dist/esm/store/open.js.map +1 -1
  309. package/dist/esm/store/preset-scope.js.map +1 -1
  310. package/dist/esm/store/readiness.js.map +1 -1
  311. package/dist/esm/store/reconcile.js.map +1 -1
  312. package/dist/esm/store/shared.js.map +1 -1
  313. package/dist/esm/store/signature.js.map +1 -1
  314. package/dist/esm/store/sql-functions.js.map +1 -1
  315. package/dist/esm/store/sqlite/connection.js.map +1 -1
  316. package/dist/esm/store/sqlite/engine.d.ts +12 -0
  317. package/dist/esm/store/sqlite/engine.js +16 -0
  318. package/dist/esm/store/sqlite/engine.js.map +1 -0
  319. package/dist/esm/store/sqlite/fieldStats.js.map +1 -1
  320. package/dist/esm/store/sqlite/lexical.js +2 -3
  321. package/dist/esm/store/sqlite/lexical.js.map +1 -1
  322. package/dist/esm/store/sqlite/open.js +38 -7
  323. package/dist/esm/store/sqlite/open.js.map +1 -1
  324. package/dist/esm/store/sqlite/reconcile.js.map +1 -1
  325. package/dist/esm/store/sqlite/sql-functions.js.map +1 -1
  326. package/dist/esm/store/sqlite/store.js.map +1 -1
  327. package/dist/esm/store/sqlite/vectors.js.map +1 -1
  328. package/dist/esm/store/stages.js.map +1 -1
  329. package/dist/esm/store/transaction.js.map +1 -1
  330. package/dist/esm/store/turso/connection.js.map +1 -1
  331. package/dist/esm/store/turso/engine.d.ts +9 -0
  332. package/dist/esm/store/turso/engine.js +14 -0
  333. package/dist/esm/store/turso/engine.js.map +1 -0
  334. package/dist/esm/store/turso/fieldStats.js.map +1 -1
  335. package/dist/esm/store/turso/lexical-text.js.map +1 -1
  336. package/dist/esm/store/turso/lexical.js +9 -9
  337. package/dist/esm/store/turso/lexical.js.map +1 -1
  338. package/dist/esm/store/turso/native.d.ts +5 -1
  339. package/dist/esm/store/turso/native.js +16 -10
  340. package/dist/esm/store/turso/native.js.map +1 -1
  341. package/dist/esm/store/turso/open.js +44 -3
  342. package/dist/esm/store/turso/open.js.map +1 -1
  343. package/dist/esm/store/turso/reconcile.d.ts +0 -1
  344. package/dist/esm/store/turso/reconcile.js +3 -8
  345. package/dist/esm/store/turso/reconcile.js.map +1 -1
  346. package/dist/esm/store/turso/sql-functions.js.map +1 -1
  347. package/dist/esm/store/turso/store.js.map +1 -1
  348. package/dist/esm/store/turso/vectors.js.map +1 -1
  349. package/dist/esm/store/types.d.ts +2 -0
  350. package/dist/esm/store/types.js.map +1 -1
  351. package/dist/esm/store/vectors.js.map +1 -1
  352. package/dist/esm/text/segment.js.map +1 -1
  353. package/dist/esm/text/strip.js.map +1 -1
  354. package/dist/esm/watch-claim.js.map +1 -1
  355. package/dist/esm/watch.js.map +1 -1
  356. package/dist/esm/workers/parse.js.map +1 -1
  357. package/dist/esm/workers/watch-heartbeat.js.map +1 -1
  358. package/package.json +21 -6
  359. package/skills/sense/SKILL.md +1 -1
  360. package/skills/sense/references/stores/duckdb.md +1 -1
  361. package/skills/sense/references/stores/sqlite.md +1 -1
  362. package/skills/sense/references/stores/turso.md +1 -1
  363. package/skills/sense-setup/references/stores/duckdb.md +1 -1
  364. package/skills/sense-setup/references/stores/sqlite.md +1 -1
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/duckdb/fieldStats.ts"],"sourcesContent":["import { quoteIdent } from '../shared.ts';\nimport type { Connection, FieldStat } from '../types.ts';\n\n// DuckDB's frontmatter columns are VARIANT (reconcile.ts), so typeof() always reports VARIANT; variant_typeof() reports the boxed native\n// type instead -- INT128, DOUBLE, VARCHAR, or a literal's narrower type like DECIMAL(p, s) -- mapped onto sqlite/turso's integer/real/text vocabulary; an unrecognized name throws (no-silent-modes).\nfunction classifyVariantType(name: string, field: string): 'integer' | 'real' | 'text' {\n if (/^U?INT\\d+$/.test(name)) return 'integer';\n if (name === 'FLOAT' || name === 'DOUBLE' || name.startsWith('DECIMAL(')) return 'real';\n if (name === 'VARCHAR') return 'text';\n throw new Error(`duckdb: variant_typeof() returned unrecognized type \"${name}\" for column \"${field}\"`);\n}\n\n// duckdb degrades superlinearly in the number of string_agg(DISTINCT variant_typeof()) aggregates\n// per projection, not the column width (301 columns: 1148ms unchunked vs 88ms at chunk size 25);\n// 16 sits inside the measured flat 10-25 range. sqlite/turso need none of this: GROUP_CONCAT over\n// 300 columns runs in 7ms.\nconst AGGREGATES_PER_QUERY = 16;\n\n// COUNT + string_agg(DISTINCT variant_typeof()) per column keeps this O(columns), not O(rows x columns).\n// '|' is the separator, not ',': DECIMAL's display name (\"DECIMAL(4, 2)\") contains a comma.\nexport async function fieldStats(conn: Connection, columns: string[], scopeWhere: string): Promise<FieldStat[]> {\n const stats: FieldStat[] = [];\n for (let start = 0; start < columns.length; start += AGGREGATES_PER_QUERY) {\n const chunk = columns.slice(start, start + AGGREGATES_PER_QUERY);\n const exprs = chunk.map((name, i) => {\n const quoted = quoteIdent(name);\n return `COUNT(${quoted}) AS n${i}, string_agg(DISTINCT variant_typeof(${quoted}), '|') FILTER (WHERE ${quoted} IS NOT NULL) AS t${i}`;\n });\n const stmt = await conn.prepare(`SELECT ${exprs.join(', ')} FROM frontmatter ${scopeWhere}`);\n const row = (await stmt.get()) as Record<string, number | string | null>;\n for (const [i, name] of chunk.entries()) {\n const raw = ((row[`t${i}`] as string) ?? '').split('|').filter(Boolean);\n const types = [...new Set(raw.map((t) => classifyVariantType(t, name)))].sort();\n stats.push({ field: name, coverage: Number(row[`n${i}`] ?? 0), type: types.join(',') });\n }\n }\n return stats;\n}\n"],"names":["quoteIdent","classifyVariantType","name","field","test","startsWith","Error","AGGREGATES_PER_QUERY","fieldStats","conn","columns","scopeWhere","stats","start","length","chunk","slice","exprs","map","i","quoted","stmt","prepare","join","row","get","entries","raw","split","filter","Boolean","types","Set","t","sort","push","coverage","Number","type"],"mappings":"AAAA,SAASA,UAAU,QAAQ,eAAe;AAG1C,yIAAyI;AACzI,sMAAsM;AACtM,SAASC,oBAAoBC,IAAY,EAAEC,KAAa;IACtD,IAAI,aAAaC,IAAI,CAACF,OAAO,OAAO;IACpC,IAAIA,SAAS,WAAWA,SAAS,YAAYA,KAAKG,UAAU,CAAC,aAAa,OAAO;IACjF,IAAIH,SAAS,WAAW,OAAO;IAC/B,MAAM,IAAII,MAAM,CAAC,qDAAqD,EAAEJ,KAAK,cAAc,EAAEC,MAAM,CAAC,CAAC;AACvG;AAEA,kGAAkG;AAClG,iGAAiG;AACjG,kGAAkG;AAClG,2BAA2B;AAC3B,MAAMI,uBAAuB;AAE7B,yGAAyG;AACzG,4FAA4F;AAC5F,OAAO,eAAeC,WAAWC,IAAgB,EAAEC,OAAiB,EAAEC,UAAkB;IACtF,MAAMC,QAAqB,EAAE;IAC7B,IAAK,IAAIC,QAAQ,GAAGA,QAAQH,QAAQI,MAAM,EAAED,SAASN,qBAAsB;QACzE,MAAMQ,QAAQL,QAAQM,KAAK,CAACH,OAAOA,QAAQN;QAC3C,MAAMU,QAAQF,MAAMG,GAAG,CAAC,CAAChB,MAAMiB;YAC7B,MAAMC,SAASpB,WAAWE;YAC1B,OAAO,CAAC,MAAM,EAAEkB,OAAO,MAAM,EAAED,EAAE,qCAAqC,EAAEC,OAAO,sBAAsB,EAAEA,OAAO,kBAAkB,EAAED,GAAG;QACvI;QACA,MAAME,OAAO,MAAMZ,KAAKa,OAAO,CAAC,CAAC,OAAO,EAAEL,MAAMM,IAAI,CAAC,MAAM,kBAAkB,EAAEZ,YAAY;QAC3F,MAAMa,MAAO,MAAMH,KAAKI,GAAG;QAC3B,KAAK,MAAM,CAACN,GAAGjB,KAAK,IAAIa,MAAMW,OAAO,GAAI;gBACzBF,OAE6BA;YAF3C,MAAMG,MAAM,EAAEH,QAAAA,GAAG,CAAC,CAAC,CAAC,EAAEL,GAAG,CAAC,cAAZK,mBAAAA,QAA2B,IAAII,KAAK,CAAC,KAAKC,MAAM,CAACC;YAC/D,MAAMC,QAAQ;mBAAI,IAAIC,IAAIL,IAAIT,GAAG,CAAC,CAACe,IAAMhC,oBAAoBgC,GAAG/B;aAAQ,CAACgC,IAAI;YAC7EtB,MAAMuB,IAAI,CAAC;gBAAEhC,OAAOD;gBAAMkC,UAAUC,QAAOb,SAAAA,GAAG,CAAC,CAAC,CAAC,EAAEL,GAAG,CAAC,cAAZK,oBAAAA,SAAgB;gBAAIc,MAAMP,MAAMR,IAAI,CAAC;YAAK;QACvF;IACF;IACA,OAAOX;AACT"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/open-source/ai/sensemaking/src/store/duckdb/fieldStats.ts"],"sourcesContent":["import { quoteIdent } from '../shared.ts';\nimport type { Connection, FieldStat } from '../types.ts';\n\n// DuckDB's frontmatter columns are VARIANT (reconcile.ts), so typeof() always reports VARIANT; variant_typeof() reports the boxed native\n// type instead -- INT128, DOUBLE, VARCHAR, or a literal's narrower type like DECIMAL(p, s) -- mapped onto sqlite/turso's integer/real/text vocabulary; an unrecognized name throws (no-silent-modes).\nfunction classifyVariantType(name: string, field: string): 'integer' | 'real' | 'text' {\n if (/^U?INT\\d+$/.test(name)) return 'integer';\n if (name === 'FLOAT' || name === 'DOUBLE' || name.startsWith('DECIMAL(')) return 'real';\n if (name === 'VARCHAR') return 'text';\n throw new Error(`duckdb: variant_typeof() returned unrecognized type \"${name}\" for column \"${field}\"`);\n}\n\n// duckdb degrades superlinearly in the number of string_agg(DISTINCT variant_typeof()) aggregates\n// per projection, not the column width (301 columns: 1148ms unchunked vs 88ms at chunk size 25);\n// 16 sits inside the measured flat 10-25 range. sqlite/turso need none of this: GROUP_CONCAT over\n// 300 columns runs in 7ms.\nconst AGGREGATES_PER_QUERY = 16;\n\n// COUNT + string_agg(DISTINCT variant_typeof()) per column keeps this O(columns), not O(rows x columns).\n// '|' is the separator, not ',': DECIMAL's display name (\"DECIMAL(4, 2)\") contains a comma.\nexport async function fieldStats(conn: Connection, columns: string[], scopeWhere: string): Promise<FieldStat[]> {\n const stats: FieldStat[] = [];\n for (let start = 0; start < columns.length; start += AGGREGATES_PER_QUERY) {\n const chunk = columns.slice(start, start + AGGREGATES_PER_QUERY);\n const exprs = chunk.map((name, i) => {\n const quoted = quoteIdent(name);\n return `COUNT(${quoted}) AS n${i}, string_agg(DISTINCT variant_typeof(${quoted}), '|') FILTER (WHERE ${quoted} IS NOT NULL) AS t${i}`;\n });\n const stmt = await conn.prepare(`SELECT ${exprs.join(', ')} FROM frontmatter ${scopeWhere}`);\n const row = (await stmt.get()) as Record<string, number | string | null>;\n for (const [i, name] of chunk.entries()) {\n const raw = ((row[`t${i}`] as string) ?? '').split('|').filter(Boolean);\n const types = [...new Set(raw.map((t) => classifyVariantType(t, name)))].sort();\n stats.push({ field: name, coverage: Number(row[`n${i}`] ?? 0), type: types.join(',') });\n }\n }\n return stats;\n}\n"],"names":["quoteIdent","classifyVariantType","name","field","test","startsWith","Error","AGGREGATES_PER_QUERY","fieldStats","conn","columns","scopeWhere","stats","start","length","chunk","slice","exprs","map","i","quoted","stmt","prepare","join","row","get","entries","raw","split","filter","Boolean","types","Set","t","sort","push","coverage","Number","type"],"mappings":"AAAA,SAASA,UAAU,QAAQ,eAAe;AAG1C,yIAAyI;AACzI,sMAAsM;AACtM,SAASC,oBAAoBC,IAAY,EAAEC,KAAa;IACtD,IAAI,aAAaC,IAAI,CAACF,OAAO,OAAO;IACpC,IAAIA,SAAS,WAAWA,SAAS,YAAYA,KAAKG,UAAU,CAAC,aAAa,OAAO;IACjF,IAAIH,SAAS,WAAW,OAAO;IAC/B,MAAM,IAAII,MAAM,CAAC,qDAAqD,EAAEJ,KAAK,cAAc,EAAEC,MAAM,CAAC,CAAC;AACvG;AAEA,kGAAkG;AAClG,iGAAiG;AACjG,kGAAkG;AAClG,2BAA2B;AAC3B,MAAMI,uBAAuB;AAE7B,yGAAyG;AACzG,4FAA4F;AAC5F,OAAO,eAAeC,WAAWC,IAAgB,EAAEC,OAAiB,EAAEC,UAAkB;IACtF,MAAMC,QAAqB,EAAE;IAC7B,IAAK,IAAIC,QAAQ,GAAGA,QAAQH,QAAQI,MAAM,EAAED,SAASN,qBAAsB;QACzE,MAAMQ,QAAQL,QAAQM,KAAK,CAACH,OAAOA,QAAQN;QAC3C,MAAMU,QAAQF,MAAMG,GAAG,CAAC,CAAChB,MAAMiB;YAC7B,MAAMC,SAASpB,WAAWE;YAC1B,OAAO,CAAC,MAAM,EAAEkB,OAAO,MAAM,EAAED,EAAE,qCAAqC,EAAEC,OAAO,sBAAsB,EAAEA,OAAO,kBAAkB,EAAED,GAAG;QACvI;QACA,MAAME,OAAO,MAAMZ,KAAKa,OAAO,CAAC,CAAC,OAAO,EAAEL,MAAMM,IAAI,CAAC,MAAM,kBAAkB,EAAEZ,YAAY;QAC3F,MAAMa,MAAO,MAAMH,KAAKI,GAAG;QAC3B,KAAK,MAAM,CAACN,GAAGjB,KAAK,IAAIa,MAAMW,OAAO,GAAI;gBACzBF,OAE6BA;YAF3C,MAAMG,MAAM,EAAEH,QAAAA,GAAG,CAAC,CAAC,CAAC,EAAEL,GAAG,CAAC,cAAZK,mBAAAA,QAA2B,IAAII,KAAK,CAAC,KAAKC,MAAM,CAACC;YAC/D,MAAMC,QAAQ;mBAAI,IAAIC,IAAIL,IAAIT,GAAG,CAAC,CAACe,IAAMhC,oBAAoBgC,GAAG/B;aAAQ,CAACgC,IAAI;YAC7EtB,MAAMuB,IAAI,CAAC;gBAAEhC,OAAOD;gBAAMkC,UAAUC,QAAOb,SAAAA,GAAG,CAAC,CAAC,CAAC,EAAEL,GAAG,CAAC,cAAZK,oBAAAA,SAAgB;gBAAIc,MAAMP,MAAMR,IAAI,CAAC;YAAK;QACvF;IACF;IACA,OAAOX;AACT"}
@@ -1,14 +1,9 @@
1
1
  import type { Connection, LexicalHit, LexicalQueryOptions } from '../types.js';
2
- interface FtsIndexState {
3
- stale: boolean;
4
- orderedMacroReady: boolean;
5
- }
6
2
  export declare function markContentStale(conn: Connection): Promise<void>;
7
3
  export declare function prepareFts(conn: Connection): Promise<void>;
8
4
  export declare function assertFtsReady(conn: Connection): Promise<void>;
9
- export declare function queryLexical(conn: Connection, terms: string, opts: LexicalQueryOptions, state: FtsIndexState): Promise<LexicalHit[]>;
5
+ export declare function queryLexical(conn: Connection, terms: string, opts: LexicalQueryOptions): Promise<LexicalHit[]>;
10
6
  export declare function createLexicalIndex(conn: Connection): {
11
7
  query: (terms: string, opts: LexicalQueryOptions) => Promise<LexicalHit[]>;
12
8
  markStale: () => Promise<void>;
13
9
  };
14
- export {};
@@ -4,14 +4,15 @@ import { SenseError } from '../../errors.js';
4
4
  import { foldForSearch, matchesSearchPhrase, searchTokens } from '../../text/segment.js';
5
5
  import { getMeta, setMeta } from '../shared.js';
6
6
  import { withTransaction } from '../transaction.js';
7
- import { ensureOrderedBm25, ORDERED_BM25_MACRO } from './ordered-bm25.js';
7
+ import { duckdbFts } from './engine.js';
8
+ import { nativeQueryTerms, orderedBm25Sql } from './ordered-bm25.js';
8
9
  const FIELDS = [
9
10
  'title',
10
11
  'summary',
11
12
  'text'
12
13
  ];
13
14
  // Mirrors sqlite's bm25(content, 10.0, 5.0, 1.0, ...) column-weight intent; DuckDB's match_bm25 has no
14
- // field_weights argument (verified 1.5.5), so each field is scored separately and combined here.
15
+ // field_weights argument (verified 1.4.5 through 1.5.6), so each field is scored separately and combined here.
15
16
  const FIELD_WEIGHT = {
16
17
  title: 10,
17
18
  summary: 5,
@@ -68,8 +69,7 @@ async function stateFor(conn) {
68
69
  // Anything but a persisted '0' (missing key, or '1') means a rebuild is owed: a cache that
69
70
  // never built an index, or one a prior process marked stale and never got to clear.
70
71
  state = {
71
- stale: await getMeta(conn, FTS_STALE_META_KEY) !== '0',
72
- orderedMacroReady: false
72
+ stale: await getMeta(conn, FTS_STALE_META_KEY) !== '0'
73
73
  };
74
74
  ftsState.set(conn, state);
75
75
  }
@@ -90,8 +90,7 @@ async function hasFtsArtifact(conn) {
90
90
  // transaction (reconcile.ts) so a crash never lands the content write without the stale mark.
91
91
  export async function markContentStale(conn) {
92
92
  ftsState.set(conn, {
93
- stale: true,
94
- orderedMacroReady: false
93
+ stale: true
95
94
  });
96
95
  await setMeta(conn, FTS_STALE_META_KEY, '1');
97
96
  }
@@ -146,11 +145,8 @@ async function rebuildFts(conn, state) {
146
145
  await conn.exec('INSTALL fts; LOAD fts;');
147
146
  if (!state.stale && await hasFtsArtifact(conn)) return;
148
147
  state.stale = true;
149
- state.orderedMacroReady = false;
150
148
  await withTransaction(conn, async ()=>{
151
- // stopwords='none': sqlite's porter/unicode61 tokenizer never removes stopwords either (verified 1.5.5); the fts extension's
152
- // default 571-word list would otherwise drop common words (e.g. "and") from the index but not match_bm25's conjunctive gate, breaking multi-word matches.
153
- await conn.exec(`PRAGMA create_fts_index('content', 'path', 'title', 'summary', 'text', stopwords='none', overwrite=1)`);
149
+ await conn.exec((await duckdbFts()).createIndex());
154
150
  // Same transaction as the rebuild: a crash before commit leaves the mark set (a redundant
155
151
  // rebuild next time), never a cleared mark serving a half-built index as fresh.
156
152
  await setMeta(conn, FTS_STALE_META_KEY, '0');
@@ -175,24 +171,23 @@ export async function assertFtsReady(conn) {
175
171
  }
176
172
  // Field-weighted fragments (title 10 / summary 5 / text 1), with params emitted in the exact left-to-right order the assembled SQL needs for positional `?` binding.
177
173
  // score/gate params stay in separate arrays until joined at the end, since the SQL puts every score part before every gate part; sharing one array would shift binds once a query has both words and substrings (PRINCIPLES: no-silent-modes).
178
- function buildScoreAndGate(words, bareWords, phraseWords, substrings) {
174
+ function buildScoreAndGate(fts, words, nativeTerms, bareWords, phraseWords, substrings) {
179
175
  const scoreParts = [];
180
176
  const gateParts = [];
181
177
  const scoreParams = [];
182
178
  const gateParams = [];
183
179
  if (words.length > 0) {
184
- const wordQuery = words.map(foldForSearch).join(' ');
185
180
  for (const field of FIELDS){
186
- scoreParts.push(`${FIELD_WEIGHT[field]}.0 * COALESCE(${ORDERED_BM25_MACRO}(content."path", ?, fields := '${field}', conjunctive := false), 0)`);
187
- scoreParams.push(wordQuery);
181
+ scoreParts.push(`${FIELD_WEIGHT[field]}.0 * COALESCE(${orderedBm25Sql(fts, field, nativeTerms.length)}, 0)`);
182
+ scoreParams.push(...nativeTerms);
188
183
  }
189
184
  if (bareWords.length > 0) {
190
- const nativeGate = `${ORDERED_BM25_MACRO}(content."path", ?, conjunctive := true) IS NOT NULL`;
185
+ const nativeGate = fts.conjunctiveGate();
191
186
  gateParts.push(nativeGate);
192
187
  gateParams.push(bareWords.map(foldForSearch).join(' '));
193
188
  }
194
189
  if (phraseWords.length > 0) {
195
- const nativeGate = `${ORDERED_BM25_MACRO}(content."path", ?, conjunctive := true) IS NOT NULL`;
190
+ const nativeGate = fts.conjunctiveGate();
196
191
  const fallback = phraseWords.map((_word)=>`(${FIELDS.map((field)=>`contains(strip_accents(lower(content.${field})), ?)`).join(' OR ')})`).join(' AND ');
197
192
  gateParts.push(`(${nativeGate} OR (${fallback}))`);
198
193
  gateParams.push(phraseWords.map(foldForSearch).join(' '));
@@ -219,7 +214,7 @@ function buildScoreAndGate(words, bareWords, phraseWords, substrings) {
219
214
  }
220
215
  // Ranked query, scoped by the caller-built SQL fragments (same fragments sqlite's queryLexical takes).
221
216
  // Phrase rows carry authored fields through the post-filter; the public result still contains paths only.
222
- export async function queryLexical(conn, terms, opts, state) {
217
+ export async function queryLexical(conn, terms, opts) {
223
218
  const unsupported = unsupportedOperator(terms);
224
219
  if (unsupported !== null) {
225
220
  throw new SenseError('STORE_CAPABILITY_MISSING', `store "duckdb" does not implement FTS5's ${unsupported.label} ("${unsupported.token}") in this build; rephrase "${terms.trim()}" without it, or set "store" to "sqlite" in this tree's config to search it as written`);
@@ -228,8 +223,9 @@ export async function queryLexical(conn, terms, opts, state) {
228
223
  const { words, bareWords, phraseWords, substrings, phrases, emptyPhrase } = splitTerms(terms);
229
224
  if (emptyPhrase || words.length === 0 && substrings.length === 0) return [];
230
225
  if (words.length > 0) await assertFtsReady(conn);
231
- if (words.length > 0) await ensureOrderedBm25(conn, state);
232
- const { scoreSql, gateSql, params } = buildScoreAndGate(words, bareWords, phraseWords, substrings);
226
+ const fts = await duckdbFts();
227
+ const nativeTerms = words.length > 0 ? await nativeQueryTerms(conn, fts, words.map(foldForSearch).join(' ')) : [];
228
+ const { scoreSql, gateSql, params } = buildScoreAndGate(fts, words, nativeTerms, bareWords, phraseWords, substrings);
233
229
  const selectFields = phrases.length > 0 ? ', content.title, content.summary, content.text' : '';
234
230
  const sqlLimit = phrases.length > 0 ? '' : ' LIMIT ?';
235
231
  const sql = `SELECT path, score${phrases.length > 0 ? ', title, summary, text' : ''} FROM (
@@ -255,7 +251,7 @@ export async function queryLexical(conn, terms, opts, state) {
255
251
  // otherwise), and reconcileContent (reconcile.ts) marks it stale again on every content change.
256
252
  export function createLexicalIndex(conn) {
257
253
  return {
258
- query: async (terms, opts)=>queryLexical(conn, terms, opts, await stateFor(conn)),
254
+ query: async (terms, opts)=>queryLexical(conn, terms, opts),
259
255
  markStale: ()=>markContentStale(conn)
260
256
  };
261
257
  }
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/duckdb/lexical.ts"],"sourcesContent":["// D1: fts BM25 for ranking, contains() scans for exact substring / phrase verification /\n// unspaced scripts, JS snippets for every doc (no snippet()).\nimport { SenseError } from '../../errors.ts';\nimport { foldForSearch, matchesSearchPhrase, type SearchToken, searchTokens } from '../../text/segment.ts';\nimport { getMeta, setMeta } from '../shared.ts';\nimport { withTransaction } from '../transaction.ts';\nimport type { Connection, LexicalHit, LexicalQueryOptions } from '../types.ts';\nimport { ensureOrderedBm25, ORDERED_BM25_MACRO } from './ordered-bm25.ts';\n\nconst FIELDS = ['title', 'summary', 'text'] as const;\ntype Field = (typeof FIELDS)[number];\n// Mirrors sqlite's bm25(content, 10.0, 5.0, 1.0, ...) column-weight intent; DuckDB's match_bm25 has no\n// field_weights argument (verified 1.5.5), so each field is scored separately and combined here.\nconst FIELD_WEIGHT: Record<Field, number> = { title: 10, summary: 5, text: 1 };\n\n// FTS5 operator syntax (sqlite.org/fts5.html sec. 3) that words/substrings below would otherwise silently\n// treat as literal terms (PRINCIPLES: no-silent-modes). Quoted contents stay ignored while their boundary remains detectable.\nconst FTS5_OPERATORS: Array<{ label: string; re: RegExp }> = [\n { label: 'prefix query', re: /[\\p{L}\\p{N}_]+\\*(?=\\s|$)/u },\n { label: 'boolean operator', re: /(?:^|\\s)(?:AND|OR|NOT)(?=\\s|$)/ },\n { label: 'NEAR operator', re: /(?:^|\\s)NEAR\\b/ },\n { label: 'initial-token operator', re: /(?:^|\\s)\\^\\S+/ },\n { label: 'column filter', re: /(?:^|\\s)[\\p{L}_]\\w*\\s*:/u },\n { label: 'grouping operator', re: /[()]/ },\n];\n\nfunction unsupportedOperator(terms: string): { label: string; token: string } | null {\n const withoutPhrases = terms.replace(/\"[^\"]*\"/g, 'phrase');\n for (const { label, re } of FTS5_OPERATORS) {\n const m = withoutPhrases.match(re);\n if (m) return { label, token: m[0].trim() };\n }\n return null;\n}\n\ninterface FtsIndexState {\n stale: boolean;\n orderedMacroReady: boolean;\n}\n\n// Durable across connections/processes (meta table, shared.ts); `stale` is the in-memory cache of\n// it for this connection's lifetime, so a watcher's repeat queries don't re-read meta every time.\nconst FTS_STALE_META_KEY = 'fts_stale';\n\n// One state per connection, not per store instance: reconcileContent (reconcile.ts) marks it\n// stale from outside createLexicalIndex's closure, whether a one-shot open or a watcher's builder ran it.\nconst ftsState = new WeakMap<Connection, FtsIndexState>();\nasync function stateFor(conn: Connection): Promise<FtsIndexState> {\n let state = ftsState.get(conn);\n if (!state) {\n // Anything but a persisted '0' (missing key, or '1') means a rebuild is owed: a cache that\n // never built an index, or one a prior process marked stale and never got to clear.\n state = { stale: (await getMeta(conn, FTS_STALE_META_KEY)) !== '0', orderedMacroReady: false };\n ftsState.set(conn, state);\n }\n return state;\n}\n\nasync function hasFtsArtifact(conn: Connection): Promise<boolean> {\n const stmt = await conn.prepare(`\n SELECT 1 AS ready\n FROM duckdb_functions()\n WHERE schema_name = 'fts_main_content'\n AND function_name = 'match_bm25'\n AND function_type = 'macro'\n LIMIT 1\n `);\n return (await stmt.get()) !== undefined;\n}\n\n// Called whenever this store's reconcileContent changes `content`; must run inside that same\n// transaction (reconcile.ts) so a crash never lands the content write without the stale mark.\nexport async function markContentStale(conn: Connection): Promise<void> {\n ftsState.set(conn, { stale: true, orderedMacroReady: false });\n await setMeta(conn, FTS_STALE_META_KEY, '1');\n}\n\n// A run whose script marks no word boundaries makes match_bm25's whitespace tokenizer index the whole run as one token,\n// so unspaced terms use contains() and quoted phrases use the shared authored-field verifier.\nfunction splitTerms(terms: string): { words: string[]; bareWords: string[]; phraseWords: string[]; substrings: string[]; phrases: SearchToken[][]; emptyPhrase: boolean } {\n const words: string[] = [];\n const bareWords: string[] = [];\n const phraseWords: string[] = [];\n const substrings: string[] = [];\n const phrases: SearchToken[][] = [];\n let emptyPhrase = false;\n const withoutPhrases = terms.replace(/\"([^\"]*)\"/g, (_m, inner: string) => {\n const phrase = inner.trim();\n const phraseTokens = searchTokens(phrase);\n if (phraseTokens.length === 0) {\n emptyPhrase = true;\n return ' ';\n }\n phrases.push(phraseTokens);\n for (const token of phraseTokens) {\n if (token.unspaced) substrings.push(token.text);\n else {\n words.push(token.text);\n phraseWords.push(token.text);\n }\n }\n return ' ';\n });\n for (const tok of withoutPhrases.split(/\\s+/)) {\n if (tok.length === 0) continue;\n for (const token of searchTokens(tok)) {\n if (token.unspaced) substrings.push(token.text);\n else {\n words.push(token.text);\n bareWords.push(token.text);\n }\n }\n }\n return { words, bareWords, phraseWords, substrings, phrases, emptyPhrase };\n}\n\n// No incremental update (verified 1.5.5: PRAGMA create_fts_index is rebuild-only), so this pays\n// the full rebuild whenever `state.stale`, which meta.fts_stale keeps true across processes until cleared below.\nasync function rebuildFts(conn: Connection, state: FtsIndexState): Promise<void> {\n await conn.exec('INSTALL fts; LOAD fts;');\n if (!state.stale && (await hasFtsArtifact(conn))) return;\n state.stale = true;\n state.orderedMacroReady = false;\n await withTransaction(conn, async () => {\n // stopwords='none': sqlite's porter/unicode61 tokenizer never removes stopwords either (verified 1.5.5); the fts extension's\n // default 571-word list would otherwise drop common words (e.g. \"and\") from the index but not match_bm25's conjunctive gate, breaking multi-word matches.\n await conn.exec(`PRAGMA create_fts_index('content', 'path', 'title', 'summary', 'text', stopwords='none', overwrite=1)`);\n // Same transaction as the rebuild: a crash before commit leaves the mark set (a redundant\n // rebuild next time), never a cleared mark serving a half-built index as fresh.\n await setMeta(conn, FTS_STALE_META_KEY, '0');\n });\n state.stale = false;\n}\n\nexport async function prepareFts(conn: Connection): Promise<void> {\n await rebuildFts(conn, await stateFor(conn));\n}\n\nexport async function assertFtsReady(conn: Connection): Promise<void> {\n const state = await stateFor(conn);\n if (state.stale) throw new SenseError('INDEX_NOT_READY', 'the DuckDB lexical index is missing or stale; run `sense build` (or library build(config)) before lexical querying with build disabled');\n try {\n await conn.exec('LOAD fts;');\n if (!(await hasFtsArtifact(conn))) {\n throw new SenseError('INDEX_NOT_READY', 'the DuckDB lexical index artifact is missing; run `sense build` (or library build(config)) before lexical querying with build disabled');\n }\n } catch (err) {\n if (err instanceof SenseError) throw err;\n throw new SenseError('INDEX_NOT_READY', `the DuckDB lexical extension is not prepared (${(err as Error).message}); run \\`sense build\\` (or library build(config))`);\n }\n}\n\n// Field-weighted fragments (title 10 / summary 5 / text 1), with params emitted in the exact left-to-right order the assembled SQL needs for positional `?` binding.\n// score/gate params stay in separate arrays until joined at the end, since the SQL puts every score part before every gate part; sharing one array would shift binds once a query has both words and substrings (PRINCIPLES: no-silent-modes).\nfunction buildScoreAndGate(words: string[], bareWords: string[], phraseWords: string[], substrings: string[]): { scoreSql: string; gateSql: string; params: unknown[] } {\n const scoreParts: string[] = [];\n const gateParts: string[] = [];\n const scoreParams: unknown[] = [];\n const gateParams: unknown[] = [];\n\n if (words.length > 0) {\n const wordQuery = words.map(foldForSearch).join(' ');\n for (const field of FIELDS) {\n scoreParts.push(`${FIELD_WEIGHT[field]}.0 * COALESCE(${ORDERED_BM25_MACRO}(content.\"path\", ?, fields := '${field}', conjunctive := false), 0)`);\n scoreParams.push(wordQuery);\n }\n if (bareWords.length > 0) {\n const nativeGate = `${ORDERED_BM25_MACRO}(content.\"path\", ?, conjunctive := true) IS NOT NULL`;\n gateParts.push(nativeGate);\n gateParams.push(bareWords.map(foldForSearch).join(' '));\n }\n if (phraseWords.length > 0) {\n const nativeGate = `${ORDERED_BM25_MACRO}(content.\"path\", ?, conjunctive := true) IS NOT NULL`;\n const fallback = phraseWords.map((_word) => `(${FIELDS.map((field) => `contains(strip_accents(lower(content.${field})), ?)`).join(' OR ')})`).join(' AND ');\n gateParts.push(`(${nativeGate} OR (${fallback}))`);\n gateParams.push(phraseWords.map(foldForSearch).join(' '));\n gateParams.push(...phraseWords.flatMap((word) => FIELDS.map(() => foldForSearch(word))));\n }\n }\n\n for (const raw of substrings) {\n const needle = raw.toLowerCase();\n for (const field of FIELDS) {\n scoreParts.push(`${FIELD_WEIGHT[field]}.0 * CAST(contains(strip_accents(lower(content.${field})), ?) AS INTEGER)`);\n scoreParams.push(foldForSearch(needle));\n }\n gateParts.push(`(${FIELDS.map((field) => `contains(strip_accents(lower(content.${field})), ?)`).join(' OR ')})`);\n gateParams.push(...FIELDS.map(() => foldForSearch(needle)));\n }\n\n return { scoreSql: scoreParts.join(' + '), gateSql: gateParts.join(' AND ') || 'TRUE', params: [...scoreParams, ...gateParams] };\n}\n\n// Ranked query, scoped by the caller-built SQL fragments (same fragments sqlite's queryLexical takes).\n// Phrase rows carry authored fields through the post-filter; the public result still contains paths only.\nexport async function queryLexical(conn: Connection, terms: string, opts: LexicalQueryOptions, state: FtsIndexState): Promise<LexicalHit[]> {\n const unsupported = unsupportedOperator(terms);\n if (unsupported !== null) {\n throw new SenseError('STORE_CAPABILITY_MISSING', `store \"duckdb\" does not implement FTS5's ${unsupported.label} (\"${unsupported.token}\") in this build; rephrase \"${terms.trim()}\" without it, or set \"store\" to \"sqlite\" in this tree's config to search it as written`);\n }\n const { whereJoin, whereCond, scopeCond, limit } = opts;\n const { words, bareWords, phraseWords, substrings, phrases, emptyPhrase } = splitTerms(terms);\n if (emptyPhrase || (words.length === 0 && substrings.length === 0)) return [];\n\n if (words.length > 0) await assertFtsReady(conn);\n if (words.length > 0) await ensureOrderedBm25(conn, state);\n\n const { scoreSql, gateSql, params } = buildScoreAndGate(words, bareWords, phraseWords, substrings);\n const selectFields = phrases.length > 0 ? ', content.title, content.summary, content.text' : '';\n const sqlLimit = phrases.length > 0 ? '' : ' LIMIT ?';\n const sql = `SELECT path, score${phrases.length > 0 ? ', title, summary, text' : ''} FROM (\n SELECT content.\"path\" AS path, (${scoreSql}) AS score${selectFields}\n FROM content\n ${whereJoin}\n WHERE ${gateSql}\n ${whereCond} ${scopeCond}\n ) sq ORDER BY score DESC, path${sqlLimit}`;\n const stmt = await conn.prepare(sql);\n const rows = (await stmt.all(...params, ...(phrases.length > 0 ? [] : [limit]))) as Array<LexicalHit & Partial<Record<Field, string>>>;\n const matches = phrases.length === 0 ? rows : rows.filter((row) => phrases.every((phrase) => FIELDS.some((field) => matchesSearchPhrase(row[field] ?? '', phrase))));\n return matches.slice(0, limit).map(({ path }) => ({ path }));\n}\n\n// `stale`'s first read for a connection comes from meta.fts_stale (a fresh connection can't know\n// otherwise), and reconcileContent (reconcile.ts) marks it stale again on every content change.\nexport function createLexicalIndex(conn: Connection): { query: (terms: string, opts: LexicalQueryOptions) => Promise<LexicalHit[]>; markStale: () => Promise<void> } {\n return {\n query: async (terms, opts) => queryLexical(conn, terms, opts, await stateFor(conn)),\n markStale: () => markContentStale(conn),\n };\n}\n"],"names":["SenseError","foldForSearch","matchesSearchPhrase","searchTokens","getMeta","setMeta","withTransaction","ensureOrderedBm25","ORDERED_BM25_MACRO","FIELDS","FIELD_WEIGHT","title","summary","text","FTS5_OPERATORS","label","re","unsupportedOperator","terms","withoutPhrases","replace","m","match","token","trim","FTS_STALE_META_KEY","ftsState","WeakMap","stateFor","conn","state","get","stale","orderedMacroReady","set","hasFtsArtifact","stmt","prepare","undefined","markContentStale","splitTerms","words","bareWords","phraseWords","substrings","phrases","emptyPhrase","_m","inner","phrase","phraseTokens","length","push","unspaced","tok","split","rebuildFts","exec","prepareFts","assertFtsReady","err","message","buildScoreAndGate","scoreParts","gateParts","scoreParams","gateParams","wordQuery","map","join","field","nativeGate","fallback","_word","flatMap","word","raw","needle","toLowerCase","scoreSql","gateSql","params","queryLexical","opts","unsupported","whereJoin","whereCond","scopeCond","limit","selectFields","sqlLimit","sql","rows","all","matches","filter","row","every","some","slice","path","createLexicalIndex","query","markStale"],"mappings":"AAAA,yFAAyF;AACzF,8DAA8D;AAC9D,SAASA,UAAU,QAAQ,kBAAkB;AAC7C,SAASC,aAAa,EAAEC,mBAAmB,EAAoBC,YAAY,QAAQ,wBAAwB;AAC3G,SAASC,OAAO,EAAEC,OAAO,QAAQ,eAAe;AAChD,SAASC,eAAe,QAAQ,oBAAoB;AAEpD,SAASC,iBAAiB,EAAEC,kBAAkB,QAAQ,oBAAoB;AAE1E,MAAMC,SAAS;IAAC;IAAS;IAAW;CAAO;AAE3C,uGAAuG;AACvG,iGAAiG;AACjG,MAAMC,eAAsC;IAAEC,OAAO;IAAIC,SAAS;IAAGC,MAAM;AAAE;AAE7E,0GAA0G;AAC1G,8HAA8H;AAC9H,MAAMC,iBAAuD;IAC3D;QAAEC,OAAO;QAAgBC,IAAI;IAA4B;IACzD;QAAED,OAAO;QAAoBC,IAAI;IAAiC;IAClE;QAAED,OAAO;QAAiBC,IAAI;IAAiB;IAC/C;QAAED,OAAO;QAA0BC,IAAI;IAAgB;IACvD;QAAED,OAAO;QAAiBC,IAAI;IAA2B;IACzD;QAAED,OAAO;QAAqBC,IAAI;IAAO;CAC1C;AAED,SAASC,oBAAoBC,KAAa;IACxC,MAAMC,iBAAiBD,MAAME,OAAO,CAAC,YAAY;IACjD,KAAK,MAAM,EAAEL,KAAK,EAAEC,EAAE,EAAE,IAAIF,eAAgB;QAC1C,MAAMO,IAAIF,eAAeG,KAAK,CAACN;QAC/B,IAAIK,GAAG,OAAO;YAAEN;YAAOQ,OAAOF,CAAC,CAAC,EAAE,CAACG,IAAI;QAAG;IAC5C;IACA,OAAO;AACT;AAOA,kGAAkG;AAClG,kGAAkG;AAClG,MAAMC,qBAAqB;AAE3B,6FAA6F;AAC7F,0GAA0G;AAC1G,MAAMC,WAAW,IAAIC;AACrB,eAAeC,SAASC,IAAgB;IACtC,IAAIC,QAAQJ,SAASK,GAAG,CAACF;IACzB,IAAI,CAACC,OAAO;QACV,2FAA2F;QAC3F,oFAAoF;QACpFA,QAAQ;YAAEE,OAAO,AAAC,MAAM5B,QAAQyB,MAAMJ,wBAAyB;YAAKQ,mBAAmB;QAAM;QAC7FP,SAASQ,GAAG,CAACL,MAAMC;IACrB;IACA,OAAOA;AACT;AAEA,eAAeK,eAAeN,IAAgB;IAC5C,MAAMO,OAAO,MAAMP,KAAKQ,OAAO,CAAC,CAAC;;;;;;;EAOjC,CAAC;IACD,OAAO,AAAC,MAAMD,KAAKL,GAAG,OAAQO;AAChC;AAEA,6FAA6F;AAC7F,8FAA8F;AAC9F,OAAO,eAAeC,iBAAiBV,IAAgB;IACrDH,SAASQ,GAAG,CAACL,MAAM;QAAEG,OAAO;QAAMC,mBAAmB;IAAM;IAC3D,MAAM5B,QAAQwB,MAAMJ,oBAAoB;AAC1C;AAEA,wHAAwH;AACxH,8FAA8F;AAC9F,SAASe,WAAWtB,KAAa;IAC/B,MAAMuB,QAAkB,EAAE;IAC1B,MAAMC,YAAsB,EAAE;IAC9B,MAAMC,cAAwB,EAAE;IAChC,MAAMC,aAAuB,EAAE;IAC/B,MAAMC,UAA2B,EAAE;IACnC,IAAIC,cAAc;IAClB,MAAM3B,iBAAiBD,MAAME,OAAO,CAAC,cAAc,CAAC2B,IAAIC;QACtD,MAAMC,SAASD,MAAMxB,IAAI;QACzB,MAAM0B,eAAe/C,aAAa8C;QAClC,IAAIC,aAAaC,MAAM,KAAK,GAAG;YAC7BL,cAAc;YACd,OAAO;QACT;QACAD,QAAQO,IAAI,CAACF;QACb,KAAK,MAAM3B,SAAS2B,aAAc;YAChC,IAAI3B,MAAM8B,QAAQ,EAAET,WAAWQ,IAAI,CAAC7B,MAAMV,IAAI;iBACzC;gBACH4B,MAAMW,IAAI,CAAC7B,MAAMV,IAAI;gBACrB8B,YAAYS,IAAI,CAAC7B,MAAMV,IAAI;YAC7B;QACF;QACA,OAAO;IACT;IACA,KAAK,MAAMyC,OAAOnC,eAAeoC,KAAK,CAAC,OAAQ;QAC7C,IAAID,IAAIH,MAAM,KAAK,GAAG;QACtB,KAAK,MAAM5B,SAASpB,aAAamD,KAAM;YACrC,IAAI/B,MAAM8B,QAAQ,EAAET,WAAWQ,IAAI,CAAC7B,MAAMV,IAAI;iBACzC;gBACH4B,MAAMW,IAAI,CAAC7B,MAAMV,IAAI;gBACrB6B,UAAUU,IAAI,CAAC7B,MAAMV,IAAI;YAC3B;QACF;IACF;IACA,OAAO;QAAE4B;QAAOC;QAAWC;QAAaC;QAAYC;QAASC;IAAY;AAC3E;AAEA,gGAAgG;AAChG,iHAAiH;AACjH,eAAeU,WAAW3B,IAAgB,EAAEC,KAAoB;IAC9D,MAAMD,KAAK4B,IAAI,CAAC;IAChB,IAAI,CAAC3B,MAAME,KAAK,IAAK,MAAMG,eAAeN,OAAQ;IAClDC,MAAME,KAAK,GAAG;IACdF,MAAMG,iBAAiB,GAAG;IAC1B,MAAM3B,gBAAgBuB,MAAM;QAC1B,6HAA6H;QAC7H,0JAA0J;QAC1J,MAAMA,KAAK4B,IAAI,CAAC,CAAC,qGAAqG,CAAC;QACvH,0FAA0F;QAC1F,gFAAgF;QAChF,MAAMpD,QAAQwB,MAAMJ,oBAAoB;IAC1C;IACAK,MAAME,KAAK,GAAG;AAChB;AAEA,OAAO,eAAe0B,WAAW7B,IAAgB;IAC/C,MAAM2B,WAAW3B,MAAM,MAAMD,SAASC;AACxC;AAEA,OAAO,eAAe8B,eAAe9B,IAAgB;IACnD,MAAMC,QAAQ,MAAMF,SAASC;IAC7B,IAAIC,MAAME,KAAK,EAAE,MAAM,IAAIhC,WAAW,mBAAmB;IACzD,IAAI;QACF,MAAM6B,KAAK4B,IAAI,CAAC;QAChB,IAAI,CAAE,MAAMtB,eAAeN,OAAQ;YACjC,MAAM,IAAI7B,WAAW,mBAAmB;QAC1C;IACF,EAAE,OAAO4D,KAAK;QACZ,IAAIA,eAAe5D,YAAY,MAAM4D;QACrC,MAAM,IAAI5D,WAAW,mBAAmB,CAAC,8CAA8C,EAAE,AAAC4D,IAAcC,OAAO,CAAC,iDAAiD,CAAC;IACpK;AACF;AAEA,qKAAqK;AACrK,+OAA+O;AAC/O,SAASC,kBAAkBrB,KAAe,EAAEC,SAAmB,EAAEC,WAAqB,EAAEC,UAAoB;IAC1G,MAAMmB,aAAuB,EAAE;IAC/B,MAAMC,YAAsB,EAAE;IAC9B,MAAMC,cAAyB,EAAE;IACjC,MAAMC,aAAwB,EAAE;IAEhC,IAAIzB,MAAMU,MAAM,GAAG,GAAG;QACpB,MAAMgB,YAAY1B,MAAM2B,GAAG,CAACnE,eAAeoE,IAAI,CAAC;QAChD,KAAK,MAAMC,SAAS7D,OAAQ;YAC1BsD,WAAWX,IAAI,CAAC,GAAG1C,YAAY,CAAC4D,MAAM,CAAC,cAAc,EAAE9D,mBAAmB,+BAA+B,EAAE8D,MAAM,4BAA4B,CAAC;YAC9IL,YAAYb,IAAI,CAACe;QACnB;QACA,IAAIzB,UAAUS,MAAM,GAAG,GAAG;YACxB,MAAMoB,aAAa,GAAG/D,mBAAmB,oDAAoD,CAAC;YAC9FwD,UAAUZ,IAAI,CAACmB;YACfL,WAAWd,IAAI,CAACV,UAAU0B,GAAG,CAACnE,eAAeoE,IAAI,CAAC;QACpD;QACA,IAAI1B,YAAYQ,MAAM,GAAG,GAAG;YAC1B,MAAMoB,aAAa,GAAG/D,mBAAmB,oDAAoD,CAAC;YAC9F,MAAMgE,WAAW7B,YAAYyB,GAAG,CAAC,CAACK,QAAU,CAAC,CAAC,EAAEhE,OAAO2D,GAAG,CAAC,CAACE,QAAU,CAAC,qCAAqC,EAAEA,MAAM,MAAM,CAAC,EAAED,IAAI,CAAC,QAAQ,CAAC,CAAC,EAAEA,IAAI,CAAC;YACnJL,UAAUZ,IAAI,CAAC,CAAC,CAAC,EAAEmB,WAAW,KAAK,EAAEC,SAAS,EAAE,CAAC;YACjDN,WAAWd,IAAI,CAACT,YAAYyB,GAAG,CAACnE,eAAeoE,IAAI,CAAC;YACpDH,WAAWd,IAAI,IAAIT,YAAY+B,OAAO,CAAC,CAACC,OAASlE,OAAO2D,GAAG,CAAC,IAAMnE,cAAc0E;QAClF;IACF;IAEA,KAAK,MAAMC,OAAOhC,WAAY;QAC5B,MAAMiC,SAASD,IAAIE,WAAW;QAC9B,KAAK,MAAMR,SAAS7D,OAAQ;YAC1BsD,WAAWX,IAAI,CAAC,GAAG1C,YAAY,CAAC4D,MAAM,CAAC,+CAA+C,EAAEA,MAAM,kBAAkB,CAAC;YACjHL,YAAYb,IAAI,CAACnD,cAAc4E;QACjC;QACAb,UAAUZ,IAAI,CAAC,CAAC,CAAC,EAAE3C,OAAO2D,GAAG,CAAC,CAACE,QAAU,CAAC,qCAAqC,EAAEA,MAAM,MAAM,CAAC,EAAED,IAAI,CAAC,QAAQ,CAAC,CAAC;QAC/GH,WAAWd,IAAI,IAAI3C,OAAO2D,GAAG,CAAC,IAAMnE,cAAc4E;IACpD;IAEA,OAAO;QAAEE,UAAUhB,WAAWM,IAAI,CAAC;QAAQW,SAAShB,UAAUK,IAAI,CAAC,YAAY;QAAQY,QAAQ;eAAIhB;eAAgBC;SAAW;IAAC;AACjI;AAEA,uGAAuG;AACvG,0GAA0G;AAC1G,OAAO,eAAegB,aAAarD,IAAgB,EAAEX,KAAa,EAAEiE,IAAyB,EAAErD,KAAoB;IACjH,MAAMsD,cAAcnE,oBAAoBC;IACxC,IAAIkE,gBAAgB,MAAM;QACxB,MAAM,IAAIpF,WAAW,4BAA4B,CAAC,yCAAyC,EAAEoF,YAAYrE,KAAK,CAAC,GAAG,EAAEqE,YAAY7D,KAAK,CAAC,4BAA4B,EAAEL,MAAMM,IAAI,GAAG,sFAAsF,CAAC;IAC1Q;IACA,MAAM,EAAE6D,SAAS,EAAEC,SAAS,EAAEC,SAAS,EAAEC,KAAK,EAAE,GAAGL;IACnD,MAAM,EAAE1C,KAAK,EAAEC,SAAS,EAAEC,WAAW,EAAEC,UAAU,EAAEC,OAAO,EAAEC,WAAW,EAAE,GAAGN,WAAWtB;IACvF,IAAI4B,eAAgBL,MAAMU,MAAM,KAAK,KAAKP,WAAWO,MAAM,KAAK,GAAI,OAAO,EAAE;IAE7E,IAAIV,MAAMU,MAAM,GAAG,GAAG,MAAMQ,eAAe9B;IAC3C,IAAIY,MAAMU,MAAM,GAAG,GAAG,MAAM5C,kBAAkBsB,MAAMC;IAEpD,MAAM,EAAEiD,QAAQ,EAAEC,OAAO,EAAEC,MAAM,EAAE,GAAGnB,kBAAkBrB,OAAOC,WAAWC,aAAaC;IACvF,MAAM6C,eAAe5C,QAAQM,MAAM,GAAG,IAAI,mDAAmD;IAC7F,MAAMuC,WAAW7C,QAAQM,MAAM,GAAG,IAAI,KAAK;IAC3C,MAAMwC,MAAM,CAAC,kBAAkB,EAAE9C,QAAQM,MAAM,GAAG,IAAI,2BAA2B,GAAG;oCAClD,EAAE4B,SAAS,UAAU,EAAEU,aAAa;;IAEpE,EAAEJ,UAAU;UACN,EAAEL,QAAQ;IAChB,EAAEM,UAAU,CAAC,EAAEC,UAAU;kCACK,EAAEG,UAAU;IAC5C,MAAMtD,OAAO,MAAMP,KAAKQ,OAAO,CAACsD;IAChC,MAAMC,OAAQ,MAAMxD,KAAKyD,GAAG,IAAIZ,WAAYpC,QAAQM,MAAM,GAAG,IAAI,EAAE,GAAG;QAACqC;KAAM;IAC7E,MAAMM,UAAUjD,QAAQM,MAAM,KAAK,IAAIyC,OAAOA,KAAKG,MAAM,CAAC,CAACC,MAAQnD,QAAQoD,KAAK,CAAC,CAAChD,SAAWxC,OAAOyF,IAAI,CAAC,CAAC5B;oBAA8B0B;uBAApB9F,qBAAoB8F,aAAAA,GAAG,CAAC1B,MAAM,cAAV0B,wBAAAA,aAAc,IAAI/C;;IAC1J,OAAO6C,QAAQK,KAAK,CAAC,GAAGX,OAAOpB,GAAG,CAAC,CAAC,EAAEgC,IAAI,EAAE,GAAM,CAAA;YAAEA;QAAK,CAAA;AAC3D;AAEA,iGAAiG;AACjG,gGAAgG;AAChG,OAAO,SAASC,mBAAmBxE,IAAgB;IACjD,OAAO;QACLyE,OAAO,OAAOpF,OAAOiE,OAASD,aAAarD,MAAMX,OAAOiE,MAAM,MAAMvD,SAASC;QAC7E0E,WAAW,IAAMhE,iBAAiBV;IACpC;AACF"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/open-source/ai/sensemaking/src/store/duckdb/lexical.ts"],"sourcesContent":["// D1: fts BM25 for ranking, contains() scans for exact substring / phrase verification /\n// unspaced scripts, JS snippets for every doc (no snippet()).\nimport { SenseError } from '../../errors.ts';\nimport { foldForSearch, matchesSearchPhrase, type SearchToken, searchTokens } from '../../text/segment.ts';\nimport { getMeta, setMeta } from '../shared.ts';\nimport { withTransaction } from '../transaction.ts';\nimport type { Connection, LexicalHit, LexicalQueryOptions } from '../types.ts';\nimport { type DuckdbFtsBuilders, duckdbFts } from './engine.ts';\nimport { nativeQueryTerms, orderedBm25Sql } from './ordered-bm25.ts';\n\nconst FIELDS = ['title', 'summary', 'text'] as const;\ntype Field = (typeof FIELDS)[number];\n// Mirrors sqlite's bm25(content, 10.0, 5.0, 1.0, ...) column-weight intent; DuckDB's match_bm25 has no\n// field_weights argument (verified 1.4.5 through 1.5.6), so each field is scored separately and combined here.\nconst FIELD_WEIGHT: Record<Field, number> = { title: 10, summary: 5, text: 1 };\n\n// FTS5 operator syntax (sqlite.org/fts5.html sec. 3) that words/substrings below would otherwise silently\n// treat as literal terms (PRINCIPLES: no-silent-modes). Quoted contents stay ignored while their boundary remains detectable.\nconst FTS5_OPERATORS: Array<{ label: string; re: RegExp }> = [\n { label: 'prefix query', re: /[\\p{L}\\p{N}_]+\\*(?=\\s|$)/u },\n { label: 'boolean operator', re: /(?:^|\\s)(?:AND|OR|NOT)(?=\\s|$)/ },\n { label: 'NEAR operator', re: /(?:^|\\s)NEAR\\b/ },\n { label: 'initial-token operator', re: /(?:^|\\s)\\^\\S+/ },\n { label: 'column filter', re: /(?:^|\\s)[\\p{L}_]\\w*\\s*:/u },\n { label: 'grouping operator', re: /[()]/ },\n];\n\nfunction unsupportedOperator(terms: string): { label: string; token: string } | null {\n const withoutPhrases = terms.replace(/\"[^\"]*\"/g, 'phrase');\n for (const { label, re } of FTS5_OPERATORS) {\n const m = withoutPhrases.match(re);\n if (m) return { label, token: m[0].trim() };\n }\n return null;\n}\n\ninterface FtsIndexState {\n stale: boolean;\n}\n\n// Durable across connections/processes (meta table, shared.ts); `stale` is the in-memory cache of\n// it for this connection's lifetime, so a watcher's repeat queries don't re-read meta every time.\nconst FTS_STALE_META_KEY = 'fts_stale';\n\n// One state per connection, not per store instance: reconcileContent (reconcile.ts) marks it\n// stale from outside createLexicalIndex's closure, whether a one-shot open or a watcher's builder ran it.\nconst ftsState = new WeakMap<Connection, FtsIndexState>();\nasync function stateFor(conn: Connection): Promise<FtsIndexState> {\n let state = ftsState.get(conn);\n if (!state) {\n // Anything but a persisted '0' (missing key, or '1') means a rebuild is owed: a cache that\n // never built an index, or one a prior process marked stale and never got to clear.\n state = { stale: (await getMeta(conn, FTS_STALE_META_KEY)) !== '0' };\n ftsState.set(conn, state);\n }\n return state;\n}\n\nasync function hasFtsArtifact(conn: Connection): Promise<boolean> {\n const stmt = await conn.prepare(`\n SELECT 1 AS ready\n FROM duckdb_functions()\n WHERE schema_name = 'fts_main_content'\n AND function_name = 'match_bm25'\n AND function_type = 'macro'\n LIMIT 1\n `);\n return (await stmt.get()) !== undefined;\n}\n\n// Called whenever this store's reconcileContent changes `content`; must run inside that same\n// transaction (reconcile.ts) so a crash never lands the content write without the stale mark.\nexport async function markContentStale(conn: Connection): Promise<void> {\n ftsState.set(conn, { stale: true });\n await setMeta(conn, FTS_STALE_META_KEY, '1');\n}\n\n// A run whose script marks no word boundaries makes match_bm25's whitespace tokenizer index the whole run as one token,\n// so unspaced terms use contains() and quoted phrases use the shared authored-field verifier.\nfunction splitTerms(terms: string): { words: string[]; bareWords: string[]; phraseWords: string[]; substrings: string[]; phrases: SearchToken[][]; emptyPhrase: boolean } {\n const words: string[] = [];\n const bareWords: string[] = [];\n const phraseWords: string[] = [];\n const substrings: string[] = [];\n const phrases: SearchToken[][] = [];\n let emptyPhrase = false;\n const withoutPhrases = terms.replace(/\"([^\"]*)\"/g, (_m, inner: string) => {\n const phrase = inner.trim();\n const phraseTokens = searchTokens(phrase);\n if (phraseTokens.length === 0) {\n emptyPhrase = true;\n return ' ';\n }\n phrases.push(phraseTokens);\n for (const token of phraseTokens) {\n if (token.unspaced) substrings.push(token.text);\n else {\n words.push(token.text);\n phraseWords.push(token.text);\n }\n }\n return ' ';\n });\n for (const tok of withoutPhrases.split(/\\s+/)) {\n if (tok.length === 0) continue;\n for (const token of searchTokens(tok)) {\n if (token.unspaced) substrings.push(token.text);\n else {\n words.push(token.text);\n bareWords.push(token.text);\n }\n }\n }\n return { words, bareWords, phraseWords, substrings, phrases, emptyPhrase };\n}\n\n// No incremental update (verified 1.5.5: PRAGMA create_fts_index is rebuild-only), so this pays\n// the full rebuild whenever `state.stale`, which meta.fts_stale keeps true across processes until cleared below.\nasync function rebuildFts(conn: Connection, state: FtsIndexState): Promise<void> {\n await conn.exec('INSTALL fts; LOAD fts;');\n if (!state.stale && (await hasFtsArtifact(conn))) return;\n state.stale = true;\n await withTransaction(conn, async () => {\n await conn.exec((await duckdbFts()).createIndex());\n // Same transaction as the rebuild: a crash before commit leaves the mark set (a redundant\n // rebuild next time), never a cleared mark serving a half-built index as fresh.\n await setMeta(conn, FTS_STALE_META_KEY, '0');\n });\n state.stale = false;\n}\n\nexport async function prepareFts(conn: Connection): Promise<void> {\n await rebuildFts(conn, await stateFor(conn));\n}\n\nexport async function assertFtsReady(conn: Connection): Promise<void> {\n const state = await stateFor(conn);\n if (state.stale) throw new SenseError('INDEX_NOT_READY', 'the DuckDB lexical index is missing or stale; run `sense build` (or library build(config)) before lexical querying with build disabled');\n try {\n await conn.exec('LOAD fts;');\n if (!(await hasFtsArtifact(conn))) {\n throw new SenseError('INDEX_NOT_READY', 'the DuckDB lexical index artifact is missing; run `sense build` (or library build(config)) before lexical querying with build disabled');\n }\n } catch (err) {\n if (err instanceof SenseError) throw err;\n throw new SenseError('INDEX_NOT_READY', `the DuckDB lexical extension is not prepared (${(err as Error).message}); run \\`sense build\\` (or library build(config))`);\n }\n}\n\n// Field-weighted fragments (title 10 / summary 5 / text 1), with params emitted in the exact left-to-right order the assembled SQL needs for positional `?` binding.\n// score/gate params stay in separate arrays until joined at the end, since the SQL puts every score part before every gate part; sharing one array would shift binds once a query has both words and substrings (PRINCIPLES: no-silent-modes).\nfunction buildScoreAndGate(fts: DuckdbFtsBuilders, words: string[], nativeTerms: string[], bareWords: string[], phraseWords: string[], substrings: string[]): { scoreSql: string; gateSql: string; params: unknown[] } {\n const scoreParts: string[] = [];\n const gateParts: string[] = [];\n const scoreParams: unknown[] = [];\n const gateParams: unknown[] = [];\n\n if (words.length > 0) {\n for (const field of FIELDS) {\n scoreParts.push(`${FIELD_WEIGHT[field]}.0 * COALESCE(${orderedBm25Sql(fts, field, nativeTerms.length)}, 0)`);\n scoreParams.push(...nativeTerms);\n }\n if (bareWords.length > 0) {\n const nativeGate = fts.conjunctiveGate();\n gateParts.push(nativeGate);\n gateParams.push(bareWords.map(foldForSearch).join(' '));\n }\n if (phraseWords.length > 0) {\n const nativeGate = fts.conjunctiveGate();\n const fallback = phraseWords.map((_word) => `(${FIELDS.map((field) => `contains(strip_accents(lower(content.${field})), ?)`).join(' OR ')})`).join(' AND ');\n gateParts.push(`(${nativeGate} OR (${fallback}))`);\n gateParams.push(phraseWords.map(foldForSearch).join(' '));\n gateParams.push(...phraseWords.flatMap((word) => FIELDS.map(() => foldForSearch(word))));\n }\n }\n\n for (const raw of substrings) {\n const needle = raw.toLowerCase();\n for (const field of FIELDS) {\n scoreParts.push(`${FIELD_WEIGHT[field]}.0 * CAST(contains(strip_accents(lower(content.${field})), ?) AS INTEGER)`);\n scoreParams.push(foldForSearch(needle));\n }\n gateParts.push(`(${FIELDS.map((field) => `contains(strip_accents(lower(content.${field})), ?)`).join(' OR ')})`);\n gateParams.push(...FIELDS.map(() => foldForSearch(needle)));\n }\n\n return { scoreSql: scoreParts.join(' + '), gateSql: gateParts.join(' AND ') || 'TRUE', params: [...scoreParams, ...gateParams] };\n}\n\n// Ranked query, scoped by the caller-built SQL fragments (same fragments sqlite's queryLexical takes).\n// Phrase rows carry authored fields through the post-filter; the public result still contains paths only.\nexport async function queryLexical(conn: Connection, terms: string, opts: LexicalQueryOptions): Promise<LexicalHit[]> {\n const unsupported = unsupportedOperator(terms);\n if (unsupported !== null) {\n throw new SenseError('STORE_CAPABILITY_MISSING', `store \"duckdb\" does not implement FTS5's ${unsupported.label} (\"${unsupported.token}\") in this build; rephrase \"${terms.trim()}\" without it, or set \"store\" to \"sqlite\" in this tree's config to search it as written`);\n }\n const { whereJoin, whereCond, scopeCond, limit } = opts;\n const { words, bareWords, phraseWords, substrings, phrases, emptyPhrase } = splitTerms(terms);\n if (emptyPhrase || (words.length === 0 && substrings.length === 0)) return [];\n\n if (words.length > 0) await assertFtsReady(conn);\n const fts = await duckdbFts();\n const nativeTerms = words.length > 0 ? await nativeQueryTerms(conn, fts, words.map(foldForSearch).join(' ')) : [];\n\n const { scoreSql, gateSql, params } = buildScoreAndGate(fts, words, nativeTerms, bareWords, phraseWords, substrings);\n const selectFields = phrases.length > 0 ? ', content.title, content.summary, content.text' : '';\n const sqlLimit = phrases.length > 0 ? '' : ' LIMIT ?';\n const sql = `SELECT path, score${phrases.length > 0 ? ', title, summary, text' : ''} FROM (\n SELECT content.\"path\" AS path, (${scoreSql}) AS score${selectFields}\n FROM content\n ${whereJoin}\n WHERE ${gateSql}\n ${whereCond} ${scopeCond}\n ) sq ORDER BY score DESC, path${sqlLimit}`;\n const stmt = await conn.prepare(sql);\n const rows = (await stmt.all(...params, ...(phrases.length > 0 ? [] : [limit]))) as Array<LexicalHit & Partial<Record<Field, string>>>;\n const matches = phrases.length === 0 ? rows : rows.filter((row) => phrases.every((phrase) => FIELDS.some((field) => matchesSearchPhrase(row[field] ?? '', phrase))));\n return matches.slice(0, limit).map(({ path }) => ({ path }));\n}\n\n// `stale`'s first read for a connection comes from meta.fts_stale (a fresh connection can't know\n// otherwise), and reconcileContent (reconcile.ts) marks it stale again on every content change.\nexport function createLexicalIndex(conn: Connection): { query: (terms: string, opts: LexicalQueryOptions) => Promise<LexicalHit[]>; markStale: () => Promise<void> } {\n return {\n query: async (terms, opts) => queryLexical(conn, terms, opts),\n markStale: () => markContentStale(conn),\n };\n}\n"],"names":["SenseError","foldForSearch","matchesSearchPhrase","searchTokens","getMeta","setMeta","withTransaction","duckdbFts","nativeQueryTerms","orderedBm25Sql","FIELDS","FIELD_WEIGHT","title","summary","text","FTS5_OPERATORS","label","re","unsupportedOperator","terms","withoutPhrases","replace","m","match","token","trim","FTS_STALE_META_KEY","ftsState","WeakMap","stateFor","conn","state","get","stale","set","hasFtsArtifact","stmt","prepare","undefined","markContentStale","splitTerms","words","bareWords","phraseWords","substrings","phrases","emptyPhrase","_m","inner","phrase","phraseTokens","length","push","unspaced","tok","split","rebuildFts","exec","createIndex","prepareFts","assertFtsReady","err","message","buildScoreAndGate","fts","nativeTerms","scoreParts","gateParts","scoreParams","gateParams","field","nativeGate","conjunctiveGate","map","join","fallback","_word","flatMap","word","raw","needle","toLowerCase","scoreSql","gateSql","params","queryLexical","opts","unsupported","whereJoin","whereCond","scopeCond","limit","selectFields","sqlLimit","sql","rows","all","matches","filter","row","every","some","slice","path","createLexicalIndex","query","markStale"],"mappings":"AAAA,yFAAyF;AACzF,8DAA8D;AAC9D,SAASA,UAAU,QAAQ,kBAAkB;AAC7C,SAASC,aAAa,EAAEC,mBAAmB,EAAoBC,YAAY,QAAQ,wBAAwB;AAC3G,SAASC,OAAO,EAAEC,OAAO,QAAQ,eAAe;AAChD,SAASC,eAAe,QAAQ,oBAAoB;AAEpD,SAAiCC,SAAS,QAAQ,cAAc;AAChE,SAASC,gBAAgB,EAAEC,cAAc,QAAQ,oBAAoB;AAErE,MAAMC,SAAS;IAAC;IAAS;IAAW;CAAO;AAE3C,uGAAuG;AACvG,+GAA+G;AAC/G,MAAMC,eAAsC;IAAEC,OAAO;IAAIC,SAAS;IAAGC,MAAM;AAAE;AAE7E,0GAA0G;AAC1G,8HAA8H;AAC9H,MAAMC,iBAAuD;IAC3D;QAAEC,OAAO;QAAgBC,IAAI;IAA4B;IACzD;QAAED,OAAO;QAAoBC,IAAI;IAAiC;IAClE;QAAED,OAAO;QAAiBC,IAAI;IAAiB;IAC/C;QAAED,OAAO;QAA0BC,IAAI;IAAgB;IACvD;QAAED,OAAO;QAAiBC,IAAI;IAA2B;IACzD;QAAED,OAAO;QAAqBC,IAAI;IAAO;CAC1C;AAED,SAASC,oBAAoBC,KAAa;IACxC,MAAMC,iBAAiBD,MAAME,OAAO,CAAC,YAAY;IACjD,KAAK,MAAM,EAAEL,KAAK,EAAEC,EAAE,EAAE,IAAIF,eAAgB;QAC1C,MAAMO,IAAIF,eAAeG,KAAK,CAACN;QAC/B,IAAIK,GAAG,OAAO;YAAEN;YAAOQ,OAAOF,CAAC,CAAC,EAAE,CAACG,IAAI;QAAG;IAC5C;IACA,OAAO;AACT;AAMA,kGAAkG;AAClG,kGAAkG;AAClG,MAAMC,qBAAqB;AAE3B,6FAA6F;AAC7F,0GAA0G;AAC1G,MAAMC,WAAW,IAAIC;AACrB,eAAeC,SAASC,IAAgB;IACtC,IAAIC,QAAQJ,SAASK,GAAG,CAACF;IACzB,IAAI,CAACC,OAAO;QACV,2FAA2F;QAC3F,oFAAoF;QACpFA,QAAQ;YAAEE,OAAO,AAAC,MAAM7B,QAAQ0B,MAAMJ,wBAAyB;QAAI;QACnEC,SAASO,GAAG,CAACJ,MAAMC;IACrB;IACA,OAAOA;AACT;AAEA,eAAeI,eAAeL,IAAgB;IAC5C,MAAMM,OAAO,MAAMN,KAAKO,OAAO,CAAC,CAAC;;;;;;;EAOjC,CAAC;IACD,OAAO,AAAC,MAAMD,KAAKJ,GAAG,OAAQM;AAChC;AAEA,6FAA6F;AAC7F,8FAA8F;AAC9F,OAAO,eAAeC,iBAAiBT,IAAgB;IACrDH,SAASO,GAAG,CAACJ,MAAM;QAAEG,OAAO;IAAK;IACjC,MAAM5B,QAAQyB,MAAMJ,oBAAoB;AAC1C;AAEA,wHAAwH;AACxH,8FAA8F;AAC9F,SAASc,WAAWrB,KAAa;IAC/B,MAAMsB,QAAkB,EAAE;IAC1B,MAAMC,YAAsB,EAAE;IAC9B,MAAMC,cAAwB,EAAE;IAChC,MAAMC,aAAuB,EAAE;IAC/B,MAAMC,UAA2B,EAAE;IACnC,IAAIC,cAAc;IAClB,MAAM1B,iBAAiBD,MAAME,OAAO,CAAC,cAAc,CAAC0B,IAAIC;QACtD,MAAMC,SAASD,MAAMvB,IAAI;QACzB,MAAMyB,eAAe/C,aAAa8C;QAClC,IAAIC,aAAaC,MAAM,KAAK,GAAG;YAC7BL,cAAc;YACd,OAAO;QACT;QACAD,QAAQO,IAAI,CAACF;QACb,KAAK,MAAM1B,SAAS0B,aAAc;YAChC,IAAI1B,MAAM6B,QAAQ,EAAET,WAAWQ,IAAI,CAAC5B,MAAMV,IAAI;iBACzC;gBACH2B,MAAMW,IAAI,CAAC5B,MAAMV,IAAI;gBACrB6B,YAAYS,IAAI,CAAC5B,MAAMV,IAAI;YAC7B;QACF;QACA,OAAO;IACT;IACA,KAAK,MAAMwC,OAAOlC,eAAemC,KAAK,CAAC,OAAQ;QAC7C,IAAID,IAAIH,MAAM,KAAK,GAAG;QACtB,KAAK,MAAM3B,SAASrB,aAAamD,KAAM;YACrC,IAAI9B,MAAM6B,QAAQ,EAAET,WAAWQ,IAAI,CAAC5B,MAAMV,IAAI;iBACzC;gBACH2B,MAAMW,IAAI,CAAC5B,MAAMV,IAAI;gBACrB4B,UAAUU,IAAI,CAAC5B,MAAMV,IAAI;YAC3B;QACF;IACF;IACA,OAAO;QAAE2B;QAAOC;QAAWC;QAAaC;QAAYC;QAASC;IAAY;AAC3E;AAEA,gGAAgG;AAChG,iHAAiH;AACjH,eAAeU,WAAW1B,IAAgB,EAAEC,KAAoB;IAC9D,MAAMD,KAAK2B,IAAI,CAAC;IAChB,IAAI,CAAC1B,MAAME,KAAK,IAAK,MAAME,eAAeL,OAAQ;IAClDC,MAAME,KAAK,GAAG;IACd,MAAM3B,gBAAgBwB,MAAM;QAC1B,MAAMA,KAAK2B,IAAI,CAAC,AAAC,CAAA,MAAMlD,WAAU,EAAGmD,WAAW;QAC/C,0FAA0F;QAC1F,gFAAgF;QAChF,MAAMrD,QAAQyB,MAAMJ,oBAAoB;IAC1C;IACAK,MAAME,KAAK,GAAG;AAChB;AAEA,OAAO,eAAe0B,WAAW7B,IAAgB;IAC/C,MAAM0B,WAAW1B,MAAM,MAAMD,SAASC;AACxC;AAEA,OAAO,eAAe8B,eAAe9B,IAAgB;IACnD,MAAMC,QAAQ,MAAMF,SAASC;IAC7B,IAAIC,MAAME,KAAK,EAAE,MAAM,IAAIjC,WAAW,mBAAmB;IACzD,IAAI;QACF,MAAM8B,KAAK2B,IAAI,CAAC;QAChB,IAAI,CAAE,MAAMtB,eAAeL,OAAQ;YACjC,MAAM,IAAI9B,WAAW,mBAAmB;QAC1C;IACF,EAAE,OAAO6D,KAAK;QACZ,IAAIA,eAAe7D,YAAY,MAAM6D;QACrC,MAAM,IAAI7D,WAAW,mBAAmB,CAAC,8CAA8C,EAAE,AAAC6D,IAAcC,OAAO,CAAC,iDAAiD,CAAC;IACpK;AACF;AAEA,qKAAqK;AACrK,+OAA+O;AAC/O,SAASC,kBAAkBC,GAAsB,EAAEvB,KAAe,EAAEwB,WAAqB,EAAEvB,SAAmB,EAAEC,WAAqB,EAAEC,UAAoB;IACzJ,MAAMsB,aAAuB,EAAE;IAC/B,MAAMC,YAAsB,EAAE;IAC9B,MAAMC,cAAyB,EAAE;IACjC,MAAMC,aAAwB,EAAE;IAEhC,IAAI5B,MAAMU,MAAM,GAAG,GAAG;QACpB,KAAK,MAAMmB,SAAS5D,OAAQ;YAC1BwD,WAAWd,IAAI,CAAC,GAAGzC,YAAY,CAAC2D,MAAM,CAAC,cAAc,EAAE7D,eAAeuD,KAAKM,OAAOL,YAAYd,MAAM,EAAE,IAAI,CAAC;YAC3GiB,YAAYhB,IAAI,IAAIa;QACtB;QACA,IAAIvB,UAAUS,MAAM,GAAG,GAAG;YACxB,MAAMoB,aAAaP,IAAIQ,eAAe;YACtCL,UAAUf,IAAI,CAACmB;YACfF,WAAWjB,IAAI,CAACV,UAAU+B,GAAG,CAACxE,eAAeyE,IAAI,CAAC;QACpD;QACA,IAAI/B,YAAYQ,MAAM,GAAG,GAAG;YAC1B,MAAMoB,aAAaP,IAAIQ,eAAe;YACtC,MAAMG,WAAWhC,YAAY8B,GAAG,CAAC,CAACG,QAAU,CAAC,CAAC,EAAElE,OAAO+D,GAAG,CAAC,CAACH,QAAU,CAAC,qCAAqC,EAAEA,MAAM,MAAM,CAAC,EAAEI,IAAI,CAAC,QAAQ,CAAC,CAAC,EAAEA,IAAI,CAAC;YACnJP,UAAUf,IAAI,CAAC,CAAC,CAAC,EAAEmB,WAAW,KAAK,EAAEI,SAAS,EAAE,CAAC;YACjDN,WAAWjB,IAAI,CAACT,YAAY8B,GAAG,CAACxE,eAAeyE,IAAI,CAAC;YACpDL,WAAWjB,IAAI,IAAIT,YAAYkC,OAAO,CAAC,CAACC,OAASpE,OAAO+D,GAAG,CAAC,IAAMxE,cAAc6E;QAClF;IACF;IAEA,KAAK,MAAMC,OAAOnC,WAAY;QAC5B,MAAMoC,SAASD,IAAIE,WAAW;QAC9B,KAAK,MAAMX,SAAS5D,OAAQ;YAC1BwD,WAAWd,IAAI,CAAC,GAAGzC,YAAY,CAAC2D,MAAM,CAAC,+CAA+C,EAAEA,MAAM,kBAAkB,CAAC;YACjHF,YAAYhB,IAAI,CAACnD,cAAc+E;QACjC;QACAb,UAAUf,IAAI,CAAC,CAAC,CAAC,EAAE1C,OAAO+D,GAAG,CAAC,CAACH,QAAU,CAAC,qCAAqC,EAAEA,MAAM,MAAM,CAAC,EAAEI,IAAI,CAAC,QAAQ,CAAC,CAAC;QAC/GL,WAAWjB,IAAI,IAAI1C,OAAO+D,GAAG,CAAC,IAAMxE,cAAc+E;IACpD;IAEA,OAAO;QAAEE,UAAUhB,WAAWQ,IAAI,CAAC;QAAQS,SAAShB,UAAUO,IAAI,CAAC,YAAY;QAAQU,QAAQ;eAAIhB;eAAgBC;SAAW;IAAC;AACjI;AAEA,uGAAuG;AACvG,0GAA0G;AAC1G,OAAO,eAAegB,aAAavD,IAAgB,EAAEX,KAAa,EAAEmE,IAAyB;IAC3F,MAAMC,cAAcrE,oBAAoBC;IACxC,IAAIoE,gBAAgB,MAAM;QACxB,MAAM,IAAIvF,WAAW,4BAA4B,CAAC,yCAAyC,EAAEuF,YAAYvE,KAAK,CAAC,GAAG,EAAEuE,YAAY/D,KAAK,CAAC,4BAA4B,EAAEL,MAAMM,IAAI,GAAG,sFAAsF,CAAC;IAC1Q;IACA,MAAM,EAAE+D,SAAS,EAAEC,SAAS,EAAEC,SAAS,EAAEC,KAAK,EAAE,GAAGL;IACnD,MAAM,EAAE7C,KAAK,EAAEC,SAAS,EAAEC,WAAW,EAAEC,UAAU,EAAEC,OAAO,EAAEC,WAAW,EAAE,GAAGN,WAAWrB;IACvF,IAAI2B,eAAgBL,MAAMU,MAAM,KAAK,KAAKP,WAAWO,MAAM,KAAK,GAAI,OAAO,EAAE;IAE7E,IAAIV,MAAMU,MAAM,GAAG,GAAG,MAAMS,eAAe9B;IAC3C,MAAMkC,MAAM,MAAMzD;IAClB,MAAM0D,cAAcxB,MAAMU,MAAM,GAAG,IAAI,MAAM3C,iBAAiBsB,MAAMkC,KAAKvB,MAAMgC,GAAG,CAACxE,eAAeyE,IAAI,CAAC,QAAQ,EAAE;IAEjH,MAAM,EAAEQ,QAAQ,EAAEC,OAAO,EAAEC,MAAM,EAAE,GAAGrB,kBAAkBC,KAAKvB,OAAOwB,aAAavB,WAAWC,aAAaC;IACzG,MAAMgD,eAAe/C,QAAQM,MAAM,GAAG,IAAI,mDAAmD;IAC7F,MAAM0C,WAAWhD,QAAQM,MAAM,GAAG,IAAI,KAAK;IAC3C,MAAM2C,MAAM,CAAC,kBAAkB,EAAEjD,QAAQM,MAAM,GAAG,IAAI,2BAA2B,GAAG;oCAClD,EAAE+B,SAAS,UAAU,EAAEU,aAAa;;IAEpE,EAAEJ,UAAU;UACN,EAAEL,QAAQ;IAChB,EAAEM,UAAU,CAAC,EAAEC,UAAU;kCACK,EAAEG,UAAU;IAC5C,MAAMzD,OAAO,MAAMN,KAAKO,OAAO,CAACyD;IAChC,MAAMC,OAAQ,MAAM3D,KAAK4D,GAAG,IAAIZ,WAAYvC,QAAQM,MAAM,GAAG,IAAI,EAAE,GAAG;QAACwC;KAAM;IAC7E,MAAMM,UAAUpD,QAAQM,MAAM,KAAK,IAAI4C,OAAOA,KAAKG,MAAM,CAAC,CAACC,MAAQtD,QAAQuD,KAAK,CAAC,CAACnD,SAAWvC,OAAO2F,IAAI,CAAC,CAAC/B;oBAA8B6B;uBAApBjG,qBAAoBiG,aAAAA,GAAG,CAAC7B,MAAM,cAAV6B,wBAAAA,aAAc,IAAIlD;;IAC1J,OAAOgD,QAAQK,KAAK,CAAC,GAAGX,OAAOlB,GAAG,CAAC,CAAC,EAAE8B,IAAI,EAAE,GAAM,CAAA;YAAEA;QAAK,CAAA;AAC3D;AAEA,iGAAiG;AACjG,gGAAgG;AAChG,OAAO,SAASC,mBAAmB1E,IAAgB;IACjD,OAAO;QACL2E,OAAO,OAAOtF,OAAOmE,OAASD,aAAavD,MAAMX,OAAOmE;QACxDoB,WAAW,IAAMnE,iBAAiBT;IACpC;AACF"}
@@ -1,4 +1,6 @@
1
+ import { type NativeModule } from '../native.js';
1
2
  export declare const DUCKDB_PACKAGE = "@duckdb/node-api";
2
- export declare const DUCKDB_VERSION = "1.5.5-r.5";
3
- export declare const DUCKDB_INSTALL_SPEC = "@duckdb/node-api@1.5.5-r.5";
4
- export declare function duckdbApi(): Promise<typeof import('@duckdb/node-api')>;
3
+ type DuckdbApi = typeof import('@duckdb/node-api');
4
+ export declare function duckdbNative(): Promise<NativeModule<DuckdbApi>>;
5
+ export declare function duckdbApi(): Promise<DuckdbApi>;
6
+ export {};
@@ -1,18 +1,20 @@
1
- import { loadOrInstall, packageNodeModules } from '../native.js';
1
+ import { loadOrInstall, packageNodeModules, peerRange } from '../native.js';
2
2
  export const DUCKDB_PACKAGE = '@duckdb/node-api';
3
- export const DUCKDB_VERSION = '1.5.5-r.5';
4
- export const DUCKDB_INSTALL_SPEC = `${DUCKDB_PACKAGE}@${DUCKDB_VERSION}`;
5
- const DUCKDB = {
6
- store: 'duckdb',
7
- pkg: DUCKDB_PACKAGE,
8
- sizeHint: '~110MB',
9
- version: DUCKDB_VERSION,
10
- installSpec: DUCKDB_INSTALL_SPEC
11
- };
12
- let duckdbApiPromise;
3
+ let duckdbNativePromise;
13
4
  // The one accessor every consumer of @duckdb/node-api must go through: after a fresh install in this process, Node keeps returning
14
5
  // the pre-install "not found" miss to any new bare-specifier lookup of the same path, so caching the resolved module here means only the first caller ever resolves it.
15
- export function duckdbApi() {
16
- if (!duckdbApiPromise) duckdbApiPromise = loadOrInstall(DUCKDB, packageNodeModules());
17
- return duckdbApiPromise;
6
+ export function duckdbNative() {
7
+ if (!duckdbNativePromise) {
8
+ const descriptor = {
9
+ store: 'duckdb',
10
+ pkg: DUCKDB_PACKAGE,
11
+ sizeHint: '~110MB',
12
+ range: peerRange(DUCKDB_PACKAGE)
13
+ };
14
+ duckdbNativePromise = loadOrInstall(descriptor, packageNodeModules());
15
+ }
16
+ return duckdbNativePromise;
17
+ }
18
+ export async function duckdbApi() {
19
+ return (await duckdbNative()).api;
18
20
  }
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/duckdb/native.ts"],"sourcesContent":["import { loadOrInstall, type NativeDescriptor, packageNodeModules } from '../native.ts';\n\nexport const DUCKDB_PACKAGE = '@duckdb/node-api';\nexport const DUCKDB_VERSION = '1.5.5-r.5';\nexport const DUCKDB_INSTALL_SPEC = `${DUCKDB_PACKAGE}@${DUCKDB_VERSION}`;\n\nconst DUCKDB: NativeDescriptor = {\n store: 'duckdb',\n pkg: DUCKDB_PACKAGE,\n sizeHint: '~110MB',\n version: DUCKDB_VERSION,\n installSpec: DUCKDB_INSTALL_SPEC,\n};\n\nlet duckdbApiPromise: Promise<typeof import('@duckdb/node-api')> | undefined;\n\n// The one accessor every consumer of @duckdb/node-api must go through: after a fresh install in this process, Node keeps returning\n// the pre-install \"not found\" miss to any new bare-specifier lookup of the same path, so caching the resolved module here means only the first caller ever resolves it.\nexport function duckdbApi(): Promise<typeof import('@duckdb/node-api')> {\n if (!duckdbApiPromise) duckdbApiPromise = loadOrInstall<typeof import('@duckdb/node-api')>(DUCKDB, packageNodeModules());\n return duckdbApiPromise;\n}\n"],"names":["loadOrInstall","packageNodeModules","DUCKDB_PACKAGE","DUCKDB_VERSION","DUCKDB_INSTALL_SPEC","DUCKDB","store","pkg","sizeHint","version","installSpec","duckdbApiPromise","duckdbApi"],"mappings":"AAAA,SAASA,aAAa,EAAyBC,kBAAkB,QAAQ,eAAe;AAExF,OAAO,MAAMC,iBAAiB,mBAAmB;AACjD,OAAO,MAAMC,iBAAiB,YAAY;AAC1C,OAAO,MAAMC,sBAAsB,GAAGF,eAAe,CAAC,EAAEC,gBAAgB,CAAC;AAEzE,MAAME,SAA2B;IAC/BC,OAAO;IACPC,KAAKL;IACLM,UAAU;IACVC,SAASN;IACTO,aAAaN;AACf;AAEA,IAAIO;AAEJ,mIAAmI;AACnI,wKAAwK;AACxK,OAAO,SAASC;IACd,IAAI,CAACD,kBAAkBA,mBAAmBX,cAAiDK,QAAQJ;IACnG,OAAOU;AACT"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/open-source/ai/sensemaking/src/store/duckdb/native.ts"],"sourcesContent":["import { loadOrInstall, type NativeDescriptor, type NativeModule, packageNodeModules, peerRange } from '../native.ts';\n\nexport const DUCKDB_PACKAGE = '@duckdb/node-api';\n\ntype DuckdbApi = typeof import('@duckdb/node-api');\n\nlet duckdbNativePromise: Promise<NativeModule<DuckdbApi>> | undefined;\n\n// The one accessor every consumer of @duckdb/node-api must go through: after a fresh install in this process, Node keeps returning\n// the pre-install \"not found\" miss to any new bare-specifier lookup of the same path, so caching the resolved module here means only the first caller ever resolves it.\nexport function duckdbNative(): Promise<NativeModule<DuckdbApi>> {\n if (!duckdbNativePromise) {\n const descriptor: NativeDescriptor = { store: 'duckdb', pkg: DUCKDB_PACKAGE, sizeHint: '~110MB', range: peerRange(DUCKDB_PACKAGE) };\n duckdbNativePromise = loadOrInstall<DuckdbApi>(descriptor, packageNodeModules());\n }\n return duckdbNativePromise;\n}\n\nexport async function duckdbApi(): Promise<DuckdbApi> {\n return (await duckdbNative()).api;\n}\n"],"names":["loadOrInstall","packageNodeModules","peerRange","DUCKDB_PACKAGE","duckdbNativePromise","duckdbNative","descriptor","store","pkg","sizeHint","range","duckdbApi","api"],"mappings":"AAAA,SAASA,aAAa,EAA4CC,kBAAkB,EAAEC,SAAS,QAAQ,eAAe;AAEtH,OAAO,MAAMC,iBAAiB,mBAAmB;AAIjD,IAAIC;AAEJ,mIAAmI;AACnI,wKAAwK;AACxK,OAAO,SAASC;IACd,IAAI,CAACD,qBAAqB;QACxB,MAAME,aAA+B;YAAEC,OAAO;YAAUC,KAAKL;YAAgBM,UAAU;YAAUC,OAAOR,UAAUC;QAAgB;QAClIC,sBAAsBJ,cAAyBM,YAAYL;IAC7D;IACA,OAAOG;AACT;AAEA,OAAO,eAAeO;IACpB,OAAO,AAAC,CAAA,MAAMN,cAAa,EAAGO,GAAG;AACnC"}
@@ -1,11 +1,13 @@
1
1
  import { STORE_DIMS } from '../../embed/types.js';
2
2
  import { SenseError } from '../../errors.js';
3
3
  import { activeFeatures } from '../../features/index.js';
4
+ import { ENGINE_CHECK_DOCS, engineIdentity, npmRemedy, checkEngine as runEngineCheck } from '../engine.js';
5
+ import { peerRange } from '../native.js';
4
6
  import { openWithDialect } from '../open.js';
5
7
  import { getMeta, setMeta } from '../shared.js';
6
8
  import { createConnection } from './connection.js';
7
- import { assertFtsReady, prepareFts } from './lexical.js';
8
- import { DUCKDB_PACKAGE, duckdbApi } from './native.js';
9
+ import { assertFtsReady, prepareFts, queryLexical } from './lexical.js';
10
+ import { DUCKDB_PACKAGE, duckdbApi, duckdbNative } from './native.js';
9
11
  import { duckdbDialect } from './reconcile.js';
10
12
  import { registerFunctions } from './sql-functions.js';
11
13
  import { createStore } from './store.js';
@@ -89,6 +91,40 @@ async function connect(dbPath, _cfg, options) {
89
91
  throw err;
90
92
  }
91
93
  }
94
+ async function duckdbEngineIdentity() {
95
+ return engineIdentity(DUCKDB_PACKAGE, (await duckdbNative()).version);
96
+ }
97
+ // The real schema, index build and lexical query over store/engine.ts's fixture, in memory.
98
+ async function checkDuckdbEngine() {
99
+ const { api, version } = await duckdbNative();
100
+ const instance = await api.DuckDBInstance.create(':memory:');
101
+ const handle = {
102
+ instance,
103
+ duckdb: await instance.connect()
104
+ };
105
+ try {
106
+ const conn = createConnection(handle.duckdb);
107
+ await conn.exec('CREATE TABLE meta (key TEXT PRIMARY KEY, value TEXT)');
108
+ await ensureCoreSchema(handle, conn);
109
+ await conn.runBatch('INSERT INTO content ("path", title, summary, text) VALUES (?, ?, ?, ?)', ENGINE_CHECK_DOCS.map((doc)=>[
110
+ doc.path,
111
+ doc.title,
112
+ doc.summary,
113
+ doc.text
114
+ ]));
115
+ await runEngineCheck('duckdb', engineIdentity(DUCKDB_PACKAGE, version), npmRemedy(DUCKDB_PACKAGE, peerRange(DUCKDB_PACKAGE)), async (terms)=>{
116
+ await prepareFts(conn);
117
+ return (await queryLexical(conn, terms, {
118
+ whereJoin: '',
119
+ whereCond: '',
120
+ scopeCond: '',
121
+ limit: 10
122
+ })).map((hit)=>hit.path);
123
+ });
124
+ } finally{
125
+ await close(handle);
126
+ }
127
+ }
92
128
  // This store's dialect (types.ts's OpenDialect) for the shared orchestration in store/open.ts.
93
129
  export const duckdbOpenDialect = {
94
130
  filename: DB_FILENAME,
@@ -100,6 +136,8 @@ export const duckdbOpenDialect = {
100
136
  // way: posix "Could not set lock on file ... Conflicting lock is held in <exe> (PID n)", Windows
101
137
  // "Cannot open file ... being used by another process". Both are the same condition.
102
138
  isLocked: (err)=>/Could not set lock on file|being used by another process/.test(err.message),
139
+ engineIdentity: duckdbEngineIdentity,
140
+ checkEngine: checkDuckdbEngine,
103
141
  ensureCoreSchema,
104
142
  ensureFeatureSchema,
105
143
  prepareConnection,
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/duckdb/open.ts"],"sourcesContent":["import type { DuckDBConnection, DuckDBInstance } from '@duckdb/node-api';\nimport type { Config, ResolvedConfig } from '../../config/index.ts';\nimport { STORE_DIMS } from '../../embed/types.ts';\nimport { SenseError } from '../../errors.ts';\nimport { activeFeatures } from '../../features/index.ts';\nimport type { InternalOpenOptions, OpenResult } from '../open.ts';\nimport { openWithDialect } from '../open.ts';\nimport { getMeta, setMeta } from '../shared.ts';\nimport type { Connection, OpenDialect } from '../types.ts';\nimport { createConnection } from './connection.ts';\nimport { assertFtsReady, prepareFts } from './lexical.ts';\nimport { DUCKDB_PACKAGE, duckdbApi } from './native.ts';\nimport { duckdbDialect } from './reconcile.ts';\nimport { registerFunctions } from './sql-functions.ts';\nimport { createStore } from './store.ts';\nimport { createVectorWriteStage } from './vectors.ts';\n\nexport const DB_FILENAME = 'cache.duckdb';\n// Independent of sqlite's SCHEMA_VERSION -- the two stores' cache shapes evolve separately (VARIANT frontmatter columns vs untyped).\n// The store name already joins the feature signature, so switching a config's `store` key rebuilds rather than reusing the other engine's cache.\nexport const SCHEMA_VERSION = '6';\n\nexport type { OpenResult };\n\n// The connection plus this store's own native handles (types.ts's OpenDialect<Handle>): the\n// instance owns the WAL and must outlive the connection borrowed from it.\ninterface DuckdbHandle {\n instance: DuckDBInstance;\n duckdb: DuckDBConnection;\n}\n\n// `content` is a plain table (not FTS-virtual): build prepares the persistent FTS index over it,\n// while contains() verification reads the table directly.\n// No tokenizer resolution: this store always uses the fts extension's default (porter) stemmer.\nasync function ensureCoreSchema(_handle: DuckdbHandle, conn: Connection): Promise<void> {\n await conn.exec(`CREATE TABLE IF NOT EXISTS frontmatter (\"path\" TEXT PRIMARY KEY, \"_mtime\" DOUBLE, \"_ctime\" DOUBLE, \"_size\" INTEGER, \"_parse_error\" TEXT)`);\n await conn.exec(`CREATE TABLE IF NOT EXISTS content (\"path\" TEXT PRIMARY KEY, title TEXT, summary TEXT, text TEXT)`);\n await conn.exec(`CREATE TABLE IF NOT EXISTS indexed_sources (\"path\" TEXT PRIMARY KEY, text TEXT NOT NULL)`);\n await conn.exec(`CREATE TABLE IF NOT EXISTS preset_files (\"path\" TEXT, preset TEXT, PRIMARY KEY (\"path\", preset))`);\n await conn.exec('CREATE INDEX IF NOT EXISTS preset_files_preset ON preset_files(preset)');\n if ((await getMeta(conn, 'schema_version')) === null) await setMeta(conn, 'schema_version', SCHEMA_VERSION);\n}\n\nasync function ensureFeatureSchema(_handle: DuckdbHandle, conn: Connection, cfg: Config): Promise<void> {\n for (const feature of activeFeatures(cfg)) {\n // Native FLOAT[STORE_DIMS] instead of the embed feature's engine-neutral BLOB+scale DDL (vectors.ts). `scale` is kept, unused,\n // so the feature's shared reconcile-time INSERT/DELETE (features/embed.ts) names a column that exists on both stores.\n if (feature.name === 'embed') {\n await conn.exec(`CREATE TABLE IF NOT EXISTS embeddings (\"path\" TEXT, chunk INTEGER, start_line INTEGER, end_line INTEGER, content_identity TEXT NOT NULL, scale REAL, vector FLOAT[${STORE_DIMS}], PRIMARY KEY (\"path\", chunk))`);\n continue;\n }\n await feature.schema(conn);\n }\n}\n\nasync function prepareConnection(handle: DuckdbHandle, _conn: Connection, _cfg: Config): Promise<void> {\n await createVectorWriteStage(handle.duckdb, STORE_DIMS);\n}\n\n// Order matters: the connection must be gone before the instance closes the WAL.\nasync function close(handle: DuckdbHandle): Promise<void> {\n handle.duckdb.disconnectSync();\n handle.instance.closeSync();\n}\n\nasync function connect(dbPath: string, _cfg: ResolvedConfig, options: { existingOnly: boolean; observational: boolean }): Promise<{ handle: DuckdbHandle; conn: Connection }> {\n // Dynamic, not a top-level import: sqlite trees must never attempt to resolve this optional peer dependency, so nothing imports\n // it as a value until a duckdb tree opens (types-only imports are erased). Installed on first use if missing, shared with sql-functions.ts and vectors.ts via native.ts's duckdbApi.\n let DuckDBInstance: typeof import('@duckdb/node-api').DuckDBInstance;\n try {\n ({ DuckDBInstance } = await duckdbApi());\n } catch (err) {\n if (err instanceof SenseError) throw err;\n throw new SenseError('STORE_DEPENDENCY_MISSING', `store \"duckdb\" needs the ${DUCKDB_PACKAGE} package (${(err as Error).message})`);\n }\n\n let duckdb: DuckDBConnection;\n let instance: DuckDBInstance | undefined;\n try {\n // On-disk files default to an older storage format for cross-version compatibility, which rejects VARIANT columns (\"VARIANT\n // columns are not supported in storage versions prior to v1.5.0\"); this store's dynamic frontmatter columns need VARIANT (reconcile.ts), so the floor is pinned explicitly.\n instance = await DuckDBInstance.create(dbPath, options.observational ? { access_mode: 'READ_ONLY' } : { storage_compatibility_version: 'v1.5.0' });\n duckdb = await instance.connect();\n } catch (err) {\n // create() may have succeeded before connect() failed: close it, or its WAL stays open.\n instance?.closeSync();\n throw new SenseError('STORE_DEPENDENCY_MISSING', `store \"duckdb\" failed to open ${dbPath}: ${(err as Error).message}`);\n }\n // A throw below would leak the open instance, whose WAL then locks the .duckdb file undeletable on Windows.\n try {\n await registerFunctions(duckdb);\n const conn = createConnection(duckdb);\n return { handle: { instance, duckdb }, conn };\n } catch (err) {\n await close({ instance, duckdb });\n throw err;\n }\n}\n\n// This store's dialect (types.ts's OpenDialect) for the shared orchestration in store/open.ts.\nexport const duckdbOpenDialect: OpenDialect<DuckdbHandle> = {\n filename: DB_FILENAME,\n schemaVersion: SCHEMA_VERSION,\n reconcileDialect: duckdbDialect,\n connect,\n close,\n // duckdb's file lock spans the connection's life, and each platform words the refusal its own\n // way: posix \"Could not set lock on file ... Conflicting lock is held in <exe> (PID n)\", Windows\n // \"Cannot open file ... being used by another process\". Both are the same condition.\n isLocked: (err) => /Could not set lock on file|being used by another process/.test(err.message),\n ensureCoreSchema,\n ensureFeatureSchema,\n prepareConnection,\n prepareLexical: prepareFts,\n assertLexicalReady: assertFtsReady,\n createStore: (handle, conn) => createStore(handle.instance, handle.duckdb, conn),\n};\n\nexport async function openDuckdb(cfg: ResolvedConfig, options?: InternalOpenOptions): Promise<OpenResult> {\n return openWithDialect(cfg, duckdbOpenDialect, options);\n}\n"],"names":["STORE_DIMS","SenseError","activeFeatures","openWithDialect","getMeta","setMeta","createConnection","assertFtsReady","prepareFts","DUCKDB_PACKAGE","duckdbApi","duckdbDialect","registerFunctions","createStore","createVectorWriteStage","DB_FILENAME","SCHEMA_VERSION","ensureCoreSchema","_handle","conn","exec","ensureFeatureSchema","cfg","feature","name","schema","prepareConnection","handle","_conn","_cfg","duckdb","close","disconnectSync","instance","closeSync","connect","dbPath","options","DuckDBInstance","err","message","create","observational","access_mode","storage_compatibility_version","duckdbOpenDialect","filename","schemaVersion","reconcileDialect","isLocked","test","prepareLexical","assertLexicalReady","openDuckdb"],"mappings":"AAEA,SAASA,UAAU,QAAQ,uBAAuB;AAClD,SAASC,UAAU,QAAQ,kBAAkB;AAC7C,SAASC,cAAc,QAAQ,0BAA0B;AAEzD,SAASC,eAAe,QAAQ,aAAa;AAC7C,SAASC,OAAO,EAAEC,OAAO,QAAQ,eAAe;AAEhD,SAASC,gBAAgB,QAAQ,kBAAkB;AACnD,SAASC,cAAc,EAAEC,UAAU,QAAQ,eAAe;AAC1D,SAASC,cAAc,EAAEC,SAAS,QAAQ,cAAc;AACxD,SAASC,aAAa,QAAQ,iBAAiB;AAC/C,SAASC,iBAAiB,QAAQ,qBAAqB;AACvD,SAASC,WAAW,QAAQ,aAAa;AACzC,SAASC,sBAAsB,QAAQ,eAAe;AAEtD,OAAO,MAAMC,cAAc,eAAe;AAC1C,qIAAqI;AACrI,iJAAiJ;AACjJ,OAAO,MAAMC,iBAAiB,IAAI;AAWlC,iGAAiG;AACjG,0DAA0D;AAC1D,gGAAgG;AAChG,eAAeC,iBAAiBC,OAAqB,EAAEC,IAAgB;IACrE,MAAMA,KAAKC,IAAI,CAAC,CAAC,wIAAwI,CAAC;IAC1J,MAAMD,KAAKC,IAAI,CAAC,CAAC,iGAAiG,CAAC;IACnH,MAAMD,KAAKC,IAAI,CAAC,CAAC,wFAAwF,CAAC;IAC1G,MAAMD,KAAKC,IAAI,CAAC,CAAC,gGAAgG,CAAC;IAClH,MAAMD,KAAKC,IAAI,CAAC;IAChB,IAAI,AAAC,MAAMhB,QAAQe,MAAM,sBAAuB,MAAM,MAAMd,QAAQc,MAAM,kBAAkBH;AAC9F;AAEA,eAAeK,oBAAoBH,OAAqB,EAAEC,IAAgB,EAAEG,GAAW;IACrF,KAAK,MAAMC,WAAWrB,eAAeoB,KAAM;QACzC,+HAA+H;QAC/H,sHAAsH;QACtH,IAAIC,QAAQC,IAAI,KAAK,SAAS;YAC5B,MAAML,KAAKC,IAAI,CAAC,CAAC,kKAAkK,EAAEpB,WAAW,+BAA+B,CAAC;YAChO;QACF;QACA,MAAMuB,QAAQE,MAAM,CAACN;IACvB;AACF;AAEA,eAAeO,kBAAkBC,MAAoB,EAAEC,KAAiB,EAAEC,IAAY;IACpF,MAAMf,uBAAuBa,OAAOG,MAAM,EAAE9B;AAC9C;AAEA,iFAAiF;AACjF,eAAe+B,MAAMJ,MAAoB;IACvCA,OAAOG,MAAM,CAACE,cAAc;IAC5BL,OAAOM,QAAQ,CAACC,SAAS;AAC3B;AAEA,eAAeC,QAAQC,MAAc,EAAEP,IAAoB,EAAEQ,OAA0D;IACrH,gIAAgI;IAChI,qLAAqL;IACrL,IAAIC;IACJ,IAAI;QACD,CAAA,EAAEA,cAAc,EAAE,GAAG,MAAM5B,WAAU;IACxC,EAAE,OAAO6B,KAAK;QACZ,IAAIA,eAAetC,YAAY,MAAMsC;QACrC,MAAM,IAAItC,WAAW,4BAA4B,CAAC,yBAAyB,EAAEQ,eAAe,UAAU,EAAE,AAAC8B,IAAcC,OAAO,CAAC,CAAC,CAAC;IACnI;IAEA,IAAIV;IACJ,IAAIG;IACJ,IAAI;QACF,4HAA4H;QAC5H,4KAA4K;QAC5KA,WAAW,MAAMK,eAAeG,MAAM,CAACL,QAAQC,QAAQK,aAAa,GAAG;YAAEC,aAAa;QAAY,IAAI;YAAEC,+BAA+B;QAAS;QAChJd,SAAS,MAAMG,SAASE,OAAO;IACjC,EAAE,OAAOI,KAAK;QACZ,wFAAwF;QACxFN,qBAAAA,+BAAAA,SAAUC,SAAS;QACnB,MAAM,IAAIjC,WAAW,4BAA4B,CAAC,8BAA8B,EAAEmC,OAAO,EAAE,EAAE,AAACG,IAAcC,OAAO,EAAE;IACvH;IACA,4GAA4G;IAC5G,IAAI;QACF,MAAM5B,kBAAkBkB;QACxB,MAAMX,OAAOb,iBAAiBwB;QAC9B,OAAO;YAAEH,QAAQ;gBAAEM;gBAAUH;YAAO;YAAGX;QAAK;IAC9C,EAAE,OAAOoB,KAAK;QACZ,MAAMR,MAAM;YAAEE;YAAUH;QAAO;QAC/B,MAAMS;IACR;AACF;AAEA,+FAA+F;AAC/F,OAAO,MAAMM,oBAA+C;IAC1DC,UAAU/B;IACVgC,eAAe/B;IACfgC,kBAAkBrC;IAClBwB;IACAJ;IACA,8FAA8F;IAC9F,iGAAiG;IACjG,qFAAqF;IACrFkB,UAAU,CAACV,MAAQ,2DAA2DW,IAAI,CAACX,IAAIC,OAAO;IAC9FvB;IACAI;IACAK;IACAyB,gBAAgB3C;IAChB4C,oBAAoB7C;IACpBM,aAAa,CAACc,QAAQR,OAASN,YAAYc,OAAOM,QAAQ,EAAEN,OAAOG,MAAM,EAAEX;AAC7E,EAAE;AAEF,OAAO,eAAekC,WAAW/B,GAAmB,EAAEe,OAA6B;IACjF,OAAOlC,gBAAgBmB,KAAKuB,mBAAmBR;AACjD"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/open-source/ai/sensemaking/src/store/duckdb/open.ts"],"sourcesContent":["import type { DuckDBConnection, DuckDBInstance } from '@duckdb/node-api';\nimport type { Config, ResolvedConfig } from '../../config/index.ts';\nimport { STORE_DIMS } from '../../embed/types.ts';\nimport { SenseError } from '../../errors.ts';\nimport { activeFeatures } from '../../features/index.ts';\nimport { ENGINE_CHECK_DOCS, engineIdentity, npmRemedy, checkEngine as runEngineCheck } from '../engine.ts';\nimport { peerRange } from '../native.ts';\nimport type { InternalOpenOptions, OpenResult } from '../open.ts';\nimport { openWithDialect } from '../open.ts';\nimport { getMeta, setMeta } from '../shared.ts';\nimport type { Connection, OpenDialect } from '../types.ts';\nimport { createConnection } from './connection.ts';\nimport { assertFtsReady, prepareFts, queryLexical } from './lexical.ts';\nimport { DUCKDB_PACKAGE, duckdbApi, duckdbNative } from './native.ts';\nimport { duckdbDialect } from './reconcile.ts';\nimport { registerFunctions } from './sql-functions.ts';\nimport { createStore } from './store.ts';\nimport { createVectorWriteStage } from './vectors.ts';\n\nexport const DB_FILENAME = 'cache.duckdb';\n// Independent of sqlite's SCHEMA_VERSION -- the two stores' cache shapes evolve separately (VARIANT frontmatter columns vs untyped).\n// The store name already joins the feature signature, so switching a config's `store` key rebuilds rather than reusing the other engine's cache.\nexport const SCHEMA_VERSION = '6';\n\nexport type { OpenResult };\n\n// The connection plus this store's own native handles (types.ts's OpenDialect<Handle>): the\n// instance owns the WAL and must outlive the connection borrowed from it.\ninterface DuckdbHandle {\n instance: DuckDBInstance;\n duckdb: DuckDBConnection;\n}\n\n// `content` is a plain table (not FTS-virtual): build prepares the persistent FTS index over it,\n// while contains() verification reads the table directly.\n// No tokenizer resolution: this store always uses the fts extension's default (porter) stemmer.\nasync function ensureCoreSchema(_handle: DuckdbHandle, conn: Connection): Promise<void> {\n await conn.exec(`CREATE TABLE IF NOT EXISTS frontmatter (\"path\" TEXT PRIMARY KEY, \"_mtime\" DOUBLE, \"_ctime\" DOUBLE, \"_size\" INTEGER, \"_parse_error\" TEXT)`);\n await conn.exec(`CREATE TABLE IF NOT EXISTS content (\"path\" TEXT PRIMARY KEY, title TEXT, summary TEXT, text TEXT)`);\n await conn.exec(`CREATE TABLE IF NOT EXISTS indexed_sources (\"path\" TEXT PRIMARY KEY, text TEXT NOT NULL)`);\n await conn.exec(`CREATE TABLE IF NOT EXISTS preset_files (\"path\" TEXT, preset TEXT, PRIMARY KEY (\"path\", preset))`);\n await conn.exec('CREATE INDEX IF NOT EXISTS preset_files_preset ON preset_files(preset)');\n if ((await getMeta(conn, 'schema_version')) === null) await setMeta(conn, 'schema_version', SCHEMA_VERSION);\n}\n\nasync function ensureFeatureSchema(_handle: DuckdbHandle, conn: Connection, cfg: Config): Promise<void> {\n for (const feature of activeFeatures(cfg)) {\n // Native FLOAT[STORE_DIMS] instead of the embed feature's engine-neutral BLOB+scale DDL (vectors.ts). `scale` is kept, unused,\n // so the feature's shared reconcile-time INSERT/DELETE (features/embed.ts) names a column that exists on both stores.\n if (feature.name === 'embed') {\n await conn.exec(`CREATE TABLE IF NOT EXISTS embeddings (\"path\" TEXT, chunk INTEGER, start_line INTEGER, end_line INTEGER, content_identity TEXT NOT NULL, scale REAL, vector FLOAT[${STORE_DIMS}], PRIMARY KEY (\"path\", chunk))`);\n continue;\n }\n await feature.schema(conn);\n }\n}\n\nasync function prepareConnection(handle: DuckdbHandle, _conn: Connection, _cfg: Config): Promise<void> {\n await createVectorWriteStage(handle.duckdb, STORE_DIMS);\n}\n\n// Order matters: the connection must be gone before the instance closes the WAL.\nasync function close(handle: DuckdbHandle): Promise<void> {\n handle.duckdb.disconnectSync();\n handle.instance.closeSync();\n}\n\nasync function connect(dbPath: string, _cfg: ResolvedConfig, options: { existingOnly: boolean; observational: boolean }): Promise<{ handle: DuckdbHandle; conn: Connection }> {\n // Dynamic, not a top-level import: sqlite trees must never attempt to resolve this optional peer dependency, so nothing imports\n // it as a value until a duckdb tree opens (types-only imports are erased). Installed on first use if missing, shared with sql-functions.ts and vectors.ts via native.ts's duckdbApi.\n let DuckDBInstance: typeof import('@duckdb/node-api').DuckDBInstance;\n try {\n ({ DuckDBInstance } = await duckdbApi());\n } catch (err) {\n if (err instanceof SenseError) throw err;\n throw new SenseError('STORE_DEPENDENCY_MISSING', `store \"duckdb\" needs the ${DUCKDB_PACKAGE} package (${(err as Error).message})`);\n }\n\n let duckdb: DuckDBConnection;\n let instance: DuckDBInstance | undefined;\n try {\n // On-disk files default to an older storage format for cross-version compatibility, which rejects VARIANT columns (\"VARIANT\n // columns are not supported in storage versions prior to v1.5.0\"); this store's dynamic frontmatter columns need VARIANT (reconcile.ts), so the floor is pinned explicitly.\n instance = await DuckDBInstance.create(dbPath, options.observational ? { access_mode: 'READ_ONLY' } : { storage_compatibility_version: 'v1.5.0' });\n duckdb = await instance.connect();\n } catch (err) {\n // create() may have succeeded before connect() failed: close it, or its WAL stays open.\n instance?.closeSync();\n throw new SenseError('STORE_DEPENDENCY_MISSING', `store \"duckdb\" failed to open ${dbPath}: ${(err as Error).message}`);\n }\n // A throw below would leak the open instance, whose WAL then locks the .duckdb file undeletable on Windows.\n try {\n await registerFunctions(duckdb);\n const conn = createConnection(duckdb);\n return { handle: { instance, duckdb }, conn };\n } catch (err) {\n await close({ instance, duckdb });\n throw err;\n }\n}\n\nasync function duckdbEngineIdentity(): Promise<string> {\n return engineIdentity(DUCKDB_PACKAGE, (await duckdbNative()).version);\n}\n\n// The real schema, index build and lexical query over store/engine.ts's fixture, in memory.\nasync function checkDuckdbEngine(): Promise<void> {\n const { api, version } = await duckdbNative();\n const instance = await api.DuckDBInstance.create(':memory:');\n const handle = { instance, duckdb: await instance.connect() };\n try {\n const conn = createConnection(handle.duckdb);\n await conn.exec('CREATE TABLE meta (key TEXT PRIMARY KEY, value TEXT)');\n await ensureCoreSchema(handle, conn);\n await conn.runBatch(\n 'INSERT INTO content (\"path\", title, summary, text) VALUES (?, ?, ?, ?)',\n ENGINE_CHECK_DOCS.map((doc) => [doc.path, doc.title, doc.summary, doc.text])\n );\n await runEngineCheck('duckdb', engineIdentity(DUCKDB_PACKAGE, version), npmRemedy(DUCKDB_PACKAGE, peerRange(DUCKDB_PACKAGE)), async (terms) => {\n await prepareFts(conn);\n return (await queryLexical(conn, terms, { whereJoin: '', whereCond: '', scopeCond: '', limit: 10 })).map((hit) => hit.path);\n });\n } finally {\n await close(handle);\n }\n}\n\n// This store's dialect (types.ts's OpenDialect) for the shared orchestration in store/open.ts.\nexport const duckdbOpenDialect: OpenDialect<DuckdbHandle> = {\n filename: DB_FILENAME,\n schemaVersion: SCHEMA_VERSION,\n reconcileDialect: duckdbDialect,\n connect,\n close,\n // duckdb's file lock spans the connection's life, and each platform words the refusal its own\n // way: posix \"Could not set lock on file ... Conflicting lock is held in <exe> (PID n)\", Windows\n // \"Cannot open file ... being used by another process\". Both are the same condition.\n isLocked: (err) => /Could not set lock on file|being used by another process/.test(err.message),\n engineIdentity: duckdbEngineIdentity,\n checkEngine: checkDuckdbEngine,\n ensureCoreSchema,\n ensureFeatureSchema,\n prepareConnection,\n prepareLexical: prepareFts,\n assertLexicalReady: assertFtsReady,\n createStore: (handle, conn) => createStore(handle.instance, handle.duckdb, conn),\n};\n\nexport async function openDuckdb(cfg: ResolvedConfig, options?: InternalOpenOptions): Promise<OpenResult> {\n return openWithDialect(cfg, duckdbOpenDialect, options);\n}\n"],"names":["STORE_DIMS","SenseError","activeFeatures","ENGINE_CHECK_DOCS","engineIdentity","npmRemedy","checkEngine","runEngineCheck","peerRange","openWithDialect","getMeta","setMeta","createConnection","assertFtsReady","prepareFts","queryLexical","DUCKDB_PACKAGE","duckdbApi","duckdbNative","duckdbDialect","registerFunctions","createStore","createVectorWriteStage","DB_FILENAME","SCHEMA_VERSION","ensureCoreSchema","_handle","conn","exec","ensureFeatureSchema","cfg","feature","name","schema","prepareConnection","handle","_conn","_cfg","duckdb","close","disconnectSync","instance","closeSync","connect","dbPath","options","DuckDBInstance","err","message","create","observational","access_mode","storage_compatibility_version","duckdbEngineIdentity","version","checkDuckdbEngine","api","runBatch","map","doc","path","title","summary","text","terms","whereJoin","whereCond","scopeCond","limit","hit","duckdbOpenDialect","filename","schemaVersion","reconcileDialect","isLocked","test","prepareLexical","assertLexicalReady","openDuckdb"],"mappings":"AAEA,SAASA,UAAU,QAAQ,uBAAuB;AAClD,SAASC,UAAU,QAAQ,kBAAkB;AAC7C,SAASC,cAAc,QAAQ,0BAA0B;AACzD,SAASC,iBAAiB,EAAEC,cAAc,EAAEC,SAAS,EAAEC,eAAeC,cAAc,QAAQ,eAAe;AAC3G,SAASC,SAAS,QAAQ,eAAe;AAEzC,SAASC,eAAe,QAAQ,aAAa;AAC7C,SAASC,OAAO,EAAEC,OAAO,QAAQ,eAAe;AAEhD,SAASC,gBAAgB,QAAQ,kBAAkB;AACnD,SAASC,cAAc,EAAEC,UAAU,EAAEC,YAAY,QAAQ,eAAe;AACxE,SAASC,cAAc,EAAEC,SAAS,EAAEC,YAAY,QAAQ,cAAc;AACtE,SAASC,aAAa,QAAQ,iBAAiB;AAC/C,SAASC,iBAAiB,QAAQ,qBAAqB;AACvD,SAASC,WAAW,QAAQ,aAAa;AACzC,SAASC,sBAAsB,QAAQ,eAAe;AAEtD,OAAO,MAAMC,cAAc,eAAe;AAC1C,qIAAqI;AACrI,iJAAiJ;AACjJ,OAAO,MAAMC,iBAAiB,IAAI;AAWlC,iGAAiG;AACjG,0DAA0D;AAC1D,gGAAgG;AAChG,eAAeC,iBAAiBC,OAAqB,EAAEC,IAAgB;IACrE,MAAMA,KAAKC,IAAI,CAAC,CAAC,wIAAwI,CAAC;IAC1J,MAAMD,KAAKC,IAAI,CAAC,CAAC,iGAAiG,CAAC;IACnH,MAAMD,KAAKC,IAAI,CAAC,CAAC,wFAAwF,CAAC;IAC1G,MAAMD,KAAKC,IAAI,CAAC,CAAC,gGAAgG,CAAC;IAClH,MAAMD,KAAKC,IAAI,CAAC;IAChB,IAAI,AAAC,MAAMlB,QAAQiB,MAAM,sBAAuB,MAAM,MAAMhB,QAAQgB,MAAM,kBAAkBH;AAC9F;AAEA,eAAeK,oBAAoBH,OAAqB,EAAEC,IAAgB,EAAEG,GAAW;IACrF,KAAK,MAAMC,WAAW7B,eAAe4B,KAAM;QACzC,+HAA+H;QAC/H,sHAAsH;QACtH,IAAIC,QAAQC,IAAI,KAAK,SAAS;YAC5B,MAAML,KAAKC,IAAI,CAAC,CAAC,kKAAkK,EAAE5B,WAAW,+BAA+B,CAAC;YAChO;QACF;QACA,MAAM+B,QAAQE,MAAM,CAACN;IACvB;AACF;AAEA,eAAeO,kBAAkBC,MAAoB,EAAEC,KAAiB,EAAEC,IAAY;IACpF,MAAMf,uBAAuBa,OAAOG,MAAM,EAAEtC;AAC9C;AAEA,iFAAiF;AACjF,eAAeuC,MAAMJ,MAAoB;IACvCA,OAAOG,MAAM,CAACE,cAAc;IAC5BL,OAAOM,QAAQ,CAACC,SAAS;AAC3B;AAEA,eAAeC,QAAQC,MAAc,EAAEP,IAAoB,EAAEQ,OAA0D;IACrH,gIAAgI;IAChI,qLAAqL;IACrL,IAAIC;IACJ,IAAI;QACD,CAAA,EAAEA,cAAc,EAAE,GAAG,MAAM7B,WAAU;IACxC,EAAE,OAAO8B,KAAK;QACZ,IAAIA,eAAe9C,YAAY,MAAM8C;QACrC,MAAM,IAAI9C,WAAW,4BAA4B,CAAC,yBAAyB,EAAEe,eAAe,UAAU,EAAE,AAAC+B,IAAcC,OAAO,CAAC,CAAC,CAAC;IACnI;IAEA,IAAIV;IACJ,IAAIG;IACJ,IAAI;QACF,4HAA4H;QAC5H,4KAA4K;QAC5KA,WAAW,MAAMK,eAAeG,MAAM,CAACL,QAAQC,QAAQK,aAAa,GAAG;YAAEC,aAAa;QAAY,IAAI;YAAEC,+BAA+B;QAAS;QAChJd,SAAS,MAAMG,SAASE,OAAO;IACjC,EAAE,OAAOI,KAAK;QACZ,wFAAwF;QACxFN,qBAAAA,+BAAAA,SAAUC,SAAS;QACnB,MAAM,IAAIzC,WAAW,4BAA4B,CAAC,8BAA8B,EAAE2C,OAAO,EAAE,EAAE,AAACG,IAAcC,OAAO,EAAE;IACvH;IACA,4GAA4G;IAC5G,IAAI;QACF,MAAM5B,kBAAkBkB;QACxB,MAAMX,OAAOf,iBAAiB0B;QAC9B,OAAO;YAAEH,QAAQ;gBAAEM;gBAAUH;YAAO;YAAGX;QAAK;IAC9C,EAAE,OAAOoB,KAAK;QACZ,MAAMR,MAAM;YAAEE;YAAUH;QAAO;QAC/B,MAAMS;IACR;AACF;AAEA,eAAeM;IACb,OAAOjD,eAAeY,gBAAgB,AAAC,CAAA,MAAME,cAAa,EAAGoC,OAAO;AACtE;AAEA,4FAA4F;AAC5F,eAAeC;IACb,MAAM,EAAEC,GAAG,EAAEF,OAAO,EAAE,GAAG,MAAMpC;IAC/B,MAAMuB,WAAW,MAAMe,IAAIV,cAAc,CAACG,MAAM,CAAC;IACjD,MAAMd,SAAS;QAAEM;QAAUH,QAAQ,MAAMG,SAASE,OAAO;IAAG;IAC5D,IAAI;QACF,MAAMhB,OAAOf,iBAAiBuB,OAAOG,MAAM;QAC3C,MAAMX,KAAKC,IAAI,CAAC;QAChB,MAAMH,iBAAiBU,QAAQR;QAC/B,MAAMA,KAAK8B,QAAQ,CACjB,0EACAtD,kBAAkBuD,GAAG,CAAC,CAACC,MAAQ;gBAACA,IAAIC,IAAI;gBAAED,IAAIE,KAAK;gBAAEF,IAAIG,OAAO;gBAAEH,IAAII,IAAI;aAAC;QAE7E,MAAMxD,eAAe,UAAUH,eAAeY,gBAAgBsC,UAAUjD,UAAUW,gBAAgBR,UAAUQ,kBAAkB,OAAOgD;YACnI,MAAMlD,WAAWa;YACjB,OAAO,AAAC,CAAA,MAAMZ,aAAaY,MAAMqC,OAAO;gBAAEC,WAAW;gBAAIC,WAAW;gBAAIC,WAAW;gBAAIC,OAAO;YAAG,EAAC,EAAGV,GAAG,CAAC,CAACW,MAAQA,IAAIT,IAAI;QAC5H;IACF,SAAU;QACR,MAAMrB,MAAMJ;IACd;AACF;AAEA,+FAA+F;AAC/F,OAAO,MAAMmC,oBAA+C;IAC1DC,UAAUhD;IACViD,eAAehD;IACfiD,kBAAkBtD;IAClBwB;IACAJ;IACA,8FAA8F;IAC9F,iGAAiG;IACjG,qFAAqF;IACrFmC,UAAU,CAAC3B,MAAQ,2DAA2D4B,IAAI,CAAC5B,IAAIC,OAAO;IAC9F5C,gBAAgBiD;IAChB/C,aAAaiD;IACb9B;IACAI;IACAK;IACA0C,gBAAgB9D;IAChB+D,oBAAoBhE;IACpBQ,aAAa,CAACc,QAAQR,OAASN,YAAYc,OAAOM,QAAQ,EAAEN,OAAOG,MAAM,EAAEX;AAC7E,EAAE;AAEF,OAAO,eAAemD,WAAWhD,GAAmB,EAAEe,OAA6B;IACjF,OAAOpC,gBAAgBqB,KAAKwC,mBAAmBzB;AACjD"}
@@ -1,17 +1,4 @@
1
1
  import type { Connection } from '../types.js';
2
- export declare const ORDERED_BM25_MACRO = "sense_ordered_match_bm25";
3
- interface NativeMacroRow {
4
- function_type: string;
5
- parameters: unknown;
6
- parameter_types: unknown;
7
- macro_definition: string;
8
- }
9
- interface NativeMacroContract {
10
- version: string;
11
- macro: NativeMacroRow;
12
- }
13
- export declare function validateNativeMacroContract({ version, macro }: NativeMacroContract): string;
14
- export declare function ensureOrderedBm25(conn: Connection, state: {
15
- orderedMacroReady: boolean;
16
- }): Promise<void>;
17
- export {};
2
+ import type { DuckdbFtsBuilders } from './engine.js';
3
+ export declare function nativeQueryTerms(conn: Connection, fts: DuckdbFtsBuilders, query: string): Promise<string[]>;
4
+ export declare function orderedBm25Sql(fts: DuckdbFtsBuilders, field: 'title' | 'summary' | 'text', termCount: number): string;
@@ -1,56 +1,15 @@
1
- import { createHash } from 'node:crypto';
2
- import { SenseError } from '../../errors.js';
3
- // This is a connection-local TEMP macro. The contract hash covers the native fts definition,
4
- // DuckDB version, and the reviewed parameter defaults captured in the retained diagnostic.
5
- export const ORDERED_BM25_MACRO = 'sense_ordered_match_bm25';
6
- const SUPPORTED_DUCKDB_VERSION = 'v1.5.5';
7
- const SUPPORTED_CONTRACT_SHA256 = '4476d543e43e7f8c89751cac0659b64e09d906cd3fefb1bfcaf4d15d3b1d9e00';
8
- const EXPECTED_PARAMETERS = [
9
- 'docname',
10
- 'query_string',
11
- 'fields',
12
- 'k',
13
- 'b',
14
- 'conjunctive'
15
- ];
16
- const EXPECTED_DEFAULTS = {
17
- fields: 'NULL',
18
- k: '1.2',
19
- b: '0.75',
20
- conjunctive: 'false'
21
- };
22
- const EXPECTED_SIGNATURE = `docname, query_string, fields := ${EXPECTED_DEFAULTS.fields}, k := ${EXPECTED_DEFAULTS.k}, b := ${EXPECTED_DEFAULTS.b}, conjunctive := ${EXPECTED_DEFAULTS.conjunctive}`;
23
- const UNORDERED_SUM = 'sum(subscore)';
24
- function unsupported(message) {
25
- return new SenseError('STORE_CAPABILITY_MISSING', `store "duckdb" cannot install its deterministic native BM25 adapter: ${message}; update the DuckDB adapter before using lexical search`);
26
- }
27
- function contractSha256(version, definition) {
28
- return createHash('sha256').update(`${version}\0${EXPECTED_SIGNATURE}\0${definition}`).digest('hex');
29
- }
30
- export function validateNativeMacroContract({ version, macro }) {
31
- if (version !== SUPPORTED_DUCKDB_VERSION) throw unsupported(`DuckDB ${version} is unsupported (validated version is ${SUPPORTED_DUCKDB_VERSION})`);
32
- if (macro.function_type !== 'macro') throw unsupported(`match_bm25 is ${macro.function_type}, not a macro`);
33
- if (JSON.stringify(macro.parameters) !== JSON.stringify(EXPECTED_PARAMETERS)) throw unsupported(`match_bm25 parameters changed (${JSON.stringify(macro.parameters)})`);
34
- if (!Array.isArray(macro.parameter_types) || macro.parameter_types.some((type)=>type !== null)) throw unsupported('match_bm25 parameter types changed');
35
- if (contractSha256(version, macro.macro_definition) !== SUPPORTED_CONTRACT_SHA256) throw unsupported('match_bm25 definition or reviewed parameter-default contract changed');
36
- if (macro.macro_definition.split(UNORDERED_SUM).length !== 2) throw unsupported('match_bm25 does not contain exactly one validated unordered subscore sum');
37
- return macro.macro_definition.replace(UNORDERED_SUM, 'sum(subscore ORDER BY subscore)');
38
- }
39
- export async function ensureOrderedBm25(conn, state) {
40
- if (state.orderedMacroReady) return;
41
- const versionStatement = await conn.prepare('SELECT version() AS version');
42
- const versionRow = await versionStatement.get();
43
- const macroStatement = await conn.prepare(`
44
- SELECT function_type, parameters, parameter_types, macro_definition
45
- FROM duckdb_functions()
46
- WHERE schema_name = 'fts_main_content' AND function_name = 'match_bm25'
47
- `);
48
- const macro = await macroStatement.get();
49
- if (!macro) throw unsupported('fts_main_content.match_bm25 is missing');
50
- const orderedDefinition = validateNativeMacroContract({
51
- version: versionRow.version,
52
- macro
53
- });
54
- await conn.exec(`CREATE OR REPLACE TEMP MACRO ${ORDERED_BM25_MACRO}(${EXPECTED_SIGNATURE}) AS ${orderedDefinition}`);
55
- state.orderedMacroReady = true;
1
+ // Keep deduplication native, but return original tokens so native scoring stems once.
2
+ export async function nativeQueryTerms(conn, fts, query) {
3
+ const stmt = await conn.prepare(fts.queryTerms());
4
+ const rows = await stmt.all(query);
5
+ return rows.map(({ term })=>term);
6
+ }
7
+ // Fixed parameters avoid a correlated document-by-term intermediate relation in DuckDB.
8
+ // Only the number of terms shapes this authored SQL; query values are always bound.
9
+ export function orderedBm25Sql(fts, field, termCount) {
10
+ if (termCount === 0) return 'NULL';
11
+ const contributions = Array.from({
12
+ length: termCount
13
+ }, ()=>fts.termScore(field));
14
+ return `(SELECT sum(score ORDER BY score) FROM (${contributions.join(' UNION ALL ')}))`;
56
15
  }
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/store/duckdb/ordered-bm25.ts"],"sourcesContent":["import { createHash } from 'node:crypto';\nimport { SenseError } from '../../errors.ts';\nimport type { Connection } from '../types.ts';\n\n// This is a connection-local TEMP macro. The contract hash covers the native fts definition,\n// DuckDB version, and the reviewed parameter defaults captured in the retained diagnostic.\nexport const ORDERED_BM25_MACRO = 'sense_ordered_match_bm25';\nconst SUPPORTED_DUCKDB_VERSION = 'v1.5.5';\nconst SUPPORTED_CONTRACT_SHA256 = '4476d543e43e7f8c89751cac0659b64e09d906cd3fefb1bfcaf4d15d3b1d9e00';\nconst EXPECTED_PARAMETERS = ['docname', 'query_string', 'fields', 'k', 'b', 'conjunctive'];\nconst EXPECTED_DEFAULTS = { fields: 'NULL', k: '1.2', b: '0.75', conjunctive: 'false' } as const;\nconst EXPECTED_SIGNATURE = `docname, query_string, fields := ${EXPECTED_DEFAULTS.fields}, k := ${EXPECTED_DEFAULTS.k}, b := ${EXPECTED_DEFAULTS.b}, conjunctive := ${EXPECTED_DEFAULTS.conjunctive}`;\nconst UNORDERED_SUM = 'sum(subscore)';\n\ninterface NativeMacroRow {\n function_type: string;\n parameters: unknown;\n parameter_types: unknown;\n macro_definition: string;\n}\n\ninterface NativeMacroContract {\n version: string;\n macro: NativeMacroRow;\n}\n\nfunction unsupported(message: string): SenseError {\n return new SenseError('STORE_CAPABILITY_MISSING', `store \"duckdb\" cannot install its deterministic native BM25 adapter: ${message}; update the DuckDB adapter before using lexical search`);\n}\n\nfunction contractSha256(version: string, definition: string): string {\n return createHash('sha256').update(`${version}\\0${EXPECTED_SIGNATURE}\\0${definition}`).digest('hex');\n}\n\nexport function validateNativeMacroContract({ version, macro }: NativeMacroContract): string {\n if (version !== SUPPORTED_DUCKDB_VERSION) throw unsupported(`DuckDB ${version} is unsupported (validated version is ${SUPPORTED_DUCKDB_VERSION})`);\n if (macro.function_type !== 'macro') throw unsupported(`match_bm25 is ${macro.function_type}, not a macro`);\n if (JSON.stringify(macro.parameters) !== JSON.stringify(EXPECTED_PARAMETERS)) throw unsupported(`match_bm25 parameters changed (${JSON.stringify(macro.parameters)})`);\n if (!Array.isArray(macro.parameter_types) || macro.parameter_types.some((type) => type !== null)) throw unsupported('match_bm25 parameter types changed');\n if (contractSha256(version, macro.macro_definition) !== SUPPORTED_CONTRACT_SHA256) throw unsupported('match_bm25 definition or reviewed parameter-default contract changed');\n if (macro.macro_definition.split(UNORDERED_SUM).length !== 2) throw unsupported('match_bm25 does not contain exactly one validated unordered subscore sum');\n return macro.macro_definition.replace(UNORDERED_SUM, 'sum(subscore ORDER BY subscore)');\n}\n\nexport async function ensureOrderedBm25(conn: Connection, state: { orderedMacroReady: boolean }): Promise<void> {\n if (state.orderedMacroReady) return;\n const versionStatement = await conn.prepare('SELECT version() AS version');\n const versionRow = (await versionStatement.get()) as { version: string };\n const macroStatement = await conn.prepare(`\n SELECT function_type, parameters, parameter_types, macro_definition\n FROM duckdb_functions()\n WHERE schema_name = 'fts_main_content' AND function_name = 'match_bm25'\n `);\n const macro = (await macroStatement.get()) as NativeMacroRow | undefined;\n if (!macro) throw unsupported('fts_main_content.match_bm25 is missing');\n const orderedDefinition = validateNativeMacroContract({ version: versionRow.version, macro });\n await conn.exec(`CREATE OR REPLACE TEMP MACRO ${ORDERED_BM25_MACRO}(${EXPECTED_SIGNATURE}) AS ${orderedDefinition}`);\n state.orderedMacroReady = true;\n}\n"],"names":["createHash","SenseError","ORDERED_BM25_MACRO","SUPPORTED_DUCKDB_VERSION","SUPPORTED_CONTRACT_SHA256","EXPECTED_PARAMETERS","EXPECTED_DEFAULTS","fields","k","b","conjunctive","EXPECTED_SIGNATURE","UNORDERED_SUM","unsupported","message","contractSha256","version","definition","update","digest","validateNativeMacroContract","macro","function_type","JSON","stringify","parameters","Array","isArray","parameter_types","some","type","macro_definition","split","length","replace","ensureOrderedBm25","conn","state","orderedMacroReady","versionStatement","prepare","versionRow","get","macroStatement","orderedDefinition","exec"],"mappings":"AAAA,SAASA,UAAU,QAAQ,cAAc;AACzC,SAASC,UAAU,QAAQ,kBAAkB;AAG7C,6FAA6F;AAC7F,2FAA2F;AAC3F,OAAO,MAAMC,qBAAqB,2BAA2B;AAC7D,MAAMC,2BAA2B;AACjC,MAAMC,4BAA4B;AAClC,MAAMC,sBAAsB;IAAC;IAAW;IAAgB;IAAU;IAAK;IAAK;CAAc;AAC1F,MAAMC,oBAAoB;IAAEC,QAAQ;IAAQC,GAAG;IAAOC,GAAG;IAAQC,aAAa;AAAQ;AACtF,MAAMC,qBAAqB,CAAC,iCAAiC,EAAEL,kBAAkBC,MAAM,CAAC,OAAO,EAAED,kBAAkBE,CAAC,CAAC,OAAO,EAAEF,kBAAkBG,CAAC,CAAC,iBAAiB,EAAEH,kBAAkBI,WAAW,EAAE;AACpM,MAAME,gBAAgB;AActB,SAASC,YAAYC,OAAe;IAClC,OAAO,IAAIb,WAAW,4BAA4B,CAAC,qEAAqE,EAAEa,QAAQ,uDAAuD,CAAC;AAC5L;AAEA,SAASC,eAAeC,OAAe,EAAEC,UAAkB;IACzD,OAAOjB,WAAW,UAAUkB,MAAM,CAAC,GAAGF,QAAQ,EAAE,EAAEL,mBAAmB,EAAE,EAAEM,YAAY,EAAEE,MAAM,CAAC;AAChG;AAEA,OAAO,SAASC,4BAA4B,EAAEJ,OAAO,EAAEK,KAAK,EAAuB;IACjF,IAAIL,YAAYb,0BAA0B,MAAMU,YAAY,CAAC,OAAO,EAAEG,QAAQ,sCAAsC,EAAEb,yBAAyB,CAAC,CAAC;IACjJ,IAAIkB,MAAMC,aAAa,KAAK,SAAS,MAAMT,YAAY,CAAC,cAAc,EAAEQ,MAAMC,aAAa,CAAC,aAAa,CAAC;IAC1G,IAAIC,KAAKC,SAAS,CAACH,MAAMI,UAAU,MAAMF,KAAKC,SAAS,CAACnB,sBAAsB,MAAMQ,YAAY,CAAC,+BAA+B,EAAEU,KAAKC,SAAS,CAACH,MAAMI,UAAU,EAAE,CAAC,CAAC;IACrK,IAAI,CAACC,MAAMC,OAAO,CAACN,MAAMO,eAAe,KAAKP,MAAMO,eAAe,CAACC,IAAI,CAAC,CAACC,OAASA,SAAS,OAAO,MAAMjB,YAAY;IACpH,IAAIE,eAAeC,SAASK,MAAMU,gBAAgB,MAAM3B,2BAA2B,MAAMS,YAAY;IACrG,IAAIQ,MAAMU,gBAAgB,CAACC,KAAK,CAACpB,eAAeqB,MAAM,KAAK,GAAG,MAAMpB,YAAY;IAChF,OAAOQ,MAAMU,gBAAgB,CAACG,OAAO,CAACtB,eAAe;AACvD;AAEA,OAAO,eAAeuB,kBAAkBC,IAAgB,EAAEC,KAAqC;IAC7F,IAAIA,MAAMC,iBAAiB,EAAE;IAC7B,MAAMC,mBAAmB,MAAMH,KAAKI,OAAO,CAAC;IAC5C,MAAMC,aAAc,MAAMF,iBAAiBG,GAAG;IAC9C,MAAMC,iBAAiB,MAAMP,KAAKI,OAAO,CAAC,CAAC;;;;EAI3C,CAAC;IACD,MAAMnB,QAAS,MAAMsB,eAAeD,GAAG;IACvC,IAAI,CAACrB,OAAO,MAAMR,YAAY;IAC9B,MAAM+B,oBAAoBxB,4BAA4B;QAAEJ,SAASyB,WAAWzB,OAAO;QAAEK;IAAM;IAC3F,MAAMe,KAAKS,IAAI,CAAC,CAAC,6BAA6B,EAAE3C,mBAAmB,CAAC,EAAES,mBAAmB,KAAK,EAAEiC,mBAAmB;IACnHP,MAAMC,iBAAiB,GAAG;AAC5B"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/open-source/ai/sensemaking/src/store/duckdb/ordered-bm25.ts"],"sourcesContent":["import type { Connection } from '../types.ts';\nimport type { DuckdbFtsBuilders } from './engine.ts';\n\n// Keep deduplication native, but return original tokens so native scoring stems once.\nexport async function nativeQueryTerms(conn: Connection, fts: DuckdbFtsBuilders, query: string): Promise<string[]> {\n const stmt = await conn.prepare(fts.queryTerms());\n const rows = (await stmt.all(query)) as Array<{ term: string }>;\n return rows.map(({ term }) => term);\n}\n\n// Fixed parameters avoid a correlated document-by-term intermediate relation in DuckDB.\n// Only the number of terms shapes this authored SQL; query values are always bound.\nexport function orderedBm25Sql(fts: DuckdbFtsBuilders, field: 'title' | 'summary' | 'text', termCount: number): string {\n if (termCount === 0) return 'NULL';\n const contributions = Array.from({ length: termCount }, () => fts.termScore(field));\n return `(SELECT sum(score ORDER BY score) FROM (${contributions.join(' UNION ALL ')}))`;\n}\n"],"names":["nativeQueryTerms","conn","fts","query","stmt","prepare","queryTerms","rows","all","map","term","orderedBm25Sql","field","termCount","contributions","Array","from","length","termScore","join"],"mappings":"AAGA,sFAAsF;AACtF,OAAO,eAAeA,iBAAiBC,IAAgB,EAAEC,GAAsB,EAAEC,KAAa;IAC5F,MAAMC,OAAO,MAAMH,KAAKI,OAAO,CAACH,IAAII,UAAU;IAC9C,MAAMC,OAAQ,MAAMH,KAAKI,GAAG,CAACL;IAC7B,OAAOI,KAAKE,GAAG,CAAC,CAAC,EAAEC,IAAI,EAAE,GAAKA;AAChC;AAEA,wFAAwF;AACxF,oFAAoF;AACpF,OAAO,SAASC,eAAeT,GAAsB,EAAEU,KAAmC,EAAEC,SAAiB;IAC3G,IAAIA,cAAc,GAAG,OAAO;IAC5B,MAAMC,gBAAgBC,MAAMC,IAAI,CAAC;QAAEC,QAAQJ;IAAU,GAAG,IAAMX,IAAIgB,SAAS,CAACN;IAC5E,OAAO,CAAC,wCAAwC,EAAEE,cAAcK,IAAI,CAAC,eAAe,EAAE,CAAC;AACzF"}