sensemaking 0.26.0 → 0.27.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (324) hide show
  1. package/README.md +36 -4
  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/download.js.map +1 -1
  11. package/dist/cjs/cli/index.d.cts +2 -2
  12. package/dist/cjs/cli/index.d.ts +2 -2
  13. package/dist/cjs/cli/index.js +2 -2
  14. package/dist/cjs/cli/index.js.map +1 -1
  15. package/dist/cjs/cli/init.js.map +1 -1
  16. package/dist/cjs/cli/map.js.map +1 -1
  17. package/dist/cjs/cli/named.js.map +1 -1
  18. package/dist/cjs/cli/path.js.map +1 -1
  19. package/dist/cjs/cli/peek.js +43 -2
  20. package/dist/cjs/cli/peek.js.map +1 -1
  21. package/dist/cjs/cli/related.js.map +1 -1
  22. package/dist/cjs/cli/search.js +38 -6
  23. package/dist/cjs/cli/search.js.map +1 -1
  24. package/dist/cjs/cli/shared.js.map +1 -1
  25. package/dist/cjs/cli/sql.js.map +1 -1
  26. package/dist/cjs/cli/status.js.map +1 -1
  27. package/dist/cjs/cli/watch.js.map +1 -1
  28. package/dist/cjs/cli.js.map +1 -1
  29. package/dist/cjs/commands/index.d.cts +2 -2
  30. package/dist/cjs/commands/index.d.ts +2 -2
  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.d.cts +9 -1
  35. package/dist/cjs/commands/peek.d.ts +9 -1
  36. package/dist/cjs/commands/peek.js +72 -29
  37. package/dist/cjs/commands/peek.js.map +1 -1
  38. package/dist/cjs/commands/related.js.map +1 -1
  39. package/dist/cjs/commands/scope.js.map +1 -1
  40. package/dist/cjs/commands/search.d.cts +6 -2
  41. package/dist/cjs/commands/search.d.ts +6 -2
  42. package/dist/cjs/commands/search.js +31 -4
  43. package/dist/cjs/commands/search.js.map +1 -1
  44. package/dist/cjs/commands/signals.d.cts +19 -5
  45. package/dist/cjs/commands/signals.d.ts +19 -5
  46. package/dist/cjs/commands/signals.js +64 -11
  47. package/dist/cjs/commands/signals.js.map +1 -1
  48. package/dist/cjs/commands/snapshot.js.map +1 -1
  49. package/dist/cjs/commands/status.js.map +1 -1
  50. package/dist/cjs/config/access.js.map +1 -1
  51. package/dist/cjs/config/index.js.map +1 -1
  52. package/dist/cjs/config/load.js +6 -0
  53. package/dist/cjs/config/load.js.map +1 -1
  54. package/dist/cjs/config/resolve.js.map +1 -1
  55. package/dist/cjs/config/signals.js.map +1 -1
  56. package/dist/cjs/config/types.js.map +1 -1
  57. package/dist/cjs/config/validate.js.map +1 -1
  58. package/dist/cjs/embed/cohere.js.map +1 -1
  59. package/dist/cjs/embed/distribution.js.map +1 -1
  60. package/dist/cjs/embed/endpoint.js.map +1 -1
  61. package/dist/cjs/embed/handoff.js.map +1 -1
  62. package/dist/cjs/embed/http.js.map +1 -1
  63. package/dist/cjs/embed/identity.js.map +1 -1
  64. package/dist/cjs/embed/langfit.js.map +1 -1
  65. package/dist/cjs/embed/languages.js.map +1 -1
  66. package/dist/cjs/embed/openai.js.map +1 -1
  67. package/dist/cjs/embed/query.js.map +1 -1
  68. package/dist/cjs/embed/registry.js.map +1 -1
  69. package/dist/cjs/embed/static.js.map +1 -1
  70. package/dist/cjs/embed/store.js.map +1 -1
  71. package/dist/cjs/embed/types.js.map +1 -1
  72. package/dist/cjs/errors.js.map +1 -1
  73. package/dist/cjs/features/embed.js.map +1 -1
  74. package/dist/cjs/features/fences.js.map +1 -1
  75. package/dist/cjs/features/index.js.map +1 -1
  76. package/dist/cjs/features/links.js.map +1 -1
  77. package/dist/cjs/features/rank.js.map +1 -1
  78. package/dist/cjs/features/sections.js.map +1 -1
  79. package/dist/cjs/features/tags.js.map +1 -1
  80. package/dist/cjs/graph/graph.js.map +1 -1
  81. package/dist/cjs/graph/traverse.js.map +1 -1
  82. package/dist/cjs/index.d.cts +1 -1
  83. package/dist/cjs/index.d.ts +1 -1
  84. package/dist/cjs/index.js.map +1 -1
  85. package/dist/cjs/lib/atomic-write.js.map +1 -1
  86. package/dist/cjs/lib/guarded-tick.js.map +1 -1
  87. package/dist/cjs/lib/serial-query.js.map +1 -1
  88. package/dist/cjs/lib/watch-notification.js.map +1 -1
  89. package/dist/cjs/lib/worker-file.js.map +1 -1
  90. package/dist/cjs/output/column-hint.js.map +1 -1
  91. package/dist/cjs/output/output.js.map +1 -1
  92. package/dist/cjs/output/progress.js.map +1 -1
  93. package/dist/cjs/output/search-error.js.map +1 -1
  94. package/dist/cjs/scan/frontmatter.js.map +1 -1
  95. package/dist/cjs/scan/index.js.map +1 -1
  96. package/dist/cjs/scan/list.js.map +1 -1
  97. package/dist/cjs/scan/pool.js.map +1 -1
  98. package/dist/cjs/scan/reparse.js.map +1 -1
  99. package/dist/cjs/scan/worker-error.js.map +1 -1
  100. package/dist/cjs/store/batch.js.map +1 -1
  101. package/dist/cjs/store/builder.js.map +1 -1
  102. package/dist/cjs/store/cache.js.map +1 -1
  103. package/dist/cjs/store/duckdb/batch.js.map +1 -1
  104. package/dist/cjs/store/duckdb/connection.js.map +1 -1
  105. package/dist/cjs/store/duckdb/fieldStats.js.map +1 -1
  106. package/dist/cjs/store/duckdb/lexical.d.cts +1 -6
  107. package/dist/cjs/store/duckdb/lexical.d.ts +1 -6
  108. package/dist/cjs/store/duckdb/lexical.js +28 -39
  109. package/dist/cjs/store/duckdb/lexical.js.map +1 -1
  110. package/dist/cjs/store/duckdb/native.d.cts +1 -2
  111. package/dist/cjs/store/duckdb/native.d.ts +1 -2
  112. package/dist/cjs/store/duckdb/native.js +1 -6
  113. package/dist/cjs/store/duckdb/native.js.map +1 -1
  114. package/dist/cjs/store/duckdb/open.js.map +1 -1
  115. package/dist/cjs/store/duckdb/ordered-bm25.d.cts +3 -16
  116. package/dist/cjs/store/duckdb/ordered-bm25.d.ts +3 -16
  117. package/dist/cjs/store/duckdb/ordered-bm25.js +27 -79
  118. package/dist/cjs/store/duckdb/ordered-bm25.js.map +1 -1
  119. package/dist/cjs/store/duckdb/reconcile.js.map +1 -1
  120. package/dist/cjs/store/duckdb/sql-functions.js.map +1 -1
  121. package/dist/cjs/store/duckdb/store.js.map +1 -1
  122. package/dist/cjs/store/duckdb/vectors.js.map +1 -1
  123. package/dist/cjs/store/embed-scope.js.map +1 -1
  124. package/dist/cjs/store/feature-scope.js.map +1 -1
  125. package/dist/cjs/store/index.js.map +1 -1
  126. package/dist/cjs/store/lock-wait.js.map +1 -1
  127. package/dist/cjs/store/native.js.map +1 -1
  128. package/dist/cjs/store/open.js.map +1 -1
  129. package/dist/cjs/store/preset-scope.js.map +1 -1
  130. package/dist/cjs/store/readiness.js.map +1 -1
  131. package/dist/cjs/store/reconcile.js.map +1 -1
  132. package/dist/cjs/store/shared.js.map +1 -1
  133. package/dist/cjs/store/signature.js.map +1 -1
  134. package/dist/cjs/store/sql-functions.js.map +1 -1
  135. package/dist/cjs/store/sqlite/connection.js.map +1 -1
  136. package/dist/cjs/store/sqlite/fieldStats.js.map +1 -1
  137. package/dist/cjs/store/sqlite/lexical.js.map +1 -1
  138. package/dist/cjs/store/sqlite/open.js.map +1 -1
  139. package/dist/cjs/store/sqlite/reconcile.js.map +1 -1
  140. package/dist/cjs/store/sqlite/sql-functions.js.map +1 -1
  141. package/dist/cjs/store/sqlite/store.js.map +1 -1
  142. package/dist/cjs/store/sqlite/vectors.js.map +1 -1
  143. package/dist/cjs/store/stages.js.map +1 -1
  144. package/dist/cjs/store/transaction.js.map +1 -1
  145. package/dist/cjs/store/turso/connection.js.map +1 -1
  146. package/dist/cjs/store/turso/fieldStats.js.map +1 -1
  147. package/dist/cjs/store/turso/lexical-text.js.map +1 -1
  148. package/dist/cjs/store/turso/lexical.js.map +1 -1
  149. package/dist/cjs/store/turso/native.js.map +1 -1
  150. package/dist/cjs/store/turso/open.js.map +1 -1
  151. package/dist/cjs/store/turso/reconcile.js.map +1 -1
  152. package/dist/cjs/store/turso/sql-functions.js.map +1 -1
  153. package/dist/cjs/store/turso/store.js.map +1 -1
  154. package/dist/cjs/store/turso/vectors.js.map +1 -1
  155. package/dist/cjs/store/types.js.map +1 -1
  156. package/dist/cjs/store/vectors.js.map +1 -1
  157. package/dist/cjs/text/segment.js.map +1 -1
  158. package/dist/cjs/text/strip.js.map +1 -1
  159. package/dist/cjs/watch-claim.js.map +1 -1
  160. package/dist/cjs/watch.js.map +1 -1
  161. package/dist/cjs/workers/parse.js.map +1 -1
  162. package/dist/cjs/workers/watch-heartbeat.js.map +1 -1
  163. package/dist/esm/chunk/extract.js.map +1 -1
  164. package/dist/esm/chunk/group.js.map +1 -1
  165. package/dist/esm/chunk/index.js.map +1 -1
  166. package/dist/esm/chunk/parse.js.map +1 -1
  167. package/dist/esm/chunk/parser.js.map +1 -1
  168. package/dist/esm/chunk/tokens.js.map +1 -1
  169. package/dist/esm/chunk/types.js.map +1 -1
  170. package/dist/esm/chunk/version.js.map +1 -1
  171. package/dist/esm/cli/build.js.map +1 -1
  172. package/dist/esm/cli/download.js.map +1 -1
  173. package/dist/esm/cli/index.d.ts +2 -2
  174. package/dist/esm/cli/index.js +2 -2
  175. package/dist/esm/cli/index.js.map +1 -1
  176. package/dist/esm/cli/init.js.map +1 -1
  177. package/dist/esm/cli/map.js.map +1 -1
  178. package/dist/esm/cli/named.js.map +1 -1
  179. package/dist/esm/cli/path.js.map +1 -1
  180. package/dist/esm/cli/peek.js +20 -2
  181. package/dist/esm/cli/peek.js.map +1 -1
  182. package/dist/esm/cli/related.js.map +1 -1
  183. package/dist/esm/cli/search.js +12 -2
  184. package/dist/esm/cli/search.js.map +1 -1
  185. package/dist/esm/cli/shared.js.map +1 -1
  186. package/dist/esm/cli/sql.js.map +1 -1
  187. package/dist/esm/cli/status.js.map +1 -1
  188. package/dist/esm/cli/types.js.map +1 -1
  189. package/dist/esm/cli/watch.js.map +1 -1
  190. package/dist/esm/cli.js.map +1 -1
  191. package/dist/esm/commands/index.d.ts +2 -2
  192. package/dist/esm/commands/index.js.map +1 -1
  193. package/dist/esm/commands/map.js.map +1 -1
  194. package/dist/esm/commands/path.js.map +1 -1
  195. package/dist/esm/commands/peek.d.ts +9 -1
  196. package/dist/esm/commands/peek.js +26 -12
  197. package/dist/esm/commands/peek.js.map +1 -1
  198. package/dist/esm/commands/related.js.map +1 -1
  199. package/dist/esm/commands/scope.js.map +1 -1
  200. package/dist/esm/commands/search.d.ts +6 -2
  201. package/dist/esm/commands/search.js +11 -3
  202. package/dist/esm/commands/search.js.map +1 -1
  203. package/dist/esm/commands/signals.d.ts +19 -5
  204. package/dist/esm/commands/signals.js +64 -14
  205. package/dist/esm/commands/signals.js.map +1 -1
  206. package/dist/esm/commands/snapshot.js.map +1 -1
  207. package/dist/esm/commands/status.js.map +1 -1
  208. package/dist/esm/config/access.js.map +1 -1
  209. package/dist/esm/config/index.js.map +1 -1
  210. package/dist/esm/config/load.js +6 -0
  211. package/dist/esm/config/load.js.map +1 -1
  212. package/dist/esm/config/resolve.js.map +1 -1
  213. package/dist/esm/config/signals.js.map +1 -1
  214. package/dist/esm/config/types.js.map +1 -1
  215. package/dist/esm/config/validate.js.map +1 -1
  216. package/dist/esm/embed/cohere.js.map +1 -1
  217. package/dist/esm/embed/distribution.js.map +1 -1
  218. package/dist/esm/embed/endpoint.js.map +1 -1
  219. package/dist/esm/embed/handoff.js.map +1 -1
  220. package/dist/esm/embed/http.js.map +1 -1
  221. package/dist/esm/embed/identity.js.map +1 -1
  222. package/dist/esm/embed/langfit.js.map +1 -1
  223. package/dist/esm/embed/languages.js.map +1 -1
  224. package/dist/esm/embed/openai.js.map +1 -1
  225. package/dist/esm/embed/query.js.map +1 -1
  226. package/dist/esm/embed/registry.js.map +1 -1
  227. package/dist/esm/embed/static.js.map +1 -1
  228. package/dist/esm/embed/store.js.map +1 -1
  229. package/dist/esm/embed/types.js.map +1 -1
  230. package/dist/esm/errors.js.map +1 -1
  231. package/dist/esm/features/embed.js.map +1 -1
  232. package/dist/esm/features/fences.js.map +1 -1
  233. package/dist/esm/features/index.js.map +1 -1
  234. package/dist/esm/features/links.js.map +1 -1
  235. package/dist/esm/features/rank.js.map +1 -1
  236. package/dist/esm/features/sections.js.map +1 -1
  237. package/dist/esm/features/tags.js.map +1 -1
  238. package/dist/esm/features/types.js.map +1 -1
  239. package/dist/esm/graph/graph.js.map +1 -1
  240. package/dist/esm/graph/traverse.js.map +1 -1
  241. package/dist/esm/index.d.ts +1 -1
  242. package/dist/esm/index.js.map +1 -1
  243. package/dist/esm/lib/atomic-write.js.map +1 -1
  244. package/dist/esm/lib/guarded-tick.js.map +1 -1
  245. package/dist/esm/lib/serial-query.js.map +1 -1
  246. package/dist/esm/lib/watch-notification.js.map +1 -1
  247. package/dist/esm/lib/worker-file.js.map +1 -1
  248. package/dist/esm/output/column-hint.js.map +1 -1
  249. package/dist/esm/output/output.js.map +1 -1
  250. package/dist/esm/output/progress.js.map +1 -1
  251. package/dist/esm/output/search-error.js.map +1 -1
  252. package/dist/esm/scan/frontmatter.js.map +1 -1
  253. package/dist/esm/scan/index.js.map +1 -1
  254. package/dist/esm/scan/list.js.map +1 -1
  255. package/dist/esm/scan/pool.js.map +1 -1
  256. package/dist/esm/scan/reparse.js.map +1 -1
  257. package/dist/esm/scan/worker-error.js.map +1 -1
  258. package/dist/esm/store/batch.js.map +1 -1
  259. package/dist/esm/store/builder.js.map +1 -1
  260. package/dist/esm/store/cache.js.map +1 -1
  261. package/dist/esm/store/duckdb/batch.js.map +1 -1
  262. package/dist/esm/store/duckdb/connection.js.map +1 -1
  263. package/dist/esm/store/duckdb/fieldStats.js.map +1 -1
  264. package/dist/esm/store/duckdb/lexical.d.ts +1 -6
  265. package/dist/esm/store/duckdb/lexical.js +13 -17
  266. package/dist/esm/store/duckdb/lexical.js.map +1 -1
  267. package/dist/esm/store/duckdb/native.d.ts +1 -2
  268. package/dist/esm/store/duckdb/native.js +1 -3
  269. package/dist/esm/store/duckdb/native.js.map +1 -1
  270. package/dist/esm/store/duckdb/open.js.map +1 -1
  271. package/dist/esm/store/duckdb/ordered-bm25.d.ts +3 -16
  272. package/dist/esm/store/duckdb/ordered-bm25.js +18 -54
  273. package/dist/esm/store/duckdb/ordered-bm25.js.map +1 -1
  274. package/dist/esm/store/duckdb/reconcile.js.map +1 -1
  275. package/dist/esm/store/duckdb/sql-functions.js.map +1 -1
  276. package/dist/esm/store/duckdb/store.js.map +1 -1
  277. package/dist/esm/store/duckdb/vectors.js.map +1 -1
  278. package/dist/esm/store/embed-scope.js.map +1 -1
  279. package/dist/esm/store/feature-scope.js.map +1 -1
  280. package/dist/esm/store/index.js.map +1 -1
  281. package/dist/esm/store/lock-wait.js.map +1 -1
  282. package/dist/esm/store/native.js.map +1 -1
  283. package/dist/esm/store/open.js.map +1 -1
  284. package/dist/esm/store/preset-scope.js.map +1 -1
  285. package/dist/esm/store/readiness.js.map +1 -1
  286. package/dist/esm/store/reconcile.js.map +1 -1
  287. package/dist/esm/store/shared.js.map +1 -1
  288. package/dist/esm/store/signature.js.map +1 -1
  289. package/dist/esm/store/sql-functions.js.map +1 -1
  290. package/dist/esm/store/sqlite/connection.js.map +1 -1
  291. package/dist/esm/store/sqlite/fieldStats.js.map +1 -1
  292. package/dist/esm/store/sqlite/lexical.js.map +1 -1
  293. package/dist/esm/store/sqlite/open.js.map +1 -1
  294. package/dist/esm/store/sqlite/reconcile.js.map +1 -1
  295. package/dist/esm/store/sqlite/sql-functions.js.map +1 -1
  296. package/dist/esm/store/sqlite/store.js.map +1 -1
  297. package/dist/esm/store/sqlite/vectors.js.map +1 -1
  298. package/dist/esm/store/stages.js.map +1 -1
  299. package/dist/esm/store/transaction.js.map +1 -1
  300. package/dist/esm/store/turso/connection.js.map +1 -1
  301. package/dist/esm/store/turso/fieldStats.js.map +1 -1
  302. package/dist/esm/store/turso/lexical-text.js.map +1 -1
  303. package/dist/esm/store/turso/lexical.js.map +1 -1
  304. package/dist/esm/store/turso/native.js.map +1 -1
  305. package/dist/esm/store/turso/open.js.map +1 -1
  306. package/dist/esm/store/turso/reconcile.js.map +1 -1
  307. package/dist/esm/store/turso/sql-functions.js.map +1 -1
  308. package/dist/esm/store/turso/store.js.map +1 -1
  309. package/dist/esm/store/turso/vectors.js.map +1 -1
  310. package/dist/esm/store/types.js.map +1 -1
  311. package/dist/esm/store/vectors.js.map +1 -1
  312. package/dist/esm/text/segment.js.map +1 -1
  313. package/dist/esm/text/strip.js.map +1 -1
  314. package/dist/esm/watch-claim.js.map +1 -1
  315. package/dist/esm/watch.js.map +1 -1
  316. package/dist/esm/workers/parse.js.map +1 -1
  317. package/dist/esm/workers/watch-heartbeat.js.map +1 -1
  318. package/package.json +2 -2
  319. package/skills/sense/EXAMPLES.md +3 -3
  320. package/skills/sense/SKILL.md +7 -1
  321. package/skills/sense/references/search.md +17 -1
  322. package/skills/sense/references/sql.md +1 -1
  323. package/skills/sense/references/terminology.md +32 -0
  324. package/skills/sense-setup/references/store-benchmarks.md +5 -5
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/config/validate.ts"],"sourcesContent":["import { SenseError } from '../errors.ts';\nimport { SIGNAL_NAMES, SIGNAL_PREREQUISITES, type SignalName } from './signals.ts';\nimport { type Config, STORE_NAMES, type StoreName } from './types.ts';\n\n// Shape check for hand-edited files, so a typo names itself instead of surfacing as a\n// TypeError. Unknown top-level keys warn (forward compat); unknown keys inside a block error.\nconst KNOWN_KEYS = new Set(['$schema', 'version', 'root', 'build', 'presets', 'features', 'embed', 'store', 'queries']);\nconst KNOWN_PRESET_KEYS = new Set(['include', 'exclude', 'k', 'signals', 'where']);\nconst KNOWN_FEATURE_KEYS = new Set(['links', 'sections', 'tags', 'rank']);\n// The one source of truth for what the embed block accepts. schema.json carries an editor-facing\n// copy for autocomplete; nothing at runtime reads it.\nexport const KNOWN_EMBED_KEYS = new Set(['model', 'provider', 'url', 'key', 'languages', 'chunkTokens']);\nconst SAVED_SEARCH_KEYS = new Set(['search', 'preset', 'include', 'exclude', 'where', 'k']);\nconst RESERVED_QUERY_NAMES = new Set(['init', 'build', 'watch', 'status', 'download', 'sql', 'search', 'map', 'peek', 'path', 'related']);\n\nexport function unknownConfigKeys(cfg: Record<string, unknown>): string[] {\n return Object.keys(cfg).filter((k) => !KNOWN_KEYS.has(k));\n}\n\nfunction isNonEmptyStringArray(value: unknown): value is string[] {\n return Array.isArray(value) && value.length > 0 && value.every((g) => typeof g === 'string' && g.length > 0);\n}\n\n// Pre-v3 shape check, just enough for migrateConfig to run safely (it reads scan.include\n// directly). The full v3 shape is checked by validateConfig once migration has produced it.\nexport function validateLegacyScan(parsed: unknown, configPath: string): void {\n const cfg = typeof parsed === 'object' && parsed !== null && !Array.isArray(parsed) ? (parsed as Record<string, unknown>) : null;\n const scan = cfg?.scan as { include?: unknown } | undefined;\n if (!cfg || !scan || !isNonEmptyStringArray(scan.include)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: scan.include must be a non-empty array of glob strings`);\n }\n}\n\nfunction validateStoreKey(value: unknown, configPath: string): void {\n if (!STORE_NAMES.includes(value as StoreName)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: store must be ${STORE_NAMES.map((name) => `\"${name}\"`).join(' or ')}`);\n }\n}\n\nfunction validateFeaturesBlock(value: unknown, configPath: string): void {\n if (typeof value !== 'object' || value === null || Array.isArray(value)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: features must be an object of name -> boolean`);\n }\n const block = value as Record<string, unknown>;\n const unknown = Object.keys(block).filter((k) => !KNOWN_FEATURE_KEYS.has(k));\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: features has unknown key(s) ${unknown.join(', ')}; embed is not a features key -- see the top-level \"embed\" block`);\n }\n for (const [name, v] of Object.entries(block)) {\n if (typeof v !== 'boolean') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: features.${name} must be a boolean`);\n }\n }\n}\n\nfunction validateEmbedBlock(value: unknown, configPath: string): void {\n if (typeof value !== 'object' || value === null || Array.isArray(value)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed must be an object of { model?, provider?: \"static\"|\"openai\"|\"cohere\", url?, key? }`);\n }\n const embed = value as Record<string, unknown>;\n const unknown = Object.keys(embed).filter((k) => !KNOWN_EMBED_KEYS.has(k));\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed has unknown key(s) ${unknown.join(', ')}`);\n }\n if (embed.model !== undefined && typeof embed.model !== 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.model must be a string`);\n }\n // The block exists to name a model; provider settings without one would read as \"vectors\n // configured\" while embedding stays off, which is the ambiguity v4 removed.\n if (embed.model === undefined || embed.model === '') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.model is required when the \"embed\" block is present (it is what turns vectors on); remove the block to index without vectors`);\n }\n if (embed.provider !== undefined && embed.provider !== 'static' && embed.provider !== 'openai' && embed.provider !== 'cohere') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.provider must be \"static\", \"openai\", or \"cohere\"`);\n }\n if (embed.url !== undefined && typeof embed.url !== 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.url must be a string`);\n }\n if (embed.key !== undefined && typeof embed.key !== 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.key must be a string`);\n }\n // openai has no default host, so a missing url would otherwise only surface at the first\n // vector search; a config error belongs at config time.\n if (embed.provider === 'openai' && (embed.url === undefined || embed.url === '')) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.provider \"openai\" requires embed.url (e.g. \"http://localhost:11434/v1\" for Ollama)`);\n }\n if (embed.languages !== undefined && (!Array.isArray(embed.languages) || embed.languages.length === 0 || embed.languages.some((l) => typeof l !== 'string' || l === ''))) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.languages must be a non-empty array of language tags (e.g. [\"en\", \"zh\"])`);\n }\n if (embed.chunkTokens !== undefined && (typeof embed.chunkTokens !== 'number' || !Number.isInteger(embed.chunkTokens) || embed.chunkTokens <= 0)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.chunkTokens must be a positive integer`);\n }\n}\n\nfunction isSignalName(value: unknown): value is SignalName {\n return typeof value === 'string' && (SIGNAL_NAMES as readonly string[]).includes(value);\n}\n\nfunction isSignalWeight(value: unknown): value is number {\n return typeof value === 'number' && Number.isFinite(value) && value > 0;\n}\n\n// Shape (an object of signal name -> weight) plus the one cross-signal prerequisite that doesn't\n// need the top-level embed block: links is seeded by word-match rows, so declaring it without words is always empty.\nfunction validatePresetSignals(name: string, value: unknown, configPath: string): void {\n if (typeof value !== 'object' || value === null || Array.isArray(value) || Object.keys(value).length === 0) {\n throw new SenseError(\n 'CONFIG_INVALID',\n `${configPath}: presets.${name}.signals must be a non-empty object of signal name -> weight (a finite number > 0), naming ${SIGNAL_NAMES.join(', ')} -- a preset that searches nothing is a mistake; name at least \"words\", or omit \"signals\" to use every signal whose prerequisites hold at weight 1`\n );\n }\n const entries = Object.entries(value as Record<string, unknown>);\n const unknown = entries.filter(([k]) => !isSignalName(k)).map(([k]) => k);\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.signals names unknown signal(s) ${unknown.join(', ')}; valid signals are ${SIGNAL_NAMES.join(', ')}`);\n }\n for (const [signal, weight] of entries) {\n if (!isSignalWeight(weight)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.signals.${signal} must be a finite number > 0`);\n }\n }\n const names = new Set(entries.map(([k]) => k));\n for (const [signal, requires] of Object.entries(SIGNAL_PREREQUISITES) as Array<[SignalName, SignalName]>) {\n if (names.has(signal) && !names.has(requires)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.signals has \"${signal}\" without \"${requires}\"; add \"${requires}\" to presets.${name}.signals, or drop \"${signal}\"`);\n }\n }\n}\n\nfunction validatePreset(name: string, value: unknown, configPath: string): void {\n if (typeof value !== 'object' || value === null || Array.isArray(value)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name} must be an object`);\n }\n const preset = value as Record<string, unknown>;\n const unknown = Object.keys(preset).filter((k) => !KNOWN_PRESET_KEYS.has(k));\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name} has unknown key(s) ${unknown.join(', ')}; a preset takes include, exclude, k, signals, where`);\n }\n if (!isNonEmptyStringArray(preset.include)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.include must be a non-empty array of glob strings`);\n }\n if (preset.exclude !== undefined && !isNonEmptyStringArray(preset.exclude)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.exclude must be a non-empty array of glob strings`);\n }\n if (preset.signals !== undefined) validatePresetSignals(name, preset.signals, configPath);\n if (preset.k !== undefined && (typeof preset.k !== 'number' || !Number.isInteger(preset.k) || preset.k <= 0)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.k must be a positive integer`);\n }\n if (preset.where !== undefined && typeof preset.where !== 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.where must be a SQL condition string`);\n }\n}\n\n// A queries.<name> entry: { sql } or a saved search { search, preset?, include?, exclude?, where?, k? }.\nfunction validateSavedQuery(name: string, value: unknown, configPath: string): void {\n if (RESERVED_QUERY_NAMES.has(name)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} uses a reserved command name and can never be run as a saved query`);\n }\n // A bare string is rejected rather than inferred to be SQL: an entry says which of the two\n // verbs it runs, the same choice the CLI makes explicit.\n if (typeof value === 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} must say which verb it runs: { \"sql\": ${JSON.stringify(value.length > 40 ? `${value.slice(0, 37)}...` : value)} } to run it as SQL, or { \"search\": \"...\" } for a ranked search`);\n }\n if (typeof value !== 'object' || value === null || Array.isArray(value)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} must be { sql } or { search, preset?, include?, exclude?, where?, k? }`);\n }\n const entry = value as Record<string, unknown>;\n if ('sql' in entry) {\n const unknown = Object.keys(entry).filter((k) => k !== 'sql');\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} has unknown key(s) ${unknown.join(', ')}; a { sql } query takes only sql`);\n }\n if (typeof entry.sql !== 'string' || entry.sql.trim() === '') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.sql must be a non-empty string`);\n }\n return;\n }\n if ('search' in entry) {\n const unknown = Object.keys(entry).filter((k) => !SAVED_SEARCH_KEYS.has(k));\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} has unknown key(s) ${unknown.join(', ')}; a saved search takes search, preset, include, exclude, where, k`);\n }\n // A saved entry saves a question; a scope without a question is just flags.\n if (typeof entry.search !== 'string' || entry.search.trim() === '') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.search must be non-empty text`);\n }\n if (entry.preset !== undefined && (typeof entry.preset !== 'string' || entry.preset.length === 0)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.preset must be a preset name`);\n }\n if (entry.include !== undefined && !isNonEmptyStringArray(entry.include)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.include must be a non-empty array of glob strings`);\n }\n if (entry.exclude !== undefined && !isNonEmptyStringArray(entry.exclude)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.exclude must be a non-empty array of glob strings`);\n }\n if (entry.where !== undefined && typeof entry.where !== 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.where must be a SQL condition string`);\n }\n if (entry.k !== undefined && (typeof entry.k !== 'number' || !Number.isInteger(entry.k) || entry.k <= 0)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.k must be a positive integer`);\n }\n return;\n }\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} must be { sql } or { search, preset?, include?, exclude?, where?, k? }`);\n}\n\nexport function validateConfig(parsed: unknown, configPath: string): Config {\n if (typeof parsed !== 'object' || parsed === null || Array.isArray(parsed)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: config must be a JSON object`);\n }\n const cfg = parsed as Record<string, unknown>;\n\n if (cfg.root !== undefined && (typeof cfg.root !== 'string' || cfg.root.length === 0)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: root must be a non-empty path string`);\n }\n if (cfg.build !== undefined && typeof cfg.build !== 'boolean') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: build must be a boolean`);\n }\n\n // `checks` is rejected by name, not warned: silence would hide that assertions are gone.\n if (cfg.checks !== undefined) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: checks was removed in v3 -- sense check no longer asserts on saved queries; a returned row set is the reader's judgment`);\n }\n\n // `content` (content.tokenize) is rejected by name for the same reason: every outcome it\n // selected (unspaced-script search, stemming, quoted-phrase substring matching) is now\n // automatic on every store, so a config that still sets it would silently get nothing back.\n if (cfg.content !== undefined) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: content.tokenize was removed -- every outcome it selected is now automatic on every store; remove the \"content\" block`);\n }\n\n const presets = cfg.presets as Record<string, unknown> | undefined;\n if (!presets || typeof presets !== 'object' || Array.isArray(presets) || Object.keys(presets).length === 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets must be a non-empty object of preset name -> { include, exclude?, k?, signals?, where? }`);\n }\n for (const [name, value] of Object.entries(presets)) {\n validatePreset(name, value, configPath);\n }\n if (presets.default === undefined) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets must include a \"default\" preset`);\n }\n\n if (cfg.queries === undefined) cfg.queries = {};\n if (typeof cfg.queries !== 'object' || cfg.queries === null || Array.isArray(cfg.queries)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries must be an object of name -> { sql } or { search }`);\n }\n const queries = cfg.queries as Record<string, unknown>;\n for (const [name, value] of Object.entries(queries)) {\n validateSavedQuery(name, value, configPath);\n }\n\n if (cfg.features !== undefined) {\n validateFeaturesBlock(cfg.features, configPath);\n }\n if (cfg.embed !== undefined) {\n validateEmbedBlock(cfg.embed, configPath);\n }\n if (cfg.store !== undefined) {\n validateStoreKey(cfg.store, configPath);\n }\n\n // vectors' prerequisite is a named model, checked here once embed is known valid (or absent).\n if (cfg.embed === undefined) {\n for (const [name, value] of Object.entries(presets)) {\n const signals = (value as Record<string, unknown>).signals as Record<string, unknown> | undefined;\n if (signals && 'vectors' in signals) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.signals includes \"vectors\", but no \"embed\" block names a model; add one, or drop \"vectors\" from presets.${name}.signals`);\n }\n }\n }\n\n return cfg as unknown as Config;\n}\n"],"names":["SenseError","SIGNAL_NAMES","SIGNAL_PREREQUISITES","STORE_NAMES","KNOWN_KEYS","Set","KNOWN_PRESET_KEYS","KNOWN_FEATURE_KEYS","KNOWN_EMBED_KEYS","SAVED_SEARCH_KEYS","RESERVED_QUERY_NAMES","unknownConfigKeys","cfg","Object","keys","filter","k","has","isNonEmptyStringArray","value","Array","isArray","length","every","g","validateLegacyScan","parsed","configPath","scan","include","validateStoreKey","includes","map","name","join","validateFeaturesBlock","block","unknown","v","entries","validateEmbedBlock","embed","model","undefined","provider","url","key","languages","some","l","chunkTokens","Number","isInteger","isSignalName","isSignalWeight","isFinite","validatePresetSignals","signal","weight","names","requires","validatePreset","preset","exclude","signals","where","validateSavedQuery","JSON","stringify","slice","entry","sql","trim","search","validateConfig","root","build","checks","content","presets","default","queries","features","store"],"mappings":"AAAA,SAASA,UAAU,QAAQ,eAAe;AAC1C,SAASC,YAAY,EAAEC,oBAAoB,QAAyB,eAAe;AACnF,SAAsBC,WAAW,QAAwB,aAAa;AAEtE,sFAAsF;AACtF,8FAA8F;AAC9F,MAAMC,aAAa,IAAIC,IAAI;IAAC;IAAW;IAAW;IAAQ;IAAS;IAAW;IAAY;IAAS;IAAS;CAAU;AACtH,MAAMC,oBAAoB,IAAID,IAAI;IAAC;IAAW;IAAW;IAAK;IAAW;CAAQ;AACjF,MAAME,qBAAqB,IAAIF,IAAI;IAAC;IAAS;IAAY;IAAQ;CAAO;AACxE,iGAAiG;AACjG,sDAAsD;AACtD,OAAO,MAAMG,mBAAmB,IAAIH,IAAI;IAAC;IAAS;IAAY;IAAO;IAAO;IAAa;CAAc,EAAE;AACzG,MAAMI,oBAAoB,IAAIJ,IAAI;IAAC;IAAU;IAAU;IAAW;IAAW;IAAS;CAAI;AAC1F,MAAMK,uBAAuB,IAAIL,IAAI;IAAC;IAAQ;IAAS;IAAS;IAAU;IAAY;IAAO;IAAU;IAAO;IAAQ;IAAQ;CAAU;AAExI,OAAO,SAASM,kBAAkBC,GAA4B;IAC5D,OAAOC,OAAOC,IAAI,CAACF,KAAKG,MAAM,CAAC,CAACC,IAAM,CAACZ,WAAWa,GAAG,CAACD;AACxD;AAEA,SAASE,sBAAsBC,KAAc;IAC3C,OAAOC,MAAMC,OAAO,CAACF,UAAUA,MAAMG,MAAM,GAAG,KAAKH,MAAMI,KAAK,CAAC,CAACC,IAAM,OAAOA,MAAM,YAAYA,EAAEF,MAAM,GAAG;AAC5G;AAEA,yFAAyF;AACzF,4FAA4F;AAC5F,OAAO,SAASG,mBAAmBC,MAAe,EAAEC,UAAkB;IACpE,MAAMf,MAAM,OAAOc,WAAW,YAAYA,WAAW,QAAQ,CAACN,MAAMC,OAAO,CAACK,UAAWA,SAAqC;IAC5H,MAAME,OAAOhB,gBAAAA,0BAAAA,IAAKgB,IAAI;IACtB,IAAI,CAAChB,OAAO,CAACgB,QAAQ,CAACV,sBAAsBU,KAAKC,OAAO,GAAG;QACzD,MAAM,IAAI7B,WAAW,kBAAkB,GAAG2B,WAAW,wDAAwD,CAAC;IAChH;AACF;AAEA,SAASG,iBAAiBX,KAAc,EAAEQ,UAAkB;IAC1D,IAAI,CAACxB,YAAY4B,QAAQ,CAACZ,QAAqB;QAC7C,MAAM,IAAInB,WAAW,kBAAkB,GAAG2B,WAAW,gBAAgB,EAAExB,YAAY6B,GAAG,CAAC,CAACC,OAAS,CAAC,CAAC,EAAEA,KAAK,CAAC,CAAC,EAAEC,IAAI,CAAC,SAAS;IAC9H;AACF;AAEA,SAASC,sBAAsBhB,KAAc,EAAEQ,UAAkB;IAC/D,IAAI,OAAOR,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,QAAQ;QACvE,MAAM,IAAInB,WAAW,kBAAkB,GAAG2B,WAAW,+CAA+C,CAAC;IACvG;IACA,MAAMS,QAAQjB;IACd,MAAMkB,UAAUxB,OAAOC,IAAI,CAACsB,OAAOrB,MAAM,CAAC,CAACC,IAAM,CAACT,mBAAmBU,GAAG,CAACD;IACzE,IAAIqB,QAAQf,MAAM,GAAG,GAAG;QACtB,MAAM,IAAItB,WAAW,kBAAkB,GAAG2B,WAAW,8BAA8B,EAAEU,QAAQH,IAAI,CAAC,MAAM,gEAAgE,CAAC;IAC3K;IACA,KAAK,MAAM,CAACD,MAAMK,EAAE,IAAIzB,OAAO0B,OAAO,CAACH,OAAQ;QAC7C,IAAI,OAAOE,MAAM,WAAW;YAC1B,MAAM,IAAItC,WAAW,kBAAkB,GAAG2B,WAAW,WAAW,EAAEM,KAAK,kBAAkB,CAAC;QAC5F;IACF;AACF;AAEA,SAASO,mBAAmBrB,KAAc,EAAEQ,UAAkB;IAC5D,IAAI,OAAOR,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,QAAQ;QACvE,MAAM,IAAInB,WAAW,kBAAkB,GAAG2B,WAAW,0FAA0F,CAAC;IAClJ;IACA,MAAMc,QAAQtB;IACd,MAAMkB,UAAUxB,OAAOC,IAAI,CAAC2B,OAAO1B,MAAM,CAAC,CAACC,IAAM,CAACR,iBAAiBS,GAAG,CAACD;IACvE,IAAIqB,QAAQf,MAAM,GAAG,GAAG;QACtB,MAAM,IAAItB,WAAW,kBAAkB,GAAG2B,WAAW,2BAA2B,EAAEU,QAAQH,IAAI,CAAC,OAAO;IACxG;IACA,IAAIO,MAAMC,KAAK,KAAKC,aAAa,OAAOF,MAAMC,KAAK,KAAK,UAAU;QAChE,MAAM,IAAI1C,WAAW,kBAAkB,GAAG2B,WAAW,8BAA8B,CAAC;IACtF;IACA,yFAAyF;IACzF,4EAA4E;IAC5E,IAAIc,MAAMC,KAAK,KAAKC,aAAaF,MAAMC,KAAK,KAAK,IAAI;QACnD,MAAM,IAAI1C,WAAW,kBAAkB,GAAG2B,WAAW,oIAAoI,CAAC;IAC5L;IACA,IAAIc,MAAMG,QAAQ,KAAKD,aAAaF,MAAMG,QAAQ,KAAK,YAAYH,MAAMG,QAAQ,KAAK,YAAYH,MAAMG,QAAQ,KAAK,UAAU;QAC7H,MAAM,IAAI5C,WAAW,kBAAkB,GAAG2B,WAAW,wDAAwD,CAAC;IAChH;IACA,IAAIc,MAAMI,GAAG,KAAKF,aAAa,OAAOF,MAAMI,GAAG,KAAK,UAAU;QAC5D,MAAM,IAAI7C,WAAW,kBAAkB,GAAG2B,WAAW,4BAA4B,CAAC;IACpF;IACA,IAAIc,MAAMK,GAAG,KAAKH,aAAa,OAAOF,MAAMK,GAAG,KAAK,UAAU;QAC5D,MAAM,IAAI9C,WAAW,kBAAkB,GAAG2B,WAAW,4BAA4B,CAAC;IACpF;IACA,yFAAyF;IACzF,wDAAwD;IACxD,IAAIc,MAAMG,QAAQ,KAAK,YAAaH,CAAAA,MAAMI,GAAG,KAAKF,aAAaF,MAAMI,GAAG,KAAK,EAAC,GAAI;QAChF,MAAM,IAAI7C,WAAW,kBAAkB,GAAG2B,WAAW,0FAA0F,CAAC;IAClJ;IACA,IAAIc,MAAMM,SAAS,KAAKJ,aAAc,CAAA,CAACvB,MAAMC,OAAO,CAACoB,MAAMM,SAAS,KAAKN,MAAMM,SAAS,CAACzB,MAAM,KAAK,KAAKmB,MAAMM,SAAS,CAACC,IAAI,CAAC,CAACC,IAAM,OAAOA,MAAM,YAAYA,MAAM,GAAE,GAAI;QACxK,MAAM,IAAIjD,WAAW,kBAAkB,GAAG2B,WAAW,gFAAgF,CAAC;IACxI;IACA,IAAIc,MAAMS,WAAW,KAAKP,aAAc,CAAA,OAAOF,MAAMS,WAAW,KAAK,YAAY,CAACC,OAAOC,SAAS,CAACX,MAAMS,WAAW,KAAKT,MAAMS,WAAW,IAAI,CAAA,GAAI;QAChJ,MAAM,IAAIlD,WAAW,kBAAkB,GAAG2B,WAAW,8CAA8C,CAAC;IACtG;AACF;AAEA,SAAS0B,aAAalC,KAAc;IAClC,OAAO,OAAOA,UAAU,YAAY,AAAClB,aAAmC8B,QAAQ,CAACZ;AACnF;AAEA,SAASmC,eAAenC,KAAc;IACpC,OAAO,OAAOA,UAAU,YAAYgC,OAAOI,QAAQ,CAACpC,UAAUA,QAAQ;AACxE;AAEA,iGAAiG;AACjG,qHAAqH;AACrH,SAASqC,sBAAsBvB,IAAY,EAAEd,KAAc,EAAEQ,UAAkB;IAC7E,IAAI,OAAOR,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,UAAUN,OAAOC,IAAI,CAACK,OAAOG,MAAM,KAAK,GAAG;QAC1G,MAAM,IAAItB,WACR,kBACA,GAAG2B,WAAW,UAAU,EAAEM,KAAK,2FAA2F,EAAEhC,aAAaiC,IAAI,CAAC,MAAM,kJAAkJ,CAAC;IAE3S;IACA,MAAMK,UAAU1B,OAAO0B,OAAO,CAACpB;IAC/B,MAAMkB,UAAUE,QAAQxB,MAAM,CAAC,CAAC,CAACC,EAAE,GAAK,CAACqC,aAAarC,IAAIgB,GAAG,CAAC,CAAC,CAAChB,EAAE,GAAKA;IACvE,IAAIqB,QAAQf,MAAM,GAAG,GAAG;QACtB,MAAM,IAAItB,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,iCAAiC,EAAEI,QAAQH,IAAI,CAAC,MAAM,oBAAoB,EAAEjC,aAAaiC,IAAI,CAAC,OAAO;IAC7K;IACA,KAAK,MAAM,CAACuB,QAAQC,OAAO,IAAInB,QAAS;QACtC,IAAI,CAACe,eAAeI,SAAS;YAC3B,MAAM,IAAI1D,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,SAAS,EAAEwB,OAAO,4BAA4B,CAAC;QACvH;IACF;IACA,MAAME,QAAQ,IAAItD,IAAIkC,QAAQP,GAAG,CAAC,CAAC,CAAChB,EAAE,GAAKA;IAC3C,KAAK,MAAM,CAACyC,QAAQG,SAAS,IAAI/C,OAAO0B,OAAO,CAACrC,sBAA0D;QACxG,IAAIyD,MAAM1C,GAAG,CAACwC,WAAW,CAACE,MAAM1C,GAAG,CAAC2C,WAAW;YAC7C,MAAM,IAAI5D,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,cAAc,EAAEwB,OAAO,WAAW,EAAEG,SAAS,QAAQ,EAAEA,SAAS,aAAa,EAAE3B,KAAK,mBAAmB,EAAEwB,OAAO,CAAC,CAAC;QAC1L;IACF;AACF;AAEA,SAASI,eAAe5B,IAAY,EAAEd,KAAc,EAAEQ,UAAkB;IACtE,IAAI,OAAOR,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,QAAQ;QACvE,MAAM,IAAInB,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,kBAAkB,CAAC;IAC3F;IACA,MAAM6B,SAAS3C;IACf,MAAMkB,UAAUxB,OAAOC,IAAI,CAACgD,QAAQ/C,MAAM,CAAC,CAACC,IAAM,CAACV,kBAAkBW,GAAG,CAACD;IACzE,IAAIqB,QAAQf,MAAM,GAAG,GAAG;QACtB,MAAM,IAAItB,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,oBAAoB,EAAEI,QAAQH,IAAI,CAAC,MAAM,oDAAoD,CAAC;IACtK;IACA,IAAI,CAAChB,sBAAsB4C,OAAOjC,OAAO,GAAG;QAC1C,MAAM,IAAI7B,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,kDAAkD,CAAC;IAC3H;IACA,IAAI6B,OAAOC,OAAO,KAAKpB,aAAa,CAACzB,sBAAsB4C,OAAOC,OAAO,GAAG;QAC1E,MAAM,IAAI/D,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,kDAAkD,CAAC;IAC3H;IACA,IAAI6B,OAAOE,OAAO,KAAKrB,WAAWa,sBAAsBvB,MAAM6B,OAAOE,OAAO,EAAErC;IAC9E,IAAImC,OAAO9C,CAAC,KAAK2B,aAAc,CAAA,OAAOmB,OAAO9C,CAAC,KAAK,YAAY,CAACmC,OAAOC,SAAS,CAACU,OAAO9C,CAAC,KAAK8C,OAAO9C,CAAC,IAAI,CAAA,GAAI;QAC5G,MAAM,IAAIhB,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,6BAA6B,CAAC;IACtG;IACA,IAAI6B,OAAOG,KAAK,KAAKtB,aAAa,OAAOmB,OAAOG,KAAK,KAAK,UAAU;QAClE,MAAM,IAAIjE,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,qCAAqC,CAAC;IAC9G;AACF;AAEA,yGAAyG;AACzG,SAASiC,mBAAmBjC,IAAY,EAAEd,KAAc,EAAEQ,UAAkB;IAC1E,IAAIjB,qBAAqBO,GAAG,CAACgB,OAAO;QAClC,MAAM,IAAIjC,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,mEAAmE,CAAC;IAC5I;IACA,2FAA2F;IAC3F,yDAAyD;IACzD,IAAI,OAAOd,UAAU,UAAU;QAC7B,MAAM,IAAInB,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,uCAAuC,EAAEkC,KAAKC,SAAS,CAACjD,MAAMG,MAAM,GAAG,KAAK,GAAGH,MAAMkD,KAAK,CAAC,GAAG,IAAI,GAAG,CAAC,GAAGlD,OAAO,+DAA+D,CAAC;IACxP;IACA,IAAI,OAAOA,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,QAAQ;QACvE,MAAM,IAAInB,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,uEAAuE,CAAC;IAChJ;IACA,MAAMqC,QAAQnD;IACd,IAAI,SAASmD,OAAO;QAClB,MAAMjC,UAAUxB,OAAOC,IAAI,CAACwD,OAAOvD,MAAM,CAAC,CAACC,IAAMA,MAAM;QACvD,IAAIqB,QAAQf,MAAM,GAAG,GAAG;YACtB,MAAM,IAAItB,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,oBAAoB,EAAEI,QAAQH,IAAI,CAAC,MAAM,gCAAgC,CAAC;QAClJ;QACA,IAAI,OAAOoC,MAAMC,GAAG,KAAK,YAAYD,MAAMC,GAAG,CAACC,IAAI,OAAO,IAAI;YAC5D,MAAM,IAAIxE,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,+BAA+B,CAAC;QACxG;QACA;IACF;IACA,IAAI,YAAYqC,OAAO;QACrB,MAAMjC,UAAUxB,OAAOC,IAAI,CAACwD,OAAOvD,MAAM,CAAC,CAACC,IAAM,CAACP,kBAAkBQ,GAAG,CAACD;QACxE,IAAIqB,QAAQf,MAAM,GAAG,GAAG;YACtB,MAAM,IAAItB,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,oBAAoB,EAAEI,QAAQH,IAAI,CAAC,MAAM,iEAAiE,CAAC;QACnL;QACA,4EAA4E;QAC5E,IAAI,OAAOoC,MAAMG,MAAM,KAAK,YAAYH,MAAMG,MAAM,CAACD,IAAI,OAAO,IAAI;YAClE,MAAM,IAAIxE,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,8BAA8B,CAAC;QACvG;QACA,IAAIqC,MAAMR,MAAM,KAAKnB,aAAc,CAAA,OAAO2B,MAAMR,MAAM,KAAK,YAAYQ,MAAMR,MAAM,CAACxC,MAAM,KAAK,CAAA,GAAI;YACjG,MAAM,IAAItB,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,6BAA6B,CAAC;QACtG;QACA,IAAIqC,MAAMzC,OAAO,KAAKc,aAAa,CAACzB,sBAAsBoD,MAAMzC,OAAO,GAAG;YACxE,MAAM,IAAI7B,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,kDAAkD,CAAC;QAC3H;QACA,IAAIqC,MAAMP,OAAO,KAAKpB,aAAa,CAACzB,sBAAsBoD,MAAMP,OAAO,GAAG;YACxE,MAAM,IAAI/D,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,kDAAkD,CAAC;QAC3H;QACA,IAAIqC,MAAML,KAAK,KAAKtB,aAAa,OAAO2B,MAAML,KAAK,KAAK,UAAU;YAChE,MAAM,IAAIjE,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,qCAAqC,CAAC;QAC9G;QACA,IAAIqC,MAAMtD,CAAC,KAAK2B,aAAc,CAAA,OAAO2B,MAAMtD,CAAC,KAAK,YAAY,CAACmC,OAAOC,SAAS,CAACkB,MAAMtD,CAAC,KAAKsD,MAAMtD,CAAC,IAAI,CAAA,GAAI;YACxG,MAAM,IAAIhB,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,6BAA6B,CAAC;QACtG;QACA;IACF;IACA,MAAM,IAAIjC,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,uEAAuE,CAAC;AAChJ;AAEA,OAAO,SAASyC,eAAehD,MAAe,EAAEC,UAAkB;IAChE,IAAI,OAAOD,WAAW,YAAYA,WAAW,QAAQN,MAAMC,OAAO,CAACK,SAAS;QAC1E,MAAM,IAAI1B,WAAW,kBAAkB,GAAG2B,WAAW,8BAA8B,CAAC;IACtF;IACA,MAAMf,MAAMc;IAEZ,IAAId,IAAI+D,IAAI,KAAKhC,aAAc,CAAA,OAAO/B,IAAI+D,IAAI,KAAK,YAAY/D,IAAI+D,IAAI,CAACrD,MAAM,KAAK,CAAA,GAAI;QACrF,MAAM,IAAItB,WAAW,kBAAkB,GAAG2B,WAAW,sCAAsC,CAAC;IAC9F;IACA,IAAIf,IAAIgE,KAAK,KAAKjC,aAAa,OAAO/B,IAAIgE,KAAK,KAAK,WAAW;QAC7D,MAAM,IAAI5E,WAAW,kBAAkB,GAAG2B,WAAW,yBAAyB,CAAC;IACjF;IAEA,yFAAyF;IACzF,IAAIf,IAAIiE,MAAM,KAAKlC,WAAW;QAC5B,MAAM,IAAI3C,WAAW,kBAAkB,GAAG2B,WAAW,yHAAyH,CAAC;IACjL;IAEA,yFAAyF;IACzF,uFAAuF;IACvF,4FAA4F;IAC5F,IAAIf,IAAIkE,OAAO,KAAKnC,WAAW;QAC7B,MAAM,IAAI3C,WAAW,kBAAkB,GAAG2B,WAAW,uHAAuH,CAAC;IAC/K;IAEA,MAAMoD,UAAUnE,IAAImE,OAAO;IAC3B,IAAI,CAACA,WAAW,OAAOA,YAAY,YAAY3D,MAAMC,OAAO,CAAC0D,YAAYlE,OAAOC,IAAI,CAACiE,SAASzD,MAAM,KAAK,GAAG;QAC1G,MAAM,IAAItB,WAAW,kBAAkB,GAAG2B,WAAW,kGAAkG,CAAC;IAC1J;IACA,KAAK,MAAM,CAACM,MAAMd,MAAM,IAAIN,OAAO0B,OAAO,CAACwC,SAAU;QACnDlB,eAAe5B,MAAMd,OAAOQ;IAC9B;IACA,IAAIoD,QAAQC,OAAO,KAAKrC,WAAW;QACjC,MAAM,IAAI3C,WAAW,kBAAkB,GAAG2B,WAAW,yCAAyC,CAAC;IACjG;IAEA,IAAIf,IAAIqE,OAAO,KAAKtC,WAAW/B,IAAIqE,OAAO,GAAG,CAAC;IAC9C,IAAI,OAAOrE,IAAIqE,OAAO,KAAK,YAAYrE,IAAIqE,OAAO,KAAK,QAAQ7D,MAAMC,OAAO,CAACT,IAAIqE,OAAO,GAAG;QACzF,MAAM,IAAIjF,WAAW,kBAAkB,GAAG2B,WAAW,4DAA4D,CAAC;IACpH;IACA,MAAMsD,UAAUrE,IAAIqE,OAAO;IAC3B,KAAK,MAAM,CAAChD,MAAMd,MAAM,IAAIN,OAAO0B,OAAO,CAAC0C,SAAU;QACnDf,mBAAmBjC,MAAMd,OAAOQ;IAClC;IAEA,IAAIf,IAAIsE,QAAQ,KAAKvC,WAAW;QAC9BR,sBAAsBvB,IAAIsE,QAAQ,EAAEvD;IACtC;IACA,IAAIf,IAAI6B,KAAK,KAAKE,WAAW;QAC3BH,mBAAmB5B,IAAI6B,KAAK,EAAEd;IAChC;IACA,IAAIf,IAAIuE,KAAK,KAAKxC,WAAW;QAC3Bb,iBAAiBlB,IAAIuE,KAAK,EAAExD;IAC9B;IAEA,8FAA8F;IAC9F,IAAIf,IAAI6B,KAAK,KAAKE,WAAW;QAC3B,KAAK,MAAM,CAACV,MAAMd,MAAM,IAAIN,OAAO0B,OAAO,CAACwC,SAAU;YACnD,MAAMf,UAAU,AAAC7C,MAAkC6C,OAAO;YAC1D,IAAIA,WAAW,aAAaA,SAAS;gBACnC,MAAM,IAAIhE,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,yGAAyG,EAAEA,KAAK,QAAQ,CAAC;YACjM;QACF;IACF;IAEA,OAAOrB;AACT"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/open-source/ai/sensemaking/src/config/validate.ts"],"sourcesContent":["import { SenseError } from '../errors.ts';\nimport { SIGNAL_NAMES, SIGNAL_PREREQUISITES, type SignalName } from './signals.ts';\nimport { type Config, STORE_NAMES, type StoreName } from './types.ts';\n\n// Shape check for hand-edited files, so a typo names itself instead of surfacing as a\n// TypeError. Unknown top-level keys warn (forward compat); unknown keys inside a block error.\nconst KNOWN_KEYS = new Set(['$schema', 'version', 'root', 'build', 'presets', 'features', 'embed', 'store', 'queries']);\nconst KNOWN_PRESET_KEYS = new Set(['include', 'exclude', 'k', 'signals', 'where']);\nconst KNOWN_FEATURE_KEYS = new Set(['links', 'sections', 'tags', 'rank']);\n// The one source of truth for what the embed block accepts. schema.json carries an editor-facing\n// copy for autocomplete; nothing at runtime reads it.\nexport const KNOWN_EMBED_KEYS = new Set(['model', 'provider', 'url', 'key', 'languages', 'chunkTokens']);\nconst SAVED_SEARCH_KEYS = new Set(['search', 'preset', 'include', 'exclude', 'where', 'k']);\nconst RESERVED_QUERY_NAMES = new Set(['init', 'build', 'watch', 'status', 'download', 'sql', 'search', 'map', 'peek', 'path', 'related']);\n\nexport function unknownConfigKeys(cfg: Record<string, unknown>): string[] {\n return Object.keys(cfg).filter((k) => !KNOWN_KEYS.has(k));\n}\n\nfunction isNonEmptyStringArray(value: unknown): value is string[] {\n return Array.isArray(value) && value.length > 0 && value.every((g) => typeof g === 'string' && g.length > 0);\n}\n\n// Pre-v3 shape check, just enough for migrateConfig to run safely (it reads scan.include\n// directly). The full v3 shape is checked by validateConfig once migration has produced it.\nexport function validateLegacyScan(parsed: unknown, configPath: string): void {\n const cfg = typeof parsed === 'object' && parsed !== null && !Array.isArray(parsed) ? (parsed as Record<string, unknown>) : null;\n const scan = cfg?.scan as { include?: unknown } | undefined;\n if (!cfg || !scan || !isNonEmptyStringArray(scan.include)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: scan.include must be a non-empty array of glob strings`);\n }\n}\n\nfunction validateStoreKey(value: unknown, configPath: string): void {\n if (!STORE_NAMES.includes(value as StoreName)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: store must be ${STORE_NAMES.map((name) => `\"${name}\"`).join(' or ')}`);\n }\n}\n\nfunction validateFeaturesBlock(value: unknown, configPath: string): void {\n if (typeof value !== 'object' || value === null || Array.isArray(value)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: features must be an object of name -> boolean`);\n }\n const block = value as Record<string, unknown>;\n const unknown = Object.keys(block).filter((k) => !KNOWN_FEATURE_KEYS.has(k));\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: features has unknown key(s) ${unknown.join(', ')}; embed is not a features key -- see the top-level \"embed\" block`);\n }\n for (const [name, v] of Object.entries(block)) {\n if (typeof v !== 'boolean') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: features.${name} must be a boolean`);\n }\n }\n}\n\nfunction validateEmbedBlock(value: unknown, configPath: string): void {\n if (typeof value !== 'object' || value === null || Array.isArray(value)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed must be an object of { model?, provider?: \"static\"|\"openai\"|\"cohere\", url?, key? }`);\n }\n const embed = value as Record<string, unknown>;\n const unknown = Object.keys(embed).filter((k) => !KNOWN_EMBED_KEYS.has(k));\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed has unknown key(s) ${unknown.join(', ')}`);\n }\n if (embed.model !== undefined && typeof embed.model !== 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.model must be a string`);\n }\n // The block exists to name a model; provider settings without one would read as \"vectors\n // configured\" while embedding stays off, which is the ambiguity v4 removed.\n if (embed.model === undefined || embed.model === '') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.model is required when the \"embed\" block is present (it is what turns vectors on); remove the block to index without vectors`);\n }\n if (embed.provider !== undefined && embed.provider !== 'static' && embed.provider !== 'openai' && embed.provider !== 'cohere') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.provider must be \"static\", \"openai\", or \"cohere\"`);\n }\n if (embed.url !== undefined && typeof embed.url !== 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.url must be a string`);\n }\n if (embed.key !== undefined && typeof embed.key !== 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.key must be a string`);\n }\n // openai has no default host, so a missing url would otherwise only surface at the first\n // vector search; a config error belongs at config time.\n if (embed.provider === 'openai' && (embed.url === undefined || embed.url === '')) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.provider \"openai\" requires embed.url (e.g. \"http://localhost:11434/v1\" for Ollama)`);\n }\n if (embed.languages !== undefined && (!Array.isArray(embed.languages) || embed.languages.length === 0 || embed.languages.some((l) => typeof l !== 'string' || l === ''))) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.languages must be a non-empty array of language tags (e.g. [\"en\", \"zh\"])`);\n }\n if (embed.chunkTokens !== undefined && (typeof embed.chunkTokens !== 'number' || !Number.isInteger(embed.chunkTokens) || embed.chunkTokens <= 0)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: embed.chunkTokens must be a positive integer`);\n }\n}\n\nfunction isSignalName(value: unknown): value is SignalName {\n return typeof value === 'string' && (SIGNAL_NAMES as readonly string[]).includes(value);\n}\n\nfunction isSignalWeight(value: unknown): value is number {\n return typeof value === 'number' && Number.isFinite(value) && value > 0;\n}\n\n// Shape (an object of signal name -> weight) plus the one cross-signal prerequisite that doesn't\n// need the top-level embed block: links is seeded by word-match rows, so declaring it without words is always empty.\nfunction validatePresetSignals(name: string, value: unknown, configPath: string): void {\n if (typeof value !== 'object' || value === null || Array.isArray(value) || Object.keys(value).length === 0) {\n throw new SenseError(\n 'CONFIG_INVALID',\n `${configPath}: presets.${name}.signals must be a non-empty object of signal name -> weight (a finite number > 0), naming ${SIGNAL_NAMES.join(', ')} -- a preset that searches nothing is a mistake; name at least \"words\", or omit \"signals\" to use every signal whose prerequisites hold at weight 1`\n );\n }\n const entries = Object.entries(value as Record<string, unknown>);\n const unknown = entries.filter(([k]) => !isSignalName(k)).map(([k]) => k);\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.signals names unknown signal(s) ${unknown.join(', ')}; valid signals are ${SIGNAL_NAMES.join(', ')}`);\n }\n for (const [signal, weight] of entries) {\n if (!isSignalWeight(weight)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.signals.${signal} must be a finite number > 0`);\n }\n }\n const names = new Set(entries.map(([k]) => k));\n for (const [signal, requires] of Object.entries(SIGNAL_PREREQUISITES) as Array<[SignalName, SignalName]>) {\n if (names.has(signal) && !names.has(requires)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.signals has \"${signal}\" without \"${requires}\"; add \"${requires}\" to presets.${name}.signals, or drop \"${signal}\"`);\n }\n }\n}\n\nfunction validatePreset(name: string, value: unknown, configPath: string): void {\n if (typeof value !== 'object' || value === null || Array.isArray(value)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name} must be an object`);\n }\n const preset = value as Record<string, unknown>;\n const unknown = Object.keys(preset).filter((k) => !KNOWN_PRESET_KEYS.has(k));\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name} has unknown key(s) ${unknown.join(', ')}; a preset takes include, exclude, k, signals, where`);\n }\n if (!isNonEmptyStringArray(preset.include)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.include must be a non-empty array of glob strings`);\n }\n if (preset.exclude !== undefined && !isNonEmptyStringArray(preset.exclude)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.exclude must be a non-empty array of glob strings`);\n }\n if (preset.signals !== undefined) validatePresetSignals(name, preset.signals, configPath);\n if (preset.k !== undefined && (typeof preset.k !== 'number' || !Number.isInteger(preset.k) || preset.k <= 0)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.k must be a positive integer`);\n }\n if (preset.where !== undefined && typeof preset.where !== 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.where must be a SQL condition string`);\n }\n}\n\n// A queries.<name> entry: { sql } or a saved search { search, preset?, include?, exclude?, where?, k? }.\nfunction validateSavedQuery(name: string, value: unknown, configPath: string): void {\n if (RESERVED_QUERY_NAMES.has(name)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} uses a reserved command name and can never be run as a saved query`);\n }\n // A bare string is rejected rather than inferred to be SQL: an entry says which of the two\n // verbs it runs, the same choice the CLI makes explicit.\n if (typeof value === 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} must say which verb it runs: { \"sql\": ${JSON.stringify(value.length > 40 ? `${value.slice(0, 37)}...` : value)} } to run it as SQL, or { \"search\": \"...\" } for a ranked search`);\n }\n if (typeof value !== 'object' || value === null || Array.isArray(value)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} must be { sql } or { search, preset?, include?, exclude?, where?, k? }`);\n }\n const entry = value as Record<string, unknown>;\n if ('sql' in entry) {\n const unknown = Object.keys(entry).filter((k) => k !== 'sql');\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} has unknown key(s) ${unknown.join(', ')}; a { sql } query takes only sql`);\n }\n if (typeof entry.sql !== 'string' || entry.sql.trim() === '') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.sql must be a non-empty string`);\n }\n return;\n }\n if ('search' in entry) {\n const unknown = Object.keys(entry).filter((k) => !SAVED_SEARCH_KEYS.has(k));\n if (unknown.length > 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} has unknown key(s) ${unknown.join(', ')}; a saved search takes search, preset, include, exclude, where, k`);\n }\n // A saved entry saves a question; a scope without a question is just flags.\n if (typeof entry.search !== 'string' || entry.search.trim() === '') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.search must be non-empty text`);\n }\n if (entry.preset !== undefined && (typeof entry.preset !== 'string' || entry.preset.length === 0)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.preset must be a preset name`);\n }\n if (entry.include !== undefined && !isNonEmptyStringArray(entry.include)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.include must be a non-empty array of glob strings`);\n }\n if (entry.exclude !== undefined && !isNonEmptyStringArray(entry.exclude)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.exclude must be a non-empty array of glob strings`);\n }\n if (entry.where !== undefined && typeof entry.where !== 'string') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.where must be a SQL condition string`);\n }\n if (entry.k !== undefined && (typeof entry.k !== 'number' || !Number.isInteger(entry.k) || entry.k <= 0)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name}.k must be a positive integer`);\n }\n return;\n }\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries.${name} must be { sql } or { search, preset?, include?, exclude?, where?, k? }`);\n}\n\nexport function validateConfig(parsed: unknown, configPath: string): Config {\n if (typeof parsed !== 'object' || parsed === null || Array.isArray(parsed)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: config must be a JSON object`);\n }\n const cfg = parsed as Record<string, unknown>;\n\n if (cfg.root !== undefined && (typeof cfg.root !== 'string' || cfg.root.length === 0)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: root must be a non-empty path string`);\n }\n if (cfg.build !== undefined && typeof cfg.build !== 'boolean') {\n throw new SenseError('CONFIG_INVALID', `${configPath}: build must be a boolean`);\n }\n\n // `checks` is rejected by name, not warned: silence would hide that assertions are gone.\n if (cfg.checks !== undefined) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: checks was removed in v3 -- sense check no longer asserts on saved queries; a returned row set is the reader's judgment`);\n }\n\n // `content` (content.tokenize) is rejected by name for the same reason: every outcome it\n // selected (unspaced-script search, stemming, quoted-phrase substring matching) is now\n // automatic on every store, so a config that still sets it would silently get nothing back.\n if (cfg.content !== undefined) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: content.tokenize was removed -- every outcome it selected is now automatic on every store; remove the \"content\" block`);\n }\n\n const presets = cfg.presets as Record<string, unknown> | undefined;\n if (!presets || typeof presets !== 'object' || Array.isArray(presets) || Object.keys(presets).length === 0) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets must be a non-empty object of preset name -> { include, exclude?, k?, signals?, where? }`);\n }\n for (const [name, value] of Object.entries(presets)) {\n validatePreset(name, value, configPath);\n }\n if (presets.default === undefined) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets must include a \"default\" preset`);\n }\n\n if (cfg.queries === undefined) cfg.queries = {};\n if (typeof cfg.queries !== 'object' || cfg.queries === null || Array.isArray(cfg.queries)) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: queries must be an object of name -> { sql } or { search }`);\n }\n const queries = cfg.queries as Record<string, unknown>;\n for (const [name, value] of Object.entries(queries)) {\n validateSavedQuery(name, value, configPath);\n }\n\n if (cfg.features !== undefined) {\n validateFeaturesBlock(cfg.features, configPath);\n }\n if (cfg.embed !== undefined) {\n validateEmbedBlock(cfg.embed, configPath);\n }\n if (cfg.store !== undefined) {\n validateStoreKey(cfg.store, configPath);\n }\n\n // vectors' prerequisite is a named model, checked here once embed is known valid (or absent).\n if (cfg.embed === undefined) {\n for (const [name, value] of Object.entries(presets)) {\n const signals = (value as Record<string, unknown>).signals as Record<string, unknown> | undefined;\n if (signals && 'vectors' in signals) {\n throw new SenseError('CONFIG_INVALID', `${configPath}: presets.${name}.signals includes \"vectors\", but no \"embed\" block names a model; add one, or drop \"vectors\" from presets.${name}.signals`);\n }\n }\n }\n\n return cfg as unknown as Config;\n}\n"],"names":["SenseError","SIGNAL_NAMES","SIGNAL_PREREQUISITES","STORE_NAMES","KNOWN_KEYS","Set","KNOWN_PRESET_KEYS","KNOWN_FEATURE_KEYS","KNOWN_EMBED_KEYS","SAVED_SEARCH_KEYS","RESERVED_QUERY_NAMES","unknownConfigKeys","cfg","Object","keys","filter","k","has","isNonEmptyStringArray","value","Array","isArray","length","every","g","validateLegacyScan","parsed","configPath","scan","include","validateStoreKey","includes","map","name","join","validateFeaturesBlock","block","unknown","v","entries","validateEmbedBlock","embed","model","undefined","provider","url","key","languages","some","l","chunkTokens","Number","isInteger","isSignalName","isSignalWeight","isFinite","validatePresetSignals","signal","weight","names","requires","validatePreset","preset","exclude","signals","where","validateSavedQuery","JSON","stringify","slice","entry","sql","trim","search","validateConfig","root","build","checks","content","presets","default","queries","features","store"],"mappings":"AAAA,SAASA,UAAU,QAAQ,eAAe;AAC1C,SAASC,YAAY,EAAEC,oBAAoB,QAAyB,eAAe;AACnF,SAAsBC,WAAW,QAAwB,aAAa;AAEtE,sFAAsF;AACtF,8FAA8F;AAC9F,MAAMC,aAAa,IAAIC,IAAI;IAAC;IAAW;IAAW;IAAQ;IAAS;IAAW;IAAY;IAAS;IAAS;CAAU;AACtH,MAAMC,oBAAoB,IAAID,IAAI;IAAC;IAAW;IAAW;IAAK;IAAW;CAAQ;AACjF,MAAME,qBAAqB,IAAIF,IAAI;IAAC;IAAS;IAAY;IAAQ;CAAO;AACxE,iGAAiG;AACjG,sDAAsD;AACtD,OAAO,MAAMG,mBAAmB,IAAIH,IAAI;IAAC;IAAS;IAAY;IAAO;IAAO;IAAa;CAAc,EAAE;AACzG,MAAMI,oBAAoB,IAAIJ,IAAI;IAAC;IAAU;IAAU;IAAW;IAAW;IAAS;CAAI;AAC1F,MAAMK,uBAAuB,IAAIL,IAAI;IAAC;IAAQ;IAAS;IAAS;IAAU;IAAY;IAAO;IAAU;IAAO;IAAQ;IAAQ;CAAU;AAExI,OAAO,SAASM,kBAAkBC,GAA4B;IAC5D,OAAOC,OAAOC,IAAI,CAACF,KAAKG,MAAM,CAAC,CAACC,IAAM,CAACZ,WAAWa,GAAG,CAACD;AACxD;AAEA,SAASE,sBAAsBC,KAAc;IAC3C,OAAOC,MAAMC,OAAO,CAACF,UAAUA,MAAMG,MAAM,GAAG,KAAKH,MAAMI,KAAK,CAAC,CAACC,IAAM,OAAOA,MAAM,YAAYA,EAAEF,MAAM,GAAG;AAC5G;AAEA,yFAAyF;AACzF,4FAA4F;AAC5F,OAAO,SAASG,mBAAmBC,MAAe,EAAEC,UAAkB;IACpE,MAAMf,MAAM,OAAOc,WAAW,YAAYA,WAAW,QAAQ,CAACN,MAAMC,OAAO,CAACK,UAAWA,SAAqC;IAC5H,MAAME,OAAOhB,gBAAAA,0BAAAA,IAAKgB,IAAI;IACtB,IAAI,CAAChB,OAAO,CAACgB,QAAQ,CAACV,sBAAsBU,KAAKC,OAAO,GAAG;QACzD,MAAM,IAAI7B,WAAW,kBAAkB,GAAG2B,WAAW,wDAAwD,CAAC;IAChH;AACF;AAEA,SAASG,iBAAiBX,KAAc,EAAEQ,UAAkB;IAC1D,IAAI,CAACxB,YAAY4B,QAAQ,CAACZ,QAAqB;QAC7C,MAAM,IAAInB,WAAW,kBAAkB,GAAG2B,WAAW,gBAAgB,EAAExB,YAAY6B,GAAG,CAAC,CAACC,OAAS,CAAC,CAAC,EAAEA,KAAK,CAAC,CAAC,EAAEC,IAAI,CAAC,SAAS;IAC9H;AACF;AAEA,SAASC,sBAAsBhB,KAAc,EAAEQ,UAAkB;IAC/D,IAAI,OAAOR,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,QAAQ;QACvE,MAAM,IAAInB,WAAW,kBAAkB,GAAG2B,WAAW,+CAA+C,CAAC;IACvG;IACA,MAAMS,QAAQjB;IACd,MAAMkB,UAAUxB,OAAOC,IAAI,CAACsB,OAAOrB,MAAM,CAAC,CAACC,IAAM,CAACT,mBAAmBU,GAAG,CAACD;IACzE,IAAIqB,QAAQf,MAAM,GAAG,GAAG;QACtB,MAAM,IAAItB,WAAW,kBAAkB,GAAG2B,WAAW,8BAA8B,EAAEU,QAAQH,IAAI,CAAC,MAAM,gEAAgE,CAAC;IAC3K;IACA,KAAK,MAAM,CAACD,MAAMK,EAAE,IAAIzB,OAAO0B,OAAO,CAACH,OAAQ;QAC7C,IAAI,OAAOE,MAAM,WAAW;YAC1B,MAAM,IAAItC,WAAW,kBAAkB,GAAG2B,WAAW,WAAW,EAAEM,KAAK,kBAAkB,CAAC;QAC5F;IACF;AACF;AAEA,SAASO,mBAAmBrB,KAAc,EAAEQ,UAAkB;IAC5D,IAAI,OAAOR,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,QAAQ;QACvE,MAAM,IAAInB,WAAW,kBAAkB,GAAG2B,WAAW,0FAA0F,CAAC;IAClJ;IACA,MAAMc,QAAQtB;IACd,MAAMkB,UAAUxB,OAAOC,IAAI,CAAC2B,OAAO1B,MAAM,CAAC,CAACC,IAAM,CAACR,iBAAiBS,GAAG,CAACD;IACvE,IAAIqB,QAAQf,MAAM,GAAG,GAAG;QACtB,MAAM,IAAItB,WAAW,kBAAkB,GAAG2B,WAAW,2BAA2B,EAAEU,QAAQH,IAAI,CAAC,OAAO;IACxG;IACA,IAAIO,MAAMC,KAAK,KAAKC,aAAa,OAAOF,MAAMC,KAAK,KAAK,UAAU;QAChE,MAAM,IAAI1C,WAAW,kBAAkB,GAAG2B,WAAW,8BAA8B,CAAC;IACtF;IACA,yFAAyF;IACzF,4EAA4E;IAC5E,IAAIc,MAAMC,KAAK,KAAKC,aAAaF,MAAMC,KAAK,KAAK,IAAI;QACnD,MAAM,IAAI1C,WAAW,kBAAkB,GAAG2B,WAAW,oIAAoI,CAAC;IAC5L;IACA,IAAIc,MAAMG,QAAQ,KAAKD,aAAaF,MAAMG,QAAQ,KAAK,YAAYH,MAAMG,QAAQ,KAAK,YAAYH,MAAMG,QAAQ,KAAK,UAAU;QAC7H,MAAM,IAAI5C,WAAW,kBAAkB,GAAG2B,WAAW,wDAAwD,CAAC;IAChH;IACA,IAAIc,MAAMI,GAAG,KAAKF,aAAa,OAAOF,MAAMI,GAAG,KAAK,UAAU;QAC5D,MAAM,IAAI7C,WAAW,kBAAkB,GAAG2B,WAAW,4BAA4B,CAAC;IACpF;IACA,IAAIc,MAAMK,GAAG,KAAKH,aAAa,OAAOF,MAAMK,GAAG,KAAK,UAAU;QAC5D,MAAM,IAAI9C,WAAW,kBAAkB,GAAG2B,WAAW,4BAA4B,CAAC;IACpF;IACA,yFAAyF;IACzF,wDAAwD;IACxD,IAAIc,MAAMG,QAAQ,KAAK,YAAaH,CAAAA,MAAMI,GAAG,KAAKF,aAAaF,MAAMI,GAAG,KAAK,EAAC,GAAI;QAChF,MAAM,IAAI7C,WAAW,kBAAkB,GAAG2B,WAAW,0FAA0F,CAAC;IAClJ;IACA,IAAIc,MAAMM,SAAS,KAAKJ,aAAc,CAAA,CAACvB,MAAMC,OAAO,CAACoB,MAAMM,SAAS,KAAKN,MAAMM,SAAS,CAACzB,MAAM,KAAK,KAAKmB,MAAMM,SAAS,CAACC,IAAI,CAAC,CAACC,IAAM,OAAOA,MAAM,YAAYA,MAAM,GAAE,GAAI;QACxK,MAAM,IAAIjD,WAAW,kBAAkB,GAAG2B,WAAW,gFAAgF,CAAC;IACxI;IACA,IAAIc,MAAMS,WAAW,KAAKP,aAAc,CAAA,OAAOF,MAAMS,WAAW,KAAK,YAAY,CAACC,OAAOC,SAAS,CAACX,MAAMS,WAAW,KAAKT,MAAMS,WAAW,IAAI,CAAA,GAAI;QAChJ,MAAM,IAAIlD,WAAW,kBAAkB,GAAG2B,WAAW,8CAA8C,CAAC;IACtG;AACF;AAEA,SAAS0B,aAAalC,KAAc;IAClC,OAAO,OAAOA,UAAU,YAAY,AAAClB,aAAmC8B,QAAQ,CAACZ;AACnF;AAEA,SAASmC,eAAenC,KAAc;IACpC,OAAO,OAAOA,UAAU,YAAYgC,OAAOI,QAAQ,CAACpC,UAAUA,QAAQ;AACxE;AAEA,iGAAiG;AACjG,qHAAqH;AACrH,SAASqC,sBAAsBvB,IAAY,EAAEd,KAAc,EAAEQ,UAAkB;IAC7E,IAAI,OAAOR,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,UAAUN,OAAOC,IAAI,CAACK,OAAOG,MAAM,KAAK,GAAG;QAC1G,MAAM,IAAItB,WACR,kBACA,GAAG2B,WAAW,UAAU,EAAEM,KAAK,2FAA2F,EAAEhC,aAAaiC,IAAI,CAAC,MAAM,kJAAkJ,CAAC;IAE3S;IACA,MAAMK,UAAU1B,OAAO0B,OAAO,CAACpB;IAC/B,MAAMkB,UAAUE,QAAQxB,MAAM,CAAC,CAAC,CAACC,EAAE,GAAK,CAACqC,aAAarC,IAAIgB,GAAG,CAAC,CAAC,CAAChB,EAAE,GAAKA;IACvE,IAAIqB,QAAQf,MAAM,GAAG,GAAG;QACtB,MAAM,IAAItB,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,iCAAiC,EAAEI,QAAQH,IAAI,CAAC,MAAM,oBAAoB,EAAEjC,aAAaiC,IAAI,CAAC,OAAO;IAC7K;IACA,KAAK,MAAM,CAACuB,QAAQC,OAAO,IAAInB,QAAS;QACtC,IAAI,CAACe,eAAeI,SAAS;YAC3B,MAAM,IAAI1D,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,SAAS,EAAEwB,OAAO,4BAA4B,CAAC;QACvH;IACF;IACA,MAAME,QAAQ,IAAItD,IAAIkC,QAAQP,GAAG,CAAC,CAAC,CAAChB,EAAE,GAAKA;IAC3C,KAAK,MAAM,CAACyC,QAAQG,SAAS,IAAI/C,OAAO0B,OAAO,CAACrC,sBAA0D;QACxG,IAAIyD,MAAM1C,GAAG,CAACwC,WAAW,CAACE,MAAM1C,GAAG,CAAC2C,WAAW;YAC7C,MAAM,IAAI5D,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,cAAc,EAAEwB,OAAO,WAAW,EAAEG,SAAS,QAAQ,EAAEA,SAAS,aAAa,EAAE3B,KAAK,mBAAmB,EAAEwB,OAAO,CAAC,CAAC;QAC1L;IACF;AACF;AAEA,SAASI,eAAe5B,IAAY,EAAEd,KAAc,EAAEQ,UAAkB;IACtE,IAAI,OAAOR,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,QAAQ;QACvE,MAAM,IAAInB,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,kBAAkB,CAAC;IAC3F;IACA,MAAM6B,SAAS3C;IACf,MAAMkB,UAAUxB,OAAOC,IAAI,CAACgD,QAAQ/C,MAAM,CAAC,CAACC,IAAM,CAACV,kBAAkBW,GAAG,CAACD;IACzE,IAAIqB,QAAQf,MAAM,GAAG,GAAG;QACtB,MAAM,IAAItB,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,oBAAoB,EAAEI,QAAQH,IAAI,CAAC,MAAM,oDAAoD,CAAC;IACtK;IACA,IAAI,CAAChB,sBAAsB4C,OAAOjC,OAAO,GAAG;QAC1C,MAAM,IAAI7B,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,kDAAkD,CAAC;IAC3H;IACA,IAAI6B,OAAOC,OAAO,KAAKpB,aAAa,CAACzB,sBAAsB4C,OAAOC,OAAO,GAAG;QAC1E,MAAM,IAAI/D,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,kDAAkD,CAAC;IAC3H;IACA,IAAI6B,OAAOE,OAAO,KAAKrB,WAAWa,sBAAsBvB,MAAM6B,OAAOE,OAAO,EAAErC;IAC9E,IAAImC,OAAO9C,CAAC,KAAK2B,aAAc,CAAA,OAAOmB,OAAO9C,CAAC,KAAK,YAAY,CAACmC,OAAOC,SAAS,CAACU,OAAO9C,CAAC,KAAK8C,OAAO9C,CAAC,IAAI,CAAA,GAAI;QAC5G,MAAM,IAAIhB,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,6BAA6B,CAAC;IACtG;IACA,IAAI6B,OAAOG,KAAK,KAAKtB,aAAa,OAAOmB,OAAOG,KAAK,KAAK,UAAU;QAClE,MAAM,IAAIjE,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,qCAAqC,CAAC;IAC9G;AACF;AAEA,yGAAyG;AACzG,SAASiC,mBAAmBjC,IAAY,EAAEd,KAAc,EAAEQ,UAAkB;IAC1E,IAAIjB,qBAAqBO,GAAG,CAACgB,OAAO;QAClC,MAAM,IAAIjC,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,mEAAmE,CAAC;IAC5I;IACA,2FAA2F;IAC3F,yDAAyD;IACzD,IAAI,OAAOd,UAAU,UAAU;QAC7B,MAAM,IAAInB,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,uCAAuC,EAAEkC,KAAKC,SAAS,CAACjD,MAAMG,MAAM,GAAG,KAAK,GAAGH,MAAMkD,KAAK,CAAC,GAAG,IAAI,GAAG,CAAC,GAAGlD,OAAO,+DAA+D,CAAC;IACxP;IACA,IAAI,OAAOA,UAAU,YAAYA,UAAU,QAAQC,MAAMC,OAAO,CAACF,QAAQ;QACvE,MAAM,IAAInB,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,uEAAuE,CAAC;IAChJ;IACA,MAAMqC,QAAQnD;IACd,IAAI,SAASmD,OAAO;QAClB,MAAMjC,UAAUxB,OAAOC,IAAI,CAACwD,OAAOvD,MAAM,CAAC,CAACC,IAAMA,MAAM;QACvD,IAAIqB,QAAQf,MAAM,GAAG,GAAG;YACtB,MAAM,IAAItB,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,oBAAoB,EAAEI,QAAQH,IAAI,CAAC,MAAM,gCAAgC,CAAC;QAClJ;QACA,IAAI,OAAOoC,MAAMC,GAAG,KAAK,YAAYD,MAAMC,GAAG,CAACC,IAAI,OAAO,IAAI;YAC5D,MAAM,IAAIxE,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,+BAA+B,CAAC;QACxG;QACA;IACF;IACA,IAAI,YAAYqC,OAAO;QACrB,MAAMjC,UAAUxB,OAAOC,IAAI,CAACwD,OAAOvD,MAAM,CAAC,CAACC,IAAM,CAACP,kBAAkBQ,GAAG,CAACD;QACxE,IAAIqB,QAAQf,MAAM,GAAG,GAAG;YACtB,MAAM,IAAItB,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,oBAAoB,EAAEI,QAAQH,IAAI,CAAC,MAAM,iEAAiE,CAAC;QACnL;QACA,4EAA4E;QAC5E,IAAI,OAAOoC,MAAMG,MAAM,KAAK,YAAYH,MAAMG,MAAM,CAACD,IAAI,OAAO,IAAI;YAClE,MAAM,IAAIxE,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,8BAA8B,CAAC;QACvG;QACA,IAAIqC,MAAMR,MAAM,KAAKnB,aAAc,CAAA,OAAO2B,MAAMR,MAAM,KAAK,YAAYQ,MAAMR,MAAM,CAACxC,MAAM,KAAK,CAAA,GAAI;YACjG,MAAM,IAAItB,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,6BAA6B,CAAC;QACtG;QACA,IAAIqC,MAAMzC,OAAO,KAAKc,aAAa,CAACzB,sBAAsBoD,MAAMzC,OAAO,GAAG;YACxE,MAAM,IAAI7B,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,kDAAkD,CAAC;QAC3H;QACA,IAAIqC,MAAMP,OAAO,KAAKpB,aAAa,CAACzB,sBAAsBoD,MAAMP,OAAO,GAAG;YACxE,MAAM,IAAI/D,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,kDAAkD,CAAC;QAC3H;QACA,IAAIqC,MAAML,KAAK,KAAKtB,aAAa,OAAO2B,MAAML,KAAK,KAAK,UAAU;YAChE,MAAM,IAAIjE,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,qCAAqC,CAAC;QAC9G;QACA,IAAIqC,MAAMtD,CAAC,KAAK2B,aAAc,CAAA,OAAO2B,MAAMtD,CAAC,KAAK,YAAY,CAACmC,OAAOC,SAAS,CAACkB,MAAMtD,CAAC,KAAKsD,MAAMtD,CAAC,IAAI,CAAA,GAAI;YACxG,MAAM,IAAIhB,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,6BAA6B,CAAC;QACtG;QACA;IACF;IACA,MAAM,IAAIjC,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,uEAAuE,CAAC;AAChJ;AAEA,OAAO,SAASyC,eAAehD,MAAe,EAAEC,UAAkB;IAChE,IAAI,OAAOD,WAAW,YAAYA,WAAW,QAAQN,MAAMC,OAAO,CAACK,SAAS;QAC1E,MAAM,IAAI1B,WAAW,kBAAkB,GAAG2B,WAAW,8BAA8B,CAAC;IACtF;IACA,MAAMf,MAAMc;IAEZ,IAAId,IAAI+D,IAAI,KAAKhC,aAAc,CAAA,OAAO/B,IAAI+D,IAAI,KAAK,YAAY/D,IAAI+D,IAAI,CAACrD,MAAM,KAAK,CAAA,GAAI;QACrF,MAAM,IAAItB,WAAW,kBAAkB,GAAG2B,WAAW,sCAAsC,CAAC;IAC9F;IACA,IAAIf,IAAIgE,KAAK,KAAKjC,aAAa,OAAO/B,IAAIgE,KAAK,KAAK,WAAW;QAC7D,MAAM,IAAI5E,WAAW,kBAAkB,GAAG2B,WAAW,yBAAyB,CAAC;IACjF;IAEA,yFAAyF;IACzF,IAAIf,IAAIiE,MAAM,KAAKlC,WAAW;QAC5B,MAAM,IAAI3C,WAAW,kBAAkB,GAAG2B,WAAW,yHAAyH,CAAC;IACjL;IAEA,yFAAyF;IACzF,uFAAuF;IACvF,4FAA4F;IAC5F,IAAIf,IAAIkE,OAAO,KAAKnC,WAAW;QAC7B,MAAM,IAAI3C,WAAW,kBAAkB,GAAG2B,WAAW,uHAAuH,CAAC;IAC/K;IAEA,MAAMoD,UAAUnE,IAAImE,OAAO;IAC3B,IAAI,CAACA,WAAW,OAAOA,YAAY,YAAY3D,MAAMC,OAAO,CAAC0D,YAAYlE,OAAOC,IAAI,CAACiE,SAASzD,MAAM,KAAK,GAAG;QAC1G,MAAM,IAAItB,WAAW,kBAAkB,GAAG2B,WAAW,kGAAkG,CAAC;IAC1J;IACA,KAAK,MAAM,CAACM,MAAMd,MAAM,IAAIN,OAAO0B,OAAO,CAACwC,SAAU;QACnDlB,eAAe5B,MAAMd,OAAOQ;IAC9B;IACA,IAAIoD,QAAQC,OAAO,KAAKrC,WAAW;QACjC,MAAM,IAAI3C,WAAW,kBAAkB,GAAG2B,WAAW,yCAAyC,CAAC;IACjG;IAEA,IAAIf,IAAIqE,OAAO,KAAKtC,WAAW/B,IAAIqE,OAAO,GAAG,CAAC;IAC9C,IAAI,OAAOrE,IAAIqE,OAAO,KAAK,YAAYrE,IAAIqE,OAAO,KAAK,QAAQ7D,MAAMC,OAAO,CAACT,IAAIqE,OAAO,GAAG;QACzF,MAAM,IAAIjF,WAAW,kBAAkB,GAAG2B,WAAW,4DAA4D,CAAC;IACpH;IACA,MAAMsD,UAAUrE,IAAIqE,OAAO;IAC3B,KAAK,MAAM,CAAChD,MAAMd,MAAM,IAAIN,OAAO0B,OAAO,CAAC0C,SAAU;QACnDf,mBAAmBjC,MAAMd,OAAOQ;IAClC;IAEA,IAAIf,IAAIsE,QAAQ,KAAKvC,WAAW;QAC9BR,sBAAsBvB,IAAIsE,QAAQ,EAAEvD;IACtC;IACA,IAAIf,IAAI6B,KAAK,KAAKE,WAAW;QAC3BH,mBAAmB5B,IAAI6B,KAAK,EAAEd;IAChC;IACA,IAAIf,IAAIuE,KAAK,KAAKxC,WAAW;QAC3Bb,iBAAiBlB,IAAIuE,KAAK,EAAExD;IAC9B;IAEA,8FAA8F;IAC9F,IAAIf,IAAI6B,KAAK,KAAKE,WAAW;QAC3B,KAAK,MAAM,CAACV,MAAMd,MAAM,IAAIN,OAAO0B,OAAO,CAACwC,SAAU;YACnD,MAAMf,UAAU,AAAC7C,MAAkC6C,OAAO;YAC1D,IAAIA,WAAW,aAAaA,SAAS;gBACnC,MAAM,IAAIhE,WAAW,kBAAkB,GAAG2B,WAAW,UAAU,EAAEM,KAAK,yGAAyG,EAAEA,KAAK,QAAQ,CAAC;YACjM;QACF;IACF;IAEA,OAAOrB;AACT"}
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/embed/cohere.ts"],"sourcesContent":["import { SenseError } from '../errors.ts';\nimport { effectiveEmbedEndpoint } from './endpoint.ts';\nimport { fetchWithRetry, HTTP_ATTEMPT_TIMEOUT_MS } from './http.ts';\nimport type { EmbedCallOptions, EmbedProvider } from './types.ts';\n\nconst BATCH_CAP = 96;\n\nfunction isNumberArray(value: unknown): value is number[] {\n return Array.isArray(value) && value.every((part) => typeof part === 'number');\n}\n\nfunction isEmbeddingArray(value: unknown): value is number[][] {\n return Array.isArray(value) && value.every(isNumberArray);\n}\n\nfunction parseBody(body: ArrayBuffer, endpoint: string): number[][] {\n let decoded: unknown;\n try {\n decoded = JSON.parse(new TextDecoder().decode(body));\n } catch {\n throw new SenseError('EMBED_MODEL', `${endpoint} returned invalid JSON`);\n }\n if (decoded === null || typeof decoded !== 'object' || !('embeddings' in decoded) || decoded.embeddings === null || typeof decoded.embeddings !== 'object' || !('float' in decoded.embeddings) || !isEmbeddingArray(decoded.embeddings.float)) {\n throw new SenseError('EMBED_MODEL', `${endpoint} returned an invalid embedding response`);\n }\n return decoded.embeddings.float;\n}\n\n// Cohere's native /v2/embed: input_type distinguishes doc vs query embeddings, which the\n// OpenAI-compatible shape cannot express.\nexport async function cohereProvider(model: string, url: string | undefined, keyEnv: string | undefined): Promise<EmbedProvider> {\n const base = effectiveEmbedEndpoint('cohere', url);\n if (!base) throw new SenseError('EMBED_MODEL', 'cohere endpoint resolution failed');\n const headers: Record<string, string> = { 'content-type': 'application/json' };\n const key = keyEnv ? process.env[keyEnv] : undefined;\n if (key) headers.authorization = `Bearer ${key}`;\n\n async function post(texts: string[], inputType: 'search_document' | 'search_query', options?: EmbedCallOptions): Promise<Float32Array[]> {\n const res = await fetchWithRetry(`${base}/v2/embed`, { method: 'POST', headers, body: JSON.stringify({ model, input_type: inputType, texts, embedding_types: ['float'] }) }, { signal: options?.signal, attemptTimeoutMs: HTTP_ATTEMPT_TIMEOUT_MS });\n options?.signal?.throwIfAborted();\n if (!res.ok) throw new SenseError('EMBED_MODEL', `${res.url} -> HTTP ${res.status}`);\n const body = parseBody(res.body, res.url);\n // Cohere returns one vector per text in order, with no index field to re-sort by; a caller\n // maps them positionally, so the count below must be trustworthy.\n const vectors = body;\n if (vectors.length !== texts.length) throw new SenseError('EMBED_MODEL', `${res.url} returned ${vectors.length} embeddings for ${texts.length} inputs`);\n return vectors.map((v) => Float32Array.from(v));\n }\n\n const dims = (await post(['dimension probe'], 'search_query'))[0].length;\n return { id: `cohere:${base}:${model}`, dims, batchCap: BATCH_CAP, embedDocuments: (texts, options) => post(texts, 'search_document', options), embedQuery: async (text, options) => (await post([text], 'search_query', options))[0] };\n}\n"],"names":["SenseError","effectiveEmbedEndpoint","fetchWithRetry","HTTP_ATTEMPT_TIMEOUT_MS","BATCH_CAP","isNumberArray","value","Array","isArray","every","part","isEmbeddingArray","parseBody","body","endpoint","decoded","JSON","parse","TextDecoder","decode","embeddings","float","cohereProvider","model","url","keyEnv","base","headers","key","process","env","undefined","authorization","post","texts","inputType","options","res","method","stringify","input_type","embedding_types","signal","attemptTimeoutMs","throwIfAborted","ok","status","vectors","length","map","v","Float32Array","from","dims","id","batchCap","embedDocuments","embedQuery","text"],"mappings":"AAAA,SAASA,UAAU,QAAQ,eAAe;AAC1C,SAASC,sBAAsB,QAAQ,gBAAgB;AACvD,SAASC,cAAc,EAAEC,uBAAuB,QAAQ,YAAY;AAGpE,MAAMC,YAAY;AAElB,SAASC,cAAcC,KAAc;IACnC,OAAOC,MAAMC,OAAO,CAACF,UAAUA,MAAMG,KAAK,CAAC,CAACC,OAAS,OAAOA,SAAS;AACvE;AAEA,SAASC,iBAAiBL,KAAc;IACtC,OAAOC,MAAMC,OAAO,CAACF,UAAUA,MAAMG,KAAK,CAACJ;AAC7C;AAEA,SAASO,UAAUC,IAAiB,EAAEC,QAAgB;IACpD,IAAIC;IACJ,IAAI;QACFA,UAAUC,KAAKC,KAAK,CAAC,IAAIC,cAAcC,MAAM,CAACN;IAChD,EAAE,OAAM;QACN,MAAM,IAAIb,WAAW,eAAe,GAAGc,SAAS,sBAAsB,CAAC;IACzE;IACA,IAAIC,YAAY,QAAQ,OAAOA,YAAY,YAAY,CAAE,CAAA,gBAAgBA,OAAM,KAAMA,QAAQK,UAAU,KAAK,QAAQ,OAAOL,QAAQK,UAAU,KAAK,YAAY,CAAE,CAAA,WAAWL,QAAQK,UAAU,AAAD,KAAM,CAACT,iBAAiBI,QAAQK,UAAU,CAACC,KAAK,GAAG;QAC7O,MAAM,IAAIrB,WAAW,eAAe,GAAGc,SAAS,uCAAuC,CAAC;IAC1F;IACA,OAAOC,QAAQK,UAAU,CAACC,KAAK;AACjC;AAEA,yFAAyF;AACzF,0CAA0C;AAC1C,OAAO,eAAeC,eAAeC,KAAa,EAAEC,GAAuB,EAAEC,MAA0B;IACrG,MAAMC,OAAOzB,uBAAuB,UAAUuB;IAC9C,IAAI,CAACE,MAAM,MAAM,IAAI1B,WAAW,eAAe;IAC/C,MAAM2B,UAAkC;QAAE,gBAAgB;IAAmB;IAC7E,MAAMC,MAAMH,SAASI,QAAQC,GAAG,CAACL,OAAO,GAAGM;IAC3C,IAAIH,KAAKD,QAAQK,aAAa,GAAG,CAAC,OAAO,EAAEJ,KAAK;IAEhD,eAAeK,KAAKC,KAAe,EAAEC,SAA6C,EAAEC,OAA0B;YAE5GA;QADA,MAAMC,MAAM,MAAMnC,eAAe,GAAGwB,KAAK,SAAS,CAAC,EAAE;YAAEY,QAAQ;YAAQX;YAASd,MAAMG,KAAKuB,SAAS,CAAC;gBAAEhB;gBAAOiB,YAAYL;gBAAWD;gBAAOO,iBAAiB;oBAAC;iBAAQ;YAAC;QAAG,GAAG;YAAEC,MAAM,EAAEN,oBAAAA,8BAAAA,QAASM,MAAM;YAAEC,kBAAkBxC;QAAwB;QAClPiC,oBAAAA,+BAAAA,kBAAAA,QAASM,MAAM,cAAfN,sCAAAA,gBAAiBQ,cAAc;QAC/B,IAAI,CAACP,IAAIQ,EAAE,EAAE,MAAM,IAAI7C,WAAW,eAAe,GAAGqC,IAAIb,GAAG,CAAC,SAAS,EAAEa,IAAIS,MAAM,EAAE;QACnF,MAAMjC,OAAOD,UAAUyB,IAAIxB,IAAI,EAAEwB,IAAIb,GAAG;QACxC,2FAA2F;QAC3F,kEAAkE;QAClE,MAAMuB,UAAUlC;QAChB,IAAIkC,QAAQC,MAAM,KAAKd,MAAMc,MAAM,EAAE,MAAM,IAAIhD,WAAW,eAAe,GAAGqC,IAAIb,GAAG,CAAC,UAAU,EAAEuB,QAAQC,MAAM,CAAC,gBAAgB,EAAEd,MAAMc,MAAM,CAAC,OAAO,CAAC;QACtJ,OAAOD,QAAQE,GAAG,CAAC,CAACC,IAAMC,aAAaC,IAAI,CAACF;IAC9C;IAEA,MAAMG,OAAO,AAAC,CAAA,MAAMpB,KAAK;QAAC;KAAkB,EAAE,eAAc,CAAE,CAAC,EAAE,CAACe,MAAM;IACxE,OAAO;QAAEM,IAAI,CAAC,OAAO,EAAE5B,KAAK,CAAC,EAAEH,OAAO;QAAE8B;QAAME,UAAUnD;QAAWoD,gBAAgB,CAACtB,OAAOE,UAAYH,KAAKC,OAAO,mBAAmBE;QAAUqB,YAAY,OAAOC,MAAMtB,UAAY,AAAC,CAAA,MAAMH,KAAK;gBAACyB;aAAK,EAAE,gBAAgBtB,QAAO,CAAE,CAAC,EAAE;IAAC;AACxO"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/open-source/ai/sensemaking/src/embed/cohere.ts"],"sourcesContent":["import { SenseError } from '../errors.ts';\nimport { effectiveEmbedEndpoint } from './endpoint.ts';\nimport { fetchWithRetry, HTTP_ATTEMPT_TIMEOUT_MS } from './http.ts';\nimport type { EmbedCallOptions, EmbedProvider } from './types.ts';\n\nconst BATCH_CAP = 96;\n\nfunction isNumberArray(value: unknown): value is number[] {\n return Array.isArray(value) && value.every((part) => typeof part === 'number');\n}\n\nfunction isEmbeddingArray(value: unknown): value is number[][] {\n return Array.isArray(value) && value.every(isNumberArray);\n}\n\nfunction parseBody(body: ArrayBuffer, endpoint: string): number[][] {\n let decoded: unknown;\n try {\n decoded = JSON.parse(new TextDecoder().decode(body));\n } catch {\n throw new SenseError('EMBED_MODEL', `${endpoint} returned invalid JSON`);\n }\n if (decoded === null || typeof decoded !== 'object' || !('embeddings' in decoded) || decoded.embeddings === null || typeof decoded.embeddings !== 'object' || !('float' in decoded.embeddings) || !isEmbeddingArray(decoded.embeddings.float)) {\n throw new SenseError('EMBED_MODEL', `${endpoint} returned an invalid embedding response`);\n }\n return decoded.embeddings.float;\n}\n\n// Cohere's native /v2/embed: input_type distinguishes doc vs query embeddings, which the\n// OpenAI-compatible shape cannot express.\nexport async function cohereProvider(model: string, url: string | undefined, keyEnv: string | undefined): Promise<EmbedProvider> {\n const base = effectiveEmbedEndpoint('cohere', url);\n if (!base) throw new SenseError('EMBED_MODEL', 'cohere endpoint resolution failed');\n const headers: Record<string, string> = { 'content-type': 'application/json' };\n const key = keyEnv ? process.env[keyEnv] : undefined;\n if (key) headers.authorization = `Bearer ${key}`;\n\n async function post(texts: string[], inputType: 'search_document' | 'search_query', options?: EmbedCallOptions): Promise<Float32Array[]> {\n const res = await fetchWithRetry(`${base}/v2/embed`, { method: 'POST', headers, body: JSON.stringify({ model, input_type: inputType, texts, embedding_types: ['float'] }) }, { signal: options?.signal, attemptTimeoutMs: HTTP_ATTEMPT_TIMEOUT_MS });\n options?.signal?.throwIfAborted();\n if (!res.ok) throw new SenseError('EMBED_MODEL', `${res.url} -> HTTP ${res.status}`);\n const body = parseBody(res.body, res.url);\n // Cohere returns one vector per text in order, with no index field to re-sort by; a caller\n // maps them positionally, so the count below must be trustworthy.\n const vectors = body;\n if (vectors.length !== texts.length) throw new SenseError('EMBED_MODEL', `${res.url} returned ${vectors.length} embeddings for ${texts.length} inputs`);\n return vectors.map((v) => Float32Array.from(v));\n }\n\n const dims = (await post(['dimension probe'], 'search_query'))[0].length;\n return { id: `cohere:${base}:${model}`, dims, batchCap: BATCH_CAP, embedDocuments: (texts, options) => post(texts, 'search_document', options), embedQuery: async (text, options) => (await post([text], 'search_query', options))[0] };\n}\n"],"names":["SenseError","effectiveEmbedEndpoint","fetchWithRetry","HTTP_ATTEMPT_TIMEOUT_MS","BATCH_CAP","isNumberArray","value","Array","isArray","every","part","isEmbeddingArray","parseBody","body","endpoint","decoded","JSON","parse","TextDecoder","decode","embeddings","float","cohereProvider","model","url","keyEnv","base","headers","key","process","env","undefined","authorization","post","texts","inputType","options","res","method","stringify","input_type","embedding_types","signal","attemptTimeoutMs","throwIfAborted","ok","status","vectors","length","map","v","Float32Array","from","dims","id","batchCap","embedDocuments","embedQuery","text"],"mappings":"AAAA,SAASA,UAAU,QAAQ,eAAe;AAC1C,SAASC,sBAAsB,QAAQ,gBAAgB;AACvD,SAASC,cAAc,EAAEC,uBAAuB,QAAQ,YAAY;AAGpE,MAAMC,YAAY;AAElB,SAASC,cAAcC,KAAc;IACnC,OAAOC,MAAMC,OAAO,CAACF,UAAUA,MAAMG,KAAK,CAAC,CAACC,OAAS,OAAOA,SAAS;AACvE;AAEA,SAASC,iBAAiBL,KAAc;IACtC,OAAOC,MAAMC,OAAO,CAACF,UAAUA,MAAMG,KAAK,CAACJ;AAC7C;AAEA,SAASO,UAAUC,IAAiB,EAAEC,QAAgB;IACpD,IAAIC;IACJ,IAAI;QACFA,UAAUC,KAAKC,KAAK,CAAC,IAAIC,cAAcC,MAAM,CAACN;IAChD,EAAE,OAAM;QACN,MAAM,IAAIb,WAAW,eAAe,GAAGc,SAAS,sBAAsB,CAAC;IACzE;IACA,IAAIC,YAAY,QAAQ,OAAOA,YAAY,YAAY,CAAE,CAAA,gBAAgBA,OAAM,KAAMA,QAAQK,UAAU,KAAK,QAAQ,OAAOL,QAAQK,UAAU,KAAK,YAAY,CAAE,CAAA,WAAWL,QAAQK,UAAU,AAAD,KAAM,CAACT,iBAAiBI,QAAQK,UAAU,CAACC,KAAK,GAAG;QAC7O,MAAM,IAAIrB,WAAW,eAAe,GAAGc,SAAS,uCAAuC,CAAC;IAC1F;IACA,OAAOC,QAAQK,UAAU,CAACC,KAAK;AACjC;AAEA,yFAAyF;AACzF,0CAA0C;AAC1C,OAAO,eAAeC,eAAeC,KAAa,EAAEC,GAAuB,EAAEC,MAA0B;IACrG,MAAMC,OAAOzB,uBAAuB,UAAUuB;IAC9C,IAAI,CAACE,MAAM,MAAM,IAAI1B,WAAW,eAAe;IAC/C,MAAM2B,UAAkC;QAAE,gBAAgB;IAAmB;IAC7E,MAAMC,MAAMH,SAASI,QAAQC,GAAG,CAACL,OAAO,GAAGM;IAC3C,IAAIH,KAAKD,QAAQK,aAAa,GAAG,CAAC,OAAO,EAAEJ,KAAK;IAEhD,eAAeK,KAAKC,KAAe,EAAEC,SAA6C,EAAEC,OAA0B;YAE5GA;QADA,MAAMC,MAAM,MAAMnC,eAAe,GAAGwB,KAAK,SAAS,CAAC,EAAE;YAAEY,QAAQ;YAAQX;YAASd,MAAMG,KAAKuB,SAAS,CAAC;gBAAEhB;gBAAOiB,YAAYL;gBAAWD;gBAAOO,iBAAiB;oBAAC;iBAAQ;YAAC;QAAG,GAAG;YAAEC,MAAM,EAAEN,oBAAAA,8BAAAA,QAASM,MAAM;YAAEC,kBAAkBxC;QAAwB;QAClPiC,oBAAAA,+BAAAA,kBAAAA,QAASM,MAAM,cAAfN,sCAAAA,gBAAiBQ,cAAc;QAC/B,IAAI,CAACP,IAAIQ,EAAE,EAAE,MAAM,IAAI7C,WAAW,eAAe,GAAGqC,IAAIb,GAAG,CAAC,SAAS,EAAEa,IAAIS,MAAM,EAAE;QACnF,MAAMjC,OAAOD,UAAUyB,IAAIxB,IAAI,EAAEwB,IAAIb,GAAG;QACxC,2FAA2F;QAC3F,kEAAkE;QAClE,MAAMuB,UAAUlC;QAChB,IAAIkC,QAAQC,MAAM,KAAKd,MAAMc,MAAM,EAAE,MAAM,IAAIhD,WAAW,eAAe,GAAGqC,IAAIb,GAAG,CAAC,UAAU,EAAEuB,QAAQC,MAAM,CAAC,gBAAgB,EAAEd,MAAMc,MAAM,CAAC,OAAO,CAAC;QACtJ,OAAOD,QAAQE,GAAG,CAAC,CAACC,IAAMC,aAAaC,IAAI,CAACF;IAC9C;IAEA,MAAMG,OAAO,AAAC,CAAA,MAAMpB,KAAK;QAAC;KAAkB,EAAE,eAAc,CAAE,CAAC,EAAE,CAACe,MAAM;IACxE,OAAO;QAAEM,IAAI,CAAC,OAAO,EAAE5B,KAAK,CAAC,EAAEH,OAAO;QAAE8B;QAAME,UAAUnD;QAAWoD,gBAAgB,CAACtB,OAAOE,UAAYH,KAAKC,OAAO,mBAAmBE;QAAUqB,YAAY,OAAOC,MAAMtB,UAAY,AAAC,CAAA,MAAMH,KAAK;gBAACyB;aAAK,EAAE,gBAAgBtB,QAAO,CAAE,CAAC,EAAE;IAAC;AACxO"}
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/embed/distribution.ts"],"sourcesContent":["import { getMeta, setMeta } from '../store/shared.ts';\nimport type { Store } from '../store/types.ts';\n\n// Split out of langfit.ts so reading the persisted counts (status.ts's only need) never pulls in\n// franc-min; only checkLanguageFit's classify path needs the classifier itself.\nconst META_KEY = 'embed_languages';\n\nexport type LangCounts = Record<string, number>;\n\nexport async function languageDistribution(store: Store): Promise<LangCounts | undefined> {\n const raw = await getMeta(store, META_KEY);\n return raw ? JSON.parse(raw) : undefined;\n}\n\nexport async function saveLanguageDistribution(store: Store, counts: LangCounts): Promise<void> {\n await setMeta(store, META_KEY, JSON.stringify(counts));\n}\n"],"names":["getMeta","setMeta","META_KEY","languageDistribution","store","raw","JSON","parse","undefined","saveLanguageDistribution","counts","stringify"],"mappings":"AAAA,SAASA,OAAO,EAAEC,OAAO,QAAQ,qBAAqB;AAGtD,iGAAiG;AACjG,gFAAgF;AAChF,MAAMC,WAAW;AAIjB,OAAO,eAAeC,qBAAqBC,KAAY;IACrD,MAAMC,MAAM,MAAML,QAAQI,OAAOF;IACjC,OAAOG,MAAMC,KAAKC,KAAK,CAACF,OAAOG;AACjC;AAEA,OAAO,eAAeC,yBAAyBL,KAAY,EAAEM,MAAkB;IAC7E,MAAMT,QAAQG,OAAOF,UAAUI,KAAKK,SAAS,CAACD;AAChD"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/open-source/ai/sensemaking/src/embed/distribution.ts"],"sourcesContent":["import { getMeta, setMeta } from '../store/shared.ts';\nimport type { Store } from '../store/types.ts';\n\n// Split out of langfit.ts so reading the persisted counts (status.ts's only need) never pulls in\n// franc-min; only checkLanguageFit's classify path needs the classifier itself.\nconst META_KEY = 'embed_languages';\n\nexport type LangCounts = Record<string, number>;\n\nexport async function languageDistribution(store: Store): Promise<LangCounts | undefined> {\n const raw = await getMeta(store, META_KEY);\n return raw ? JSON.parse(raw) : undefined;\n}\n\nexport async function saveLanguageDistribution(store: Store, counts: LangCounts): Promise<void> {\n await setMeta(store, META_KEY, JSON.stringify(counts));\n}\n"],"names":["getMeta","setMeta","META_KEY","languageDistribution","store","raw","JSON","parse","undefined","saveLanguageDistribution","counts","stringify"],"mappings":"AAAA,SAASA,OAAO,EAAEC,OAAO,QAAQ,qBAAqB;AAGtD,iGAAiG;AACjG,gFAAgF;AAChF,MAAMC,WAAW;AAIjB,OAAO,eAAeC,qBAAqBC,KAAY;IACrD,MAAMC,MAAM,MAAML,QAAQI,OAAOF;IACjC,OAAOG,MAAMC,KAAKC,KAAK,CAACF,OAAOG;AACjC;AAEA,OAAO,eAAeC,yBAAyBL,KAAY,EAAEM,MAAkB;IAC7E,MAAMT,QAAQG,OAAOF,UAAUI,KAAKK,SAAS,CAACD;AAChD"}
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/embed/endpoint.ts"],"sourcesContent":["import { createHash } from 'node:crypto';\n\nconst DEFAULT_COHERE_ENDPOINT = 'https://api.cohere.com';\n\nexport function effectiveEmbedEndpoint(provider: 'static' | 'openai' | 'cohere', url: string | undefined): string | undefined {\n if (provider === 'static') return undefined;\n const endpoint = provider === 'cohere' ? (url ?? DEFAULT_COHERE_ENDPOINT) : url;\n return endpoint?.replace(/\\/+$/, '');\n}\n\nexport function embedEndpointIdentity(provider: 'static' | 'openai' | 'cohere', url: string | undefined): string | undefined {\n const endpoint = effectiveEmbedEndpoint(provider, url);\n return endpoint === undefined ? undefined : createHash('sha256').update(endpoint).digest('hex');\n}\n"],"names":["createHash","DEFAULT_COHERE_ENDPOINT","effectiveEmbedEndpoint","provider","url","undefined","endpoint","replace","embedEndpointIdentity","update","digest"],"mappings":"AAAA,SAASA,UAAU,QAAQ,cAAc;AAEzC,MAAMC,0BAA0B;AAEhC,OAAO,SAASC,uBAAuBC,QAAwC,EAAEC,GAAuB;IACtG,IAAID,aAAa,UAAU,OAAOE;IAClC,MAAMC,WAAWH,aAAa,WAAYC,gBAAAA,iBAAAA,MAAOH,0BAA2BG;IAC5E,OAAOE,qBAAAA,+BAAAA,SAAUC,OAAO,CAAC,QAAQ;AACnC;AAEA,OAAO,SAASC,sBAAsBL,QAAwC,EAAEC,GAAuB;IACrG,MAAME,WAAWJ,uBAAuBC,UAAUC;IAClD,OAAOE,aAAaD,YAAYA,YAAYL,WAAW,UAAUS,MAAM,CAACH,UAAUI,MAAM,CAAC;AAC3F"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/open-source/ai/sensemaking/src/embed/endpoint.ts"],"sourcesContent":["import { createHash } from 'node:crypto';\n\nconst DEFAULT_COHERE_ENDPOINT = 'https://api.cohere.com';\n\nexport function effectiveEmbedEndpoint(provider: 'static' | 'openai' | 'cohere', url: string | undefined): string | undefined {\n if (provider === 'static') return undefined;\n const endpoint = provider === 'cohere' ? (url ?? DEFAULT_COHERE_ENDPOINT) : url;\n return endpoint?.replace(/\\/+$/, '');\n}\n\nexport function embedEndpointIdentity(provider: 'static' | 'openai' | 'cohere', url: string | undefined): string | undefined {\n const endpoint = effectiveEmbedEndpoint(provider, url);\n return endpoint === undefined ? undefined : createHash('sha256').update(endpoint).digest('hex');\n}\n"],"names":["createHash","DEFAULT_COHERE_ENDPOINT","effectiveEmbedEndpoint","provider","url","undefined","endpoint","replace","embedEndpointIdentity","update","digest"],"mappings":"AAAA,SAASA,UAAU,QAAQ,cAAc;AAEzC,MAAMC,0BAA0B;AAEhC,OAAO,SAASC,uBAAuBC,QAAwC,EAAEC,GAAuB;IACtG,IAAID,aAAa,UAAU,OAAOE;IAClC,MAAMC,WAAWH,aAAa,WAAYC,gBAAAA,iBAAAA,MAAOH,0BAA2BG;IAC5E,OAAOE,qBAAAA,+BAAAA,SAAUC,OAAO,CAAC,QAAQ;AACnC;AAEA,OAAO,SAASC,sBAAsBL,QAAwC,EAAEC,GAAuB;IACrG,MAAME,WAAWJ,uBAAuBC,UAAUC;IAClD,OAAOE,aAAaD,YAAYA,YAAYL,WAAW,UAAUS,MAAM,CAACH,UAAUI,MAAM,CAAC;AAC3F"}
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/embed/handoff.ts"],"sourcesContent":["import { createHash } from 'node:crypto';\n\nexport interface ChunkText {\n text: string;\n identity: string;\n}\n\nexport function identifyChunkText(text: string): ChunkText {\n return { text, identity: createHash('sha256').update(text).digest('hex') };\n}\n\n// Chunk text from reconcile to embed time when both run in one process, so the text is neither\n// re-derived nor persisted. Keyed by the store it was stashed against, so an entry dies with it.\nconst HANDOFF = new WeakMap<object, Map<string, ChunkText[]>>();\n\nexport function stashChunkText(key: object, texts: Map<string, ChunkText[]>): void {\n if (texts.size > 0) HANDOFF.set(key, texts);\n}\n\n// Moves an entry from one key to another: reconcile runs before the store object exists, so it\n// stashes against the connection and open() re-keys onto the store it then builds.\nexport function rekeyChunkText(from: object, to: object): void {\n const texts = HANDOFF.get(from);\n if (!texts) return;\n HANDOFF.delete(from);\n HANDOFF.set(to, texts);\n}\n\n// Read once and drop: the text is consumed by the embed pass that follows reconcile, and holding\n// it past that would pin the tree's prose in memory for the life of the store.\nexport function takeChunkText(key: object): Map<string, ChunkText[]> | undefined {\n const texts = HANDOFF.get(key);\n HANDOFF.delete(key);\n return texts;\n}\n"],"names":["createHash","identifyChunkText","text","identity","update","digest","HANDOFF","WeakMap","stashChunkText","key","texts","size","set","rekeyChunkText","from","to","get","delete","takeChunkText"],"mappings":"AAAA,SAASA,UAAU,QAAQ,cAAc;AAOzC,OAAO,SAASC,kBAAkBC,IAAY;IAC5C,OAAO;QAAEA;QAAMC,UAAUH,WAAW,UAAUI,MAAM,CAACF,MAAMG,MAAM,CAAC;IAAO;AAC3E;AAEA,+FAA+F;AAC/F,iGAAiG;AACjG,MAAMC,UAAU,IAAIC;AAEpB,OAAO,SAASC,eAAeC,GAAW,EAAEC,KAA+B;IACzE,IAAIA,MAAMC,IAAI,GAAG,GAAGL,QAAQM,GAAG,CAACH,KAAKC;AACvC;AAEA,+FAA+F;AAC/F,mFAAmF;AACnF,OAAO,SAASG,eAAeC,IAAY,EAAEC,EAAU;IACrD,MAAML,QAAQJ,QAAQU,GAAG,CAACF;IAC1B,IAAI,CAACJ,OAAO;IACZJ,QAAQW,MAAM,CAACH;IACfR,QAAQM,GAAG,CAACG,IAAIL;AAClB;AAEA,iGAAiG;AACjG,+EAA+E;AAC/E,OAAO,SAASQ,cAAcT,GAAW;IACvC,MAAMC,QAAQJ,QAAQU,GAAG,CAACP;IAC1BH,QAAQW,MAAM,CAACR;IACf,OAAOC;AACT"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/open-source/ai/sensemaking/src/embed/handoff.ts"],"sourcesContent":["import { createHash } from 'node:crypto';\n\nexport interface ChunkText {\n text: string;\n identity: string;\n}\n\nexport function identifyChunkText(text: string): ChunkText {\n return { text, identity: createHash('sha256').update(text).digest('hex') };\n}\n\n// Chunk text from reconcile to embed time when both run in one process, so the text is neither\n// re-derived nor persisted. Keyed by the store it was stashed against, so an entry dies with it.\nconst HANDOFF = new WeakMap<object, Map<string, ChunkText[]>>();\n\nexport function stashChunkText(key: object, texts: Map<string, ChunkText[]>): void {\n if (texts.size > 0) HANDOFF.set(key, texts);\n}\n\n// Moves an entry from one key to another: reconcile runs before the store object exists, so it\n// stashes against the connection and open() re-keys onto the store it then builds.\nexport function rekeyChunkText(from: object, to: object): void {\n const texts = HANDOFF.get(from);\n if (!texts) return;\n HANDOFF.delete(from);\n HANDOFF.set(to, texts);\n}\n\n// Read once and drop: the text is consumed by the embed pass that follows reconcile, and holding\n// it past that would pin the tree's prose in memory for the life of the store.\nexport function takeChunkText(key: object): Map<string, ChunkText[]> | undefined {\n const texts = HANDOFF.get(key);\n HANDOFF.delete(key);\n return texts;\n}\n"],"names":["createHash","identifyChunkText","text","identity","update","digest","HANDOFF","WeakMap","stashChunkText","key","texts","size","set","rekeyChunkText","from","to","get","delete","takeChunkText"],"mappings":"AAAA,SAASA,UAAU,QAAQ,cAAc;AAOzC,OAAO,SAASC,kBAAkBC,IAAY;IAC5C,OAAO;QAAEA;QAAMC,UAAUH,WAAW,UAAUI,MAAM,CAACF,MAAMG,MAAM,CAAC;IAAO;AAC3E;AAEA,+FAA+F;AAC/F,iGAAiG;AACjG,MAAMC,UAAU,IAAIC;AAEpB,OAAO,SAASC,eAAeC,GAAW,EAAEC,KAA+B;IACzE,IAAIA,MAAMC,IAAI,GAAG,GAAGL,QAAQM,GAAG,CAACH,KAAKC;AACvC;AAEA,+FAA+F;AAC/F,mFAAmF;AACnF,OAAO,SAASG,eAAeC,IAAY,EAAEC,EAAU;IACrD,MAAML,QAAQJ,QAAQU,GAAG,CAACF;IAC1B,IAAI,CAACJ,OAAO;IACZJ,QAAQW,MAAM,CAACH;IACfR,QAAQM,GAAG,CAACG,IAAIL;AAClB;AAEA,iGAAiG;AACjG,+EAA+E;AAC/E,OAAO,SAASQ,cAAcT,GAAW;IACvC,MAAMC,QAAQJ,QAAQU,GAAG,CAACP;IAC1BH,QAAQW,MAAM,CAACR;IACf,OAAOC;AACT"}
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/embed/http.ts"],"sourcesContent":["import { SenseError } from '../errors.ts';\n\nconst MAX_ATTEMPTS = 3;\nconst MAX_RETRY_AFTER_MS = 10_000;\nconst DEFAULT_BASE_DELAY_MS = 200;\n\n// Initial policy bounds small API/metadata responses separately from model-file downloads.\nexport const HTTP_ATTEMPT_TIMEOUT_MS = 30_000;\nexport const MODEL_FILE_ATTEMPT_TIMEOUT_MS = 5 * 60_000;\n\nexport type FetchRequestInit = Omit<RequestInit, 'signal'> & { signal?: never };\n\nexport interface FetchRetryOptions {\n signal?: AbortSignal;\n attemptTimeoutMs?: number;\n baseDelayMs?: number;\n}\n\nexport interface BufferedResponse {\n readonly ok: boolean;\n readonly status: number;\n readonly headers: Headers;\n readonly body: ArrayBuffer;\n readonly url: string;\n}\n\ntype ResponsePhase = 'response headers' | 'response body';\n\nfunction safeUrl(url: string): string {\n try {\n const parsed = new URL(url);\n parsed.username = '';\n parsed.password = '';\n parsed.search = '';\n parsed.hash = '';\n return parsed.href;\n } catch {\n return 'the configured embedding endpoint';\n }\n}\n\nfunction attemptController(\n callerSignal: AbortSignal | undefined,\n timeoutMs: number\n): {\n controller: AbortController;\n timeoutReason: object;\n cleanup: () => void;\n} {\n callerSignal?.throwIfAborted();\n const controller = new AbortController();\n const timeoutReason = {};\n const relayAbort = () => controller.abort(callerSignal?.reason);\n callerSignal?.addEventListener('abort', relayAbort, { once: true });\n if (callerSignal?.aborted) relayAbort();\n const timer = setTimeout(() => controller.abort(timeoutReason), timeoutMs);\n return {\n controller,\n timeoutReason,\n cleanup: () => {\n clearTimeout(timer);\n callerSignal?.removeEventListener('abort', relayAbort);\n },\n };\n}\n\nfunction disposeResponse(response: Response, controller: AbortController): void {\n const cancellation = response.body && !response.bodyUsed ? response.body.cancel() : undefined;\n controller.abort();\n // The controller owns socket cleanup; observe stream cancellation without delaying the attempt.\n if (cancellation)\n void cancellation.then(\n () => undefined,\n () => undefined\n );\n}\n\nfunction sleep(ms: number, signal: AbortSignal | undefined): Promise<void> {\n signal?.throwIfAborted();\n if (!signal) return new Promise((resolve) => setTimeout(resolve, ms));\n return new Promise((resolve, reject) => {\n let timer: ReturnType<typeof setTimeout> | undefined;\n const cleanup = () => {\n if (timer) clearTimeout(timer);\n signal.removeEventListener('abort', onAbort);\n };\n const onAbort = () => {\n cleanup();\n reject(signal.reason);\n };\n signal.addEventListener('abort', onAbort, { once: true });\n timer = setTimeout(() => {\n cleanup();\n resolve();\n }, ms);\n if (signal.aborted) onAbort();\n });\n}\n\nfunction failedRequest(url: string, phase: ResponsePhase, attempts: number, timeoutMs: number, timedOut: boolean): SenseError {\n const endpoint = safeUrl(url);\n if (timedOut) {\n return new SenseError('EMBED_MODEL', `${endpoint} timed out after ${timeoutMs}ms while waiting for ${phase} (attempt ${attempts}/${MAX_ATTEMPTS}); check the embedding service and network`);\n }\n return new SenseError('EMBED_MODEL', `${endpoint} failed while waiting for ${phase} after ${attempts} attempts; check the embedding service and network`);\n}\n\nfunction buffered(response: Response, body: ArrayBuffer, requestUrl: string): BufferedResponse {\n return {\n ok: response.ok,\n status: response.status,\n headers: response.headers,\n body,\n url: safeUrl(response.url || requestUrl),\n };\n}\n\n// Each attempt covers response headers and the complete successful body. Retry cleanup and\n// backoff honor one caller signal without transferring ownership of that signal.\nexport async function fetchWithRetry(url: string, init: FetchRequestInit, options: FetchRetryOptions = {}): Promise<BufferedResponse> {\n const timeoutMs = options.attemptTimeoutMs ?? HTTP_ATTEMPT_TIMEOUT_MS;\n const baseDelayMs = options.baseDelayMs ?? DEFAULT_BASE_DELAY_MS;\n for (let attempt = 0; attempt < MAX_ATTEMPTS; attempt++) {\n options.signal?.throwIfAborted();\n const { controller, timeoutReason, cleanup } = attemptController(options.signal, timeoutMs);\n let response: Response | undefined;\n let phase: ResponsePhase = 'response headers';\n let retryDelayMs: number | undefined;\n try {\n response = await fetch(url, { ...init, signal: controller.signal });\n if (!response.ok) {\n const retryable = response.status === 429 || response.status >= 500;\n if (retryable && attempt < MAX_ATTEMPTS - 1) {\n const retryAfter = Number(response.headers.get('retry-after'));\n retryDelayMs = retryAfter > 0 ? Math.min(retryAfter * 1000, MAX_RETRY_AFTER_MS) : baseDelayMs * 2 ** attempt;\n disposeResponse(response, controller);\n } else {\n disposeResponse(response, controller);\n options.signal?.throwIfAborted();\n return buffered(response, new ArrayBuffer(0), url);\n }\n } else {\n phase = 'response body';\n const body = await response.arrayBuffer();\n options.signal?.throwIfAborted();\n return buffered(response, body, url);\n }\n } catch {\n try {\n options.signal?.throwIfAborted();\n } finally {\n if (response) disposeResponse(response, controller);\n }\n const timedOut = controller.signal.reason === timeoutReason;\n if (attempt === MAX_ATTEMPTS - 1) throw failedRequest(url, phase, attempt + 1, timeoutMs, timedOut);\n retryDelayMs = baseDelayMs * 2 ** attempt;\n } finally {\n cleanup();\n }\n await sleep(retryDelayMs ?? baseDelayMs * 2 ** attempt, options.signal);\n }\n throw new Error('unreachable');\n}\n\n// Any HTTP response means the endpoint is up. Abort it immediately because status does not\n// consume its body, and bound the headers wait for an air-gapped machine.\nexport async function probeReachable(url: string, timeoutMs = 1500): Promise<boolean> {\n const { controller, cleanup } = attemptController(undefined, timeoutMs);\n try {\n const response = await fetch(url, { signal: controller.signal });\n disposeResponse(response, controller);\n return true;\n } catch {\n return false;\n } finally {\n cleanup();\n }\n}\n"],"names":["SenseError","MAX_ATTEMPTS","MAX_RETRY_AFTER_MS","DEFAULT_BASE_DELAY_MS","HTTP_ATTEMPT_TIMEOUT_MS","MODEL_FILE_ATTEMPT_TIMEOUT_MS","safeUrl","url","parsed","URL","username","password","search","hash","href","attemptController","callerSignal","timeoutMs","throwIfAborted","controller","AbortController","timeoutReason","relayAbort","abort","reason","addEventListener","once","aborted","timer","setTimeout","cleanup","clearTimeout","removeEventListener","disposeResponse","response","cancellation","body","bodyUsed","cancel","undefined","then","sleep","ms","signal","Promise","resolve","reject","onAbort","failedRequest","phase","attempts","timedOut","endpoint","buffered","requestUrl","ok","status","headers","fetchWithRetry","init","options","attemptTimeoutMs","baseDelayMs","attempt","retryDelayMs","fetch","retryable","retryAfter","Number","get","Math","min","ArrayBuffer","arrayBuffer","Error","probeReachable"],"mappings":"AAAA,SAASA,UAAU,QAAQ,eAAe;AAE1C,MAAMC,eAAe;AACrB,MAAMC,qBAAqB;AAC3B,MAAMC,wBAAwB;AAE9B,2FAA2F;AAC3F,OAAO,MAAMC,0BAA0B,MAAO;AAC9C,OAAO,MAAMC,gCAAgC,IAAI,MAAO;AAoBxD,SAASC,QAAQC,GAAW;IAC1B,IAAI;QACF,MAAMC,SAAS,IAAIC,IAAIF;QACvBC,OAAOE,QAAQ,GAAG;QAClBF,OAAOG,QAAQ,GAAG;QAClBH,OAAOI,MAAM,GAAG;QAChBJ,OAAOK,IAAI,GAAG;QACd,OAAOL,OAAOM,IAAI;IACpB,EAAE,OAAM;QACN,OAAO;IACT;AACF;AAEA,SAASC,kBACPC,YAAqC,EACrCC,SAAiB;IAMjBD,yBAAAA,mCAAAA,aAAcE,cAAc;IAC5B,MAAMC,aAAa,IAAIC;IACvB,MAAMC,gBAAgB,CAAC;IACvB,MAAMC,aAAa,IAAMH,WAAWI,KAAK,CAACP,yBAAAA,mCAAAA,aAAcQ,MAAM;IAC9DR,yBAAAA,mCAAAA,aAAcS,gBAAgB,CAAC,SAASH,YAAY;QAAEI,MAAM;IAAK;IACjE,IAAIV,yBAAAA,mCAAAA,aAAcW,OAAO,EAAEL;IAC3B,MAAMM,QAAQC,WAAW,IAAMV,WAAWI,KAAK,CAACF,gBAAgBJ;IAChE,OAAO;QACLE;QACAE;QACAS,SAAS;YACPC,aAAaH;YACbZ,yBAAAA,mCAAAA,aAAcgB,mBAAmB,CAAC,SAASV;QAC7C;IACF;AACF;AAEA,SAASW,gBAAgBC,QAAkB,EAAEf,UAA2B;IACtE,MAAMgB,eAAeD,SAASE,IAAI,IAAI,CAACF,SAASG,QAAQ,GAAGH,SAASE,IAAI,CAACE,MAAM,KAAKC;IACpFpB,WAAWI,KAAK;IAChB,gGAAgG;IAChG,IAAIY,cACF,KAAKA,aAAaK,IAAI,CACpB,IAAMD,WACN,IAAMA;AAEZ;AAEA,SAASE,MAAMC,EAAU,EAAEC,MAA+B;IACxDA,mBAAAA,6BAAAA,OAAQzB,cAAc;IACtB,IAAI,CAACyB,QAAQ,OAAO,IAAIC,QAAQ,CAACC,UAAYhB,WAAWgB,SAASH;IACjE,OAAO,IAAIE,QAAQ,CAACC,SAASC;QAC3B,IAAIlB;QACJ,MAAME,UAAU;YACd,IAAIF,OAAOG,aAAaH;YACxBe,OAAOX,mBAAmB,CAAC,SAASe;QACtC;QACA,MAAMA,UAAU;YACdjB;YACAgB,OAAOH,OAAOnB,MAAM;QACtB;QACAmB,OAAOlB,gBAAgB,CAAC,SAASsB,SAAS;YAAErB,MAAM;QAAK;QACvDE,QAAQC,WAAW;YACjBC;YACAe;QACF,GAAGH;QACH,IAAIC,OAAOhB,OAAO,EAAEoB;IACtB;AACF;AAEA,SAASC,cAAczC,GAAW,EAAE0C,KAAoB,EAAEC,QAAgB,EAAEjC,SAAiB,EAAEkC,QAAiB;IAC9G,MAAMC,WAAW9C,QAAQC;IACzB,IAAI4C,UAAU;QACZ,OAAO,IAAInD,WAAW,eAAe,GAAGoD,SAAS,iBAAiB,EAAEnC,UAAU,qBAAqB,EAAEgC,MAAM,UAAU,EAAEC,SAAS,CAAC,EAAEjD,aAAa,0CAA0C,CAAC;IAC7L;IACA,OAAO,IAAID,WAAW,eAAe,GAAGoD,SAAS,0BAA0B,EAAEH,MAAM,OAAO,EAAEC,SAAS,kDAAkD,CAAC;AAC1J;AAEA,SAASG,SAASnB,QAAkB,EAAEE,IAAiB,EAAEkB,UAAkB;IACzE,OAAO;QACLC,IAAIrB,SAASqB,EAAE;QACfC,QAAQtB,SAASsB,MAAM;QACvBC,SAASvB,SAASuB,OAAO;QACzBrB;QACA7B,KAAKD,QAAQ4B,SAAS3B,GAAG,IAAI+C;IAC/B;AACF;AAEA,2FAA2F;AAC3F,iFAAiF;AACjF,OAAO,eAAeI,eAAenD,GAAW,EAAEoD,IAAsB,EAAEC,UAA6B,CAAC,CAAC;QACrFA,2BACEA;IADpB,MAAM3C,aAAY2C,4BAAAA,QAAQC,gBAAgB,cAAxBD,uCAAAA,4BAA4BxD;IAC9C,MAAM0D,eAAcF,uBAAAA,QAAQE,WAAW,cAAnBF,kCAAAA,uBAAuBzD;IAC3C,IAAK,IAAI4D,UAAU,GAAGA,UAAU9D,cAAc8D,UAAW;YACvDH;SAAAA,kBAAAA,QAAQjB,MAAM,cAAdiB,sCAAAA,gBAAgB1C,cAAc;QAC9B,MAAM,EAAEC,UAAU,EAAEE,aAAa,EAAES,OAAO,EAAE,GAAGf,kBAAkB6C,QAAQjB,MAAM,EAAE1B;QACjF,IAAIiB;QACJ,IAAIe,QAAuB;QAC3B,IAAIe;QACJ,IAAI;YACF9B,WAAW,MAAM+B,MAAM1D,KAAK;gBAAE,GAAGoD,IAAI;gBAAEhB,QAAQxB,WAAWwB,MAAM;YAAC;YACjE,IAAI,CAACT,SAASqB,EAAE,EAAE;gBAChB,MAAMW,YAAYhC,SAASsB,MAAM,KAAK,OAAOtB,SAASsB,MAAM,IAAI;gBAChE,IAAIU,aAAaH,UAAU9D,eAAe,GAAG;oBAC3C,MAAMkE,aAAaC,OAAOlC,SAASuB,OAAO,CAACY,GAAG,CAAC;oBAC/CL,eAAeG,aAAa,IAAIG,KAAKC,GAAG,CAACJ,aAAa,MAAMjE,sBAAsB4D,cAAc,KAAKC;oBACrG9B,gBAAgBC,UAAUf;gBAC5B,OAAO;wBAELyC;oBADA3B,gBAAgBC,UAAUf;qBAC1ByC,mBAAAA,QAAQjB,MAAM,cAAdiB,uCAAAA,iBAAgB1C,cAAc;oBAC9B,OAAOmC,SAASnB,UAAU,IAAIsC,YAAY,IAAIjE;gBAChD;YACF,OAAO;oBAGLqD;gBAFAX,QAAQ;gBACR,MAAMb,OAAO,MAAMF,SAASuC,WAAW;iBACvCb,mBAAAA,QAAQjB,MAAM,cAAdiB,uCAAAA,iBAAgB1C,cAAc;gBAC9B,OAAOmC,SAASnB,UAAUE,MAAM7B;YAClC;QACF,EAAE,OAAM;YACN,IAAI;oBACFqD;iBAAAA,mBAAAA,QAAQjB,MAAM,cAAdiB,uCAAAA,iBAAgB1C,cAAc;YAChC,SAAU;gBACR,IAAIgB,UAAUD,gBAAgBC,UAAUf;YAC1C;YACA,MAAMgC,WAAWhC,WAAWwB,MAAM,CAACnB,MAAM,KAAKH;YAC9C,IAAI0C,YAAY9D,eAAe,GAAG,MAAM+C,cAAczC,KAAK0C,OAAOc,UAAU,GAAG9C,WAAWkC;YAC1Fa,eAAeF,cAAc,KAAKC;QACpC,SAAU;YACRjC;QACF;QACA,MAAMW,MAAMuB,yBAAAA,0BAAAA,eAAgBF,cAAc,KAAKC,SAASH,QAAQjB,MAAM;IACxE;IACA,MAAM,IAAI+B,MAAM;AAClB;AAEA,2FAA2F;AAC3F,0EAA0E;AAC1E,OAAO,eAAeC,eAAepE,GAAW,EAAEU,YAAY,IAAI;IAChE,MAAM,EAAEE,UAAU,EAAEW,OAAO,EAAE,GAAGf,kBAAkBwB,WAAWtB;IAC7D,IAAI;QACF,MAAMiB,WAAW,MAAM+B,MAAM1D,KAAK;YAAEoC,QAAQxB,WAAWwB,MAAM;QAAC;QAC9DV,gBAAgBC,UAAUf;QAC1B,OAAO;IACT,EAAE,OAAM;QACN,OAAO;IACT,SAAU;QACRW;IACF;AACF"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/open-source/ai/sensemaking/src/embed/http.ts"],"sourcesContent":["import { SenseError } from '../errors.ts';\n\nconst MAX_ATTEMPTS = 3;\nconst MAX_RETRY_AFTER_MS = 10_000;\nconst DEFAULT_BASE_DELAY_MS = 200;\n\n// Initial policy bounds small API/metadata responses separately from model-file downloads.\nexport const HTTP_ATTEMPT_TIMEOUT_MS = 30_000;\nexport const MODEL_FILE_ATTEMPT_TIMEOUT_MS = 5 * 60_000;\n\nexport type FetchRequestInit = Omit<RequestInit, 'signal'> & { signal?: never };\n\nexport interface FetchRetryOptions {\n signal?: AbortSignal;\n attemptTimeoutMs?: number;\n baseDelayMs?: number;\n}\n\nexport interface BufferedResponse {\n readonly ok: boolean;\n readonly status: number;\n readonly headers: Headers;\n readonly body: ArrayBuffer;\n readonly url: string;\n}\n\ntype ResponsePhase = 'response headers' | 'response body';\n\nfunction safeUrl(url: string): string {\n try {\n const parsed = new URL(url);\n parsed.username = '';\n parsed.password = '';\n parsed.search = '';\n parsed.hash = '';\n return parsed.href;\n } catch {\n return 'the configured embedding endpoint';\n }\n}\n\nfunction attemptController(\n callerSignal: AbortSignal | undefined,\n timeoutMs: number\n): {\n controller: AbortController;\n timeoutReason: object;\n cleanup: () => void;\n} {\n callerSignal?.throwIfAborted();\n const controller = new AbortController();\n const timeoutReason = {};\n const relayAbort = () => controller.abort(callerSignal?.reason);\n callerSignal?.addEventListener('abort', relayAbort, { once: true });\n if (callerSignal?.aborted) relayAbort();\n const timer = setTimeout(() => controller.abort(timeoutReason), timeoutMs);\n return {\n controller,\n timeoutReason,\n cleanup: () => {\n clearTimeout(timer);\n callerSignal?.removeEventListener('abort', relayAbort);\n },\n };\n}\n\nfunction disposeResponse(response: Response, controller: AbortController): void {\n const cancellation = response.body && !response.bodyUsed ? response.body.cancel() : undefined;\n controller.abort();\n // The controller owns socket cleanup; observe stream cancellation without delaying the attempt.\n if (cancellation)\n void cancellation.then(\n () => undefined,\n () => undefined\n );\n}\n\nfunction sleep(ms: number, signal: AbortSignal | undefined): Promise<void> {\n signal?.throwIfAborted();\n if (!signal) return new Promise((resolve) => setTimeout(resolve, ms));\n return new Promise((resolve, reject) => {\n let timer: ReturnType<typeof setTimeout> | undefined;\n const cleanup = () => {\n if (timer) clearTimeout(timer);\n signal.removeEventListener('abort', onAbort);\n };\n const onAbort = () => {\n cleanup();\n reject(signal.reason);\n };\n signal.addEventListener('abort', onAbort, { once: true });\n timer = setTimeout(() => {\n cleanup();\n resolve();\n }, ms);\n if (signal.aborted) onAbort();\n });\n}\n\nfunction failedRequest(url: string, phase: ResponsePhase, attempts: number, timeoutMs: number, timedOut: boolean): SenseError {\n const endpoint = safeUrl(url);\n if (timedOut) {\n return new SenseError('EMBED_MODEL', `${endpoint} timed out after ${timeoutMs}ms while waiting for ${phase} (attempt ${attempts}/${MAX_ATTEMPTS}); check the embedding service and network`);\n }\n return new SenseError('EMBED_MODEL', `${endpoint} failed while waiting for ${phase} after ${attempts} attempts; check the embedding service and network`);\n}\n\nfunction buffered(response: Response, body: ArrayBuffer, requestUrl: string): BufferedResponse {\n return {\n ok: response.ok,\n status: response.status,\n headers: response.headers,\n body,\n url: safeUrl(response.url || requestUrl),\n };\n}\n\n// Each attempt covers response headers and the complete successful body. Retry cleanup and\n// backoff honor one caller signal without transferring ownership of that signal.\nexport async function fetchWithRetry(url: string, init: FetchRequestInit, options: FetchRetryOptions = {}): Promise<BufferedResponse> {\n const timeoutMs = options.attemptTimeoutMs ?? HTTP_ATTEMPT_TIMEOUT_MS;\n const baseDelayMs = options.baseDelayMs ?? DEFAULT_BASE_DELAY_MS;\n for (let attempt = 0; attempt < MAX_ATTEMPTS; attempt++) {\n options.signal?.throwIfAborted();\n const { controller, timeoutReason, cleanup } = attemptController(options.signal, timeoutMs);\n let response: Response | undefined;\n let phase: ResponsePhase = 'response headers';\n let retryDelayMs: number | undefined;\n try {\n response = await fetch(url, { ...init, signal: controller.signal });\n if (!response.ok) {\n const retryable = response.status === 429 || response.status >= 500;\n if (retryable && attempt < MAX_ATTEMPTS - 1) {\n const retryAfter = Number(response.headers.get('retry-after'));\n retryDelayMs = retryAfter > 0 ? Math.min(retryAfter * 1000, MAX_RETRY_AFTER_MS) : baseDelayMs * 2 ** attempt;\n disposeResponse(response, controller);\n } else {\n disposeResponse(response, controller);\n options.signal?.throwIfAborted();\n return buffered(response, new ArrayBuffer(0), url);\n }\n } else {\n phase = 'response body';\n const body = await response.arrayBuffer();\n options.signal?.throwIfAborted();\n return buffered(response, body, url);\n }\n } catch {\n try {\n options.signal?.throwIfAborted();\n } finally {\n if (response) disposeResponse(response, controller);\n }\n const timedOut = controller.signal.reason === timeoutReason;\n if (attempt === MAX_ATTEMPTS - 1) throw failedRequest(url, phase, attempt + 1, timeoutMs, timedOut);\n retryDelayMs = baseDelayMs * 2 ** attempt;\n } finally {\n cleanup();\n }\n await sleep(retryDelayMs ?? baseDelayMs * 2 ** attempt, options.signal);\n }\n throw new Error('unreachable');\n}\n\n// Any HTTP response means the endpoint is up. Abort it immediately because status does not\n// consume its body, and bound the headers wait for an air-gapped machine.\nexport async function probeReachable(url: string, timeoutMs = 1500): Promise<boolean> {\n const { controller, cleanup } = attemptController(undefined, timeoutMs);\n try {\n const response = await fetch(url, { signal: controller.signal });\n disposeResponse(response, controller);\n return true;\n } catch {\n return false;\n } finally {\n cleanup();\n }\n}\n"],"names":["SenseError","MAX_ATTEMPTS","MAX_RETRY_AFTER_MS","DEFAULT_BASE_DELAY_MS","HTTP_ATTEMPT_TIMEOUT_MS","MODEL_FILE_ATTEMPT_TIMEOUT_MS","safeUrl","url","parsed","URL","username","password","search","hash","href","attemptController","callerSignal","timeoutMs","throwIfAborted","controller","AbortController","timeoutReason","relayAbort","abort","reason","addEventListener","once","aborted","timer","setTimeout","cleanup","clearTimeout","removeEventListener","disposeResponse","response","cancellation","body","bodyUsed","cancel","undefined","then","sleep","ms","signal","Promise","resolve","reject","onAbort","failedRequest","phase","attempts","timedOut","endpoint","buffered","requestUrl","ok","status","headers","fetchWithRetry","init","options","attemptTimeoutMs","baseDelayMs","attempt","retryDelayMs","fetch","retryable","retryAfter","Number","get","Math","min","ArrayBuffer","arrayBuffer","Error","probeReachable"],"mappings":"AAAA,SAASA,UAAU,QAAQ,eAAe;AAE1C,MAAMC,eAAe;AACrB,MAAMC,qBAAqB;AAC3B,MAAMC,wBAAwB;AAE9B,2FAA2F;AAC3F,OAAO,MAAMC,0BAA0B,MAAO;AAC9C,OAAO,MAAMC,gCAAgC,IAAI,MAAO;AAoBxD,SAASC,QAAQC,GAAW;IAC1B,IAAI;QACF,MAAMC,SAAS,IAAIC,IAAIF;QACvBC,OAAOE,QAAQ,GAAG;QAClBF,OAAOG,QAAQ,GAAG;QAClBH,OAAOI,MAAM,GAAG;QAChBJ,OAAOK,IAAI,GAAG;QACd,OAAOL,OAAOM,IAAI;IACpB,EAAE,OAAM;QACN,OAAO;IACT;AACF;AAEA,SAASC,kBACPC,YAAqC,EACrCC,SAAiB;IAMjBD,yBAAAA,mCAAAA,aAAcE,cAAc;IAC5B,MAAMC,aAAa,IAAIC;IACvB,MAAMC,gBAAgB,CAAC;IACvB,MAAMC,aAAa,IAAMH,WAAWI,KAAK,CAACP,yBAAAA,mCAAAA,aAAcQ,MAAM;IAC9DR,yBAAAA,mCAAAA,aAAcS,gBAAgB,CAAC,SAASH,YAAY;QAAEI,MAAM;IAAK;IACjE,IAAIV,yBAAAA,mCAAAA,aAAcW,OAAO,EAAEL;IAC3B,MAAMM,QAAQC,WAAW,IAAMV,WAAWI,KAAK,CAACF,gBAAgBJ;IAChE,OAAO;QACLE;QACAE;QACAS,SAAS;YACPC,aAAaH;YACbZ,yBAAAA,mCAAAA,aAAcgB,mBAAmB,CAAC,SAASV;QAC7C;IACF;AACF;AAEA,SAASW,gBAAgBC,QAAkB,EAAEf,UAA2B;IACtE,MAAMgB,eAAeD,SAASE,IAAI,IAAI,CAACF,SAASG,QAAQ,GAAGH,SAASE,IAAI,CAACE,MAAM,KAAKC;IACpFpB,WAAWI,KAAK;IAChB,gGAAgG;IAChG,IAAIY,cACF,KAAKA,aAAaK,IAAI,CACpB,IAAMD,WACN,IAAMA;AAEZ;AAEA,SAASE,MAAMC,EAAU,EAAEC,MAA+B;IACxDA,mBAAAA,6BAAAA,OAAQzB,cAAc;IACtB,IAAI,CAACyB,QAAQ,OAAO,IAAIC,QAAQ,CAACC,UAAYhB,WAAWgB,SAASH;IACjE,OAAO,IAAIE,QAAQ,CAACC,SAASC;QAC3B,IAAIlB;QACJ,MAAME,UAAU;YACd,IAAIF,OAAOG,aAAaH;YACxBe,OAAOX,mBAAmB,CAAC,SAASe;QACtC;QACA,MAAMA,UAAU;YACdjB;YACAgB,OAAOH,OAAOnB,MAAM;QACtB;QACAmB,OAAOlB,gBAAgB,CAAC,SAASsB,SAAS;YAAErB,MAAM;QAAK;QACvDE,QAAQC,WAAW;YACjBC;YACAe;QACF,GAAGH;QACH,IAAIC,OAAOhB,OAAO,EAAEoB;IACtB;AACF;AAEA,SAASC,cAAczC,GAAW,EAAE0C,KAAoB,EAAEC,QAAgB,EAAEjC,SAAiB,EAAEkC,QAAiB;IAC9G,MAAMC,WAAW9C,QAAQC;IACzB,IAAI4C,UAAU;QACZ,OAAO,IAAInD,WAAW,eAAe,GAAGoD,SAAS,iBAAiB,EAAEnC,UAAU,qBAAqB,EAAEgC,MAAM,UAAU,EAAEC,SAAS,CAAC,EAAEjD,aAAa,0CAA0C,CAAC;IAC7L;IACA,OAAO,IAAID,WAAW,eAAe,GAAGoD,SAAS,0BAA0B,EAAEH,MAAM,OAAO,EAAEC,SAAS,kDAAkD,CAAC;AAC1J;AAEA,SAASG,SAASnB,QAAkB,EAAEE,IAAiB,EAAEkB,UAAkB;IACzE,OAAO;QACLC,IAAIrB,SAASqB,EAAE;QACfC,QAAQtB,SAASsB,MAAM;QACvBC,SAASvB,SAASuB,OAAO;QACzBrB;QACA7B,KAAKD,QAAQ4B,SAAS3B,GAAG,IAAI+C;IAC/B;AACF;AAEA,2FAA2F;AAC3F,iFAAiF;AACjF,OAAO,eAAeI,eAAenD,GAAW,EAAEoD,IAAsB,EAAEC,UAA6B,CAAC,CAAC;QACrFA,2BACEA;IADpB,MAAM3C,aAAY2C,4BAAAA,QAAQC,gBAAgB,cAAxBD,uCAAAA,4BAA4BxD;IAC9C,MAAM0D,eAAcF,uBAAAA,QAAQE,WAAW,cAAnBF,kCAAAA,uBAAuBzD;IAC3C,IAAK,IAAI4D,UAAU,GAAGA,UAAU9D,cAAc8D,UAAW;YACvDH;SAAAA,kBAAAA,QAAQjB,MAAM,cAAdiB,sCAAAA,gBAAgB1C,cAAc;QAC9B,MAAM,EAAEC,UAAU,EAAEE,aAAa,EAAES,OAAO,EAAE,GAAGf,kBAAkB6C,QAAQjB,MAAM,EAAE1B;QACjF,IAAIiB;QACJ,IAAIe,QAAuB;QAC3B,IAAIe;QACJ,IAAI;YACF9B,WAAW,MAAM+B,MAAM1D,KAAK;gBAAE,GAAGoD,IAAI;gBAAEhB,QAAQxB,WAAWwB,MAAM;YAAC;YACjE,IAAI,CAACT,SAASqB,EAAE,EAAE;gBAChB,MAAMW,YAAYhC,SAASsB,MAAM,KAAK,OAAOtB,SAASsB,MAAM,IAAI;gBAChE,IAAIU,aAAaH,UAAU9D,eAAe,GAAG;oBAC3C,MAAMkE,aAAaC,OAAOlC,SAASuB,OAAO,CAACY,GAAG,CAAC;oBAC/CL,eAAeG,aAAa,IAAIG,KAAKC,GAAG,CAACJ,aAAa,MAAMjE,sBAAsB4D,cAAc,KAAKC;oBACrG9B,gBAAgBC,UAAUf;gBAC5B,OAAO;wBAELyC;oBADA3B,gBAAgBC,UAAUf;qBAC1ByC,mBAAAA,QAAQjB,MAAM,cAAdiB,uCAAAA,iBAAgB1C,cAAc;oBAC9B,OAAOmC,SAASnB,UAAU,IAAIsC,YAAY,IAAIjE;gBAChD;YACF,OAAO;oBAGLqD;gBAFAX,QAAQ;gBACR,MAAMb,OAAO,MAAMF,SAASuC,WAAW;iBACvCb,mBAAAA,QAAQjB,MAAM,cAAdiB,uCAAAA,iBAAgB1C,cAAc;gBAC9B,OAAOmC,SAASnB,UAAUE,MAAM7B;YAClC;QACF,EAAE,OAAM;YACN,IAAI;oBACFqD;iBAAAA,mBAAAA,QAAQjB,MAAM,cAAdiB,uCAAAA,iBAAgB1C,cAAc;YAChC,SAAU;gBACR,IAAIgB,UAAUD,gBAAgBC,UAAUf;YAC1C;YACA,MAAMgC,WAAWhC,WAAWwB,MAAM,CAACnB,MAAM,KAAKH;YAC9C,IAAI0C,YAAY9D,eAAe,GAAG,MAAM+C,cAAczC,KAAK0C,OAAOc,UAAU,GAAG9C,WAAWkC;YAC1Fa,eAAeF,cAAc,KAAKC;QACpC,SAAU;YACRjC;QACF;QACA,MAAMW,MAAMuB,yBAAAA,0BAAAA,eAAgBF,cAAc,KAAKC,SAASH,QAAQjB,MAAM;IACxE;IACA,MAAM,IAAI+B,MAAM;AAClB;AAEA,2FAA2F;AAC3F,0EAA0E;AAC1E,OAAO,eAAeC,eAAepE,GAAW,EAAEU,YAAY,IAAI;IAChE,MAAM,EAAEE,UAAU,EAAEW,OAAO,EAAE,GAAGf,kBAAkBwB,WAAWtB;IAC7D,IAAI;QACF,MAAMiB,WAAW,MAAM+B,MAAM1D,KAAK;YAAEoC,QAAQxB,WAAWwB,MAAM;QAAC;QAC9DV,gBAAgBC,UAAUf;QAC1B,OAAO;IACT,EAAE,OAAM;QACN,OAAO;IACT,SAAU;QACRW;IACF;AACF"}
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/embed/identity.ts"],"sourcesContent":["import { existsSync, mkdirSync, readFileSync, statSync } from 'node:fs';\nimport { homedir } from 'node:os';\nimport { join } from 'node:path';\nimport { writeFileAtomic } from '../lib/atomic-write.ts';\n\n// HF cache layout, no network: ~/.sense/models/models--<org>--<name>/{refs/main,\n// snapshots/<sha>/<files>}, files stored directly (hf_hub's blobs/symlinks layer skipped).\n\nexport const MODEL_FILES = ['model.safetensors', 'tokenizer.json'];\n// The pair, for messages that name what a local model directory must contain.\nexport const MODEL_FILENAMES = MODEL_FILES.join(' and ');\n\n// A Hugging Face repo id: one slash, HF's charset. Anything else is a path, used as one rather\n// than resolved against the HF cache.\nconst HF_MODEL_ID = /^[A-Za-z0-9][\\w.-]*\\/[A-Za-z0-9][\\w.-]*$/;\n\nexport function isDownloadable(model: string): boolean {\n return HF_MODEL_ID.test(model);\n}\n\n// The one production default; tests pass their own scratch root through the same parameter\n// rather than seeding fixture ids into this machine cache.\nexport const DEFAULT_MODELS_ROOT = join(homedir(), '.sense', 'models');\n\nfunction repoDir(model: string, root: string = DEFAULT_MODELS_ROOT): string {\n const [org, name] = model.split('/');\n return join(root, `models--${org}--${name}`);\n}\n\nexport function refsMainPath(model: string, root: string = DEFAULT_MODELS_ROOT): string {\n return join(repoDir(model, root), 'refs', 'main');\n}\n\n// What `main` last resolved to, read from disk -- no network. Undefined until some resolution\n// (a download or an explicit prefetch) has written it once.\nexport function readRef(model: string, root: string = DEFAULT_MODELS_ROOT): string | undefined {\n const p = refsMainPath(model, root);\n if (!existsSync(p)) return undefined;\n const sha = readFileSync(p, 'utf8').trim();\n return sha.length > 0 ? sha : undefined;\n}\n\n// A recorded ref pins until its snapshot directory is removed; nothing here re-resolves it.\nexport function writeRef(model: string, sha: string, root: string = DEFAULT_MODELS_ROOT): void {\n mkdirSync(join(repoDir(model, root), 'refs'), { recursive: true });\n writeFileAtomic(refsMainPath(model, root), sha);\n}\n\nexport function snapshotDir(model: string, sha: string, root: string = DEFAULT_MODELS_ROOT): string {\n return join(repoDir(model, root), 'snapshots', sha);\n}\n\n// Languages describe the repo, not one weight snapshot, so this lives beside refs/, not\n// inside a snapshot dir.\nexport function languagesPath(model: string, root: string = DEFAULT_MODELS_ROOT): string {\n return join(repoDir(model, root), 'languages.json');\n}\n\n// undefined means \"never resolved yet\"; an empty array is a resolved fact (checked, none\n// declared) and is cached the same way, so a languageless model is not re-fetched every run.\nexport function readLanguages(model: string, root: string = DEFAULT_MODELS_ROOT): string[] | undefined {\n const p = languagesPath(model, root);\n if (!existsSync(p)) return undefined;\n try {\n return JSON.parse(readFileSync(p, 'utf8'));\n } catch {\n return undefined;\n }\n}\n\nexport function writeLanguages(model: string, languages: string[], root: string = DEFAULT_MODELS_ROOT): void {\n mkdirSync(repoDir(model, root), { recursive: true });\n writeFileAtomic(languagesPath(model, root), JSON.stringify(languages));\n}\n\n// A local directory the caller manages, or the resolved snapshot once `main` is pinned; an HF\n// id with no recorded ref yet has no known snapshot, so this names the (fileless) repo dir.\nexport function modelDir(model: string, root: string = DEFAULT_MODELS_ROOT): string {\n if (!isDownloadable(model)) return model;\n const sha = readRef(model, root);\n return sha ? snapshotDir(model, sha, root) : repoDir(model, root);\n}\n\n// Both files, so an interrupted download reads as absent and the next one resumes it.\nexport function hasModelFiles(model: string, root: string = DEFAULT_MODELS_ROOT): boolean {\n const dir = modelDir(model, root);\n return MODEL_FILES.every((file) => existsSync(join(dir, file)));\n}\n\n// Weight identity for the signature, no network: a HF id's resolved sha, or the two\n// files' size+mtime for a local path. Undefined when nothing is known yet.\nexport function modelIdentity(model: string, root: string = DEFAULT_MODELS_ROOT): string | undefined {\n if (isDownloadable(model)) return readRef(model, root);\n const dir = modelDir(model, root);\n const parts: string[] = [];\n for (const file of MODEL_FILES) {\n try {\n const st = statSync(join(dir, file));\n parts.push(`${st.size}:${Math.trunc(st.mtimeMs)}`);\n } catch {\n return undefined;\n }\n }\n return parts.join(',');\n}\n"],"names":["existsSync","mkdirSync","readFileSync","statSync","homedir","join","writeFileAtomic","MODEL_FILES","MODEL_FILENAMES","HF_MODEL_ID","isDownloadable","model","test","DEFAULT_MODELS_ROOT","repoDir","root","org","name","split","refsMainPath","readRef","p","undefined","sha","trim","length","writeRef","recursive","snapshotDir","languagesPath","readLanguages","JSON","parse","writeLanguages","languages","stringify","modelDir","hasModelFiles","dir","every","file","modelIdentity","parts","st","push","size","Math","trunc","mtimeMs"],"mappings":"AAAA,SAASA,UAAU,EAAEC,SAAS,EAAEC,YAAY,EAAEC,QAAQ,QAAQ,UAAU;AACxE,SAASC,OAAO,QAAQ,UAAU;AAClC,SAASC,IAAI,QAAQ,YAAY;AACjC,SAASC,eAAe,QAAQ,yBAAyB;AAEzD,iFAAiF;AACjF,2FAA2F;AAE3F,OAAO,MAAMC,cAAc;IAAC;IAAqB;CAAiB,CAAC;AACnE,8EAA8E;AAC9E,OAAO,MAAMC,kBAAkBD,YAAYF,IAAI,CAAC,SAAS;AAEzD,+FAA+F;AAC/F,sCAAsC;AACtC,MAAMI,cAAc;AAEpB,OAAO,SAASC,eAAeC,KAAa;IAC1C,OAAOF,YAAYG,IAAI,CAACD;AAC1B;AAEA,2FAA2F;AAC3F,2DAA2D;AAC3D,OAAO,MAAME,sBAAsBR,KAAKD,WAAW,UAAU,UAAU;AAEvE,SAASU,QAAQH,KAAa,EAAEI,OAAeF,mBAAmB;IAChE,MAAM,CAACG,KAAKC,KAAK,GAAGN,MAAMO,KAAK,CAAC;IAChC,OAAOb,KAAKU,MAAM,CAAC,QAAQ,EAAEC,IAAI,EAAE,EAAEC,MAAM;AAC7C;AAEA,OAAO,SAASE,aAAaR,KAAa,EAAEI,OAAeF,mBAAmB;IAC5E,OAAOR,KAAKS,QAAQH,OAAOI,OAAO,QAAQ;AAC5C;AAEA,8FAA8F;AAC9F,4DAA4D;AAC5D,OAAO,SAASK,QAAQT,KAAa,EAAEI,OAAeF,mBAAmB;IACvE,MAAMQ,IAAIF,aAAaR,OAAOI;IAC9B,IAAI,CAACf,WAAWqB,IAAI,OAAOC;IAC3B,MAAMC,MAAMrB,aAAamB,GAAG,QAAQG,IAAI;IACxC,OAAOD,IAAIE,MAAM,GAAG,IAAIF,MAAMD;AAChC;AAEA,4FAA4F;AAC5F,OAAO,SAASI,SAASf,KAAa,EAAEY,GAAW,EAAER,OAAeF,mBAAmB;IACrFZ,UAAUI,KAAKS,QAAQH,OAAOI,OAAO,SAAS;QAAEY,WAAW;IAAK;IAChErB,gBAAgBa,aAAaR,OAAOI,OAAOQ;AAC7C;AAEA,OAAO,SAASK,YAAYjB,KAAa,EAAEY,GAAW,EAAER,OAAeF,mBAAmB;IACxF,OAAOR,KAAKS,QAAQH,OAAOI,OAAO,aAAaQ;AACjD;AAEA,wFAAwF;AACxF,yBAAyB;AACzB,OAAO,SAASM,cAAclB,KAAa,EAAEI,OAAeF,mBAAmB;IAC7E,OAAOR,KAAKS,QAAQH,OAAOI,OAAO;AACpC;AAEA,yFAAyF;AACzF,6FAA6F;AAC7F,OAAO,SAASe,cAAcnB,KAAa,EAAEI,OAAeF,mBAAmB;IAC7E,MAAMQ,IAAIQ,cAAclB,OAAOI;IAC/B,IAAI,CAACf,WAAWqB,IAAI,OAAOC;IAC3B,IAAI;QACF,OAAOS,KAAKC,KAAK,CAAC9B,aAAamB,GAAG;IACpC,EAAE,OAAM;QACN,OAAOC;IACT;AACF;AAEA,OAAO,SAASW,eAAetB,KAAa,EAAEuB,SAAmB,EAAEnB,OAAeF,mBAAmB;IACnGZ,UAAUa,QAAQH,OAAOI,OAAO;QAAEY,WAAW;IAAK;IAClDrB,gBAAgBuB,cAAclB,OAAOI,OAAOgB,KAAKI,SAAS,CAACD;AAC7D;AAEA,8FAA8F;AAC9F,4FAA4F;AAC5F,OAAO,SAASE,SAASzB,KAAa,EAAEI,OAAeF,mBAAmB;IACxE,IAAI,CAACH,eAAeC,QAAQ,OAAOA;IACnC,MAAMY,MAAMH,QAAQT,OAAOI;IAC3B,OAAOQ,MAAMK,YAAYjB,OAAOY,KAAKR,QAAQD,QAAQH,OAAOI;AAC9D;AAEA,sFAAsF;AACtF,OAAO,SAASsB,cAAc1B,KAAa,EAAEI,OAAeF,mBAAmB;IAC7E,MAAMyB,MAAMF,SAASzB,OAAOI;IAC5B,OAAOR,YAAYgC,KAAK,CAAC,CAACC,OAASxC,WAAWK,KAAKiC,KAAKE;AAC1D;AAEA,oFAAoF;AACpF,2EAA2E;AAC3E,OAAO,SAASC,cAAc9B,KAAa,EAAEI,OAAeF,mBAAmB;IAC7E,IAAIH,eAAeC,QAAQ,OAAOS,QAAQT,OAAOI;IACjD,MAAMuB,MAAMF,SAASzB,OAAOI;IAC5B,MAAM2B,QAAkB,EAAE;IAC1B,KAAK,MAAMF,QAAQjC,YAAa;QAC9B,IAAI;YACF,MAAMoC,KAAKxC,SAASE,KAAKiC,KAAKE;YAC9BE,MAAME,IAAI,CAAC,GAAGD,GAAGE,IAAI,CAAC,CAAC,EAAEC,KAAKC,KAAK,CAACJ,GAAGK,OAAO,GAAG;QACnD,EAAE,OAAM;YACN,OAAO1B;QACT;IACF;IACA,OAAOoB,MAAMrC,IAAI,CAAC;AACpB"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/open-source/ai/sensemaking/src/embed/identity.ts"],"sourcesContent":["import { existsSync, mkdirSync, readFileSync, statSync } from 'node:fs';\nimport { homedir } from 'node:os';\nimport { join } from 'node:path';\nimport { writeFileAtomic } from '../lib/atomic-write.ts';\n\n// HF cache layout, no network: ~/.sense/models/models--<org>--<name>/{refs/main,\n// snapshots/<sha>/<files>}, files stored directly (hf_hub's blobs/symlinks layer skipped).\n\nexport const MODEL_FILES = ['model.safetensors', 'tokenizer.json'];\n// The pair, for messages that name what a local model directory must contain.\nexport const MODEL_FILENAMES = MODEL_FILES.join(' and ');\n\n// A Hugging Face repo id: one slash, HF's charset. Anything else is a path, used as one rather\n// than resolved against the HF cache.\nconst HF_MODEL_ID = /^[A-Za-z0-9][\\w.-]*\\/[A-Za-z0-9][\\w.-]*$/;\n\nexport function isDownloadable(model: string): boolean {\n return HF_MODEL_ID.test(model);\n}\n\n// The one production default; tests pass their own scratch root through the same parameter\n// rather than seeding fixture ids into this machine cache.\nexport const DEFAULT_MODELS_ROOT = join(homedir(), '.sense', 'models');\n\nfunction repoDir(model: string, root: string = DEFAULT_MODELS_ROOT): string {\n const [org, name] = model.split('/');\n return join(root, `models--${org}--${name}`);\n}\n\nexport function refsMainPath(model: string, root: string = DEFAULT_MODELS_ROOT): string {\n return join(repoDir(model, root), 'refs', 'main');\n}\n\n// What `main` last resolved to, read from disk -- no network. Undefined until some resolution\n// (a download or an explicit prefetch) has written it once.\nexport function readRef(model: string, root: string = DEFAULT_MODELS_ROOT): string | undefined {\n const p = refsMainPath(model, root);\n if (!existsSync(p)) return undefined;\n const sha = readFileSync(p, 'utf8').trim();\n return sha.length > 0 ? sha : undefined;\n}\n\n// A recorded ref pins until its snapshot directory is removed; nothing here re-resolves it.\nexport function writeRef(model: string, sha: string, root: string = DEFAULT_MODELS_ROOT): void {\n mkdirSync(join(repoDir(model, root), 'refs'), { recursive: true });\n writeFileAtomic(refsMainPath(model, root), sha);\n}\n\nexport function snapshotDir(model: string, sha: string, root: string = DEFAULT_MODELS_ROOT): string {\n return join(repoDir(model, root), 'snapshots', sha);\n}\n\n// Languages describe the repo, not one weight snapshot, so this lives beside refs/, not\n// inside a snapshot dir.\nexport function languagesPath(model: string, root: string = DEFAULT_MODELS_ROOT): string {\n return join(repoDir(model, root), 'languages.json');\n}\n\n// undefined means \"never resolved yet\"; an empty array is a resolved fact (checked, none\n// declared) and is cached the same way, so a languageless model is not re-fetched every run.\nexport function readLanguages(model: string, root: string = DEFAULT_MODELS_ROOT): string[] | undefined {\n const p = languagesPath(model, root);\n if (!existsSync(p)) return undefined;\n try {\n return JSON.parse(readFileSync(p, 'utf8'));\n } catch {\n return undefined;\n }\n}\n\nexport function writeLanguages(model: string, languages: string[], root: string = DEFAULT_MODELS_ROOT): void {\n mkdirSync(repoDir(model, root), { recursive: true });\n writeFileAtomic(languagesPath(model, root), JSON.stringify(languages));\n}\n\n// A local directory the caller manages, or the resolved snapshot once `main` is pinned; an HF\n// id with no recorded ref yet has no known snapshot, so this names the (fileless) repo dir.\nexport function modelDir(model: string, root: string = DEFAULT_MODELS_ROOT): string {\n if (!isDownloadable(model)) return model;\n const sha = readRef(model, root);\n return sha ? snapshotDir(model, sha, root) : repoDir(model, root);\n}\n\n// Both files, so an interrupted download reads as absent and the next one resumes it.\nexport function hasModelFiles(model: string, root: string = DEFAULT_MODELS_ROOT): boolean {\n const dir = modelDir(model, root);\n return MODEL_FILES.every((file) => existsSync(join(dir, file)));\n}\n\n// Weight identity for the signature, no network: a HF id's resolved sha, or the two\n// files' size+mtime for a local path. Undefined when nothing is known yet.\nexport function modelIdentity(model: string, root: string = DEFAULT_MODELS_ROOT): string | undefined {\n if (isDownloadable(model)) return readRef(model, root);\n const dir = modelDir(model, root);\n const parts: string[] = [];\n for (const file of MODEL_FILES) {\n try {\n const st = statSync(join(dir, file));\n parts.push(`${st.size}:${Math.trunc(st.mtimeMs)}`);\n } catch {\n return undefined;\n }\n }\n return parts.join(',');\n}\n"],"names":["existsSync","mkdirSync","readFileSync","statSync","homedir","join","writeFileAtomic","MODEL_FILES","MODEL_FILENAMES","HF_MODEL_ID","isDownloadable","model","test","DEFAULT_MODELS_ROOT","repoDir","root","org","name","split","refsMainPath","readRef","p","undefined","sha","trim","length","writeRef","recursive","snapshotDir","languagesPath","readLanguages","JSON","parse","writeLanguages","languages","stringify","modelDir","hasModelFiles","dir","every","file","modelIdentity","parts","st","push","size","Math","trunc","mtimeMs"],"mappings":"AAAA,SAASA,UAAU,EAAEC,SAAS,EAAEC,YAAY,EAAEC,QAAQ,QAAQ,UAAU;AACxE,SAASC,OAAO,QAAQ,UAAU;AAClC,SAASC,IAAI,QAAQ,YAAY;AACjC,SAASC,eAAe,QAAQ,yBAAyB;AAEzD,iFAAiF;AACjF,2FAA2F;AAE3F,OAAO,MAAMC,cAAc;IAAC;IAAqB;CAAiB,CAAC;AACnE,8EAA8E;AAC9E,OAAO,MAAMC,kBAAkBD,YAAYF,IAAI,CAAC,SAAS;AAEzD,+FAA+F;AAC/F,sCAAsC;AACtC,MAAMI,cAAc;AAEpB,OAAO,SAASC,eAAeC,KAAa;IAC1C,OAAOF,YAAYG,IAAI,CAACD;AAC1B;AAEA,2FAA2F;AAC3F,2DAA2D;AAC3D,OAAO,MAAME,sBAAsBR,KAAKD,WAAW,UAAU,UAAU;AAEvE,SAASU,QAAQH,KAAa,EAAEI,OAAeF,mBAAmB;IAChE,MAAM,CAACG,KAAKC,KAAK,GAAGN,MAAMO,KAAK,CAAC;IAChC,OAAOb,KAAKU,MAAM,CAAC,QAAQ,EAAEC,IAAI,EAAE,EAAEC,MAAM;AAC7C;AAEA,OAAO,SAASE,aAAaR,KAAa,EAAEI,OAAeF,mBAAmB;IAC5E,OAAOR,KAAKS,QAAQH,OAAOI,OAAO,QAAQ;AAC5C;AAEA,8FAA8F;AAC9F,4DAA4D;AAC5D,OAAO,SAASK,QAAQT,KAAa,EAAEI,OAAeF,mBAAmB;IACvE,MAAMQ,IAAIF,aAAaR,OAAOI;IAC9B,IAAI,CAACf,WAAWqB,IAAI,OAAOC;IAC3B,MAAMC,MAAMrB,aAAamB,GAAG,QAAQG,IAAI;IACxC,OAAOD,IAAIE,MAAM,GAAG,IAAIF,MAAMD;AAChC;AAEA,4FAA4F;AAC5F,OAAO,SAASI,SAASf,KAAa,EAAEY,GAAW,EAAER,OAAeF,mBAAmB;IACrFZ,UAAUI,KAAKS,QAAQH,OAAOI,OAAO,SAAS;QAAEY,WAAW;IAAK;IAChErB,gBAAgBa,aAAaR,OAAOI,OAAOQ;AAC7C;AAEA,OAAO,SAASK,YAAYjB,KAAa,EAAEY,GAAW,EAAER,OAAeF,mBAAmB;IACxF,OAAOR,KAAKS,QAAQH,OAAOI,OAAO,aAAaQ;AACjD;AAEA,wFAAwF;AACxF,yBAAyB;AACzB,OAAO,SAASM,cAAclB,KAAa,EAAEI,OAAeF,mBAAmB;IAC7E,OAAOR,KAAKS,QAAQH,OAAOI,OAAO;AACpC;AAEA,yFAAyF;AACzF,6FAA6F;AAC7F,OAAO,SAASe,cAAcnB,KAAa,EAAEI,OAAeF,mBAAmB;IAC7E,MAAMQ,IAAIQ,cAAclB,OAAOI;IAC/B,IAAI,CAACf,WAAWqB,IAAI,OAAOC;IAC3B,IAAI;QACF,OAAOS,KAAKC,KAAK,CAAC9B,aAAamB,GAAG;IACpC,EAAE,OAAM;QACN,OAAOC;IACT;AACF;AAEA,OAAO,SAASW,eAAetB,KAAa,EAAEuB,SAAmB,EAAEnB,OAAeF,mBAAmB;IACnGZ,UAAUa,QAAQH,OAAOI,OAAO;QAAEY,WAAW;IAAK;IAClDrB,gBAAgBuB,cAAclB,OAAOI,OAAOgB,KAAKI,SAAS,CAACD;AAC7D;AAEA,8FAA8F;AAC9F,4FAA4F;AAC5F,OAAO,SAASE,SAASzB,KAAa,EAAEI,OAAeF,mBAAmB;IACxE,IAAI,CAACH,eAAeC,QAAQ,OAAOA;IACnC,MAAMY,MAAMH,QAAQT,OAAOI;IAC3B,OAAOQ,MAAMK,YAAYjB,OAAOY,KAAKR,QAAQD,QAAQH,OAAOI;AAC9D;AAEA,sFAAsF;AACtF,OAAO,SAASsB,cAAc1B,KAAa,EAAEI,OAAeF,mBAAmB;IAC7E,MAAMyB,MAAMF,SAASzB,OAAOI;IAC5B,OAAOR,YAAYgC,KAAK,CAAC,CAACC,OAASxC,WAAWK,KAAKiC,KAAKE;AAC1D;AAEA,oFAAoF;AACpF,2EAA2E;AAC3E,OAAO,SAASC,cAAc9B,KAAa,EAAEI,OAAeF,mBAAmB;IAC7E,IAAIH,eAAeC,QAAQ,OAAOS,QAAQT,OAAOI;IACjD,MAAMuB,MAAMF,SAASzB,OAAOI;IAC5B,MAAM2B,QAAkB,EAAE;IAC1B,KAAK,MAAMF,QAAQjC,YAAa;QAC9B,IAAI;YACF,MAAMoC,KAAKxC,SAASE,KAAKiC,KAAKE;YAC9BE,MAAME,IAAI,CAAC,GAAGD,GAAGE,IAAI,CAAC,CAAC,EAAEC,KAAKC,KAAK,CAACJ,GAAGK,OAAO,GAAG;QACnD,EAAE,OAAM;YACN,OAAO1B;QACT;IACF;IACA,OAAOoB,MAAMrC,IAAI,CAAC;AACpB"}
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/embed/langfit.ts"],"sourcesContent":["import Module from 'node:module';\nimport { SenseError } from '../errors.ts';\nimport type { Store } from '../store/types.ts';\nimport { languageDistribution, saveLanguageDistribution } from './distribution.ts';\nimport { toIso3 } from './languages.ts';\nimport type { EmbedProvider } from './types.ts';\n\n// franc's own reliability floor is a handful of characters; this is a cheap upper bound, not a\n// per-chunk minimum -- a short chunk is simply more likely to come back 'und' (unclassified).\nconst SAMPLE_CHARS = 300;\n// A majority computed over a handful of chunks is noise, not a corpus signal; the decision rule\n// never fires below this many classified chunks, declared languages or not.\nconst MIN_CLASSIFIED = 10;\n\n// franc-min is a 148K static import; a top-level import would put it on every command's module\n// graph regardless of whether the semantic path ran, so it's deferred to a synchronous require.\nconst _require = typeof require === 'undefined' ? Module.createRequire(import.meta.url) : require;\n\nfunction classify(franc: typeof import('franc-min')['franc'], text: string): string | undefined {\n const code = franc(text.slice(0, SAMPLE_CHARS));\n return code === 'und' ? undefined : code;\n}\n\n// Merges chunk texts into the persisted distribution; throws (without persisting) if the model\n// declares languages and a clear majority classified so far is not among them.\nexport async function checkLanguageFit(store: Store, provider: EmbedProvider, texts: string[]): Promise<void> {\n if (texts.length === 0) return;\n const { franc } = _require('franc-min') as typeof import('franc-min');\n const persisted = (await languageDistribution(store)) ?? {};\n const merged = { ...persisted };\n for (const text of texts) {\n const code = classify(franc, text);\n if (code) merged[code] = (merged[code] ?? 0) + 1;\n }\n\n const total = Object.values(merged).reduce((a, b) => a + b, 0);\n if (provider.languages && provider.languages.length > 0 && total >= MIN_CLASSIFIED) {\n const declared = new Set(provider.languages.map(toIso3));\n const [majorityLang, majorityCount] = Object.entries(merged).sort((a, b) => b[1] - a[1])[0];\n if (majorityCount / total >= 0.5 && !declared.has(majorityLang)) {\n const pct = Math.round((majorityCount / total) * 100);\n throw new SenseError(\n 'EMBED_MODEL_MISMATCH',\n `embed model ${provider.id} declares languages [${provider.languages.join(', ')}], but ${pct}% of this tree's classified text is \"${majorityLang}\" (ISO 639-3), which is not among them; choose a model for this tree's languages: \\`sense init --model ...\\`, see the sense-setup skill or INTEGRATIONS.md`\n );\n }\n }\n await saveLanguageDistribution(store, merged);\n}\n"],"names":["Module","SenseError","languageDistribution","saveLanguageDistribution","toIso3","SAMPLE_CHARS","MIN_CLASSIFIED","_require","require","createRequire","url","classify","franc","text","code","slice","undefined","checkLanguageFit","store","provider","texts","length","persisted","merged","total","Object","values","reduce","a","b","languages","declared","Set","map","majorityLang","majorityCount","entries","sort","has","pct","Math","round","id","join"],"mappings":"AAAA,OAAOA,YAAY,cAAc;AACjC,SAASC,UAAU,QAAQ,eAAe;AAE1C,SAASC,oBAAoB,EAAEC,wBAAwB,QAAQ,oBAAoB;AACnF,SAASC,MAAM,QAAQ,iBAAiB;AAGxC,+FAA+F;AAC/F,8FAA8F;AAC9F,MAAMC,eAAe;AACrB,gGAAgG;AAChG,4EAA4E;AAC5E,MAAMC,iBAAiB;AAEvB,+FAA+F;AAC/F,gGAAgG;AAChG,MAAMC,WAAW,OAAOC,YAAY,cAAcR,OAAOS,aAAa,CAAC,YAAYC,GAAG,IAAIF;AAE1F,SAASG,SAASC,KAA0C,EAAEC,IAAY;IACxE,MAAMC,OAAOF,MAAMC,KAAKE,KAAK,CAAC,GAAGV;IACjC,OAAOS,SAAS,QAAQE,YAAYF;AACtC;AAEA,+FAA+F;AAC/F,+EAA+E;AAC/E,OAAO,eAAeG,iBAAiBC,KAAY,EAAEC,QAAuB,EAAEC,KAAe;QAGxE;IAFnB,IAAIA,MAAMC,MAAM,KAAK,GAAG;IACxB,MAAM,EAAET,KAAK,EAAE,GAAGL,SAAS;IAC3B,MAAMe,aAAa,OAAA,MAAMpB,qBAAqBgB,oBAA3B,kBAAA,OAAsC,CAAC;IAC1D,MAAMK,SAAS;QAAE,GAAGD,SAAS;IAAC;IAC9B,KAAK,MAAMT,QAAQO,MAAO;YAEEG;QAD1B,MAAMT,OAAOH,SAASC,OAAOC;QAC7B,IAAIC,MAAMS,MAAM,CAACT,KAAK,GAAG,EAACS,eAAAA,MAAM,CAACT,KAAK,cAAZS,0BAAAA,eAAgB,KAAK;IACjD;IAEA,MAAMC,QAAQC,OAAOC,MAAM,CAACH,QAAQI,MAAM,CAAC,CAACC,GAAGC,IAAMD,IAAIC,GAAG;IAC5D,IAAIV,SAASW,SAAS,IAAIX,SAASW,SAAS,CAACT,MAAM,GAAG,KAAKG,SAASlB,gBAAgB;QAClF,MAAMyB,WAAW,IAAIC,IAAIb,SAASW,SAAS,CAACG,GAAG,CAAC7B;QAChD,MAAM,CAAC8B,cAAcC,cAAc,GAAGV,OAAOW,OAAO,CAACb,QAAQc,IAAI,CAAC,CAACT,GAAGC,IAAMA,CAAC,CAAC,EAAE,GAAGD,CAAC,CAAC,EAAE,CAAC,CAAC,EAAE;QAC3F,IAAIO,gBAAgBX,SAAS,OAAO,CAACO,SAASO,GAAG,CAACJ,eAAe;YAC/D,MAAMK,MAAMC,KAAKC,KAAK,CAAC,AAACN,gBAAgBX,QAAS;YACjD,MAAM,IAAIvB,WACR,wBACA,CAAC,YAAY,EAAEkB,SAASuB,EAAE,CAAC,qBAAqB,EAAEvB,SAASW,SAAS,CAACa,IAAI,CAAC,MAAM,OAAO,EAAEJ,IAAI,qCAAqC,EAAEL,aAAa,0JAA0J,CAAC;QAEhT;IACF;IACA,MAAM/B,yBAAyBe,OAAOK;AACxC"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/open-source/ai/sensemaking/src/embed/langfit.ts"],"sourcesContent":["import Module from 'node:module';\nimport { SenseError } from '../errors.ts';\nimport type { Store } from '../store/types.ts';\nimport { languageDistribution, saveLanguageDistribution } from './distribution.ts';\nimport { toIso3 } from './languages.ts';\nimport type { EmbedProvider } from './types.ts';\n\n// franc's own reliability floor is a handful of characters; this is a cheap upper bound, not a\n// per-chunk minimum -- a short chunk is simply more likely to come back 'und' (unclassified).\nconst SAMPLE_CHARS = 300;\n// A majority computed over a handful of chunks is noise, not a corpus signal; the decision rule\n// never fires below this many classified chunks, declared languages or not.\nconst MIN_CLASSIFIED = 10;\n\n// franc-min is a 148K static import; a top-level import would put it on every command's module\n// graph regardless of whether the semantic path ran, so it's deferred to a synchronous require.\nconst _require = typeof require === 'undefined' ? Module.createRequire(import.meta.url) : require;\n\nfunction classify(franc: typeof import('franc-min')['franc'], text: string): string | undefined {\n const code = franc(text.slice(0, SAMPLE_CHARS));\n return code === 'und' ? undefined : code;\n}\n\n// Merges chunk texts into the persisted distribution; throws (without persisting) if the model\n// declares languages and a clear majority classified so far is not among them.\nexport async function checkLanguageFit(store: Store, provider: EmbedProvider, texts: string[]): Promise<void> {\n if (texts.length === 0) return;\n const { franc } = _require('franc-min') as typeof import('franc-min');\n const persisted = (await languageDistribution(store)) ?? {};\n const merged = { ...persisted };\n for (const text of texts) {\n const code = classify(franc, text);\n if (code) merged[code] = (merged[code] ?? 0) + 1;\n }\n\n const total = Object.values(merged).reduce((a, b) => a + b, 0);\n if (provider.languages && provider.languages.length > 0 && total >= MIN_CLASSIFIED) {\n const declared = new Set(provider.languages.map(toIso3));\n const [majorityLang, majorityCount] = Object.entries(merged).sort((a, b) => b[1] - a[1])[0];\n if (majorityCount / total >= 0.5 && !declared.has(majorityLang)) {\n const pct = Math.round((majorityCount / total) * 100);\n throw new SenseError(\n 'EMBED_MODEL_MISMATCH',\n `embed model ${provider.id} declares languages [${provider.languages.join(', ')}], but ${pct}% of this tree's classified text is \"${majorityLang}\" (ISO 639-3), which is not among them; choose a model for this tree's languages: \\`sense init --model ...\\`, see the sense-setup skill or INTEGRATIONS.md`\n );\n }\n }\n await saveLanguageDistribution(store, merged);\n}\n"],"names":["Module","SenseError","languageDistribution","saveLanguageDistribution","toIso3","SAMPLE_CHARS","MIN_CLASSIFIED","_require","require","createRequire","url","classify","franc","text","code","slice","undefined","checkLanguageFit","store","provider","texts","length","persisted","merged","total","Object","values","reduce","a","b","languages","declared","Set","map","majorityLang","majorityCount","entries","sort","has","pct","Math","round","id","join"],"mappings":"AAAA,OAAOA,YAAY,cAAc;AACjC,SAASC,UAAU,QAAQ,eAAe;AAE1C,SAASC,oBAAoB,EAAEC,wBAAwB,QAAQ,oBAAoB;AACnF,SAASC,MAAM,QAAQ,iBAAiB;AAGxC,+FAA+F;AAC/F,8FAA8F;AAC9F,MAAMC,eAAe;AACrB,gGAAgG;AAChG,4EAA4E;AAC5E,MAAMC,iBAAiB;AAEvB,+FAA+F;AAC/F,gGAAgG;AAChG,MAAMC,WAAW,OAAOC,YAAY,cAAcR,OAAOS,aAAa,CAAC,YAAYC,GAAG,IAAIF;AAE1F,SAASG,SAASC,KAA0C,EAAEC,IAAY;IACxE,MAAMC,OAAOF,MAAMC,KAAKE,KAAK,CAAC,GAAGV;IACjC,OAAOS,SAAS,QAAQE,YAAYF;AACtC;AAEA,+FAA+F;AAC/F,+EAA+E;AAC/E,OAAO,eAAeG,iBAAiBC,KAAY,EAAEC,QAAuB,EAAEC,KAAe;QAGxE;IAFnB,IAAIA,MAAMC,MAAM,KAAK,GAAG;IACxB,MAAM,EAAET,KAAK,EAAE,GAAGL,SAAS;IAC3B,MAAMe,aAAa,OAAA,MAAMpB,qBAAqBgB,oBAA3B,kBAAA,OAAsC,CAAC;IAC1D,MAAMK,SAAS;QAAE,GAAGD,SAAS;IAAC;IAC9B,KAAK,MAAMT,QAAQO,MAAO;YAEEG;QAD1B,MAAMT,OAAOH,SAASC,OAAOC;QAC7B,IAAIC,MAAMS,MAAM,CAACT,KAAK,GAAG,EAACS,eAAAA,MAAM,CAACT,KAAK,cAAZS,0BAAAA,eAAgB,KAAK;IACjD;IAEA,MAAMC,QAAQC,OAAOC,MAAM,CAACH,QAAQI,MAAM,CAAC,CAACC,GAAGC,IAAMD,IAAIC,GAAG;IAC5D,IAAIV,SAASW,SAAS,IAAIX,SAASW,SAAS,CAACT,MAAM,GAAG,KAAKG,SAASlB,gBAAgB;QAClF,MAAMyB,WAAW,IAAIC,IAAIb,SAASW,SAAS,CAACG,GAAG,CAAC7B;QAChD,MAAM,CAAC8B,cAAcC,cAAc,GAAGV,OAAOW,OAAO,CAACb,QAAQc,IAAI,CAAC,CAACT,GAAGC,IAAMA,CAAC,CAAC,EAAE,GAAGD,CAAC,CAAC,EAAE,CAAC,CAAC,EAAE;QAC3F,IAAIO,gBAAgBX,SAAS,OAAO,CAACO,SAASO,GAAG,CAACJ,eAAe;YAC/D,MAAMK,MAAMC,KAAKC,KAAK,CAAC,AAACN,gBAAgBX,QAAS;YACjD,MAAM,IAAIvB,WACR,wBACA,CAAC,YAAY,EAAEkB,SAASuB,EAAE,CAAC,qBAAqB,EAAEvB,SAASW,SAAS,CAACa,IAAI,CAAC,MAAM,OAAO,EAAEJ,IAAI,qCAAqC,EAAEL,aAAa,0JAA0J,CAAC;QAEhT;IACF;IACA,MAAM/B,yBAAyBe,OAAOK;AACxC"}
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/embed/languages.ts"],"sourcesContent":["// HF model cards declare languages as ISO 639-1 (e.g. \"zh\"); franc-min reports ISO 639-3 (e.g.\n// \"cmn\"). Limited to franc-min's own 82-language table (https://github.com/wooorm/franc).\nconst ISO_639_1_TO_3: Record<string, string> = {\n ar: 'arb',\n am: 'amh',\n az: 'azj',\n be: 'bel',\n bg: 'bul',\n bn: 'ben',\n bs: 'bos',\n cs: 'ces',\n de: 'deu',\n el: 'ell',\n en: 'eng',\n es: 'spa',\n fa: 'pes',\n fr: 'fra',\n gu: 'guj',\n ha: 'hau',\n hi: 'hin',\n hr: 'hrv',\n hu: 'hun',\n id: 'ind',\n ig: 'ibo',\n it: 'ita',\n ja: 'jpn',\n jv: 'jav',\n kk: 'kaz',\n kn: 'kan',\n ko: 'kor',\n ml: 'mal',\n mr: 'mar',\n ms: 'zlm',\n my: 'mya',\n ne: 'npi',\n nl: 'nld',\n pa: 'pan',\n pl: 'pol',\n pt: 'por',\n ro: 'ron',\n ru: 'rus',\n rw: 'kin',\n si: 'sin',\n so: 'som',\n sr: 'srp',\n su: 'sun',\n sv: 'swe',\n sw: 'swh',\n ta: 'tam',\n te: 'tel',\n th: 'tha',\n tl: 'tgl',\n tr: 'tur',\n uk: 'ukr',\n ur: 'urd',\n uz: 'uzn',\n vi: 'vie',\n yo: 'yor',\n zh: 'cmn',\n zu: 'zul',\n};\n\nexport function isKnownLanguageTag(tag: string): boolean {\n return tag in ISO_639_1_TO_3;\n}\n\n// Normalizes a declared or detected language code to ISO 639-3 for comparison; strips a\n// region/script subtag (\"zh-cn\" -> \"zh\") first. Codes already in 639-3 form pass through.\nexport function toIso3(code: string): string {\n const base = code.toLowerCase().split('-')[0];\n return ISO_639_1_TO_3[base] ?? base;\n}\n\n// Languages this project measured for models whose cards declare none, so the fit check is\n// not silently off for the shipped default; asserted here because it was measured, not attested.\nexport const MEASURED_MODEL_LANGUAGES: Record<string, string[]> = {\n 'minishlab/potion-retrieval-32M': ['en'], // benchmark/reports/2026-08-27-embedding-model-selection.md\n};\n"],"names":["ISO_639_1_TO_3","ar","am","az","be","bg","bn","bs","cs","de","el","en","es","fa","fr","gu","ha","hi","hr","hu","id","ig","it","ja","jv","kk","kn","ko","ml","mr","ms","my","ne","nl","pa","pl","pt","ro","ru","rw","si","so","sr","su","sv","sw","ta","te","th","tl","tr","uk","ur","uz","vi","yo","zh","zu","isKnownLanguageTag","tag","toIso3","code","base","toLowerCase","split","MEASURED_MODEL_LANGUAGES"],"mappings":"AAAA,+FAA+F;AAC/F,0FAA0F;AAC1F,MAAMA,iBAAyC;IAC7CC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;AACN;AAEA,OAAO,SAASC,mBAAmBC,GAAW;IAC5C,OAAOA,OAAO3D;AAChB;AAEA,wFAAwF;AACxF,0FAA0F;AAC1F,OAAO,SAAS4D,OAAOC,IAAY;QAE1B7D;IADP,MAAM8D,OAAOD,KAAKE,WAAW,GAAGC,KAAK,CAAC,IAAI,CAAC,EAAE;IAC7C,QAAOhE,uBAAAA,cAAc,CAAC8D,KAAK,cAApB9D,kCAAAA,uBAAwB8D;AACjC;AAEA,2FAA2F;AAC3F,iGAAiG;AACjG,OAAO,MAAMG,2BAAqD;IAChE,kCAAkC;QAAC;KAAK;AAC1C,EAAE"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/open-source/ai/sensemaking/src/embed/languages.ts"],"sourcesContent":["// HF model cards declare languages as ISO 639-1 (e.g. \"zh\"); franc-min reports ISO 639-3 (e.g.\n// \"cmn\"). Limited to franc-min's own 82-language table (https://github.com/wooorm/franc).\nconst ISO_639_1_TO_3: Record<string, string> = {\n ar: 'arb',\n am: 'amh',\n az: 'azj',\n be: 'bel',\n bg: 'bul',\n bn: 'ben',\n bs: 'bos',\n cs: 'ces',\n de: 'deu',\n el: 'ell',\n en: 'eng',\n es: 'spa',\n fa: 'pes',\n fr: 'fra',\n gu: 'guj',\n ha: 'hau',\n hi: 'hin',\n hr: 'hrv',\n hu: 'hun',\n id: 'ind',\n ig: 'ibo',\n it: 'ita',\n ja: 'jpn',\n jv: 'jav',\n kk: 'kaz',\n kn: 'kan',\n ko: 'kor',\n ml: 'mal',\n mr: 'mar',\n ms: 'zlm',\n my: 'mya',\n ne: 'npi',\n nl: 'nld',\n pa: 'pan',\n pl: 'pol',\n pt: 'por',\n ro: 'ron',\n ru: 'rus',\n rw: 'kin',\n si: 'sin',\n so: 'som',\n sr: 'srp',\n su: 'sun',\n sv: 'swe',\n sw: 'swh',\n ta: 'tam',\n te: 'tel',\n th: 'tha',\n tl: 'tgl',\n tr: 'tur',\n uk: 'ukr',\n ur: 'urd',\n uz: 'uzn',\n vi: 'vie',\n yo: 'yor',\n zh: 'cmn',\n zu: 'zul',\n};\n\nexport function isKnownLanguageTag(tag: string): boolean {\n return tag in ISO_639_1_TO_3;\n}\n\n// Normalizes a declared or detected language code to ISO 639-3 for comparison; strips a\n// region/script subtag (\"zh-cn\" -> \"zh\") first. Codes already in 639-3 form pass through.\nexport function toIso3(code: string): string {\n const base = code.toLowerCase().split('-')[0];\n return ISO_639_1_TO_3[base] ?? base;\n}\n\n// Languages this project measured for models whose cards declare none, so the fit check is\n// not silently off for the shipped default; asserted here because it was measured, not attested.\nexport const MEASURED_MODEL_LANGUAGES: Record<string, string[]> = {\n 'minishlab/potion-retrieval-32M': ['en'], // benchmark/reports/2026-08-27-embedding-model-selection.md\n};\n"],"names":["ISO_639_1_TO_3","ar","am","az","be","bg","bn","bs","cs","de","el","en","es","fa","fr","gu","ha","hi","hr","hu","id","ig","it","ja","jv","kk","kn","ko","ml","mr","ms","my","ne","nl","pa","pl","pt","ro","ru","rw","si","so","sr","su","sv","sw","ta","te","th","tl","tr","uk","ur","uz","vi","yo","zh","zu","isKnownLanguageTag","tag","toIso3","code","base","toLowerCase","split","MEASURED_MODEL_LANGUAGES"],"mappings":"AAAA,+FAA+F;AAC/F,0FAA0F;AAC1F,MAAMA,iBAAyC;IAC7CC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;IACJC,IAAI;AACN;AAEA,OAAO,SAASC,mBAAmBC,GAAW;IAC5C,OAAOA,OAAO3D;AAChB;AAEA,wFAAwF;AACxF,0FAA0F;AAC1F,OAAO,SAAS4D,OAAOC,IAAY;QAE1B7D;IADP,MAAM8D,OAAOD,KAAKE,WAAW,GAAGC,KAAK,CAAC,IAAI,CAAC,EAAE;IAC7C,QAAOhE,uBAAAA,cAAc,CAAC8D,KAAK,cAApB9D,kCAAAA,uBAAwB8D;AACjC;AAEA,2FAA2F;AAC3F,iGAAiG;AACjG,OAAO,MAAMG,2BAAqD;IAChE,kCAAkC;QAAC;KAAK;AAC1C,EAAE"}
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/embed/openai.ts"],"sourcesContent":["import { SenseError } from '../errors.ts';\nimport { effectiveEmbedEndpoint } from './endpoint.ts';\nimport { fetchWithRetry, HTTP_ATTEMPT_TIMEOUT_MS } from './http.ts';\nimport type { EmbedCallOptions, EmbedProvider } from './types.ts';\n\nconst BATCH_CAP = 64;\n\ninterface OpenaiEmbedding {\n embedding: number[];\n index?: number;\n}\n\nfunction isEmbedding(value: unknown): value is OpenaiEmbedding {\n if (value === null || typeof value !== 'object' || !('embedding' in value) || !Array.isArray(value.embedding) || !value.embedding.every((part) => typeof part === 'number')) return false;\n return !('index' in value) || value.index === undefined || typeof value.index === 'number';\n}\n\nfunction parseBody(body: ArrayBuffer, endpoint: string): OpenaiEmbedding[] {\n let decoded: unknown;\n try {\n decoded = JSON.parse(new TextDecoder().decode(body));\n } catch {\n throw new SenseError('EMBED_MODEL', `${endpoint} returned invalid JSON`);\n }\n if (decoded === null || typeof decoded !== 'object' || !('data' in decoded) || !Array.isArray(decoded.data) || !decoded.data.every(isEmbedding)) {\n throw new SenseError('EMBED_MODEL', `${endpoint} returned an invalid embedding response`);\n }\n return decoded.data;\n}\n\n// One POST against any OpenAI-compatible /embeddings endpoint: Ollama, LM Studio, Cloudflare\n// Workers AI, Jina, Voyage, and Gemini's compat base URL all serve this shape.\nexport async function openaiProvider(model: string, url: string | undefined, keyEnv: string | undefined): Promise<EmbedProvider> {\n const base = effectiveEmbedEndpoint('openai', url);\n if (!base) throw new SenseError('EMBED_MODEL', 'embed.provider \"openai\" requires a url');\n const headers: Record<string, string> = { 'content-type': 'application/json' };\n const key = keyEnv ? process.env[keyEnv] : undefined;\n if (key) headers.authorization = `Bearer ${key}`;\n\n async function post(texts: string[], options?: EmbedCallOptions): Promise<Float32Array[]> {\n const res = await fetchWithRetry(`${base}/embeddings`, { method: 'POST', headers, body: JSON.stringify({ model, input: texts }) }, { signal: options?.signal, attemptTimeoutMs: HTTP_ATTEMPT_TIMEOUT_MS });\n options?.signal?.throwIfAborted();\n if (!res.ok) throw new SenseError('EMBED_MODEL', `${res.url} -> HTTP ${res.status}`);\n const body = parseBody(res.body, res.url);\n // The schema carries `index` because response order is not promised, and the caller maps\n // vectors back to chunks positionally: sort by it whenever every element has one.\n const data = body.every((d): d is OpenaiEmbedding & { index: number } => typeof d.index === 'number') ? [...body].sort((a, b) => a.index - b.index) : body;\n // A server that drops an input (empty or over-length strings are the usual cause) would\n // otherwise shift every later vector onto the wrong chunk, silently.\n if (data.length !== texts.length) throw new SenseError('EMBED_MODEL', `${res.url} returned ${data.length} embeddings for ${texts.length} inputs`);\n return data.map((d) => Float32Array.from(d.embedding));\n }\n\n const dims = (await post(['dimension probe']))[0].length;\n // embedQuery and embedDocuments are symmetric here; per-model prefixes like nomic's\n // search_document/search_query are not applied.\n return { id: `openai:${base}:${model}`, dims, batchCap: BATCH_CAP, embedDocuments: post, embedQuery: async (text, options) => (await post([text], options))[0] };\n}\n"],"names":["SenseError","effectiveEmbedEndpoint","fetchWithRetry","HTTP_ATTEMPT_TIMEOUT_MS","BATCH_CAP","isEmbedding","value","Array","isArray","embedding","every","part","index","undefined","parseBody","body","endpoint","decoded","JSON","parse","TextDecoder","decode","data","openaiProvider","model","url","keyEnv","base","headers","key","process","env","authorization","post","texts","options","res","method","stringify","input","signal","attemptTimeoutMs","throwIfAborted","ok","status","d","sort","a","b","length","map","Float32Array","from","dims","id","batchCap","embedDocuments","embedQuery","text"],"mappings":"AAAA,SAASA,UAAU,QAAQ,eAAe;AAC1C,SAASC,sBAAsB,QAAQ,gBAAgB;AACvD,SAASC,cAAc,EAAEC,uBAAuB,QAAQ,YAAY;AAGpE,MAAMC,YAAY;AAOlB,SAASC,YAAYC,KAAc;IACjC,IAAIA,UAAU,QAAQ,OAAOA,UAAU,YAAY,CAAE,CAAA,eAAeA,KAAI,KAAM,CAACC,MAAMC,OAAO,CAACF,MAAMG,SAAS,KAAK,CAACH,MAAMG,SAAS,CAACC,KAAK,CAAC,CAACC,OAAS,OAAOA,SAAS,WAAW,OAAO;IACpL,OAAO,CAAE,CAAA,WAAWL,KAAI,KAAMA,MAAMM,KAAK,KAAKC,aAAa,OAAOP,MAAMM,KAAK,KAAK;AACpF;AAEA,SAASE,UAAUC,IAAiB,EAAEC,QAAgB;IACpD,IAAIC;IACJ,IAAI;QACFA,UAAUC,KAAKC,KAAK,CAAC,IAAIC,cAAcC,MAAM,CAACN;IAChD,EAAE,OAAM;QACN,MAAM,IAAIf,WAAW,eAAe,GAAGgB,SAAS,sBAAsB,CAAC;IACzE;IACA,IAAIC,YAAY,QAAQ,OAAOA,YAAY,YAAY,CAAE,CAAA,UAAUA,OAAM,KAAM,CAACV,MAAMC,OAAO,CAACS,QAAQK,IAAI,KAAK,CAACL,QAAQK,IAAI,CAACZ,KAAK,CAACL,cAAc;QAC/I,MAAM,IAAIL,WAAW,eAAe,GAAGgB,SAAS,uCAAuC,CAAC;IAC1F;IACA,OAAOC,QAAQK,IAAI;AACrB;AAEA,6FAA6F;AAC7F,+EAA+E;AAC/E,OAAO,eAAeC,eAAeC,KAAa,EAAEC,GAAuB,EAAEC,MAA0B;IACrG,MAAMC,OAAO1B,uBAAuB,UAAUwB;IAC9C,IAAI,CAACE,MAAM,MAAM,IAAI3B,WAAW,eAAe;IAC/C,MAAM4B,UAAkC;QAAE,gBAAgB;IAAmB;IAC7E,MAAMC,MAAMH,SAASI,QAAQC,GAAG,CAACL,OAAO,GAAGb;IAC3C,IAAIgB,KAAKD,QAAQI,aAAa,GAAG,CAAC,OAAO,EAAEH,KAAK;IAEhD,eAAeI,KAAKC,KAAe,EAAEC,OAA0B;YAE7DA;QADA,MAAMC,MAAM,MAAMlC,eAAe,GAAGyB,KAAK,WAAW,CAAC,EAAE;YAAEU,QAAQ;YAAQT;YAASb,MAAMG,KAAKoB,SAAS,CAAC;gBAAEd;gBAAOe,OAAOL;YAAM;QAAG,GAAG;YAAEM,MAAM,EAAEL,oBAAAA,8BAAAA,QAASK,MAAM;YAAEC,kBAAkBtC;QAAwB;QACxMgC,oBAAAA,+BAAAA,kBAAAA,QAASK,MAAM,cAAfL,sCAAAA,gBAAiBO,cAAc;QAC/B,IAAI,CAACN,IAAIO,EAAE,EAAE,MAAM,IAAI3C,WAAW,eAAe,GAAGoC,IAAIX,GAAG,CAAC,SAAS,EAAEW,IAAIQ,MAAM,EAAE;QACnF,MAAM7B,OAAOD,UAAUsB,IAAIrB,IAAI,EAAEqB,IAAIX,GAAG;QACxC,yFAAyF;QACzF,kFAAkF;QAClF,MAAMH,OAAOP,KAAKL,KAAK,CAAC,CAACmC,IAAgD,OAAOA,EAAEjC,KAAK,KAAK,YAAY;eAAIG;SAAK,CAAC+B,IAAI,CAAC,CAACC,GAAGC,IAAMD,EAAEnC,KAAK,GAAGoC,EAAEpC,KAAK,IAAIG;QACtJ,wFAAwF;QACxF,qEAAqE;QACrE,IAAIO,KAAK2B,MAAM,KAAKf,MAAMe,MAAM,EAAE,MAAM,IAAIjD,WAAW,eAAe,GAAGoC,IAAIX,GAAG,CAAC,UAAU,EAAEH,KAAK2B,MAAM,CAAC,gBAAgB,EAAEf,MAAMe,MAAM,CAAC,OAAO,CAAC;QAChJ,OAAO3B,KAAK4B,GAAG,CAAC,CAACL,IAAMM,aAAaC,IAAI,CAACP,EAAEpC,SAAS;IACtD;IAEA,MAAM4C,OAAO,AAAC,CAAA,MAAMpB,KAAK;QAAC;KAAkB,CAAA,CAAE,CAAC,EAAE,CAACgB,MAAM;IACxD,oFAAoF;IACpF,gDAAgD;IAChD,OAAO;QAAEK,IAAI,CAAC,OAAO,EAAE3B,KAAK,CAAC,EAAEH,OAAO;QAAE6B;QAAME,UAAUnD;QAAWoD,gBAAgBvB;QAAMwB,YAAY,OAAOC,MAAMvB,UAAY,AAAC,CAAA,MAAMF,KAAK;gBAACyB;aAAK,EAAEvB,QAAO,CAAE,CAAC,EAAE;IAAC;AACjK"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/open-source/ai/sensemaking/src/embed/openai.ts"],"sourcesContent":["import { SenseError } from '../errors.ts';\nimport { effectiveEmbedEndpoint } from './endpoint.ts';\nimport { fetchWithRetry, HTTP_ATTEMPT_TIMEOUT_MS } from './http.ts';\nimport type { EmbedCallOptions, EmbedProvider } from './types.ts';\n\nconst BATCH_CAP = 64;\n\ninterface OpenaiEmbedding {\n embedding: number[];\n index?: number;\n}\n\nfunction isEmbedding(value: unknown): value is OpenaiEmbedding {\n if (value === null || typeof value !== 'object' || !('embedding' in value) || !Array.isArray(value.embedding) || !value.embedding.every((part) => typeof part === 'number')) return false;\n return !('index' in value) || value.index === undefined || typeof value.index === 'number';\n}\n\nfunction parseBody(body: ArrayBuffer, endpoint: string): OpenaiEmbedding[] {\n let decoded: unknown;\n try {\n decoded = JSON.parse(new TextDecoder().decode(body));\n } catch {\n throw new SenseError('EMBED_MODEL', `${endpoint} returned invalid JSON`);\n }\n if (decoded === null || typeof decoded !== 'object' || !('data' in decoded) || !Array.isArray(decoded.data) || !decoded.data.every(isEmbedding)) {\n throw new SenseError('EMBED_MODEL', `${endpoint} returned an invalid embedding response`);\n }\n return decoded.data;\n}\n\n// One POST against any OpenAI-compatible /embeddings endpoint: Ollama, LM Studio, Cloudflare\n// Workers AI, Jina, Voyage, and Gemini's compat base URL all serve this shape.\nexport async function openaiProvider(model: string, url: string | undefined, keyEnv: string | undefined): Promise<EmbedProvider> {\n const base = effectiveEmbedEndpoint('openai', url);\n if (!base) throw new SenseError('EMBED_MODEL', 'embed.provider \"openai\" requires a url');\n const headers: Record<string, string> = { 'content-type': 'application/json' };\n const key = keyEnv ? process.env[keyEnv] : undefined;\n if (key) headers.authorization = `Bearer ${key}`;\n\n async function post(texts: string[], options?: EmbedCallOptions): Promise<Float32Array[]> {\n const res = await fetchWithRetry(`${base}/embeddings`, { method: 'POST', headers, body: JSON.stringify({ model, input: texts }) }, { signal: options?.signal, attemptTimeoutMs: HTTP_ATTEMPT_TIMEOUT_MS });\n options?.signal?.throwIfAborted();\n if (!res.ok) throw new SenseError('EMBED_MODEL', `${res.url} -> HTTP ${res.status}`);\n const body = parseBody(res.body, res.url);\n // The schema carries `index` because response order is not promised, and the caller maps\n // vectors back to chunks positionally: sort by it whenever every element has one.\n const data = body.every((d): d is OpenaiEmbedding & { index: number } => typeof d.index === 'number') ? [...body].sort((a, b) => a.index - b.index) : body;\n // A server that drops an input (empty or over-length strings are the usual cause) would\n // otherwise shift every later vector onto the wrong chunk, silently.\n if (data.length !== texts.length) throw new SenseError('EMBED_MODEL', `${res.url} returned ${data.length} embeddings for ${texts.length} inputs`);\n return data.map((d) => Float32Array.from(d.embedding));\n }\n\n const dims = (await post(['dimension probe']))[0].length;\n // embedQuery and embedDocuments are symmetric here; per-model prefixes like nomic's\n // search_document/search_query are not applied.\n return { id: `openai:${base}:${model}`, dims, batchCap: BATCH_CAP, embedDocuments: post, embedQuery: async (text, options) => (await post([text], options))[0] };\n}\n"],"names":["SenseError","effectiveEmbedEndpoint","fetchWithRetry","HTTP_ATTEMPT_TIMEOUT_MS","BATCH_CAP","isEmbedding","value","Array","isArray","embedding","every","part","index","undefined","parseBody","body","endpoint","decoded","JSON","parse","TextDecoder","decode","data","openaiProvider","model","url","keyEnv","base","headers","key","process","env","authorization","post","texts","options","res","method","stringify","input","signal","attemptTimeoutMs","throwIfAborted","ok","status","d","sort","a","b","length","map","Float32Array","from","dims","id","batchCap","embedDocuments","embedQuery","text"],"mappings":"AAAA,SAASA,UAAU,QAAQ,eAAe;AAC1C,SAASC,sBAAsB,QAAQ,gBAAgB;AACvD,SAASC,cAAc,EAAEC,uBAAuB,QAAQ,YAAY;AAGpE,MAAMC,YAAY;AAOlB,SAASC,YAAYC,KAAc;IACjC,IAAIA,UAAU,QAAQ,OAAOA,UAAU,YAAY,CAAE,CAAA,eAAeA,KAAI,KAAM,CAACC,MAAMC,OAAO,CAACF,MAAMG,SAAS,KAAK,CAACH,MAAMG,SAAS,CAACC,KAAK,CAAC,CAACC,OAAS,OAAOA,SAAS,WAAW,OAAO;IACpL,OAAO,CAAE,CAAA,WAAWL,KAAI,KAAMA,MAAMM,KAAK,KAAKC,aAAa,OAAOP,MAAMM,KAAK,KAAK;AACpF;AAEA,SAASE,UAAUC,IAAiB,EAAEC,QAAgB;IACpD,IAAIC;IACJ,IAAI;QACFA,UAAUC,KAAKC,KAAK,CAAC,IAAIC,cAAcC,MAAM,CAACN;IAChD,EAAE,OAAM;QACN,MAAM,IAAIf,WAAW,eAAe,GAAGgB,SAAS,sBAAsB,CAAC;IACzE;IACA,IAAIC,YAAY,QAAQ,OAAOA,YAAY,YAAY,CAAE,CAAA,UAAUA,OAAM,KAAM,CAACV,MAAMC,OAAO,CAACS,QAAQK,IAAI,KAAK,CAACL,QAAQK,IAAI,CAACZ,KAAK,CAACL,cAAc;QAC/I,MAAM,IAAIL,WAAW,eAAe,GAAGgB,SAAS,uCAAuC,CAAC;IAC1F;IACA,OAAOC,QAAQK,IAAI;AACrB;AAEA,6FAA6F;AAC7F,+EAA+E;AAC/E,OAAO,eAAeC,eAAeC,KAAa,EAAEC,GAAuB,EAAEC,MAA0B;IACrG,MAAMC,OAAO1B,uBAAuB,UAAUwB;IAC9C,IAAI,CAACE,MAAM,MAAM,IAAI3B,WAAW,eAAe;IAC/C,MAAM4B,UAAkC;QAAE,gBAAgB;IAAmB;IAC7E,MAAMC,MAAMH,SAASI,QAAQC,GAAG,CAACL,OAAO,GAAGb;IAC3C,IAAIgB,KAAKD,QAAQI,aAAa,GAAG,CAAC,OAAO,EAAEH,KAAK;IAEhD,eAAeI,KAAKC,KAAe,EAAEC,OAA0B;YAE7DA;QADA,MAAMC,MAAM,MAAMlC,eAAe,GAAGyB,KAAK,WAAW,CAAC,EAAE;YAAEU,QAAQ;YAAQT;YAASb,MAAMG,KAAKoB,SAAS,CAAC;gBAAEd;gBAAOe,OAAOL;YAAM;QAAG,GAAG;YAAEM,MAAM,EAAEL,oBAAAA,8BAAAA,QAASK,MAAM;YAAEC,kBAAkBtC;QAAwB;QACxMgC,oBAAAA,+BAAAA,kBAAAA,QAASK,MAAM,cAAfL,sCAAAA,gBAAiBO,cAAc;QAC/B,IAAI,CAACN,IAAIO,EAAE,EAAE,MAAM,IAAI3C,WAAW,eAAe,GAAGoC,IAAIX,GAAG,CAAC,SAAS,EAAEW,IAAIQ,MAAM,EAAE;QACnF,MAAM7B,OAAOD,UAAUsB,IAAIrB,IAAI,EAAEqB,IAAIX,GAAG;QACxC,yFAAyF;QACzF,kFAAkF;QAClF,MAAMH,OAAOP,KAAKL,KAAK,CAAC,CAACmC,IAAgD,OAAOA,EAAEjC,KAAK,KAAK,YAAY;eAAIG;SAAK,CAAC+B,IAAI,CAAC,CAACC,GAAGC,IAAMD,EAAEnC,KAAK,GAAGoC,EAAEpC,KAAK,IAAIG;QACtJ,wFAAwF;QACxF,qEAAqE;QACrE,IAAIO,KAAK2B,MAAM,KAAKf,MAAMe,MAAM,EAAE,MAAM,IAAIjD,WAAW,eAAe,GAAGoC,IAAIX,GAAG,CAAC,UAAU,EAAEH,KAAK2B,MAAM,CAAC,gBAAgB,EAAEf,MAAMe,MAAM,CAAC,OAAO,CAAC;QAChJ,OAAO3B,KAAK4B,GAAG,CAAC,CAACL,IAAMM,aAAaC,IAAI,CAACP,EAAEpC,SAAS;IACtD;IAEA,MAAM4C,OAAO,AAAC,CAAA,MAAMpB,KAAK;QAAC;KAAkB,CAAA,CAAE,CAAC,EAAE,CAACgB,MAAM;IACxD,oFAAoF;IACpF,gDAAgD;IAChD,OAAO;QAAEK,IAAI,CAAC,OAAO,EAAE3B,KAAK,CAAC,EAAEH,OAAO;QAAE6B;QAAME,UAAUnD;QAAWoD,gBAAgBvB;QAAMwB,YAAY,OAAOC,MAAMvB,UAAY,AAAC,CAAA,MAAMF,KAAK;gBAACyB;aAAK,EAAEvB,QAAO,CAAE,CAAC,EAAE;IAAC;AACjK"}
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/embed/query.ts"],"sourcesContent":["import { type Config, featureSignature, type ResolvedConfig } from '../config/index.ts';\nimport { SenseError } from '../errors.ts';\nimport type { Chunk } from '../features/embed.ts';\nimport { embed } from '../features/embed.ts';\nimport { FEATURES } from '../features/index.ts';\nimport { progress } from '../output/progress.ts';\nimport { parseSource } from '../scan/index.ts';\nimport { FEATURE_SIGNATURE_META_KEY } from '../store/readiness.ts';\nimport { getMeta } from '../store/shared.ts';\nimport { embeddingSignatureCompatible } from '../store/signature.ts';\nimport type { Store, VectorCandidate, VectorSimilar } from '../store/types.ts';\nimport { identifyChunkText, takeChunkText } from './handoff.ts';\nimport { checkLanguageFit } from './langfit.ts';\nimport { getProvider } from './registry.ts';\nimport { type EmbedCallOptions, STORE_DIMS } from './types.ts';\n\n// Slice + re-normalize (Matryoshka); optionally round through int8 storage. Exported for\n// benchmark/lib/embed.mjs, which scores the same lever math offline.\nexport function toStore(full: Float32Array, dims: number, int8: boolean): { v: Float32Array; scale: number } {\n const v = new Float32Array(dims);\n let norm = 0;\n for (let d = 0; d < dims; d++) norm += full[d] * full[d];\n norm = Math.sqrt(norm) + 1e-32;\n for (let d = 0; d < dims; d++) v[d] = full[d] / norm;\n if (!int8) return { v, scale: 1 };\n let max = 0;\n for (let d = 0; d < dims; d++) max = Math.max(max, Math.abs(v[d]));\n return { v, scale: max / 127 || 1 };\n}\n\n// Embed rows whose vector is NULL. Reconcile hands its chunk text over in memory when it ran in\n// this process (handoff.ts); a row left pending by an earlier command re-derives from the exact\n// source string persisted with that indexed generation.\nfunction staleEmbeddingWork(reason: string): SenseError {\n return new SenseError('INDEX_NOT_READY', `${reason}; run \\`sense build\\` (or library build(config)) again to prepare the current index generation`);\n}\n\n// expectedSignature is captured by open() before provider work. Direct internal callers omit it\n// and capture the durable value here, before selecting chunks or constructing a provider.\nexport async function embedPending(store: Store, cfg: Config, allowed?: ReadonlySet<string>, expectedSignature?: string | null, options: EmbedCallOptions = {}): Promise<void> {\n options.signal?.throwIfAborted();\n const expectedDurableSignature = expectedSignature === undefined ? await getMeta(store, FEATURE_SIGNATURE_META_KEY) : expectedSignature;\n options.signal?.throwIfAborted();\n if (!embeddingSignatureCompatible(expectedDurableSignature, featureSignature(cfg, FEATURES))) {\n throw staleEmbeddingWork('the index configuration changed before vectors could be prepared');\n }\n\n const pendingStmt = await store.prepare('SELECT \"path\", chunk, content_identity FROM embeddings WHERE vector IS NULL ORDER BY \"path\", chunk');\n const dirty = ((await pendingStmt.all()) as Array<{ path: string; chunk: number; content_identity: string }>).filter((row) => allowed?.has(row.path) ?? true);\n options.signal?.throwIfAborted();\n if (dirty.length === 0) return;\n\n const provider = await getProvider(cfg, options);\n options.signal?.throwIfAborted();\n const submittedSignature = featureSignature(cfg, FEATURES);\n if (!embeddingSignatureCompatible(expectedDurableSignature, submittedSignature)) {\n throw staleEmbeddingWork('the index configuration changed while the embedding provider was being prepared');\n }\n const storeDims = Math.min(STORE_DIMS, provider.dims);\n\n const byPath = new Map<string, typeof dirty>();\n for (const row of dirty) {\n const list = byPath.get(row.path) ?? [];\n list.push(row);\n byPath.set(row.path, list);\n }\n\n const textByPath = takeChunkText(store) ?? new Map();\n const sourceStmt = await store.prepare('SELECT text FROM indexed_sources WHERE \"path\" = ?');\n\n const jobs: Array<{ path: string; chunk: number; text: string; identity: string }> = [];\n for (const [path, pending] of byPath) {\n const texts = textByPath.get(path);\n if (texts && pending.every((row) => texts[row.chunk]?.identity === row.content_identity)) {\n for (const row of pending) jobs.push({ path, chunk: row.chunk, ...texts[row.chunk] });\n continue;\n }\n // A prior process may have reconciled the row, so its in-memory handoff is gone. Re-derive\n // from that indexed generation's stored source, never from a newer live file.\n const source = (await sourceStmt.get(path)) as { text: string } | undefined;\n options.signal?.throwIfAborted();\n if (!source) throw new SenseError('INDEX_NOT_READY', `indexed source text is missing for ${path}; run \\`sense build --force\\` (or library build(config, { force: true })) before preparing vectors`);\n const chunks = parseSource({ relPath: path, absPath: path, mtimeMs: 0, ctimeMs: 0, size: Buffer.byteLength(source.text), presets: [], embed: true }, source.text, [embed], cfg).doc.extracted.embed as Chunk[];\n for (const row of pending) {\n const chunk = chunks[row.chunk];\n if (!chunk) throw new SenseError('INDEX_NOT_READY', `indexed vector metadata for ${path} does not match its indexed source; run \\`sense build --force\\` (or library build(config, { force: true }))`);\n const identified = identifyChunkText(chunk.text);\n if (identified.identity !== row.content_identity) throw staleEmbeddingWork(`indexed chunk ${path}#${row.chunk} changed before vectors could be prepared`);\n jobs.push({ path, chunk: row.chunk, ...identified });\n }\n }\n\n // Over the exact texts about to be embedded, before any of them are: a mismatch fails\n // this run loudly instead of quietly storing vectors from the wrong model.\n options.signal?.throwIfAborted();\n await checkLanguageFit(\n store,\n provider,\n jobs.map((j) => j.text)\n );\n options.signal?.throwIfAborted();\n\n // Preparing a large configured scope can take long enough to look stalled, so report each\n // completed provider batch.\n const report = progress('embedding chunks', jobs.length);\n for (let i = 0; i < jobs.length; i += provider.batchCap) {\n const batch = jobs.slice(i, i + provider.batchCap);\n options.signal?.throwIfAborted();\n if (featureSignature(cfg, FEATURES) !== submittedSignature) throw staleEmbeddingWork('the embedding configuration changed before a provider batch was submitted');\n const vectors = await provider.embedDocuments(\n batch.map((j) => j.text),\n options\n );\n options.signal?.throwIfAborted();\n if (featureSignature(cfg, FEATURES) !== submittedSignature) throw staleEmbeddingWork('the embedding configuration changed while a provider batch was running');\n const writes = batch.map((job, j) => {\n const { v, scale } = toStore(vectors[j], storeDims, true);\n const q = new Int8Array(storeDims);\n for (let d = 0; d < storeDims; d++) q[d] = Math.round(v[d] / scale);\n return { path: job.path, chunk: job.chunk, scale, vector: Buffer.from(q.buffer) };\n });\n options.signal?.throwIfAborted();\n await store.transaction(async () => {\n // Write first so SQLite owns the write lock before the fence reads. Any mismatch throws\n // inside this transaction and rolls every row in the provider batch back together.\n options.signal?.throwIfAborted();\n await store.vectors.writeVectors(writes);\n options.signal?.throwIfAborted();\n if ((await getMeta(store, FEATURE_SIGNATURE_META_KEY)) !== expectedDurableSignature) throw staleEmbeddingWork('the durable embedding configuration changed while a provider batch was running');\n const currentStmt = await store.prepare(`SELECT \"path\", chunk, content_identity FROM embeddings WHERE ${batch.map(() => '(\"path\" = ? AND chunk = ?)').join(' OR ')}`);\n const current = (await currentStmt.all(...batch.flatMap((job) => [job.path, job.chunk]))) as Array<{ path: string; chunk: number; content_identity: string }>;\n const identities = new Map(current.map((row) => [`${row.path}\\0${Number(row.chunk)}`, row.content_identity]));\n if (batch.some((job) => identities.get(`${job.path}\\0${job.chunk}`) !== job.identity)) throw staleEmbeddingWork('indexed chunks changed while a provider batch was running');\n options.signal?.throwIfAborted();\n });\n report.tick(Math.min(i + provider.batchCap, jobs.length));\n }\n report.finish();\n}\n\nexport async function ensureDocumentEmbeddings(store: Store, _cfg: Config, allowed: ReadonlySet<string>): Promise<void> {\n const pending = (await store.vectors.pending()).filter((row) => allowed.has(row.path));\n if (pending.length > 0) {\n throw new SenseError('INDEX_NOT_READY', `${pending.length} indexed chunk(s) in the requested scope still need vectors; run \\`sense build\\` (or library build(config)) before querying with build disabled`);\n }\n}\n\nexport interface PreparedSemanticQuery {\n readonly vector: Float32Array;\n readonly storeDims: number;\n}\n\n// Provider/model work is independent of the index snapshot. FTS5 operators are stripped as\n// lexical syntax before the query is embedded, matching the prior semantic query behavior.\nexport async function prepareSemanticQuery(cfg: Config, terms: string, options: EmbedCallOptions = {}): Promise<PreparedSemanticQuery> {\n const rootDir = (cfg as Partial<ResolvedConfig>).rootDir ?? (cfg as Partial<ResolvedConfig>).baseDir;\n if (!rootDir) throw new SenseError('EMBED_MODEL', 'semantic expansion needs a resolved config (use loadConfig/open)');\n const text = (terms.match(/[\\p{L}\\p{N}]+/gu) ?? []).filter((t) => !['AND', 'OR', 'NOT', 'NEAR'].includes(t)).join(' ');\n options.signal?.throwIfAborted();\n const provider = await getProvider(cfg, options);\n options.signal?.throwIfAborted();\n const storeDims = Math.min(STORE_DIMS, provider.dims);\n const queryVector = await provider.embedQuery(text, options);\n options.signal?.throwIfAborted();\n const { v: qv } = toStore(queryVector, storeDims, false);\n return { vector: qv, storeDims };\n}\n\n// Best chunk per file by cosine, its line range riding along. The caller owns the read snapshot\n// and the prepared vector, so this performs no provider work and only delegates to the store.\nexport function semanticCandidates(store: Store, prepared: PreparedSemanticQuery, fetch: number, allowed?: Set<string>): Promise<VectorCandidate[]> {\n return store.vectors.candidates(prepared.vector, prepared.storeDims, fetch, allowed);\n}\n\n// Whether a note has any chunk with a vector. Distinguishes \"nothing is near this note\" from\n// \"this note has no text\", which look the same in an empty result.\nexport async function hasEmbedding(store: Store, path: string): Promise<boolean> {\n return store.vectors.hasVector(path);\n}\n\n// Note-to-note similarity is the max cosine over (target chunk, other chunk) pairs, one linear\n// scan of stored vectors.\nexport async function similarNotes(store: Store, _cfg: Config, path: string, opts: { exclude: Set<string>; allowed?: Set<string>; k: number }): Promise<VectorSimilar[]> {\n return store.vectors.similar(path, opts);\n}\n"],"names":["featureSignature","SenseError","embed","FEATURES","progress","parseSource","FEATURE_SIGNATURE_META_KEY","getMeta","embeddingSignatureCompatible","identifyChunkText","takeChunkText","checkLanguageFit","getProvider","STORE_DIMS","toStore","full","dims","int8","v","Float32Array","norm","d","Math","sqrt","scale","max","abs","staleEmbeddingWork","reason","embedPending","store","cfg","allowed","expectedSignature","options","signal","throwIfAborted","expectedDurableSignature","undefined","pendingStmt","prepare","dirty","all","filter","row","has","path","length","provider","submittedSignature","storeDims","min","byPath","Map","list","get","push","set","textByPath","sourceStmt","jobs","pending","texts","every","chunk","identity","content_identity","source","chunks","relPath","absPath","mtimeMs","ctimeMs","size","Buffer","byteLength","text","presets","doc","extracted","identified","map","j","report","i","batchCap","batch","slice","vectors","embedDocuments","writes","job","q","Int8Array","round","vector","from","buffer","transaction","writeVectors","currentStmt","join","current","flatMap","identities","Number","some","tick","finish","ensureDocumentEmbeddings","_cfg","prepareSemanticQuery","terms","rootDir","baseDir","match","t","includes","queryVector","embedQuery","qv","semanticCandidates","prepared","fetch","candidates","hasEmbedding","hasVector","similarNotes","opts","similar"],"mappings":"AAAA,SAAsBA,gBAAgB,QAA6B,qBAAqB;AACxF,SAASC,UAAU,QAAQ,eAAe;AAE1C,SAASC,KAAK,QAAQ,uBAAuB;AAC7C,SAASC,QAAQ,QAAQ,uBAAuB;AAChD,SAASC,QAAQ,QAAQ,wBAAwB;AACjD,SAASC,WAAW,QAAQ,mBAAmB;AAC/C,SAASC,0BAA0B,QAAQ,wBAAwB;AACnE,SAASC,OAAO,QAAQ,qBAAqB;AAC7C,SAASC,4BAA4B,QAAQ,wBAAwB;AAErE,SAASC,iBAAiB,EAAEC,aAAa,QAAQ,eAAe;AAChE,SAASC,gBAAgB,QAAQ,eAAe;AAChD,SAASC,WAAW,QAAQ,gBAAgB;AAC5C,SAAgCC,UAAU,QAAQ,aAAa;AAE/D,yFAAyF;AACzF,qEAAqE;AACrE,OAAO,SAASC,QAAQC,IAAkB,EAAEC,IAAY,EAAEC,IAAa;IACrE,MAAMC,IAAI,IAAIC,aAAaH;IAC3B,IAAII,OAAO;IACX,IAAK,IAAIC,IAAI,GAAGA,IAAIL,MAAMK,IAAKD,QAAQL,IAAI,CAACM,EAAE,GAAGN,IAAI,CAACM,EAAE;IACxDD,OAAOE,KAAKC,IAAI,CAACH,QAAQ;IACzB,IAAK,IAAIC,IAAI,GAAGA,IAAIL,MAAMK,IAAKH,CAAC,CAACG,EAAE,GAAGN,IAAI,CAACM,EAAE,GAAGD;IAChD,IAAI,CAACH,MAAM,OAAO;QAAEC;QAAGM,OAAO;IAAE;IAChC,IAAIC,MAAM;IACV,IAAK,IAAIJ,IAAI,GAAGA,IAAIL,MAAMK,IAAKI,MAAMH,KAAKG,GAAG,CAACA,KAAKH,KAAKI,GAAG,CAACR,CAAC,CAACG,EAAE;IAChE,OAAO;QAAEH;QAAGM,OAAOC,MAAM,OAAO;IAAE;AACpC;AAEA,gGAAgG;AAChG,gGAAgG;AAChG,wDAAwD;AACxD,SAASE,mBAAmBC,MAAc;IACxC,OAAO,IAAI3B,WAAW,mBAAmB,GAAG2B,OAAO,8FAA8F,CAAC;AACpJ;AAEA,gGAAgG;AAChG,0FAA0F;AAC1F,OAAO,eAAeC,aAAaC,KAAY,EAAEC,GAAW,EAAEC,OAA6B,EAAEC,iBAAiC,EAAEC,UAA4B,CAAC,CAAC;QA4BzIxB;QA3BnBwB,iBAEAA,kBAOAA,kBAIAA,kBAuCA,sFAAsF;IACtF,2EAA2E;IAC3EA,kBAMAA;KA5DAA,kBAAAA,QAAQC,MAAM,cAAdD,sCAAAA,gBAAgBE,cAAc;IAC9B,MAAMC,2BAA2BJ,sBAAsBK,YAAY,MAAM/B,QAAQuB,OAAOxB,8BAA8B2B;KACtHC,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;IAC9B,IAAI,CAAC5B,6BAA6B6B,0BAA0BrC,iBAAiB+B,KAAK5B,YAAY;QAC5F,MAAMwB,mBAAmB;IAC3B;IAEA,MAAMY,cAAc,MAAMT,MAAMU,OAAO,CAAC;IACxC,MAAMC,QAAQ,AAAE,CAAA,MAAMF,YAAYG,GAAG,EAAC,EAAwEC,MAAM,CAAC,CAACC;;uBAAQZ,oBAAAA,8BAAAA,QAASa,GAAG,CAACD,IAAIE,IAAI,wCAAK;;KACxJZ,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;IAC9B,IAAIK,MAAMM,MAAM,KAAK,GAAG;IAExB,MAAMC,WAAW,MAAMpC,YAAYmB,KAAKG;KACxCA,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;IAC9B,MAAMa,qBAAqBjD,iBAAiB+B,KAAK5B;IACjD,IAAI,CAACK,6BAA6B6B,0BAA0BY,qBAAqB;QAC/E,MAAMtB,mBAAmB;IAC3B;IACA,MAAMuB,YAAY5B,KAAK6B,GAAG,CAACtC,YAAYmC,SAAShC,IAAI;IAEpD,MAAMoC,SAAS,IAAIC;IACnB,KAAK,MAAMT,OAAOH,MAAO;YACVW;QAAb,MAAME,QAAOF,cAAAA,OAAOG,GAAG,CAACX,IAAIE,IAAI,eAAnBM,yBAAAA,cAAwB,EAAE;QACvCE,KAAKE,IAAI,CAACZ;QACVQ,OAAOK,GAAG,CAACb,IAAIE,IAAI,EAAEQ;IACvB;IAEA,MAAMI,cAAahD,iBAAAA,cAAcoB,oBAAdpB,4BAAAA,iBAAwB,IAAI2C;IAC/C,MAAMM,aAAa,MAAM7B,MAAMU,OAAO,CAAC;IAEvC,MAAMoB,OAA+E,EAAE;IACvF,KAAK,MAAM,CAACd,MAAMe,QAAQ,IAAIT,OAAQ;YASpClB;QARA,MAAM4B,QAAQJ,WAAWH,GAAG,CAACT;QAC7B,IAAIgB,SAASD,QAAQE,KAAK,CAAC,CAACnB;gBAAQkB;mBAAAA,EAAAA,mBAAAA,KAAK,CAAClB,IAAIoB,KAAK,CAAC,cAAhBF,uCAAAA,iBAAkBG,QAAQ,MAAKrB,IAAIsB,gBAAgB;YAAG;YACxF,KAAK,MAAMtB,OAAOiB,QAASD,KAAKJ,IAAI,CAAC;gBAAEV;gBAAMkB,OAAOpB,IAAIoB,KAAK;gBAAE,GAAGF,KAAK,CAAClB,IAAIoB,KAAK,CAAC;YAAC;YACnF;QACF;QACA,2FAA2F;QAC3F,8EAA8E;QAC9E,MAAMG,SAAU,MAAMR,WAAWJ,GAAG,CAACT;SACrCZ,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;QAC9B,IAAI,CAAC+B,QAAQ,MAAM,IAAIlE,WAAW,mBAAmB,CAAC,mCAAmC,EAAE6C,KAAK,kGAAkG,CAAC;QACnM,MAAMsB,SAAS/D,YAAY;YAAEgE,SAASvB;YAAMwB,SAASxB;YAAMyB,SAAS;YAAGC,SAAS;YAAGC,MAAMC,OAAOC,UAAU,CAACR,OAAOS,IAAI;YAAGC,SAAS,EAAE;YAAE3E,OAAO;QAAK,GAAGiE,OAAOS,IAAI,EAAE;YAAC1E;SAAM,EAAE6B,KAAK+C,GAAG,CAACC,SAAS,CAAC7E,KAAK;QACnM,KAAK,MAAM0C,OAAOiB,QAAS;YACzB,MAAMG,QAAQI,MAAM,CAACxB,IAAIoB,KAAK,CAAC;YAC/B,IAAI,CAACA,OAAO,MAAM,IAAI/D,WAAW,mBAAmB,CAAC,4BAA4B,EAAE6C,KAAK,2GAA2G,CAAC;YACpM,MAAMkC,aAAavE,kBAAkBuD,MAAMY,IAAI;YAC/C,IAAII,WAAWf,QAAQ,KAAKrB,IAAIsB,gBAAgB,EAAE,MAAMvC,mBAAmB,CAAC,cAAc,EAAEmB,KAAK,CAAC,EAAEF,IAAIoB,KAAK,CAAC,yCAAyC,CAAC;YACxJJ,KAAKJ,IAAI,CAAC;gBAAEV;gBAAMkB,OAAOpB,IAAIoB,KAAK;gBAAE,GAAGgB,UAAU;YAAC;QACpD;IACF;KAIA9C,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;IAC9B,MAAMzB,iBACJmB,OACAkB,UACAY,KAAKqB,GAAG,CAAC,CAACC,IAAMA,EAAEN,IAAI;KAExB1C,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;IAE9B,0FAA0F;IAC1F,4BAA4B;IAC5B,MAAM+C,SAAS/E,SAAS,oBAAoBwD,KAAKb,MAAM;IACvD,IAAK,IAAIqC,IAAI,GAAGA,IAAIxB,KAAKb,MAAM,EAAEqC,KAAKpC,SAASqC,QAAQ,CAAE;YAEvDnD,kBAMAA,kBAQAA;QAfA,MAAMoD,QAAQ1B,KAAK2B,KAAK,CAACH,GAAGA,IAAIpC,SAASqC,QAAQ;SACjDnD,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;QAC9B,IAAIpC,iBAAiB+B,KAAK5B,cAAc8C,oBAAoB,MAAMtB,mBAAmB;QACrF,MAAM6D,UAAU,MAAMxC,SAASyC,cAAc,CAC3CH,MAAML,GAAG,CAAC,CAACC,IAAMA,EAAEN,IAAI,GACvB1C;SAEFA,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;QAC9B,IAAIpC,iBAAiB+B,KAAK5B,cAAc8C,oBAAoB,MAAMtB,mBAAmB;QACrF,MAAM+D,SAASJ,MAAML,GAAG,CAAC,CAACU,KAAKT;YAC7B,MAAM,EAAEhE,CAAC,EAAEM,KAAK,EAAE,GAAGV,QAAQ0E,OAAO,CAACN,EAAE,EAAEhC,WAAW;YACpD,MAAM0C,IAAI,IAAIC,UAAU3C;YACxB,IAAK,IAAI7B,IAAI,GAAGA,IAAI6B,WAAW7B,IAAKuE,CAAC,CAACvE,EAAE,GAAGC,KAAKwE,KAAK,CAAC5E,CAAC,CAACG,EAAE,GAAGG;YAC7D,OAAO;gBAAEsB,MAAM6C,IAAI7C,IAAI;gBAAEkB,OAAO2B,IAAI3B,KAAK;gBAAExC;gBAAOuE,QAAQrB,OAAOsB,IAAI,CAACJ,EAAEK,MAAM;YAAE;QAClF;SACA/D,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;QAC9B,MAAMN,MAAMoE,WAAW,CAAC;gBACtB,wFAAwF;YACxF,mFAAmF;YACnFhE,iBAEAA,kBAMAA;aARAA,kBAAAA,QAAQC,MAAM,cAAdD,sCAAAA,gBAAgBE,cAAc;YAC9B,MAAMN,MAAM0D,OAAO,CAACW,YAAY,CAACT;aACjCxD,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;YAC9B,IAAI,AAAC,MAAM7B,QAAQuB,OAAOxB,gCAAiC+B,0BAA0B,MAAMV,mBAAmB;YAC9G,MAAMyE,cAAc,MAAMtE,MAAMU,OAAO,CAAC,CAAC,6DAA6D,EAAE8C,MAAML,GAAG,CAAC,IAAM,8BAA8BoB,IAAI,CAAC,SAAS;YACpK,MAAMC,UAAW,MAAMF,YAAY1D,GAAG,IAAI4C,MAAMiB,OAAO,CAAC,CAACZ,MAAQ;oBAACA,IAAI7C,IAAI;oBAAE6C,IAAI3B,KAAK;iBAAC;YACtF,MAAMwC,aAAa,IAAInD,IAAIiD,QAAQrB,GAAG,CAAC,CAACrC,MAAQ;oBAAC,GAAGA,IAAIE,IAAI,CAAC,EAAE,EAAE2D,OAAO7D,IAAIoB,KAAK,GAAG;oBAAEpB,IAAIsB,gBAAgB;iBAAC;YAC3G,IAAIoB,MAAMoB,IAAI,CAAC,CAACf,MAAQa,WAAWjD,GAAG,CAAC,GAAGoC,IAAI7C,IAAI,CAAC,EAAE,EAAE6C,IAAI3B,KAAK,EAAE,MAAM2B,IAAI1B,QAAQ,GAAG,MAAMtC,mBAAmB;aAChHO,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;QAChC;QACA+C,OAAOwB,IAAI,CAACrF,KAAK6B,GAAG,CAACiC,IAAIpC,SAASqC,QAAQ,EAAEzB,KAAKb,MAAM;IACzD;IACAoC,OAAOyB,MAAM;AACf;AAEA,OAAO,eAAeC,yBAAyB/E,KAAY,EAAEgF,IAAY,EAAE9E,OAA4B;IACrG,MAAM6B,UAAU,AAAC,CAAA,MAAM/B,MAAM0D,OAAO,CAAC3B,OAAO,EAAC,EAAGlB,MAAM,CAAC,CAACC,MAAQZ,QAAQa,GAAG,CAACD,IAAIE,IAAI;IACpF,IAAIe,QAAQd,MAAM,GAAG,GAAG;QACtB,MAAM,IAAI9C,WAAW,mBAAmB,GAAG4D,QAAQd,MAAM,CAAC,+IAA+I,CAAC;IAC5M;AACF;AAOA,2FAA2F;AAC3F,2FAA2F;AAC3F,OAAO,eAAegE,qBAAqBhF,GAAW,EAAEiF,KAAa,EAAE9E,UAA4B,CAAC,CAAC;QACnF,cAEF8E;QACd9E,iBAEAA,kBAGAA;IARA,MAAM+E,WAAU,eAAA,AAAClF,IAAgCkF,OAAO,cAAxC,0BAAA,eAA4C,AAAClF,IAAgCmF,OAAO;IACpG,IAAI,CAACD,SAAS,MAAM,IAAIhH,WAAW,eAAe;IAClD,MAAM2E,OAAO,EAACoC,eAAAA,MAAMG,KAAK,CAAC,gCAAZH,0BAAAA,eAAkC,EAAE,EAAErE,MAAM,CAAC,CAACyE,IAAM,CAAC;YAAC;YAAO;YAAM;YAAO;SAAO,CAACC,QAAQ,CAACD,IAAIf,IAAI,CAAC;KAClHnE,kBAAAA,QAAQC,MAAM,cAAdD,sCAAAA,gBAAgBE,cAAc;IAC9B,MAAMY,WAAW,MAAMpC,YAAYmB,KAAKG;KACxCA,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;IAC9B,MAAMc,YAAY5B,KAAK6B,GAAG,CAACtC,YAAYmC,SAAShC,IAAI;IACpD,MAAMsG,cAAc,MAAMtE,SAASuE,UAAU,CAAC3C,MAAM1C;KACpDA,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;IAC9B,MAAM,EAAElB,GAAGsG,EAAE,EAAE,GAAG1G,QAAQwG,aAAapE,WAAW;IAClD,OAAO;QAAE6C,QAAQyB;QAAItE;IAAU;AACjC;AAEA,gGAAgG;AAChG,8FAA8F;AAC9F,OAAO,SAASuE,mBAAmB3F,KAAY,EAAE4F,QAA+B,EAAEC,KAAa,EAAE3F,OAAqB;IACpH,OAAOF,MAAM0D,OAAO,CAACoC,UAAU,CAACF,SAAS3B,MAAM,EAAE2B,SAASxE,SAAS,EAAEyE,OAAO3F;AAC9E;AAEA,6FAA6F;AAC7F,mEAAmE;AACnE,OAAO,eAAe6F,aAAa/F,KAAY,EAAEgB,IAAY;IAC3D,OAAOhB,MAAM0D,OAAO,CAACsC,SAAS,CAAChF;AACjC;AAEA,+FAA+F;AAC/F,0BAA0B;AAC1B,OAAO,eAAeiF,aAAajG,KAAY,EAAEgF,IAAY,EAAEhE,IAAY,EAAEkF,IAAgE;IAC3I,OAAOlG,MAAM0D,OAAO,CAACyC,OAAO,CAACnF,MAAMkF;AACrC"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/open-source/ai/sensemaking/src/embed/query.ts"],"sourcesContent":["import { type Config, featureSignature, type ResolvedConfig } from '../config/index.ts';\nimport { SenseError } from '../errors.ts';\nimport type { Chunk } from '../features/embed.ts';\nimport { embed } from '../features/embed.ts';\nimport { FEATURES } from '../features/index.ts';\nimport { progress } from '../output/progress.ts';\nimport { parseSource } from '../scan/index.ts';\nimport { FEATURE_SIGNATURE_META_KEY } from '../store/readiness.ts';\nimport { getMeta } from '../store/shared.ts';\nimport { embeddingSignatureCompatible } from '../store/signature.ts';\nimport type { Store, VectorCandidate, VectorSimilar } from '../store/types.ts';\nimport { identifyChunkText, takeChunkText } from './handoff.ts';\nimport { checkLanguageFit } from './langfit.ts';\nimport { getProvider } from './registry.ts';\nimport { type EmbedCallOptions, STORE_DIMS } from './types.ts';\n\n// Slice + re-normalize (Matryoshka); optionally round through int8 storage. Exported for\n// benchmark/lib/embed.mjs, which scores the same lever math offline.\nexport function toStore(full: Float32Array, dims: number, int8: boolean): { v: Float32Array; scale: number } {\n const v = new Float32Array(dims);\n let norm = 0;\n for (let d = 0; d < dims; d++) norm += full[d] * full[d];\n norm = Math.sqrt(norm) + 1e-32;\n for (let d = 0; d < dims; d++) v[d] = full[d] / norm;\n if (!int8) return { v, scale: 1 };\n let max = 0;\n for (let d = 0; d < dims; d++) max = Math.max(max, Math.abs(v[d]));\n return { v, scale: max / 127 || 1 };\n}\n\n// Embed rows whose vector is NULL. Reconcile hands its chunk text over in memory when it ran in\n// this process (handoff.ts); a row left pending by an earlier command re-derives from the exact\n// source string persisted with that indexed generation.\nfunction staleEmbeddingWork(reason: string): SenseError {\n return new SenseError('INDEX_NOT_READY', `${reason}; run \\`sense build\\` (or library build(config)) again to prepare the current index generation`);\n}\n\n// expectedSignature is captured by open() before provider work. Direct internal callers omit it\n// and capture the durable value here, before selecting chunks or constructing a provider.\nexport async function embedPending(store: Store, cfg: Config, allowed?: ReadonlySet<string>, expectedSignature?: string | null, options: EmbedCallOptions = {}): Promise<void> {\n options.signal?.throwIfAborted();\n const expectedDurableSignature = expectedSignature === undefined ? await getMeta(store, FEATURE_SIGNATURE_META_KEY) : expectedSignature;\n options.signal?.throwIfAborted();\n if (!embeddingSignatureCompatible(expectedDurableSignature, featureSignature(cfg, FEATURES))) {\n throw staleEmbeddingWork('the index configuration changed before vectors could be prepared');\n }\n\n const pendingStmt = await store.prepare('SELECT \"path\", chunk, content_identity FROM embeddings WHERE vector IS NULL ORDER BY \"path\", chunk');\n const dirty = ((await pendingStmt.all()) as Array<{ path: string; chunk: number; content_identity: string }>).filter((row) => allowed?.has(row.path) ?? true);\n options.signal?.throwIfAborted();\n if (dirty.length === 0) return;\n\n const provider = await getProvider(cfg, options);\n options.signal?.throwIfAborted();\n const submittedSignature = featureSignature(cfg, FEATURES);\n if (!embeddingSignatureCompatible(expectedDurableSignature, submittedSignature)) {\n throw staleEmbeddingWork('the index configuration changed while the embedding provider was being prepared');\n }\n const storeDims = Math.min(STORE_DIMS, provider.dims);\n\n const byPath = new Map<string, typeof dirty>();\n for (const row of dirty) {\n const list = byPath.get(row.path) ?? [];\n list.push(row);\n byPath.set(row.path, list);\n }\n\n const textByPath = takeChunkText(store) ?? new Map();\n const sourceStmt = await store.prepare('SELECT text FROM indexed_sources WHERE \"path\" = ?');\n\n const jobs: Array<{ path: string; chunk: number; text: string; identity: string }> = [];\n for (const [path, pending] of byPath) {\n const texts = textByPath.get(path);\n if (texts && pending.every((row) => texts[row.chunk]?.identity === row.content_identity)) {\n for (const row of pending) jobs.push({ path, chunk: row.chunk, ...texts[row.chunk] });\n continue;\n }\n // A prior process may have reconciled the row, so its in-memory handoff is gone. Re-derive\n // from that indexed generation's stored source, never from a newer live file.\n const source = (await sourceStmt.get(path)) as { text: string } | undefined;\n options.signal?.throwIfAborted();\n if (!source) throw new SenseError('INDEX_NOT_READY', `indexed source text is missing for ${path}; run \\`sense build --force\\` (or library build(config, { force: true })) before preparing vectors`);\n const chunks = parseSource({ relPath: path, absPath: path, mtimeMs: 0, ctimeMs: 0, size: Buffer.byteLength(source.text), presets: [], embed: true }, source.text, [embed], cfg).doc.extracted.embed as Chunk[];\n for (const row of pending) {\n const chunk = chunks[row.chunk];\n if (!chunk) throw new SenseError('INDEX_NOT_READY', `indexed vector metadata for ${path} does not match its indexed source; run \\`sense build --force\\` (or library build(config, { force: true }))`);\n const identified = identifyChunkText(chunk.text);\n if (identified.identity !== row.content_identity) throw staleEmbeddingWork(`indexed chunk ${path}#${row.chunk} changed before vectors could be prepared`);\n jobs.push({ path, chunk: row.chunk, ...identified });\n }\n }\n\n // Over the exact texts about to be embedded, before any of them are: a mismatch fails\n // this run loudly instead of quietly storing vectors from the wrong model.\n options.signal?.throwIfAborted();\n await checkLanguageFit(\n store,\n provider,\n jobs.map((j) => j.text)\n );\n options.signal?.throwIfAborted();\n\n // Preparing a large configured scope can take long enough to look stalled, so report each\n // completed provider batch.\n const report = progress('embedding chunks', jobs.length);\n for (let i = 0; i < jobs.length; i += provider.batchCap) {\n const batch = jobs.slice(i, i + provider.batchCap);\n options.signal?.throwIfAborted();\n if (featureSignature(cfg, FEATURES) !== submittedSignature) throw staleEmbeddingWork('the embedding configuration changed before a provider batch was submitted');\n const vectors = await provider.embedDocuments(\n batch.map((j) => j.text),\n options\n );\n options.signal?.throwIfAborted();\n if (featureSignature(cfg, FEATURES) !== submittedSignature) throw staleEmbeddingWork('the embedding configuration changed while a provider batch was running');\n const writes = batch.map((job, j) => {\n const { v, scale } = toStore(vectors[j], storeDims, true);\n const q = new Int8Array(storeDims);\n for (let d = 0; d < storeDims; d++) q[d] = Math.round(v[d] / scale);\n return { path: job.path, chunk: job.chunk, scale, vector: Buffer.from(q.buffer) };\n });\n options.signal?.throwIfAborted();\n await store.transaction(async () => {\n // Write first so SQLite owns the write lock before the fence reads. Any mismatch throws\n // inside this transaction and rolls every row in the provider batch back together.\n options.signal?.throwIfAborted();\n await store.vectors.writeVectors(writes);\n options.signal?.throwIfAborted();\n if ((await getMeta(store, FEATURE_SIGNATURE_META_KEY)) !== expectedDurableSignature) throw staleEmbeddingWork('the durable embedding configuration changed while a provider batch was running');\n const currentStmt = await store.prepare(`SELECT \"path\", chunk, content_identity FROM embeddings WHERE ${batch.map(() => '(\"path\" = ? AND chunk = ?)').join(' OR ')}`);\n const current = (await currentStmt.all(...batch.flatMap((job) => [job.path, job.chunk]))) as Array<{ path: string; chunk: number; content_identity: string }>;\n const identities = new Map(current.map((row) => [`${row.path}\\0${Number(row.chunk)}`, row.content_identity]));\n if (batch.some((job) => identities.get(`${job.path}\\0${job.chunk}`) !== job.identity)) throw staleEmbeddingWork('indexed chunks changed while a provider batch was running');\n options.signal?.throwIfAborted();\n });\n report.tick(Math.min(i + provider.batchCap, jobs.length));\n }\n report.finish();\n}\n\nexport async function ensureDocumentEmbeddings(store: Store, _cfg: Config, allowed: ReadonlySet<string>): Promise<void> {\n const pending = (await store.vectors.pending()).filter((row) => allowed.has(row.path));\n if (pending.length > 0) {\n throw new SenseError('INDEX_NOT_READY', `${pending.length} indexed chunk(s) in the requested scope still need vectors; run \\`sense build\\` (or library build(config)) before querying with build disabled`);\n }\n}\n\nexport interface PreparedSemanticQuery {\n readonly vector: Float32Array;\n readonly storeDims: number;\n}\n\n// Provider/model work is independent of the index snapshot. FTS5 operators are stripped as\n// lexical syntax before the query is embedded, matching the prior semantic query behavior.\nexport async function prepareSemanticQuery(cfg: Config, terms: string, options: EmbedCallOptions = {}): Promise<PreparedSemanticQuery> {\n const rootDir = (cfg as Partial<ResolvedConfig>).rootDir ?? (cfg as Partial<ResolvedConfig>).baseDir;\n if (!rootDir) throw new SenseError('EMBED_MODEL', 'semantic expansion needs a resolved config (use loadConfig/open)');\n const text = (terms.match(/[\\p{L}\\p{N}]+/gu) ?? []).filter((t) => !['AND', 'OR', 'NOT', 'NEAR'].includes(t)).join(' ');\n options.signal?.throwIfAborted();\n const provider = await getProvider(cfg, options);\n options.signal?.throwIfAborted();\n const storeDims = Math.min(STORE_DIMS, provider.dims);\n const queryVector = await provider.embedQuery(text, options);\n options.signal?.throwIfAborted();\n const { v: qv } = toStore(queryVector, storeDims, false);\n return { vector: qv, storeDims };\n}\n\n// Best chunk per file by cosine, its line range riding along. The caller owns the read snapshot\n// and the prepared vector, so this performs no provider work and only delegates to the store.\nexport function semanticCandidates(store: Store, prepared: PreparedSemanticQuery, fetch: number, allowed?: Set<string>): Promise<VectorCandidate[]> {\n return store.vectors.candidates(prepared.vector, prepared.storeDims, fetch, allowed);\n}\n\n// Whether a note has any chunk with a vector. Distinguishes \"nothing is near this note\" from\n// \"this note has no text\", which look the same in an empty result.\nexport async function hasEmbedding(store: Store, path: string): Promise<boolean> {\n return store.vectors.hasVector(path);\n}\n\n// Note-to-note similarity is the max cosine over (target chunk, other chunk) pairs, one linear\n// scan of stored vectors.\nexport async function similarNotes(store: Store, _cfg: Config, path: string, opts: { exclude: Set<string>; allowed?: Set<string>; k: number }): Promise<VectorSimilar[]> {\n return store.vectors.similar(path, opts);\n}\n"],"names":["featureSignature","SenseError","embed","FEATURES","progress","parseSource","FEATURE_SIGNATURE_META_KEY","getMeta","embeddingSignatureCompatible","identifyChunkText","takeChunkText","checkLanguageFit","getProvider","STORE_DIMS","toStore","full","dims","int8","v","Float32Array","norm","d","Math","sqrt","scale","max","abs","staleEmbeddingWork","reason","embedPending","store","cfg","allowed","expectedSignature","options","signal","throwIfAborted","expectedDurableSignature","undefined","pendingStmt","prepare","dirty","all","filter","row","has","path","length","provider","submittedSignature","storeDims","min","byPath","Map","list","get","push","set","textByPath","sourceStmt","jobs","pending","texts","every","chunk","identity","content_identity","source","chunks","relPath","absPath","mtimeMs","ctimeMs","size","Buffer","byteLength","text","presets","doc","extracted","identified","map","j","report","i","batchCap","batch","slice","vectors","embedDocuments","writes","job","q","Int8Array","round","vector","from","buffer","transaction","writeVectors","currentStmt","join","current","flatMap","identities","Number","some","tick","finish","ensureDocumentEmbeddings","_cfg","prepareSemanticQuery","terms","rootDir","baseDir","match","t","includes","queryVector","embedQuery","qv","semanticCandidates","prepared","fetch","candidates","hasEmbedding","hasVector","similarNotes","opts","similar"],"mappings":"AAAA,SAAsBA,gBAAgB,QAA6B,qBAAqB;AACxF,SAASC,UAAU,QAAQ,eAAe;AAE1C,SAASC,KAAK,QAAQ,uBAAuB;AAC7C,SAASC,QAAQ,QAAQ,uBAAuB;AAChD,SAASC,QAAQ,QAAQ,wBAAwB;AACjD,SAASC,WAAW,QAAQ,mBAAmB;AAC/C,SAASC,0BAA0B,QAAQ,wBAAwB;AACnE,SAASC,OAAO,QAAQ,qBAAqB;AAC7C,SAASC,4BAA4B,QAAQ,wBAAwB;AAErE,SAASC,iBAAiB,EAAEC,aAAa,QAAQ,eAAe;AAChE,SAASC,gBAAgB,QAAQ,eAAe;AAChD,SAASC,WAAW,QAAQ,gBAAgB;AAC5C,SAAgCC,UAAU,QAAQ,aAAa;AAE/D,yFAAyF;AACzF,qEAAqE;AACrE,OAAO,SAASC,QAAQC,IAAkB,EAAEC,IAAY,EAAEC,IAAa;IACrE,MAAMC,IAAI,IAAIC,aAAaH;IAC3B,IAAII,OAAO;IACX,IAAK,IAAIC,IAAI,GAAGA,IAAIL,MAAMK,IAAKD,QAAQL,IAAI,CAACM,EAAE,GAAGN,IAAI,CAACM,EAAE;IACxDD,OAAOE,KAAKC,IAAI,CAACH,QAAQ;IACzB,IAAK,IAAIC,IAAI,GAAGA,IAAIL,MAAMK,IAAKH,CAAC,CAACG,EAAE,GAAGN,IAAI,CAACM,EAAE,GAAGD;IAChD,IAAI,CAACH,MAAM,OAAO;QAAEC;QAAGM,OAAO;IAAE;IAChC,IAAIC,MAAM;IACV,IAAK,IAAIJ,IAAI,GAAGA,IAAIL,MAAMK,IAAKI,MAAMH,KAAKG,GAAG,CAACA,KAAKH,KAAKI,GAAG,CAACR,CAAC,CAACG,EAAE;IAChE,OAAO;QAAEH;QAAGM,OAAOC,MAAM,OAAO;IAAE;AACpC;AAEA,gGAAgG;AAChG,gGAAgG;AAChG,wDAAwD;AACxD,SAASE,mBAAmBC,MAAc;IACxC,OAAO,IAAI3B,WAAW,mBAAmB,GAAG2B,OAAO,8FAA8F,CAAC;AACpJ;AAEA,gGAAgG;AAChG,0FAA0F;AAC1F,OAAO,eAAeC,aAAaC,KAAY,EAAEC,GAAW,EAAEC,OAA6B,EAAEC,iBAAiC,EAAEC,UAA4B,CAAC,CAAC;QA4BzIxB;QA3BnBwB,iBAEAA,kBAOAA,kBAIAA,kBAuCA,sFAAsF;IACtF,2EAA2E;IAC3EA,kBAMAA;KA5DAA,kBAAAA,QAAQC,MAAM,cAAdD,sCAAAA,gBAAgBE,cAAc;IAC9B,MAAMC,2BAA2BJ,sBAAsBK,YAAY,MAAM/B,QAAQuB,OAAOxB,8BAA8B2B;KACtHC,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;IAC9B,IAAI,CAAC5B,6BAA6B6B,0BAA0BrC,iBAAiB+B,KAAK5B,YAAY;QAC5F,MAAMwB,mBAAmB;IAC3B;IAEA,MAAMY,cAAc,MAAMT,MAAMU,OAAO,CAAC;IACxC,MAAMC,QAAQ,AAAE,CAAA,MAAMF,YAAYG,GAAG,EAAC,EAAwEC,MAAM,CAAC,CAACC;;uBAAQZ,oBAAAA,8BAAAA,QAASa,GAAG,CAACD,IAAIE,IAAI,wCAAK;;KACxJZ,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;IAC9B,IAAIK,MAAMM,MAAM,KAAK,GAAG;IAExB,MAAMC,WAAW,MAAMpC,YAAYmB,KAAKG;KACxCA,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;IAC9B,MAAMa,qBAAqBjD,iBAAiB+B,KAAK5B;IACjD,IAAI,CAACK,6BAA6B6B,0BAA0BY,qBAAqB;QAC/E,MAAMtB,mBAAmB;IAC3B;IACA,MAAMuB,YAAY5B,KAAK6B,GAAG,CAACtC,YAAYmC,SAAShC,IAAI;IAEpD,MAAMoC,SAAS,IAAIC;IACnB,KAAK,MAAMT,OAAOH,MAAO;YACVW;QAAb,MAAME,QAAOF,cAAAA,OAAOG,GAAG,CAACX,IAAIE,IAAI,eAAnBM,yBAAAA,cAAwB,EAAE;QACvCE,KAAKE,IAAI,CAACZ;QACVQ,OAAOK,GAAG,CAACb,IAAIE,IAAI,EAAEQ;IACvB;IAEA,MAAMI,cAAahD,iBAAAA,cAAcoB,oBAAdpB,4BAAAA,iBAAwB,IAAI2C;IAC/C,MAAMM,aAAa,MAAM7B,MAAMU,OAAO,CAAC;IAEvC,MAAMoB,OAA+E,EAAE;IACvF,KAAK,MAAM,CAACd,MAAMe,QAAQ,IAAIT,OAAQ;YASpClB;QARA,MAAM4B,QAAQJ,WAAWH,GAAG,CAACT;QAC7B,IAAIgB,SAASD,QAAQE,KAAK,CAAC,CAACnB;gBAAQkB;mBAAAA,EAAAA,mBAAAA,KAAK,CAAClB,IAAIoB,KAAK,CAAC,cAAhBF,uCAAAA,iBAAkBG,QAAQ,MAAKrB,IAAIsB,gBAAgB;YAAG;YACxF,KAAK,MAAMtB,OAAOiB,QAASD,KAAKJ,IAAI,CAAC;gBAAEV;gBAAMkB,OAAOpB,IAAIoB,KAAK;gBAAE,GAAGF,KAAK,CAAClB,IAAIoB,KAAK,CAAC;YAAC;YACnF;QACF;QACA,2FAA2F;QAC3F,8EAA8E;QAC9E,MAAMG,SAAU,MAAMR,WAAWJ,GAAG,CAACT;SACrCZ,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;QAC9B,IAAI,CAAC+B,QAAQ,MAAM,IAAIlE,WAAW,mBAAmB,CAAC,mCAAmC,EAAE6C,KAAK,kGAAkG,CAAC;QACnM,MAAMsB,SAAS/D,YAAY;YAAEgE,SAASvB;YAAMwB,SAASxB;YAAMyB,SAAS;YAAGC,SAAS;YAAGC,MAAMC,OAAOC,UAAU,CAACR,OAAOS,IAAI;YAAGC,SAAS,EAAE;YAAE3E,OAAO;QAAK,GAAGiE,OAAOS,IAAI,EAAE;YAAC1E;SAAM,EAAE6B,KAAK+C,GAAG,CAACC,SAAS,CAAC7E,KAAK;QACnM,KAAK,MAAM0C,OAAOiB,QAAS;YACzB,MAAMG,QAAQI,MAAM,CAACxB,IAAIoB,KAAK,CAAC;YAC/B,IAAI,CAACA,OAAO,MAAM,IAAI/D,WAAW,mBAAmB,CAAC,4BAA4B,EAAE6C,KAAK,2GAA2G,CAAC;YACpM,MAAMkC,aAAavE,kBAAkBuD,MAAMY,IAAI;YAC/C,IAAII,WAAWf,QAAQ,KAAKrB,IAAIsB,gBAAgB,EAAE,MAAMvC,mBAAmB,CAAC,cAAc,EAAEmB,KAAK,CAAC,EAAEF,IAAIoB,KAAK,CAAC,yCAAyC,CAAC;YACxJJ,KAAKJ,IAAI,CAAC;gBAAEV;gBAAMkB,OAAOpB,IAAIoB,KAAK;gBAAE,GAAGgB,UAAU;YAAC;QACpD;IACF;KAIA9C,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;IAC9B,MAAMzB,iBACJmB,OACAkB,UACAY,KAAKqB,GAAG,CAAC,CAACC,IAAMA,EAAEN,IAAI;KAExB1C,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;IAE9B,0FAA0F;IAC1F,4BAA4B;IAC5B,MAAM+C,SAAS/E,SAAS,oBAAoBwD,KAAKb,MAAM;IACvD,IAAK,IAAIqC,IAAI,GAAGA,IAAIxB,KAAKb,MAAM,EAAEqC,KAAKpC,SAASqC,QAAQ,CAAE;YAEvDnD,kBAMAA,kBAQAA;QAfA,MAAMoD,QAAQ1B,KAAK2B,KAAK,CAACH,GAAGA,IAAIpC,SAASqC,QAAQ;SACjDnD,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;QAC9B,IAAIpC,iBAAiB+B,KAAK5B,cAAc8C,oBAAoB,MAAMtB,mBAAmB;QACrF,MAAM6D,UAAU,MAAMxC,SAASyC,cAAc,CAC3CH,MAAML,GAAG,CAAC,CAACC,IAAMA,EAAEN,IAAI,GACvB1C;SAEFA,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;QAC9B,IAAIpC,iBAAiB+B,KAAK5B,cAAc8C,oBAAoB,MAAMtB,mBAAmB;QACrF,MAAM+D,SAASJ,MAAML,GAAG,CAAC,CAACU,KAAKT;YAC7B,MAAM,EAAEhE,CAAC,EAAEM,KAAK,EAAE,GAAGV,QAAQ0E,OAAO,CAACN,EAAE,EAAEhC,WAAW;YACpD,MAAM0C,IAAI,IAAIC,UAAU3C;YACxB,IAAK,IAAI7B,IAAI,GAAGA,IAAI6B,WAAW7B,IAAKuE,CAAC,CAACvE,EAAE,GAAGC,KAAKwE,KAAK,CAAC5E,CAAC,CAACG,EAAE,GAAGG;YAC7D,OAAO;gBAAEsB,MAAM6C,IAAI7C,IAAI;gBAAEkB,OAAO2B,IAAI3B,KAAK;gBAAExC;gBAAOuE,QAAQrB,OAAOsB,IAAI,CAACJ,EAAEK,MAAM;YAAE;QAClF;SACA/D,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;QAC9B,MAAMN,MAAMoE,WAAW,CAAC;gBACtB,wFAAwF;YACxF,mFAAmF;YACnFhE,iBAEAA,kBAMAA;aARAA,kBAAAA,QAAQC,MAAM,cAAdD,sCAAAA,gBAAgBE,cAAc;YAC9B,MAAMN,MAAM0D,OAAO,CAACW,YAAY,CAACT;aACjCxD,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;YAC9B,IAAI,AAAC,MAAM7B,QAAQuB,OAAOxB,gCAAiC+B,0BAA0B,MAAMV,mBAAmB;YAC9G,MAAMyE,cAAc,MAAMtE,MAAMU,OAAO,CAAC,CAAC,6DAA6D,EAAE8C,MAAML,GAAG,CAAC,IAAM,8BAA8BoB,IAAI,CAAC,SAAS;YACpK,MAAMC,UAAW,MAAMF,YAAY1D,GAAG,IAAI4C,MAAMiB,OAAO,CAAC,CAACZ,MAAQ;oBAACA,IAAI7C,IAAI;oBAAE6C,IAAI3B,KAAK;iBAAC;YACtF,MAAMwC,aAAa,IAAInD,IAAIiD,QAAQrB,GAAG,CAAC,CAACrC,MAAQ;oBAAC,GAAGA,IAAIE,IAAI,CAAC,EAAE,EAAE2D,OAAO7D,IAAIoB,KAAK,GAAG;oBAAEpB,IAAIsB,gBAAgB;iBAAC;YAC3G,IAAIoB,MAAMoB,IAAI,CAAC,CAACf,MAAQa,WAAWjD,GAAG,CAAC,GAAGoC,IAAI7C,IAAI,CAAC,EAAE,EAAE6C,IAAI3B,KAAK,EAAE,MAAM2B,IAAI1B,QAAQ,GAAG,MAAMtC,mBAAmB;aAChHO,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;QAChC;QACA+C,OAAOwB,IAAI,CAACrF,KAAK6B,GAAG,CAACiC,IAAIpC,SAASqC,QAAQ,EAAEzB,KAAKb,MAAM;IACzD;IACAoC,OAAOyB,MAAM;AACf;AAEA,OAAO,eAAeC,yBAAyB/E,KAAY,EAAEgF,IAAY,EAAE9E,OAA4B;IACrG,MAAM6B,UAAU,AAAC,CAAA,MAAM/B,MAAM0D,OAAO,CAAC3B,OAAO,EAAC,EAAGlB,MAAM,CAAC,CAACC,MAAQZ,QAAQa,GAAG,CAACD,IAAIE,IAAI;IACpF,IAAIe,QAAQd,MAAM,GAAG,GAAG;QACtB,MAAM,IAAI9C,WAAW,mBAAmB,GAAG4D,QAAQd,MAAM,CAAC,+IAA+I,CAAC;IAC5M;AACF;AAOA,2FAA2F;AAC3F,2FAA2F;AAC3F,OAAO,eAAegE,qBAAqBhF,GAAW,EAAEiF,KAAa,EAAE9E,UAA4B,CAAC,CAAC;QACnF,cAEF8E;QACd9E,iBAEAA,kBAGAA;IARA,MAAM+E,WAAU,eAAA,AAAClF,IAAgCkF,OAAO,cAAxC,0BAAA,eAA4C,AAAClF,IAAgCmF,OAAO;IACpG,IAAI,CAACD,SAAS,MAAM,IAAIhH,WAAW,eAAe;IAClD,MAAM2E,OAAO,EAACoC,eAAAA,MAAMG,KAAK,CAAC,gCAAZH,0BAAAA,eAAkC,EAAE,EAAErE,MAAM,CAAC,CAACyE,IAAM,CAAC;YAAC;YAAO;YAAM;YAAO;SAAO,CAACC,QAAQ,CAACD,IAAIf,IAAI,CAAC;KAClHnE,kBAAAA,QAAQC,MAAM,cAAdD,sCAAAA,gBAAgBE,cAAc;IAC9B,MAAMY,WAAW,MAAMpC,YAAYmB,KAAKG;KACxCA,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;IAC9B,MAAMc,YAAY5B,KAAK6B,GAAG,CAACtC,YAAYmC,SAAShC,IAAI;IACpD,MAAMsG,cAAc,MAAMtE,SAASuE,UAAU,CAAC3C,MAAM1C;KACpDA,mBAAAA,QAAQC,MAAM,cAAdD,uCAAAA,iBAAgBE,cAAc;IAC9B,MAAM,EAAElB,GAAGsG,EAAE,EAAE,GAAG1G,QAAQwG,aAAapE,WAAW;IAClD,OAAO;QAAE6C,QAAQyB;QAAItE;IAAU;AACjC;AAEA,gGAAgG;AAChG,8FAA8F;AAC9F,OAAO,SAASuE,mBAAmB3F,KAAY,EAAE4F,QAA+B,EAAEC,KAAa,EAAE3F,OAAqB;IACpH,OAAOF,MAAM0D,OAAO,CAACoC,UAAU,CAACF,SAAS3B,MAAM,EAAE2B,SAASxE,SAAS,EAAEyE,OAAO3F;AAC9E;AAEA,6FAA6F;AAC7F,mEAAmE;AACnE,OAAO,eAAe6F,aAAa/F,KAAY,EAAEgB,IAAY;IAC3D,OAAOhB,MAAM0D,OAAO,CAACsC,SAAS,CAAChF;AACjC;AAEA,+FAA+F;AAC/F,0BAA0B;AAC1B,OAAO,eAAeiF,aAAajG,KAAY,EAAEgF,IAAY,EAAEhE,IAAY,EAAEkF,IAAgE;IAC3I,OAAOlG,MAAM0D,OAAO,CAACyC,OAAO,CAACnF,MAAMkF;AACrC"}
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/kevin/Dev/OpenSource/ai/sensemaking/src/embed/registry.ts"],"sourcesContent":["import type { Config } from '../config/index.ts';\nimport { embedConfig } from '../config/index.ts';\nimport { SenseError } from '../errors.ts';\nimport { cohereProvider } from './cohere.ts';\nimport { effectiveEmbedEndpoint } from './endpoint.ts';\nimport { openaiProvider } from './openai.ts';\nimport { staticProvider } from './static.ts';\nimport type { EmbedCallOptions, EmbedProvider } from './types.ts';\n\ntype ProviderFactory = (model: string, url: string | undefined, key: string | undefined) => Promise<EmbedProvider>;\n\n// One factory per wire protocol: adding a protocol is a new file plus an entry here, not\n// an edit to a dispatch ternary.\nconst REGISTRY: Record<string, ProviderFactory> = {\n static: (model) => staticProvider(model),\n openai: (model, url, key) => openaiProvider(model, url, key),\n cohere: (model, url, key) => cohereProvider(model, url, key),\n};\n\nconst providers = new Map<string, Promise<EmbedProvider>>();\n\nfunction aborted(reason: unknown): Promise<never> {\n return Promise.reject(reason);\n}\n\nfunction waitForProvider(provider: Promise<EmbedProvider>, signal: AbortSignal): Promise<EmbedProvider> {\n if (signal.aborted) return aborted(signal.reason);\n return new Promise((resolve, reject) => {\n const cleanup = () => signal.removeEventListener('abort', onAbort);\n const onAbort = () => {\n cleanup();\n reject(signal.reason);\n };\n signal.addEventListener('abort', onAbort, { once: true });\n provider.then(\n (value) => {\n cleanup();\n resolve(value);\n },\n (err: unknown) => {\n cleanup();\n reject(err);\n }\n );\n if (signal.aborted) onAbort();\n });\n}\n\nexport function getProvider(cfg: Config, options: EmbedCallOptions = {}): Promise<EmbedProvider> {\n // A caller that has already left must not start shared model initialization.\n if (options.signal?.aborted) return aborted(options.signal.reason);\n const e = embedConfig(cfg);\n if (!e) throw new SenseError('EMBED_DISABLED', 'this tree has no embedding model: add an \"embed\" block naming one to sense.config.json; a Hugging Face id fetches automatically at first use, or run `sense download` to prefetch it');\n const factory = REGISTRY[e.provider];\n if (!factory) throw new SenseError('EMBED_MODEL', `embed.provider \"${e.provider}\" has no provider implementation in this build`);\n const endpoint = effectiveEmbedEndpoint(e.provider, e.url);\n // Every input that shapes the provider is in the key: two trees sharing an endpoint and\n // model but declaring different languages or key env vars are two providers, not one.\n const sig = JSON.stringify([e.provider, e.model, endpoint ?? '', e.key ?? '', e.languages ?? []]);\n let p = providers.get(sig);\n if (!p) {\n p = factory(e.model, endpoint, e.key);\n // Owner-declared languages override a provider's own: the config is the one home for\n // what no model card can state (http models have no fetchable card here).\n if (e.languages) p = p.then((prov) => ({ ...prov, languages: e.languages }));\n providers.set(sig, p);\n // A rejected construction is evicted, so a transient failure retries instead of\n // poisoning this config for the process lifetime.\n p.catch(() => providers.delete(sig));\n }\n return options.signal ? waitForProvider(p, options.signal) : p;\n}\n"],"names":["embedConfig","SenseError","cohereProvider","effectiveEmbedEndpoint","openaiProvider","staticProvider","REGISTRY","static","model","openai","url","key","cohere","providers","Map","aborted","reason","Promise","reject","waitForProvider","provider","signal","resolve","cleanup","removeEventListener","onAbort","addEventListener","once","then","value","err","getProvider","cfg","options","e","factory","endpoint","sig","JSON","stringify","languages","p","get","prov","set","catch","delete"],"mappings":"AACA,SAASA,WAAW,QAAQ,qBAAqB;AACjD,SAASC,UAAU,QAAQ,eAAe;AAC1C,SAASC,cAAc,QAAQ,cAAc;AAC7C,SAASC,sBAAsB,QAAQ,gBAAgB;AACvD,SAASC,cAAc,QAAQ,cAAc;AAC7C,SAASC,cAAc,QAAQ,cAAc;AAK7C,yFAAyF;AACzF,iCAAiC;AACjC,MAAMC,WAA4C;IAChDC,QAAQ,CAACC,QAAUH,eAAeG;IAClCC,QAAQ,CAACD,OAAOE,KAAKC,MAAQP,eAAeI,OAAOE,KAAKC;IACxDC,QAAQ,CAACJ,OAAOE,KAAKC,MAAQT,eAAeM,OAAOE,KAAKC;AAC1D;AAEA,MAAME,YAAY,IAAIC;AAEtB,SAASC,QAAQC,MAAe;IAC9B,OAAOC,QAAQC,MAAM,CAACF;AACxB;AAEA,SAASG,gBAAgBC,QAAgC,EAAEC,MAAmB;IAC5E,IAAIA,OAAON,OAAO,EAAE,OAAOA,QAAQM,OAAOL,MAAM;IAChD,OAAO,IAAIC,QAAQ,CAACK,SAASJ;QAC3B,MAAMK,UAAU,IAAMF,OAAOG,mBAAmB,CAAC,SAASC;QAC1D,MAAMA,UAAU;YACdF;YACAL,OAAOG,OAAOL,MAAM;QACtB;QACAK,OAAOK,gBAAgB,CAAC,SAASD,SAAS;YAAEE,MAAM;QAAK;QACvDP,SAASQ,IAAI,CACX,CAACC;YACCN;YACAD,QAAQO;QACV,GACA,CAACC;YACCP;YACAL,OAAOY;QACT;QAEF,IAAIT,OAAON,OAAO,EAAEU;IACtB;AACF;AAEA,OAAO,SAASM,YAAYC,GAAW,EAAEC,UAA4B,CAAC,CAAC;QAUJC,QAAaA;QAR1ED;IADJ,6EAA6E;IAC7E,KAAIA,kBAAAA,QAAQZ,MAAM,cAAdY,sCAAAA,gBAAgBlB,OAAO,EAAE,OAAOA,QAAQkB,QAAQZ,MAAM,CAACL,MAAM;IACjE,MAAMkB,IAAIlC,YAAYgC;IACtB,IAAI,CAACE,GAAG,MAAM,IAAIjC,WAAW,kBAAkB;IAC/C,MAAMkC,UAAU7B,QAAQ,CAAC4B,EAAEd,QAAQ,CAAC;IACpC,IAAI,CAACe,SAAS,MAAM,IAAIlC,WAAW,eAAe,CAAC,gBAAgB,EAAEiC,EAAEd,QAAQ,CAAC,8CAA8C,CAAC;IAC/H,MAAMgB,WAAWjC,uBAAuB+B,EAAEd,QAAQ,EAAEc,EAAExB,GAAG;IACzD,wFAAwF;IACxF,sFAAsF;IACtF,MAAM2B,MAAMC,KAAKC,SAAS,CAAC;QAACL,EAAEd,QAAQ;QAAEc,EAAE1B,KAAK;QAAE4B,qBAAAA,sBAAAA,WAAY;SAAIF,SAAAA,EAAEvB,GAAG,cAALuB,oBAAAA,SAAS;SAAIA,eAAAA,EAAEM,SAAS,cAAXN,0BAAAA,eAAe,EAAE;KAAC;IAChG,IAAIO,IAAI5B,UAAU6B,GAAG,CAACL;IACtB,IAAI,CAACI,GAAG;QACNA,IAAIN,QAAQD,EAAE1B,KAAK,EAAE4B,UAAUF,EAAEvB,GAAG;QACpC,qFAAqF;QACrF,0EAA0E;QAC1E,IAAIuB,EAAEM,SAAS,EAAEC,IAAIA,EAAEb,IAAI,CAAC,CAACe,OAAU,CAAA;gBAAE,GAAGA,IAAI;gBAAEH,WAAWN,EAAEM,SAAS;YAAC,CAAA;QACzE3B,UAAU+B,GAAG,CAACP,KAAKI;QACnB,gFAAgF;QAChF,kDAAkD;QAClDA,EAAEI,KAAK,CAAC,IAAMhC,UAAUiC,MAAM,CAACT;IACjC;IACA,OAAOJ,QAAQZ,MAAM,GAAGF,gBAAgBsB,GAAGR,QAAQZ,MAAM,IAAIoB;AAC/D"}
1
+ {"version":3,"sources":["/Users/kevin/Dev/open-source/ai/sensemaking/src/embed/registry.ts"],"sourcesContent":["import type { Config } from '../config/index.ts';\nimport { embedConfig } from '../config/index.ts';\nimport { SenseError } from '../errors.ts';\nimport { cohereProvider } from './cohere.ts';\nimport { effectiveEmbedEndpoint } from './endpoint.ts';\nimport { openaiProvider } from './openai.ts';\nimport { staticProvider } from './static.ts';\nimport type { EmbedCallOptions, EmbedProvider } from './types.ts';\n\ntype ProviderFactory = (model: string, url: string | undefined, key: string | undefined) => Promise<EmbedProvider>;\n\n// One factory per wire protocol: adding a protocol is a new file plus an entry here, not\n// an edit to a dispatch ternary.\nconst REGISTRY: Record<string, ProviderFactory> = {\n static: (model) => staticProvider(model),\n openai: (model, url, key) => openaiProvider(model, url, key),\n cohere: (model, url, key) => cohereProvider(model, url, key),\n};\n\nconst providers = new Map<string, Promise<EmbedProvider>>();\n\nfunction aborted(reason: unknown): Promise<never> {\n return Promise.reject(reason);\n}\n\nfunction waitForProvider(provider: Promise<EmbedProvider>, signal: AbortSignal): Promise<EmbedProvider> {\n if (signal.aborted) return aborted(signal.reason);\n return new Promise((resolve, reject) => {\n const cleanup = () => signal.removeEventListener('abort', onAbort);\n const onAbort = () => {\n cleanup();\n reject(signal.reason);\n };\n signal.addEventListener('abort', onAbort, { once: true });\n provider.then(\n (value) => {\n cleanup();\n resolve(value);\n },\n (err: unknown) => {\n cleanup();\n reject(err);\n }\n );\n if (signal.aborted) onAbort();\n });\n}\n\nexport function getProvider(cfg: Config, options: EmbedCallOptions = {}): Promise<EmbedProvider> {\n // A caller that has already left must not start shared model initialization.\n if (options.signal?.aborted) return aborted(options.signal.reason);\n const e = embedConfig(cfg);\n if (!e) throw new SenseError('EMBED_DISABLED', 'this tree has no embedding model: add an \"embed\" block naming one to sense.config.json; a Hugging Face id fetches automatically at first use, or run `sense download` to prefetch it');\n const factory = REGISTRY[e.provider];\n if (!factory) throw new SenseError('EMBED_MODEL', `embed.provider \"${e.provider}\" has no provider implementation in this build`);\n const endpoint = effectiveEmbedEndpoint(e.provider, e.url);\n // Every input that shapes the provider is in the key: two trees sharing an endpoint and\n // model but declaring different languages or key env vars are two providers, not one.\n const sig = JSON.stringify([e.provider, e.model, endpoint ?? '', e.key ?? '', e.languages ?? []]);\n let p = providers.get(sig);\n if (!p) {\n p = factory(e.model, endpoint, e.key);\n // Owner-declared languages override a provider's own: the config is the one home for\n // what no model card can state (http models have no fetchable card here).\n if (e.languages) p = p.then((prov) => ({ ...prov, languages: e.languages }));\n providers.set(sig, p);\n // A rejected construction is evicted, so a transient failure retries instead of\n // poisoning this config for the process lifetime.\n p.catch(() => providers.delete(sig));\n }\n return options.signal ? waitForProvider(p, options.signal) : p;\n}\n"],"names":["embedConfig","SenseError","cohereProvider","effectiveEmbedEndpoint","openaiProvider","staticProvider","REGISTRY","static","model","openai","url","key","cohere","providers","Map","aborted","reason","Promise","reject","waitForProvider","provider","signal","resolve","cleanup","removeEventListener","onAbort","addEventListener","once","then","value","err","getProvider","cfg","options","e","factory","endpoint","sig","JSON","stringify","languages","p","get","prov","set","catch","delete"],"mappings":"AACA,SAASA,WAAW,QAAQ,qBAAqB;AACjD,SAASC,UAAU,QAAQ,eAAe;AAC1C,SAASC,cAAc,QAAQ,cAAc;AAC7C,SAASC,sBAAsB,QAAQ,gBAAgB;AACvD,SAASC,cAAc,QAAQ,cAAc;AAC7C,SAASC,cAAc,QAAQ,cAAc;AAK7C,yFAAyF;AACzF,iCAAiC;AACjC,MAAMC,WAA4C;IAChDC,QAAQ,CAACC,QAAUH,eAAeG;IAClCC,QAAQ,CAACD,OAAOE,KAAKC,MAAQP,eAAeI,OAAOE,KAAKC;IACxDC,QAAQ,CAACJ,OAAOE,KAAKC,MAAQT,eAAeM,OAAOE,KAAKC;AAC1D;AAEA,MAAME,YAAY,IAAIC;AAEtB,SAASC,QAAQC,MAAe;IAC9B,OAAOC,QAAQC,MAAM,CAACF;AACxB;AAEA,SAASG,gBAAgBC,QAAgC,EAAEC,MAAmB;IAC5E,IAAIA,OAAON,OAAO,EAAE,OAAOA,QAAQM,OAAOL,MAAM;IAChD,OAAO,IAAIC,QAAQ,CAACK,SAASJ;QAC3B,MAAMK,UAAU,IAAMF,OAAOG,mBAAmB,CAAC,SAASC;QAC1D,MAAMA,UAAU;YACdF;YACAL,OAAOG,OAAOL,MAAM;QACtB;QACAK,OAAOK,gBAAgB,CAAC,SAASD,SAAS;YAAEE,MAAM;QAAK;QACvDP,SAASQ,IAAI,CACX,CAACC;YACCN;YACAD,QAAQO;QACV,GACA,CAACC;YACCP;YACAL,OAAOY;QACT;QAEF,IAAIT,OAAON,OAAO,EAAEU;IACtB;AACF;AAEA,OAAO,SAASM,YAAYC,GAAW,EAAEC,UAA4B,CAAC,CAAC;QAUJC,QAAaA;QAR1ED;IADJ,6EAA6E;IAC7E,KAAIA,kBAAAA,QAAQZ,MAAM,cAAdY,sCAAAA,gBAAgBlB,OAAO,EAAE,OAAOA,QAAQkB,QAAQZ,MAAM,CAACL,MAAM;IACjE,MAAMkB,IAAIlC,YAAYgC;IACtB,IAAI,CAACE,GAAG,MAAM,IAAIjC,WAAW,kBAAkB;IAC/C,MAAMkC,UAAU7B,QAAQ,CAAC4B,EAAEd,QAAQ,CAAC;IACpC,IAAI,CAACe,SAAS,MAAM,IAAIlC,WAAW,eAAe,CAAC,gBAAgB,EAAEiC,EAAEd,QAAQ,CAAC,8CAA8C,CAAC;IAC/H,MAAMgB,WAAWjC,uBAAuB+B,EAAEd,QAAQ,EAAEc,EAAExB,GAAG;IACzD,wFAAwF;IACxF,sFAAsF;IACtF,MAAM2B,MAAMC,KAAKC,SAAS,CAAC;QAACL,EAAEd,QAAQ;QAAEc,EAAE1B,KAAK;QAAE4B,qBAAAA,sBAAAA,WAAY;SAAIF,SAAAA,EAAEvB,GAAG,cAALuB,oBAAAA,SAAS;SAAIA,eAAAA,EAAEM,SAAS,cAAXN,0BAAAA,eAAe,EAAE;KAAC;IAChG,IAAIO,IAAI5B,UAAU6B,GAAG,CAACL;IACtB,IAAI,CAACI,GAAG;QACNA,IAAIN,QAAQD,EAAE1B,KAAK,EAAE4B,UAAUF,EAAEvB,GAAG;QACpC,qFAAqF;QACrF,0EAA0E;QAC1E,IAAIuB,EAAEM,SAAS,EAAEC,IAAIA,EAAEb,IAAI,CAAC,CAACe,OAAU,CAAA;gBAAE,GAAGA,IAAI;gBAAEH,WAAWN,EAAEM,SAAS;YAAC,CAAA;QACzE3B,UAAU+B,GAAG,CAACP,KAAKI;QACnB,gFAAgF;QAChF,kDAAkD;QAClDA,EAAEI,KAAK,CAAC,IAAMhC,UAAUiC,MAAM,CAACT;IACjC;IACA,OAAOJ,QAAQZ,MAAM,GAAGF,gBAAgBsB,GAAGR,QAAQZ,MAAM,IAAIoB;AAC/D"}