kreuzberg 4.0.0.pre.rc.6

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 (330) hide show
  1. checksums.yaml +7 -0
  2. data/.gitignore +14 -0
  3. data/.rspec +3 -0
  4. data/.rubocop.yaml +1 -0
  5. data/.rubocop.yml +538 -0
  6. data/Gemfile +8 -0
  7. data/Gemfile.lock +157 -0
  8. data/README.md +426 -0
  9. data/Rakefile +25 -0
  10. data/Steepfile +47 -0
  11. data/examples/async_patterns.rb +341 -0
  12. data/ext/kreuzberg_rb/extconf.rb +45 -0
  13. data/ext/kreuzberg_rb/native/Cargo.lock +6535 -0
  14. data/ext/kreuzberg_rb/native/Cargo.toml +44 -0
  15. data/ext/kreuzberg_rb/native/README.md +425 -0
  16. data/ext/kreuzberg_rb/native/build.rs +15 -0
  17. data/ext/kreuzberg_rb/native/include/ieeefp.h +11 -0
  18. data/ext/kreuzberg_rb/native/include/msvc_compat/strings.h +14 -0
  19. data/ext/kreuzberg_rb/native/include/strings.h +20 -0
  20. data/ext/kreuzberg_rb/native/include/unistd.h +47 -0
  21. data/ext/kreuzberg_rb/native/src/lib.rs +2998 -0
  22. data/extconf.rb +28 -0
  23. data/kreuzberg.gemspec +148 -0
  24. data/lib/kreuzberg/api_proxy.rb +142 -0
  25. data/lib/kreuzberg/cache_api.rb +46 -0
  26. data/lib/kreuzberg/cli.rb +55 -0
  27. data/lib/kreuzberg/cli_proxy.rb +127 -0
  28. data/lib/kreuzberg/config.rb +691 -0
  29. data/lib/kreuzberg/error_context.rb +32 -0
  30. data/lib/kreuzberg/errors.rb +118 -0
  31. data/lib/kreuzberg/extraction_api.rb +85 -0
  32. data/lib/kreuzberg/mcp_proxy.rb +186 -0
  33. data/lib/kreuzberg/ocr_backend_protocol.rb +113 -0
  34. data/lib/kreuzberg/post_processor_protocol.rb +86 -0
  35. data/lib/kreuzberg/result.rb +216 -0
  36. data/lib/kreuzberg/setup_lib_path.rb +80 -0
  37. data/lib/kreuzberg/validator_protocol.rb +89 -0
  38. data/lib/kreuzberg/version.rb +5 -0
  39. data/lib/kreuzberg.rb +103 -0
  40. data/sig/kreuzberg/internal.rbs +184 -0
  41. data/sig/kreuzberg.rbs +520 -0
  42. data/spec/binding/cache_spec.rb +227 -0
  43. data/spec/binding/cli_proxy_spec.rb +85 -0
  44. data/spec/binding/cli_spec.rb +55 -0
  45. data/spec/binding/config_spec.rb +345 -0
  46. data/spec/binding/config_validation_spec.rb +283 -0
  47. data/spec/binding/error_handling_spec.rb +213 -0
  48. data/spec/binding/errors_spec.rb +66 -0
  49. data/spec/binding/plugins/ocr_backend_spec.rb +307 -0
  50. data/spec/binding/plugins/postprocessor_spec.rb +269 -0
  51. data/spec/binding/plugins/validator_spec.rb +274 -0
  52. data/spec/fixtures/config.toml +39 -0
  53. data/spec/fixtures/config.yaml +41 -0
  54. data/spec/fixtures/invalid_config.toml +4 -0
  55. data/spec/smoke/package_spec.rb +178 -0
  56. data/spec/spec_helper.rb +42 -0
  57. data/vendor/kreuzberg/Cargo.toml +204 -0
  58. data/vendor/kreuzberg/README.md +175 -0
  59. data/vendor/kreuzberg/benches/otel_overhead.rs +48 -0
  60. data/vendor/kreuzberg/build.rs +474 -0
  61. data/vendor/kreuzberg/src/api/error.rs +81 -0
  62. data/vendor/kreuzberg/src/api/handlers.rs +199 -0
  63. data/vendor/kreuzberg/src/api/mod.rs +79 -0
  64. data/vendor/kreuzberg/src/api/server.rs +353 -0
  65. data/vendor/kreuzberg/src/api/types.rs +170 -0
  66. data/vendor/kreuzberg/src/cache/mod.rs +1167 -0
  67. data/vendor/kreuzberg/src/chunking/mod.rs +677 -0
  68. data/vendor/kreuzberg/src/core/batch_mode.rs +95 -0
  69. data/vendor/kreuzberg/src/core/config.rs +1032 -0
  70. data/vendor/kreuzberg/src/core/extractor.rs +1024 -0
  71. data/vendor/kreuzberg/src/core/io.rs +329 -0
  72. data/vendor/kreuzberg/src/core/mime.rs +605 -0
  73. data/vendor/kreuzberg/src/core/mod.rs +45 -0
  74. data/vendor/kreuzberg/src/core/pipeline.rs +984 -0
  75. data/vendor/kreuzberg/src/embeddings.rs +432 -0
  76. data/vendor/kreuzberg/src/error.rs +431 -0
  77. data/vendor/kreuzberg/src/extraction/archive.rs +954 -0
  78. data/vendor/kreuzberg/src/extraction/docx.rs +40 -0
  79. data/vendor/kreuzberg/src/extraction/email.rs +854 -0
  80. data/vendor/kreuzberg/src/extraction/excel.rs +688 -0
  81. data/vendor/kreuzberg/src/extraction/html.rs +553 -0
  82. data/vendor/kreuzberg/src/extraction/image.rs +368 -0
  83. data/vendor/kreuzberg/src/extraction/libreoffice.rs +563 -0
  84. data/vendor/kreuzberg/src/extraction/markdown.rs +213 -0
  85. data/vendor/kreuzberg/src/extraction/mod.rs +81 -0
  86. data/vendor/kreuzberg/src/extraction/office_metadata/app_properties.rs +398 -0
  87. data/vendor/kreuzberg/src/extraction/office_metadata/core_properties.rs +247 -0
  88. data/vendor/kreuzberg/src/extraction/office_metadata/custom_properties.rs +240 -0
  89. data/vendor/kreuzberg/src/extraction/office_metadata/mod.rs +130 -0
  90. data/vendor/kreuzberg/src/extraction/office_metadata/odt_properties.rs +287 -0
  91. data/vendor/kreuzberg/src/extraction/pptx.rs +3000 -0
  92. data/vendor/kreuzberg/src/extraction/structured.rs +490 -0
  93. data/vendor/kreuzberg/src/extraction/table.rs +328 -0
  94. data/vendor/kreuzberg/src/extraction/text.rs +269 -0
  95. data/vendor/kreuzberg/src/extraction/xml.rs +333 -0
  96. data/vendor/kreuzberg/src/extractors/archive.rs +446 -0
  97. data/vendor/kreuzberg/src/extractors/bibtex.rs +469 -0
  98. data/vendor/kreuzberg/src/extractors/docbook.rs +502 -0
  99. data/vendor/kreuzberg/src/extractors/docx.rs +367 -0
  100. data/vendor/kreuzberg/src/extractors/email.rs +143 -0
  101. data/vendor/kreuzberg/src/extractors/epub.rs +707 -0
  102. data/vendor/kreuzberg/src/extractors/excel.rs +343 -0
  103. data/vendor/kreuzberg/src/extractors/fictionbook.rs +491 -0
  104. data/vendor/kreuzberg/src/extractors/fictionbook.rs.backup2 +738 -0
  105. data/vendor/kreuzberg/src/extractors/html.rs +393 -0
  106. data/vendor/kreuzberg/src/extractors/image.rs +198 -0
  107. data/vendor/kreuzberg/src/extractors/jats.rs +1051 -0
  108. data/vendor/kreuzberg/src/extractors/jupyter.rs +367 -0
  109. data/vendor/kreuzberg/src/extractors/latex.rs +652 -0
  110. data/vendor/kreuzberg/src/extractors/markdown.rs +700 -0
  111. data/vendor/kreuzberg/src/extractors/mod.rs +365 -0
  112. data/vendor/kreuzberg/src/extractors/odt.rs +628 -0
  113. data/vendor/kreuzberg/src/extractors/opml.rs +634 -0
  114. data/vendor/kreuzberg/src/extractors/orgmode.rs +528 -0
  115. data/vendor/kreuzberg/src/extractors/pdf.rs +493 -0
  116. data/vendor/kreuzberg/src/extractors/pptx.rs +248 -0
  117. data/vendor/kreuzberg/src/extractors/rst.rs +576 -0
  118. data/vendor/kreuzberg/src/extractors/rtf.rs +810 -0
  119. data/vendor/kreuzberg/src/extractors/security.rs +484 -0
  120. data/vendor/kreuzberg/src/extractors/security_tests.rs +367 -0
  121. data/vendor/kreuzberg/src/extractors/structured.rs +140 -0
  122. data/vendor/kreuzberg/src/extractors/text.rs +260 -0
  123. data/vendor/kreuzberg/src/extractors/typst.rs +650 -0
  124. data/vendor/kreuzberg/src/extractors/xml.rs +135 -0
  125. data/vendor/kreuzberg/src/image/dpi.rs +164 -0
  126. data/vendor/kreuzberg/src/image/mod.rs +6 -0
  127. data/vendor/kreuzberg/src/image/preprocessing.rs +417 -0
  128. data/vendor/kreuzberg/src/image/resize.rs +89 -0
  129. data/vendor/kreuzberg/src/keywords/config.rs +154 -0
  130. data/vendor/kreuzberg/src/keywords/mod.rs +237 -0
  131. data/vendor/kreuzberg/src/keywords/processor.rs +267 -0
  132. data/vendor/kreuzberg/src/keywords/rake.rs +293 -0
  133. data/vendor/kreuzberg/src/keywords/types.rs +68 -0
  134. data/vendor/kreuzberg/src/keywords/yake.rs +163 -0
  135. data/vendor/kreuzberg/src/language_detection/mod.rs +942 -0
  136. data/vendor/kreuzberg/src/lib.rs +105 -0
  137. data/vendor/kreuzberg/src/mcp/mod.rs +32 -0
  138. data/vendor/kreuzberg/src/mcp/server.rs +1968 -0
  139. data/vendor/kreuzberg/src/ocr/cache.rs +469 -0
  140. data/vendor/kreuzberg/src/ocr/error.rs +37 -0
  141. data/vendor/kreuzberg/src/ocr/hocr.rs +216 -0
  142. data/vendor/kreuzberg/src/ocr/mod.rs +58 -0
  143. data/vendor/kreuzberg/src/ocr/processor.rs +863 -0
  144. data/vendor/kreuzberg/src/ocr/table/mod.rs +4 -0
  145. data/vendor/kreuzberg/src/ocr/table/tsv_parser.rs +144 -0
  146. data/vendor/kreuzberg/src/ocr/tesseract_backend.rs +450 -0
  147. data/vendor/kreuzberg/src/ocr/types.rs +393 -0
  148. data/vendor/kreuzberg/src/ocr/utils.rs +47 -0
  149. data/vendor/kreuzberg/src/ocr/validation.rs +206 -0
  150. data/vendor/kreuzberg/src/panic_context.rs +154 -0
  151. data/vendor/kreuzberg/src/pdf/error.rs +122 -0
  152. data/vendor/kreuzberg/src/pdf/images.rs +139 -0
  153. data/vendor/kreuzberg/src/pdf/metadata.rs +346 -0
  154. data/vendor/kreuzberg/src/pdf/mod.rs +50 -0
  155. data/vendor/kreuzberg/src/pdf/rendering.rs +369 -0
  156. data/vendor/kreuzberg/src/pdf/table.rs +393 -0
  157. data/vendor/kreuzberg/src/pdf/text.rs +158 -0
  158. data/vendor/kreuzberg/src/plugins/extractor.rs +1013 -0
  159. data/vendor/kreuzberg/src/plugins/mod.rs +209 -0
  160. data/vendor/kreuzberg/src/plugins/ocr.rs +620 -0
  161. data/vendor/kreuzberg/src/plugins/processor.rs +642 -0
  162. data/vendor/kreuzberg/src/plugins/registry.rs +1337 -0
  163. data/vendor/kreuzberg/src/plugins/traits.rs +258 -0
  164. data/vendor/kreuzberg/src/plugins/validator.rs +956 -0
  165. data/vendor/kreuzberg/src/stopwords/mod.rs +1470 -0
  166. data/vendor/kreuzberg/src/text/mod.rs +19 -0
  167. data/vendor/kreuzberg/src/text/quality.rs +697 -0
  168. data/vendor/kreuzberg/src/text/string_utils.rs +217 -0
  169. data/vendor/kreuzberg/src/text/token_reduction/cjk_utils.rs +164 -0
  170. data/vendor/kreuzberg/src/text/token_reduction/config.rs +100 -0
  171. data/vendor/kreuzberg/src/text/token_reduction/core.rs +796 -0
  172. data/vendor/kreuzberg/src/text/token_reduction/filters.rs +902 -0
  173. data/vendor/kreuzberg/src/text/token_reduction/mod.rs +160 -0
  174. data/vendor/kreuzberg/src/text/token_reduction/semantic.rs +619 -0
  175. data/vendor/kreuzberg/src/text/token_reduction/simd_text.rs +147 -0
  176. data/vendor/kreuzberg/src/types.rs +903 -0
  177. data/vendor/kreuzberg/src/utils/mod.rs +17 -0
  178. data/vendor/kreuzberg/src/utils/quality.rs +959 -0
  179. data/vendor/kreuzberg/src/utils/string_utils.rs +381 -0
  180. data/vendor/kreuzberg/stopwords/af_stopwords.json +53 -0
  181. data/vendor/kreuzberg/stopwords/ar_stopwords.json +482 -0
  182. data/vendor/kreuzberg/stopwords/bg_stopwords.json +261 -0
  183. data/vendor/kreuzberg/stopwords/bn_stopwords.json +400 -0
  184. data/vendor/kreuzberg/stopwords/br_stopwords.json +1205 -0
  185. data/vendor/kreuzberg/stopwords/ca_stopwords.json +280 -0
  186. data/vendor/kreuzberg/stopwords/cs_stopwords.json +425 -0
  187. data/vendor/kreuzberg/stopwords/da_stopwords.json +172 -0
  188. data/vendor/kreuzberg/stopwords/de_stopwords.json +622 -0
  189. data/vendor/kreuzberg/stopwords/el_stopwords.json +849 -0
  190. data/vendor/kreuzberg/stopwords/en_stopwords.json +1300 -0
  191. data/vendor/kreuzberg/stopwords/eo_stopwords.json +175 -0
  192. data/vendor/kreuzberg/stopwords/es_stopwords.json +734 -0
  193. data/vendor/kreuzberg/stopwords/et_stopwords.json +37 -0
  194. data/vendor/kreuzberg/stopwords/eu_stopwords.json +100 -0
  195. data/vendor/kreuzberg/stopwords/fa_stopwords.json +801 -0
  196. data/vendor/kreuzberg/stopwords/fi_stopwords.json +849 -0
  197. data/vendor/kreuzberg/stopwords/fr_stopwords.json +693 -0
  198. data/vendor/kreuzberg/stopwords/ga_stopwords.json +111 -0
  199. data/vendor/kreuzberg/stopwords/gl_stopwords.json +162 -0
  200. data/vendor/kreuzberg/stopwords/gu_stopwords.json +226 -0
  201. data/vendor/kreuzberg/stopwords/ha_stopwords.json +41 -0
  202. data/vendor/kreuzberg/stopwords/he_stopwords.json +196 -0
  203. data/vendor/kreuzberg/stopwords/hi_stopwords.json +227 -0
  204. data/vendor/kreuzberg/stopwords/hr_stopwords.json +181 -0
  205. data/vendor/kreuzberg/stopwords/hu_stopwords.json +791 -0
  206. data/vendor/kreuzberg/stopwords/hy_stopwords.json +47 -0
  207. data/vendor/kreuzberg/stopwords/id_stopwords.json +760 -0
  208. data/vendor/kreuzberg/stopwords/it_stopwords.json +634 -0
  209. data/vendor/kreuzberg/stopwords/ja_stopwords.json +136 -0
  210. data/vendor/kreuzberg/stopwords/kn_stopwords.json +84 -0
  211. data/vendor/kreuzberg/stopwords/ko_stopwords.json +681 -0
  212. data/vendor/kreuzberg/stopwords/ku_stopwords.json +64 -0
  213. data/vendor/kreuzberg/stopwords/la_stopwords.json +51 -0
  214. data/vendor/kreuzberg/stopwords/lt_stopwords.json +476 -0
  215. data/vendor/kreuzberg/stopwords/lv_stopwords.json +163 -0
  216. data/vendor/kreuzberg/stopwords/ml_stopwords.json +1 -0
  217. data/vendor/kreuzberg/stopwords/mr_stopwords.json +101 -0
  218. data/vendor/kreuzberg/stopwords/ms_stopwords.json +477 -0
  219. data/vendor/kreuzberg/stopwords/ne_stopwords.json +490 -0
  220. data/vendor/kreuzberg/stopwords/nl_stopwords.json +415 -0
  221. data/vendor/kreuzberg/stopwords/no_stopwords.json +223 -0
  222. data/vendor/kreuzberg/stopwords/pl_stopwords.json +331 -0
  223. data/vendor/kreuzberg/stopwords/pt_stopwords.json +562 -0
  224. data/vendor/kreuzberg/stopwords/ro_stopwords.json +436 -0
  225. data/vendor/kreuzberg/stopwords/ru_stopwords.json +561 -0
  226. data/vendor/kreuzberg/stopwords/si_stopwords.json +193 -0
  227. data/vendor/kreuzberg/stopwords/sk_stopwords.json +420 -0
  228. data/vendor/kreuzberg/stopwords/sl_stopwords.json +448 -0
  229. data/vendor/kreuzberg/stopwords/so_stopwords.json +32 -0
  230. data/vendor/kreuzberg/stopwords/st_stopwords.json +33 -0
  231. data/vendor/kreuzberg/stopwords/sv_stopwords.json +420 -0
  232. data/vendor/kreuzberg/stopwords/sw_stopwords.json +76 -0
  233. data/vendor/kreuzberg/stopwords/ta_stopwords.json +129 -0
  234. data/vendor/kreuzberg/stopwords/te_stopwords.json +54 -0
  235. data/vendor/kreuzberg/stopwords/th_stopwords.json +118 -0
  236. data/vendor/kreuzberg/stopwords/tl_stopwords.json +149 -0
  237. data/vendor/kreuzberg/stopwords/tr_stopwords.json +506 -0
  238. data/vendor/kreuzberg/stopwords/uk_stopwords.json +75 -0
  239. data/vendor/kreuzberg/stopwords/ur_stopwords.json +519 -0
  240. data/vendor/kreuzberg/stopwords/vi_stopwords.json +647 -0
  241. data/vendor/kreuzberg/stopwords/yo_stopwords.json +62 -0
  242. data/vendor/kreuzberg/stopwords/zh_stopwords.json +796 -0
  243. data/vendor/kreuzberg/stopwords/zu_stopwords.json +31 -0
  244. data/vendor/kreuzberg/tests/api_extract_multipart.rs +52 -0
  245. data/vendor/kreuzberg/tests/api_tests.rs +966 -0
  246. data/vendor/kreuzberg/tests/archive_integration.rs +543 -0
  247. data/vendor/kreuzberg/tests/batch_orchestration.rs +556 -0
  248. data/vendor/kreuzberg/tests/batch_processing.rs +316 -0
  249. data/vendor/kreuzberg/tests/bibtex_parity_test.rs +421 -0
  250. data/vendor/kreuzberg/tests/concurrency_stress.rs +525 -0
  251. data/vendor/kreuzberg/tests/config_features.rs +598 -0
  252. data/vendor/kreuzberg/tests/config_loading_tests.rs +415 -0
  253. data/vendor/kreuzberg/tests/core_integration.rs +510 -0
  254. data/vendor/kreuzberg/tests/csv_integration.rs +414 -0
  255. data/vendor/kreuzberg/tests/docbook_extractor_tests.rs +498 -0
  256. data/vendor/kreuzberg/tests/docx_metadata_extraction_test.rs +122 -0
  257. data/vendor/kreuzberg/tests/docx_vs_pandoc_comparison.rs +370 -0
  258. data/vendor/kreuzberg/tests/email_integration.rs +325 -0
  259. data/vendor/kreuzberg/tests/epub_native_extractor_tests.rs +275 -0
  260. data/vendor/kreuzberg/tests/error_handling.rs +393 -0
  261. data/vendor/kreuzberg/tests/fictionbook_extractor_tests.rs +228 -0
  262. data/vendor/kreuzberg/tests/format_integration.rs +159 -0
  263. data/vendor/kreuzberg/tests/helpers/mod.rs +142 -0
  264. data/vendor/kreuzberg/tests/html_table_test.rs +551 -0
  265. data/vendor/kreuzberg/tests/image_integration.rs +253 -0
  266. data/vendor/kreuzberg/tests/instrumentation_test.rs +139 -0
  267. data/vendor/kreuzberg/tests/jats_extractor_tests.rs +639 -0
  268. data/vendor/kreuzberg/tests/jupyter_extractor_tests.rs +704 -0
  269. data/vendor/kreuzberg/tests/keywords_integration.rs +479 -0
  270. data/vendor/kreuzberg/tests/keywords_quality.rs +509 -0
  271. data/vendor/kreuzberg/tests/latex_extractor_tests.rs +496 -0
  272. data/vendor/kreuzberg/tests/markdown_extractor_tests.rs +490 -0
  273. data/vendor/kreuzberg/tests/mime_detection.rs +428 -0
  274. data/vendor/kreuzberg/tests/ocr_configuration.rs +510 -0
  275. data/vendor/kreuzberg/tests/ocr_errors.rs +676 -0
  276. data/vendor/kreuzberg/tests/ocr_quality.rs +627 -0
  277. data/vendor/kreuzberg/tests/ocr_stress.rs +469 -0
  278. data/vendor/kreuzberg/tests/odt_extractor_tests.rs +695 -0
  279. data/vendor/kreuzberg/tests/opml_extractor_tests.rs +616 -0
  280. data/vendor/kreuzberg/tests/orgmode_extractor_tests.rs +822 -0
  281. data/vendor/kreuzberg/tests/pdf_integration.rs +43 -0
  282. data/vendor/kreuzberg/tests/pipeline_integration.rs +1411 -0
  283. data/vendor/kreuzberg/tests/plugin_ocr_backend_test.rs +771 -0
  284. data/vendor/kreuzberg/tests/plugin_postprocessor_test.rs +560 -0
  285. data/vendor/kreuzberg/tests/plugin_system.rs +921 -0
  286. data/vendor/kreuzberg/tests/plugin_validator_test.rs +783 -0
  287. data/vendor/kreuzberg/tests/registry_integration_tests.rs +586 -0
  288. data/vendor/kreuzberg/tests/rst_extractor_tests.rs +692 -0
  289. data/vendor/kreuzberg/tests/rtf_extractor_tests.rs +776 -0
  290. data/vendor/kreuzberg/tests/security_validation.rs +415 -0
  291. data/vendor/kreuzberg/tests/stopwords_integration_test.rs +888 -0
  292. data/vendor/kreuzberg/tests/test_fastembed.rs +609 -0
  293. data/vendor/kreuzberg/tests/typst_behavioral_tests.rs +1259 -0
  294. data/vendor/kreuzberg/tests/typst_extractor_tests.rs +647 -0
  295. data/vendor/kreuzberg/tests/xlsx_metadata_extraction_test.rs +87 -0
  296. data/vendor/rb-sys/.cargo-ok +1 -0
  297. data/vendor/rb-sys/.cargo_vcs_info.json +6 -0
  298. data/vendor/rb-sys/Cargo.lock +393 -0
  299. data/vendor/rb-sys/Cargo.toml +70 -0
  300. data/vendor/rb-sys/Cargo.toml.orig +57 -0
  301. data/vendor/rb-sys/LICENSE-APACHE +190 -0
  302. data/vendor/rb-sys/LICENSE-MIT +21 -0
  303. data/vendor/rb-sys/bin/release.sh +21 -0
  304. data/vendor/rb-sys/build/features.rs +108 -0
  305. data/vendor/rb-sys/build/main.rs +246 -0
  306. data/vendor/rb-sys/build/stable_api_config.rs +153 -0
  307. data/vendor/rb-sys/build/version.rs +48 -0
  308. data/vendor/rb-sys/readme.md +36 -0
  309. data/vendor/rb-sys/src/bindings.rs +21 -0
  310. data/vendor/rb-sys/src/hidden.rs +11 -0
  311. data/vendor/rb-sys/src/lib.rs +34 -0
  312. data/vendor/rb-sys/src/macros.rs +371 -0
  313. data/vendor/rb-sys/src/memory.rs +53 -0
  314. data/vendor/rb-sys/src/ruby_abi_version.rs +38 -0
  315. data/vendor/rb-sys/src/special_consts.rs +31 -0
  316. data/vendor/rb-sys/src/stable_api/compiled.c +179 -0
  317. data/vendor/rb-sys/src/stable_api/compiled.rs +257 -0
  318. data/vendor/rb-sys/src/stable_api/ruby_2_6.rs +316 -0
  319. data/vendor/rb-sys/src/stable_api/ruby_2_7.rs +316 -0
  320. data/vendor/rb-sys/src/stable_api/ruby_3_0.rs +324 -0
  321. data/vendor/rb-sys/src/stable_api/ruby_3_1.rs +317 -0
  322. data/vendor/rb-sys/src/stable_api/ruby_3_2.rs +315 -0
  323. data/vendor/rb-sys/src/stable_api/ruby_3_3.rs +326 -0
  324. data/vendor/rb-sys/src/stable_api/ruby_3_4.rs +327 -0
  325. data/vendor/rb-sys/src/stable_api.rs +261 -0
  326. data/vendor/rb-sys/src/symbol.rs +31 -0
  327. data/vendor/rb-sys/src/tracking_allocator.rs +332 -0
  328. data/vendor/rb-sys/src/utils.rs +89 -0
  329. data/vendor/rb-sys/src/value_type.rs +7 -0
  330. metadata +536 -0
@@ -0,0 +1,956 @@
1
+ //! Validator plugin trait.
2
+ //!
3
+ //! This module defines the trait for implementing custom validation logic.
4
+
5
+ use crate::Result;
6
+ use crate::core::config::ExtractionConfig;
7
+ use crate::plugins::Plugin;
8
+ use crate::types::ExtractionResult;
9
+ use async_trait::async_trait;
10
+ use std::sync::Arc;
11
+
12
+ /// Trait for validator plugins.
13
+ ///
14
+ /// Validators check extraction results for quality, completeness, or correctness.
15
+ /// Unlike post-processors, validator errors **fail fast** - if a validator returns
16
+ /// an error, the extraction fails immediately.
17
+ ///
18
+ /// # Use Cases
19
+ ///
20
+ /// - **Quality Gates**: Ensure extracted content meets minimum quality standards
21
+ /// - **Compliance**: Verify content meets regulatory requirements
22
+ /// - **Content Filtering**: Reject documents containing unwanted content
23
+ /// - **Format Validation**: Verify extracted content structure
24
+ /// - **Security Checks**: Scan for malicious content
25
+ ///
26
+ /// # Error Handling
27
+ ///
28
+ /// Validator errors are **fatal** - they cause the extraction to fail and bubble up
29
+ /// to the caller. Use validators for hard requirements that must be met.
30
+ ///
31
+ /// For non-fatal checks, use post-processors instead.
32
+ ///
33
+ /// # Thread Safety
34
+ ///
35
+ /// Validators must be thread-safe (`Send + Sync`).
36
+ ///
37
+ /// # Example
38
+ ///
39
+ /// ```rust
40
+ /// use kreuzberg::plugins::{Plugin, Validator};
41
+ /// use kreuzberg::{Result, ExtractionResult, ExtractionConfig, KreuzbergError};
42
+ /// use async_trait::async_trait;
43
+ ///
44
+ /// /// Validate that extracted content has minimum length
45
+ /// struct MinimumLengthValidator {
46
+ /// min_length: usize,
47
+ /// }
48
+ ///
49
+ /// impl Plugin for MinimumLengthValidator {
50
+ /// fn name(&self) -> &str { "min-length-validator" }
51
+ /// fn version(&self) -> String { "1.0.0".to_string() }
52
+ /// fn initialize(&self) -> Result<()> { Ok(()) }
53
+ /// fn shutdown(&self) -> Result<()> { Ok(()) }
54
+ /// }
55
+ ///
56
+ /// #[async_trait]
57
+ /// impl Validator for MinimumLengthValidator {
58
+ /// async fn validate(&self, result: &ExtractionResult, config: &ExtractionConfig)
59
+ /// -> Result<()> {
60
+ /// if result.content.len() < self.min_length {
61
+ /// return Err(KreuzbergError::validation(format!(
62
+ /// "Content too short: {} < {} characters",
63
+ /// result.content.len(),
64
+ /// self.min_length
65
+ /// )));
66
+ /// }
67
+ /// Ok(())
68
+ /// }
69
+ /// }
70
+ /// ```
71
+ #[cfg_attr(not(target_arch = "wasm32"), async_trait)]
72
+ #[cfg_attr(target_arch = "wasm32", async_trait(?Send))]
73
+ pub trait Validator: Plugin {
74
+ /// Validate an extraction result.
75
+ ///
76
+ /// Check the extraction result and return `Ok(())` if valid, or an error
77
+ /// if validation fails.
78
+ ///
79
+ /// # Arguments
80
+ ///
81
+ /// * `result` - The extraction result to validate
82
+ /// * `config` - Extraction configuration
83
+ ///
84
+ /// # Returns
85
+ ///
86
+ /// - `Ok(())` if validation passes
87
+ /// - `Err(...)` if validation fails (extraction will fail)
88
+ ///
89
+ /// # Errors
90
+ ///
91
+ /// - `KreuzbergError::Validation` - Validation failed
92
+ /// - Any other error type appropriate for the failure
93
+ ///
94
+ /// # Example - Content Length Validation
95
+ ///
96
+ /// ```rust
97
+ /// # use kreuzberg::plugins::{Plugin, Validator};
98
+ /// # use kreuzberg::{Result, ExtractionResult, ExtractionConfig, KreuzbergError};
99
+ /// # use async_trait::async_trait;
100
+ /// # struct ContentLengthValidator { min: usize, max: usize }
101
+ /// # impl Plugin for ContentLengthValidator {
102
+ /// # fn name(&self) -> &str { "length-validator" }
103
+ /// # fn version(&self) -> String { "1.0.0".to_string() }
104
+ /// # fn initialize(&self) -> Result<()> { Ok(()) }
105
+ /// # fn shutdown(&self) -> Result<()> { Ok(()) }
106
+ /// # }
107
+ /// # #[async_trait]
108
+ /// # impl Validator for ContentLengthValidator {
109
+ /// async fn validate(&self, result: &ExtractionResult, config: &ExtractionConfig)
110
+ /// -> Result<()> {
111
+ /// let length = result.content.len();
112
+ ///
113
+ /// if length < self.min {
114
+ /// return Err(KreuzbergError::validation(format!(
115
+ /// "Content too short: {} < {} characters",
116
+ /// length, self.min
117
+ /// )));
118
+ /// }
119
+ ///
120
+ /// if length > self.max {
121
+ /// return Err(KreuzbergError::validation(format!(
122
+ /// "Content too long: {} > {} characters",
123
+ /// length, self.max
124
+ /// )));
125
+ /// }
126
+ ///
127
+ /// Ok(())
128
+ /// }
129
+ /// # }
130
+ /// ```
131
+ ///
132
+ /// # Example - Quality Score Validation
133
+ ///
134
+ /// ```rust
135
+ /// # use kreuzberg::plugins::{Plugin, Validator};
136
+ /// # use kreuzberg::{Result, ExtractionResult, ExtractionConfig, KreuzbergError};
137
+ /// # use async_trait::async_trait;
138
+ /// # struct QualityValidator { min_score: f64 }
139
+ /// # impl Plugin for QualityValidator {
140
+ /// # fn name(&self) -> &str { "quality-validator" }
141
+ /// # fn version(&self) -> String { "1.0.0".to_string() }
142
+ /// # fn initialize(&self) -> Result<()> { Ok(()) }
143
+ /// # fn shutdown(&self) -> Result<()> { Ok(()) }
144
+ /// # }
145
+ /// # #[async_trait]
146
+ /// # impl Validator for QualityValidator {
147
+ /// async fn validate(&self, result: &ExtractionResult, config: &ExtractionConfig)
148
+ /// -> Result<()> {
149
+ /// // Check if quality_score exists in metadata
150
+ /// let score = result.metadata
151
+ /// .additional
152
+ /// .get("quality_score")
153
+ /// .and_then(|v| v.as_f64())
154
+ /// .unwrap_or(0.0);
155
+ ///
156
+ /// if score < self.min_score {
157
+ /// return Err(KreuzbergError::validation(format!(
158
+ /// "Quality score too low: {} < {}",
159
+ /// score, self.min_score
160
+ /// )));
161
+ /// }
162
+ ///
163
+ /// Ok(())
164
+ /// }
165
+ /// # }
166
+ /// ```
167
+ ///
168
+ /// # Example - Security Validation
169
+ ///
170
+ /// ```rust
171
+ /// # use kreuzberg::plugins::{Plugin, Validator};
172
+ /// # use kreuzberg::{Result, ExtractionResult, ExtractionConfig, KreuzbergError};
173
+ /// # use async_trait::async_trait;
174
+ /// # struct SecurityValidator { blocked_patterns: Vec<String> }
175
+ /// # impl Plugin for SecurityValidator {
176
+ /// # fn name(&self) -> &str { "security-validator" }
177
+ /// # fn version(&self) -> String { "1.0.0".to_string() }
178
+ /// # fn initialize(&self) -> Result<()> { Ok(()) }
179
+ /// # fn shutdown(&self) -> Result<()> { Ok(()) }
180
+ /// # }
181
+ /// # #[async_trait]
182
+ /// # impl Validator for SecurityValidator {
183
+ /// async fn validate(&self, result: &ExtractionResult, config: &ExtractionConfig)
184
+ /// -> Result<()> {
185
+ /// // Check for blocked patterns
186
+ /// for pattern in &self.blocked_patterns {
187
+ /// if result.content.contains(pattern) {
188
+ /// return Err(KreuzbergError::validation(format!(
189
+ /// "Content contains blocked pattern: {}",
190
+ /// pattern
191
+ /// )));
192
+ /// }
193
+ /// }
194
+ ///
195
+ /// Ok(())
196
+ /// }
197
+ /// # }
198
+ /// ```
199
+ async fn validate(&self, result: &ExtractionResult, config: &ExtractionConfig) -> Result<()>;
200
+
201
+ /// Optional: Check if this validator should run for a given result.
202
+ ///
203
+ /// Allows conditional validation based on MIME type, metadata, or content.
204
+ /// Defaults to `true` (always run).
205
+ ///
206
+ /// # Arguments
207
+ ///
208
+ /// * `result` - The extraction result to check
209
+ /// * `config` - Extraction configuration
210
+ ///
211
+ /// # Returns
212
+ ///
213
+ /// `true` if the validator should run, `false` to skip.
214
+ ///
215
+ /// # Example
216
+ ///
217
+ /// ```rust
218
+ /// # use kreuzberg::plugins::{Plugin, Validator};
219
+ /// # use kreuzberg::{Result, ExtractionResult, ExtractionConfig};
220
+ /// # use async_trait::async_trait;
221
+ /// # struct PdfValidator;
222
+ /// # impl Plugin for PdfValidator {
223
+ /// # fn name(&self) -> &str { "pdf-validator" }
224
+ /// # fn version(&self) -> String { "1.0.0".to_string() }
225
+ /// # fn initialize(&self) -> Result<()> { Ok(()) }
226
+ /// # fn shutdown(&self) -> Result<()> { Ok(()) }
227
+ /// # }
228
+ /// # #[async_trait]
229
+ /// # impl Validator for PdfValidator {
230
+ /// # async fn validate(&self, _: &ExtractionResult, _: &ExtractionConfig) -> Result<()> { Ok(()) }
231
+ /// /// Only validate PDF documents
232
+ /// fn should_validate(&self, result: &ExtractionResult, config: &ExtractionConfig) -> bool {
233
+ /// result.mime_type == "application/pdf"
234
+ /// }
235
+ /// # }
236
+ /// ```
237
+ fn should_validate(&self, _result: &ExtractionResult, _config: &ExtractionConfig) -> bool {
238
+ true
239
+ }
240
+
241
+ /// Optional: Get the validation priority.
242
+ ///
243
+ /// Higher priority validators run first. Useful for ordering validation checks
244
+ /// (e.g., run cheap validations before expensive ones).
245
+ ///
246
+ /// Default priority is 50.
247
+ ///
248
+ /// # Returns
249
+ ///
250
+ /// Priority value (higher = runs earlier).
251
+ ///
252
+ /// # Example
253
+ ///
254
+ /// ```rust
255
+ /// # use kreuzberg::plugins::{Plugin, Validator};
256
+ /// # use kreuzberg::{Result, ExtractionResult, ExtractionConfig};
257
+ /// # use async_trait::async_trait;
258
+ /// # struct FastValidator;
259
+ /// # impl Plugin for FastValidator {
260
+ /// # fn name(&self) -> &str { "fast-validator" }
261
+ /// # fn version(&self) -> String { "1.0.0".to_string() }
262
+ /// # fn initialize(&self) -> Result<()> { Ok(()) }
263
+ /// # fn shutdown(&self) -> Result<()> { Ok(()) }
264
+ /// # }
265
+ /// # #[async_trait]
266
+ /// # impl Validator for FastValidator {
267
+ /// # async fn validate(&self, _: &ExtractionResult, _: &ExtractionConfig) -> Result<()> { Ok(()) }
268
+ /// /// Run this validator first (it's fast)
269
+ /// fn priority(&self) -> i32 {
270
+ /// 100
271
+ /// }
272
+ /// # }
273
+ /// ```
274
+ fn priority(&self) -> i32 {
275
+ 50
276
+ }
277
+ }
278
+
279
+ /// Register a validator with the global registry.
280
+ ///
281
+ /// The validator will be registered with its default priority and will be called
282
+ /// during extraction validation. The validator's `name()` method is used as the
283
+ /// registration name.
284
+ ///
285
+ /// # Arguments
286
+ ///
287
+ /// * `validator` - The validator implementation wrapped in Arc
288
+ ///
289
+ /// # Returns
290
+ ///
291
+ /// - `Ok(())` if registration succeeded
292
+ /// - `Err(...)` if validation failed or initialization failed
293
+ ///
294
+ /// # Errors
295
+ ///
296
+ /// - `KreuzbergError::Validation` - Invalid validator name (empty or contains whitespace)
297
+ /// - Any error from the validator's `initialize()` method
298
+ ///
299
+ /// # Example
300
+ ///
301
+ /// ```rust
302
+ /// use kreuzberg::plugins::{Plugin, Validator, register_validator};
303
+ /// use kreuzberg::{Result, ExtractionResult, ExtractionConfig, KreuzbergError};
304
+ /// use async_trait::async_trait;
305
+ /// use std::sync::Arc;
306
+ ///
307
+ /// struct MinLengthValidator { min_length: usize }
308
+ ///
309
+ /// impl Plugin for MinLengthValidator {
310
+ /// fn name(&self) -> &str { "min-length" }
311
+ /// fn version(&self) -> String { "1.0.0".to_string() }
312
+ /// fn initialize(&self) -> Result<()> { Ok(()) }
313
+ /// fn shutdown(&self) -> Result<()> { Ok(()) }
314
+ /// }
315
+ ///
316
+ /// #[async_trait]
317
+ /// impl Validator for MinLengthValidator {
318
+ /// async fn validate(&self, result: &ExtractionResult, _: &ExtractionConfig) -> Result<()> {
319
+ /// if result.content.len() < self.min_length {
320
+ /// return Err(KreuzbergError::validation(
321
+ /// format!("Content too short: {} < {}", result.content.len(), self.min_length)
322
+ /// ));
323
+ /// }
324
+ /// Ok(())
325
+ /// }
326
+ /// }
327
+ ///
328
+ /// # tokio_test::block_on(async {
329
+ /// let validator = Arc::new(MinLengthValidator { min_length: 10 });
330
+ /// register_validator(validator)?;
331
+ /// # Ok::<(), KreuzbergError>(())
332
+ /// # });
333
+ /// ```
334
+ pub fn register_validator(validator: Arc<dyn Validator>) -> crate::Result<()> {
335
+ use crate::plugins::registry::get_validator_registry;
336
+
337
+ let registry = get_validator_registry();
338
+ let mut registry = registry
339
+ .write()
340
+ .expect("~keep Failed to acquire write lock on validator registry"); // ~keep
341
+
342
+ registry.register(validator)
343
+ }
344
+
345
+ /// Unregister a validator by name.
346
+ ///
347
+ /// Removes the validator from the global registry and calls its `shutdown()` method.
348
+ ///
349
+ /// # Arguments
350
+ ///
351
+ /// * `name` - Name of the validator to unregister
352
+ ///
353
+ /// # Returns
354
+ ///
355
+ /// - `Ok(())` if the validator was unregistered or didn't exist
356
+ /// - `Err(...)` if the shutdown method failed
357
+ ///
358
+ /// # Example
359
+ ///
360
+ /// ```rust
361
+ /// use kreuzberg::plugins::unregister_validator;
362
+ ///
363
+ /// # tokio_test::block_on(async {
364
+ /// unregister_validator("min-length")?;
365
+ /// # Ok::<(), kreuzberg::KreuzbergError>(())
366
+ /// # });
367
+ /// ```
368
+ pub fn unregister_validator(name: &str) -> crate::Result<()> {
369
+ use crate::plugins::registry::get_validator_registry;
370
+
371
+ let registry = get_validator_registry();
372
+ let mut registry = registry
373
+ .write()
374
+ .expect("~keep Failed to acquire write lock on validator registry"); // ~keep
375
+
376
+ registry.remove(name)
377
+ }
378
+
379
+ /// List all registered validators.
380
+ ///
381
+ /// Returns the names of all validators currently registered in the global registry.
382
+ ///
383
+ /// # Returns
384
+ ///
385
+ /// A vector of validator names.
386
+ ///
387
+ /// # Example
388
+ ///
389
+ /// ```rust
390
+ /// use kreuzberg::plugins::list_validators;
391
+ ///
392
+ /// # tokio_test::block_on(async {
393
+ /// let validators = list_validators()?;
394
+ /// for name in validators {
395
+ /// println!("Registered validator: {}", name);
396
+ /// }
397
+ /// # Ok::<(), kreuzberg::KreuzbergError>(())
398
+ /// # });
399
+ /// ```
400
+ pub fn list_validators() -> crate::Result<Vec<String>> {
401
+ use crate::plugins::registry::get_validator_registry;
402
+
403
+ let registry = get_validator_registry();
404
+ let registry = registry
405
+ .read()
406
+ .expect("~keep Failed to acquire read lock on validator registry"); // ~keep
407
+
408
+ Ok(registry.list())
409
+ }
410
+
411
+ /// Clear all validators from the global registry.
412
+ ///
413
+ /// Removes all validators and calls their `shutdown()` methods.
414
+ ///
415
+ /// # Returns
416
+ ///
417
+ /// - `Ok(())` if all validators were cleared successfully
418
+ /// - `Err(...)` if any shutdown method failed
419
+ ///
420
+ /// # Example
421
+ ///
422
+ /// ```rust
423
+ /// use kreuzberg::plugins::clear_validators;
424
+ ///
425
+ /// # tokio_test::block_on(async {
426
+ /// clear_validators()?;
427
+ /// # Ok::<(), kreuzberg::KreuzbergError>(())
428
+ /// # });
429
+ /// ```
430
+ pub fn clear_validators() -> crate::Result<()> {
431
+ use crate::plugins::registry::get_validator_registry;
432
+
433
+ let registry = get_validator_registry();
434
+ let mut registry = registry
435
+ .write()
436
+ .expect("~keep Failed to acquire write lock on validator registry"); // ~keep
437
+
438
+ registry.shutdown_all()
439
+ }
440
+
441
+ #[cfg(test)]
442
+ mod tests {
443
+ use super::*;
444
+ use crate::KreuzbergError;
445
+ use std::collections::HashMap;
446
+
447
+ struct MockValidator {
448
+ should_fail: bool,
449
+ }
450
+
451
+ impl Plugin for MockValidator {
452
+ fn name(&self) -> &str {
453
+ "mock-validator"
454
+ }
455
+
456
+ fn version(&self) -> String {
457
+ "1.0.0".to_string()
458
+ }
459
+
460
+ fn initialize(&self) -> Result<()> {
461
+ Ok(())
462
+ }
463
+
464
+ fn shutdown(&self) -> Result<()> {
465
+ Ok(())
466
+ }
467
+ }
468
+
469
+ #[async_trait]
470
+ impl Validator for MockValidator {
471
+ async fn validate(&self, _result: &ExtractionResult, _config: &ExtractionConfig) -> Result<()> {
472
+ if self.should_fail {
473
+ Err(KreuzbergError::validation("Validation failed".to_string()))
474
+ } else {
475
+ Ok(())
476
+ }
477
+ }
478
+ }
479
+
480
+ #[tokio::test]
481
+ async fn test_validator_success() {
482
+ let validator = MockValidator { should_fail: false };
483
+
484
+ let result = ExtractionResult {
485
+ content: "test content".to_string(),
486
+ mime_type: "text/plain".to_string(),
487
+ metadata: crate::types::Metadata::default(),
488
+ tables: vec![],
489
+ detected_languages: None,
490
+ chunks: None,
491
+ images: None,
492
+ };
493
+
494
+ let config = ExtractionConfig::default();
495
+ assert!(validator.validate(&result, &config).await.is_ok());
496
+ }
497
+
498
+ #[tokio::test]
499
+ async fn test_validator_failure() {
500
+ let validator = MockValidator { should_fail: true };
501
+
502
+ let result = ExtractionResult {
503
+ content: "test content".to_string(),
504
+ mime_type: "text/plain".to_string(),
505
+ metadata: crate::types::Metadata::default(),
506
+ tables: vec![],
507
+ detected_languages: None,
508
+ chunks: None,
509
+ images: None,
510
+ };
511
+
512
+ let config = ExtractionConfig::default();
513
+ let validation_result = validator.validate(&result, &config).await;
514
+
515
+ assert!(matches!(validation_result, Err(KreuzbergError::Validation { .. })));
516
+ }
517
+
518
+ #[test]
519
+ fn test_validator_should_validate_default() {
520
+ let validator = MockValidator { should_fail: false };
521
+
522
+ let result = ExtractionResult {
523
+ content: "test".to_string(),
524
+ mime_type: "text/plain".to_string(),
525
+ metadata: crate::types::Metadata::default(),
526
+ tables: vec![],
527
+ detected_languages: None,
528
+ chunks: None,
529
+ images: None,
530
+ };
531
+
532
+ let config = ExtractionConfig::default();
533
+
534
+ assert!(validator.should_validate(&result, &config));
535
+ }
536
+
537
+ #[test]
538
+ fn test_validator_priority_default() {
539
+ let validator = MockValidator { should_fail: false };
540
+ assert_eq!(validator.priority(), 50);
541
+ }
542
+
543
+ #[tokio::test]
544
+ async fn test_validator_plugin_interface() {
545
+ let validator = MockValidator { should_fail: false };
546
+
547
+ assert_eq!(validator.name(), "mock-validator");
548
+ assert_eq!(validator.version(), "1.0.0");
549
+ assert!(validator.initialize().is_ok());
550
+ assert!(validator.shutdown().is_ok());
551
+ }
552
+
553
+ #[tokio::test]
554
+ async fn test_validator_empty_content() {
555
+ let validator = MockValidator { should_fail: false };
556
+
557
+ let result = ExtractionResult {
558
+ content: String::new(),
559
+ mime_type: "text/plain".to_string(),
560
+ metadata: crate::types::Metadata::default(),
561
+ tables: vec![],
562
+ detected_languages: None,
563
+ chunks: None,
564
+ images: None,
565
+ };
566
+
567
+ let config = ExtractionConfig::default();
568
+ assert!(validator.validate(&result, &config).await.is_ok());
569
+ }
570
+
571
+ #[test]
572
+ fn test_validator_should_validate_conditional() {
573
+ struct PdfOnlyValidator;
574
+
575
+ impl Plugin for PdfOnlyValidator {
576
+ fn name(&self) -> &str {
577
+ "pdf-only"
578
+ }
579
+ fn version(&self) -> String {
580
+ "1.0.0".to_string()
581
+ }
582
+ fn initialize(&self) -> Result<()> {
583
+ Ok(())
584
+ }
585
+ fn shutdown(&self) -> Result<()> {
586
+ Ok(())
587
+ }
588
+ }
589
+
590
+ #[async_trait]
591
+ impl Validator for PdfOnlyValidator {
592
+ async fn validate(&self, _result: &ExtractionResult, _config: &ExtractionConfig) -> Result<()> {
593
+ Ok(())
594
+ }
595
+
596
+ fn should_validate(&self, result: &ExtractionResult, _config: &ExtractionConfig) -> bool {
597
+ result.mime_type == "application/pdf"
598
+ }
599
+ }
600
+
601
+ let validator = PdfOnlyValidator;
602
+ let config = ExtractionConfig::default();
603
+
604
+ let pdf_result = ExtractionResult {
605
+ content: "test".to_string(),
606
+ mime_type: "application/pdf".to_string(),
607
+ metadata: crate::types::Metadata::default(),
608
+ tables: vec![],
609
+ detected_languages: None,
610
+ chunks: None,
611
+ images: None,
612
+ };
613
+
614
+ let txt_result = ExtractionResult {
615
+ content: "test".to_string(),
616
+ mime_type: "text/plain".to_string(),
617
+ metadata: crate::types::Metadata::default(),
618
+ tables: vec![],
619
+ detected_languages: None,
620
+ chunks: None,
621
+ images: None,
622
+ };
623
+
624
+ assert!(validator.should_validate(&pdf_result, &config));
625
+ assert!(!validator.should_validate(&txt_result, &config));
626
+ }
627
+
628
+ #[test]
629
+ fn test_validator_priority_ranges() {
630
+ struct HighPriorityValidator;
631
+ struct LowPriorityValidator;
632
+
633
+ impl Plugin for HighPriorityValidator {
634
+ fn name(&self) -> &str {
635
+ "high-priority"
636
+ }
637
+ fn version(&self) -> String {
638
+ "1.0.0".to_string()
639
+ }
640
+ fn initialize(&self) -> Result<()> {
641
+ Ok(())
642
+ }
643
+ fn shutdown(&self) -> Result<()> {
644
+ Ok(())
645
+ }
646
+ }
647
+
648
+ impl Plugin for LowPriorityValidator {
649
+ fn name(&self) -> &str {
650
+ "low-priority"
651
+ }
652
+ fn version(&self) -> String {
653
+ "1.0.0".to_string()
654
+ }
655
+ fn initialize(&self) -> Result<()> {
656
+ Ok(())
657
+ }
658
+ fn shutdown(&self) -> Result<()> {
659
+ Ok(())
660
+ }
661
+ }
662
+
663
+ #[async_trait]
664
+ impl Validator for HighPriorityValidator {
665
+ async fn validate(&self, _result: &ExtractionResult, _config: &ExtractionConfig) -> Result<()> {
666
+ Ok(())
667
+ }
668
+
669
+ fn priority(&self) -> i32 {
670
+ 100
671
+ }
672
+ }
673
+
674
+ #[async_trait]
675
+ impl Validator for LowPriorityValidator {
676
+ async fn validate(&self, _result: &ExtractionResult, _config: &ExtractionConfig) -> Result<()> {
677
+ Ok(())
678
+ }
679
+
680
+ fn priority(&self) -> i32 {
681
+ 10
682
+ }
683
+ }
684
+
685
+ let high = HighPriorityValidator;
686
+ let low = LowPriorityValidator;
687
+
688
+ assert_eq!(high.priority(), 100);
689
+ assert_eq!(low.priority(), 10);
690
+ assert!(high.priority() > low.priority());
691
+ }
692
+
693
+ #[tokio::test]
694
+ async fn test_validator_error_message() {
695
+ let validator = MockValidator { should_fail: true };
696
+
697
+ let result = ExtractionResult {
698
+ content: "test".to_string(),
699
+ mime_type: "text/plain".to_string(),
700
+ metadata: crate::types::Metadata::default(),
701
+ tables: vec![],
702
+ detected_languages: None,
703
+ chunks: None,
704
+ images: None,
705
+ };
706
+
707
+ let config = ExtractionConfig::default();
708
+ let err = validator.validate(&result, &config).await.unwrap_err();
709
+
710
+ match err {
711
+ KreuzbergError::Validation { message: msg, .. } => {
712
+ assert_eq!(msg, "Validation failed");
713
+ }
714
+ _ => panic!("Expected Validation error"),
715
+ }
716
+ }
717
+
718
+ #[tokio::test]
719
+ async fn test_validator_with_metadata() {
720
+ let validator = MockValidator { should_fail: false };
721
+
722
+ let mut additional = HashMap::new();
723
+ additional.insert("quality_score".to_string(), serde_json::json!(0.95));
724
+
725
+ let result = ExtractionResult {
726
+ content: "test".to_string(),
727
+ mime_type: "text/plain".to_string(),
728
+ metadata: crate::types::Metadata {
729
+ additional,
730
+ ..Default::default()
731
+ },
732
+ tables: vec![],
733
+ detected_languages: None,
734
+ chunks: None,
735
+ images: None,
736
+ };
737
+
738
+ let config = ExtractionConfig::default();
739
+ assert!(validator.validate(&result, &config).await.is_ok());
740
+ }
741
+
742
+ #[tokio::test]
743
+ async fn test_validator_with_tables() {
744
+ use crate::types::Table;
745
+
746
+ let validator = MockValidator { should_fail: false };
747
+
748
+ let table = Table {
749
+ cells: vec![vec!["A".to_string(), "B".to_string()]],
750
+ markdown: "| A | B |".to_string(),
751
+ page_number: 0,
752
+ };
753
+
754
+ let result = ExtractionResult {
755
+ content: "test".to_string(),
756
+ mime_type: "text/plain".to_string(),
757
+ metadata: crate::types::Metadata::default(),
758
+ tables: vec![table],
759
+ detected_languages: None,
760
+ chunks: None,
761
+ images: None,
762
+ };
763
+
764
+ let config = ExtractionConfig::default();
765
+ assert!(validator.validate(&result, &config).await.is_ok());
766
+ }
767
+
768
+ #[tokio::test]
769
+ async fn test_validator_different_mime_types() {
770
+ let validator = MockValidator { should_fail: false };
771
+ let config = ExtractionConfig::default();
772
+
773
+ let mime_types = vec![
774
+ "text/plain",
775
+ "application/pdf",
776
+ "application/json",
777
+ "text/html",
778
+ "image/png",
779
+ ];
780
+
781
+ for mime_type in mime_types {
782
+ let result = ExtractionResult {
783
+ content: "test".to_string(),
784
+ mime_type: mime_type.to_string(),
785
+ metadata: crate::types::Metadata::default(),
786
+ tables: vec![],
787
+ detected_languages: None,
788
+ chunks: None,
789
+ images: None,
790
+ };
791
+
792
+ assert!(validator.validate(&result, &config).await.is_ok());
793
+ }
794
+ }
795
+
796
+ #[tokio::test]
797
+ async fn test_validator_long_content() {
798
+ let validator = MockValidator { should_fail: false };
799
+
800
+ let result = ExtractionResult {
801
+ content: "test content ".repeat(10000),
802
+ mime_type: "text/plain".to_string(),
803
+ metadata: crate::types::Metadata::default(),
804
+ tables: vec![],
805
+ detected_languages: None,
806
+ chunks: None,
807
+ images: None,
808
+ };
809
+
810
+ let config = ExtractionConfig::default();
811
+ assert!(validator.validate(&result, &config).await.is_ok());
812
+ }
813
+
814
+ #[test]
815
+ #[serial_test::serial]
816
+ fn test_register_validator() {
817
+ use std::sync::Arc;
818
+
819
+ let validator = Arc::new(MockValidator { should_fail: false });
820
+ let result = super::register_validator(validator);
821
+ assert!(result.is_ok());
822
+
823
+ let _ = super::unregister_validator("mock-validator");
824
+ }
825
+
826
+ #[test]
827
+ #[serial_test::serial]
828
+ fn test_unregister_validator() {
829
+ use std::sync::Arc;
830
+
831
+ let validator = Arc::new(MockValidator { should_fail: false });
832
+ super::register_validator(validator).unwrap();
833
+
834
+ let result = super::unregister_validator("mock-validator");
835
+ assert!(result.is_ok());
836
+ }
837
+
838
+ #[test]
839
+ #[serial_test::serial]
840
+ fn test_unregister_nonexistent_validator() {
841
+ let result = super::unregister_validator("nonexistent-validator-xyz");
842
+ assert!(result.is_ok());
843
+ }
844
+
845
+ #[test]
846
+ #[serial_test::serial]
847
+ fn test_list_validators() {
848
+ use std::sync::Arc;
849
+
850
+ super::clear_validators().unwrap();
851
+
852
+ let validator1 = Arc::new(MockValidator { should_fail: false });
853
+ let validator2 = Arc::new(MockValidator { should_fail: false });
854
+
855
+ let list_before = super::list_validators().unwrap();
856
+ assert_eq!(list_before.len(), 0);
857
+
858
+ super::register_validator(validator1).unwrap();
859
+ super::register_validator(validator2).unwrap();
860
+
861
+ let list = super::list_validators().unwrap();
862
+ assert_eq!(list.len(), 1);
863
+ assert!(list.contains(&"mock-validator".to_string()));
864
+
865
+ super::unregister_validator("mock-validator").unwrap();
866
+ }
867
+
868
+ #[test]
869
+ #[serial_test::serial]
870
+ fn test_clear_validators() {
871
+ use std::sync::Arc;
872
+
873
+ super::clear_validators().unwrap();
874
+
875
+ let validator1 = Arc::new(MockValidator { should_fail: false });
876
+ let validator2 = Arc::new(MockValidator { should_fail: false });
877
+
878
+ super::register_validator(validator1).unwrap();
879
+ super::register_validator(validator2).unwrap();
880
+
881
+ let list_before = super::list_validators().unwrap();
882
+ assert!(!list_before.is_empty());
883
+
884
+ let result = super::clear_validators();
885
+ assert!(result.is_ok());
886
+
887
+ let list = super::list_validators().unwrap();
888
+ assert_eq!(list.len(), 0);
889
+ }
890
+
891
+ #[test]
892
+ #[serial_test::serial]
893
+ fn test_register_validator_with_invalid_name() {
894
+ use std::sync::Arc;
895
+
896
+ struct InvalidNameValidator;
897
+ impl Plugin for InvalidNameValidator {
898
+ fn name(&self) -> &str {
899
+ "invalid name with spaces"
900
+ }
901
+ fn version(&self) -> String {
902
+ "1.0.0".to_string()
903
+ }
904
+ fn initialize(&self) -> Result<()> {
905
+ Ok(())
906
+ }
907
+ fn shutdown(&self) -> Result<()> {
908
+ Ok(())
909
+ }
910
+ }
911
+
912
+ #[async_trait]
913
+ impl Validator for InvalidNameValidator {
914
+ async fn validate(&self, _: &ExtractionResult, _: &ExtractionConfig) -> Result<()> {
915
+ Ok(())
916
+ }
917
+ }
918
+
919
+ let validator = Arc::new(InvalidNameValidator);
920
+ let result = super::register_validator(validator);
921
+ assert!(matches!(result, Err(KreuzbergError::Validation { .. })));
922
+ }
923
+
924
+ #[test]
925
+ #[serial_test::serial]
926
+ fn test_register_validator_with_empty_name() {
927
+ use std::sync::Arc;
928
+
929
+ struct EmptyNameValidator;
930
+ impl Plugin for EmptyNameValidator {
931
+ fn name(&self) -> &str {
932
+ ""
933
+ }
934
+ fn version(&self) -> String {
935
+ "1.0.0".to_string()
936
+ }
937
+ fn initialize(&self) -> Result<()> {
938
+ Ok(())
939
+ }
940
+ fn shutdown(&self) -> Result<()> {
941
+ Ok(())
942
+ }
943
+ }
944
+
945
+ #[async_trait]
946
+ impl Validator for EmptyNameValidator {
947
+ async fn validate(&self, _: &ExtractionResult, _: &ExtractionConfig) -> Result<()> {
948
+ Ok(())
949
+ }
950
+ }
951
+
952
+ let validator = Arc::new(EmptyNameValidator);
953
+ let result = super::register_validator(validator);
954
+ assert!(matches!(result, Err(KreuzbergError::Validation { .. })));
955
+ }
956
+ }