kreuzberg 4.0.0.rc2 → 4.0.1

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 (446) hide show
  1. checksums.yaml +4 -4
  2. data/.gitignore +14 -14
  3. data/.rspec +3 -3
  4. data/.rubocop.yaml +1 -1
  5. data/.rubocop.yml +543 -538
  6. data/Gemfile +8 -8
  7. data/Gemfile.lock +194 -6
  8. data/README.md +391 -426
  9. data/Rakefile +34 -25
  10. data/Steepfile +51 -47
  11. data/examples/async_patterns.rb +283 -341
  12. data/ext/kreuzberg_rb/extconf.rb +65 -45
  13. data/ext/kreuzberg_rb/native/.cargo/config.toml +23 -0
  14. data/ext/kreuzberg_rb/native/Cargo.lock +7619 -6535
  15. data/ext/kreuzberg_rb/native/Cargo.toml +75 -44
  16. data/ext/kreuzberg_rb/native/README.md +425 -425
  17. data/ext/kreuzberg_rb/native/build.rs +15 -15
  18. data/ext/kreuzberg_rb/native/include/ieeefp.h +11 -11
  19. data/ext/kreuzberg_rb/native/include/msvc_compat/strings.h +14 -14
  20. data/ext/kreuzberg_rb/native/include/strings.h +20 -20
  21. data/ext/kreuzberg_rb/native/include/unistd.h +47 -47
  22. data/ext/kreuzberg_rb/native/src/lib.rs +3802 -2998
  23. data/extconf.rb +60 -28
  24. data/kreuzberg.gemspec +199 -148
  25. data/lib/kreuzberg/api_proxy.rb +126 -142
  26. data/lib/kreuzberg/cache_api.rb +67 -46
  27. data/lib/kreuzberg/cli.rb +47 -55
  28. data/lib/kreuzberg/cli_proxy.rb +117 -127
  29. data/lib/kreuzberg/config.rb +936 -691
  30. data/lib/kreuzberg/error_context.rb +136 -32
  31. data/lib/kreuzberg/errors.rb +116 -118
  32. data/lib/kreuzberg/extraction_api.rb +313 -85
  33. data/lib/kreuzberg/mcp_proxy.rb +177 -186
  34. data/lib/kreuzberg/ocr_backend_protocol.rb +40 -113
  35. data/lib/kreuzberg/post_processor_protocol.rb +15 -86
  36. data/lib/kreuzberg/result.rb +334 -216
  37. data/lib/kreuzberg/setup_lib_path.rb +99 -80
  38. data/lib/kreuzberg/types.rb +170 -0
  39. data/lib/kreuzberg/validator_protocol.rb +16 -89
  40. data/lib/kreuzberg/version.rb +5 -5
  41. data/lib/kreuzberg.rb +96 -103
  42. data/lib/libpdfium.so +0 -0
  43. data/sig/kreuzberg/internal.rbs +184 -184
  44. data/sig/kreuzberg.rbs +561 -520
  45. data/spec/binding/async_operations_spec.rb +473 -0
  46. data/spec/binding/batch_operations_spec.rb +595 -0
  47. data/spec/binding/batch_spec.rb +359 -0
  48. data/spec/binding/cache_spec.rb +227 -227
  49. data/spec/binding/cli_proxy_spec.rb +85 -85
  50. data/spec/binding/cli_spec.rb +55 -55
  51. data/spec/binding/config_result_spec.rb +377 -0
  52. data/spec/binding/config_spec.rb +419 -345
  53. data/spec/binding/config_validation_spec.rb +377 -283
  54. data/spec/binding/embeddings_spec.rb +816 -0
  55. data/spec/binding/error_handling_spec.rb +399 -213
  56. data/spec/binding/error_recovery_spec.rb +488 -0
  57. data/spec/binding/errors_spec.rb +66 -66
  58. data/spec/binding/font_config_spec.rb +220 -0
  59. data/spec/binding/images_spec.rb +738 -0
  60. data/spec/binding/keywords_extraction_spec.rb +600 -0
  61. data/spec/binding/metadata_types_spec.rb +1228 -0
  62. data/spec/binding/pages_extraction_spec.rb +471 -0
  63. data/spec/binding/plugins/ocr_backend_spec.rb +307 -307
  64. data/spec/binding/plugins/postprocessor_spec.rb +269 -269
  65. data/spec/binding/plugins/validator_spec.rb +273 -274
  66. data/spec/binding/tables_spec.rb +641 -0
  67. data/spec/fixtures/config.toml +38 -39
  68. data/spec/fixtures/config.yaml +41 -41
  69. data/spec/fixtures/invalid_config.toml +3 -4
  70. data/spec/smoke/package_spec.rb +177 -178
  71. data/spec/spec_helper.rb +40 -42
  72. data/spec/unit/config/chunking_config_spec.rb +213 -0
  73. data/spec/unit/config/embedding_config_spec.rb +343 -0
  74. data/spec/unit/config/extraction_config_spec.rb +438 -0
  75. data/spec/unit/config/font_config_spec.rb +285 -0
  76. data/spec/unit/config/hierarchy_config_spec.rb +314 -0
  77. data/spec/unit/config/image_extraction_config_spec.rb +209 -0
  78. data/spec/unit/config/image_preprocessing_config_spec.rb +249 -0
  79. data/spec/unit/config/keyword_config_spec.rb +229 -0
  80. data/spec/unit/config/language_detection_config_spec.rb +258 -0
  81. data/spec/unit/config/ocr_config_spec.rb +171 -0
  82. data/spec/unit/config/page_config_spec.rb +221 -0
  83. data/spec/unit/config/pdf_config_spec.rb +267 -0
  84. data/spec/unit/config/postprocessor_config_spec.rb +290 -0
  85. data/spec/unit/config/tesseract_config_spec.rb +181 -0
  86. data/spec/unit/config/token_reduction_config_spec.rb +251 -0
  87. data/test/metadata_types_test.rb +959 -0
  88. data/vendor/Cargo.toml +61 -0
  89. data/vendor/kreuzberg/Cargo.toml +259 -204
  90. data/vendor/kreuzberg/README.md +263 -175
  91. data/vendor/kreuzberg/build.rs +782 -474
  92. data/vendor/kreuzberg/examples/bench_fixes.rs +71 -0
  93. data/vendor/kreuzberg/examples/test_pdfium_fork.rs +62 -0
  94. data/vendor/kreuzberg/src/api/error.rs +81 -81
  95. data/vendor/kreuzberg/src/api/handlers.rs +320 -199
  96. data/vendor/kreuzberg/src/api/mod.rs +94 -79
  97. data/vendor/kreuzberg/src/api/server.rs +518 -353
  98. data/vendor/kreuzberg/src/api/types.rs +206 -170
  99. data/vendor/kreuzberg/src/cache/mod.rs +1167 -1167
  100. data/vendor/kreuzberg/src/chunking/mod.rs +2303 -677
  101. data/vendor/kreuzberg/src/chunking/processor.rs +219 -0
  102. data/vendor/kreuzberg/src/core/batch_mode.rs +95 -95
  103. data/vendor/kreuzberg/src/core/batch_optimizations.rs +385 -0
  104. data/vendor/kreuzberg/src/core/config.rs +1914 -1032
  105. data/vendor/kreuzberg/src/core/config_validation.rs +949 -0
  106. data/vendor/kreuzberg/src/core/extractor.rs +1200 -1024
  107. data/vendor/kreuzberg/src/core/formats.rs +235 -0
  108. data/vendor/kreuzberg/src/core/io.rs +329 -329
  109. data/vendor/kreuzberg/src/core/mime.rs +605 -605
  110. data/vendor/kreuzberg/src/core/mod.rs +61 -45
  111. data/vendor/kreuzberg/src/core/pipeline.rs +1223 -984
  112. data/vendor/kreuzberg/src/core/server_config.rs +1220 -0
  113. data/vendor/kreuzberg/src/embeddings.rs +471 -432
  114. data/vendor/kreuzberg/src/error.rs +431 -431
  115. data/vendor/kreuzberg/src/extraction/archive.rs +959 -954
  116. data/vendor/kreuzberg/src/extraction/capacity.rs +263 -0
  117. data/vendor/kreuzberg/src/extraction/docx.rs +404 -40
  118. data/vendor/kreuzberg/src/extraction/email.rs +855 -854
  119. data/vendor/kreuzberg/src/extraction/excel.rs +697 -688
  120. data/vendor/kreuzberg/src/extraction/html.rs +1830 -553
  121. data/vendor/kreuzberg/src/extraction/image.rs +492 -368
  122. data/vendor/kreuzberg/src/extraction/libreoffice.rs +574 -563
  123. data/vendor/kreuzberg/src/extraction/markdown.rs +216 -213
  124. data/vendor/kreuzberg/src/extraction/mod.rs +93 -81
  125. data/vendor/kreuzberg/src/extraction/office_metadata/app_properties.rs +398 -398
  126. data/vendor/kreuzberg/src/extraction/office_metadata/core_properties.rs +247 -247
  127. data/vendor/kreuzberg/src/extraction/office_metadata/custom_properties.rs +240 -240
  128. data/vendor/kreuzberg/src/extraction/office_metadata/mod.rs +130 -130
  129. data/vendor/kreuzberg/src/extraction/office_metadata/odt_properties.rs +284 -287
  130. data/vendor/kreuzberg/src/extraction/pptx.rs +3102 -3000
  131. data/vendor/kreuzberg/src/extraction/structured.rs +491 -490
  132. data/vendor/kreuzberg/src/extraction/table.rs +329 -328
  133. data/vendor/kreuzberg/src/extraction/text.rs +277 -269
  134. data/vendor/kreuzberg/src/extraction/xml.rs +333 -333
  135. data/vendor/kreuzberg/src/extractors/archive.rs +447 -446
  136. data/vendor/kreuzberg/src/extractors/bibtex.rs +470 -469
  137. data/vendor/kreuzberg/src/extractors/docbook.rs +504 -502
  138. data/vendor/kreuzberg/src/extractors/docx.rs +400 -367
  139. data/vendor/kreuzberg/src/extractors/email.rs +157 -143
  140. data/vendor/kreuzberg/src/extractors/epub.rs +696 -707
  141. data/vendor/kreuzberg/src/extractors/excel.rs +385 -343
  142. data/vendor/kreuzberg/src/extractors/fictionbook.rs +492 -491
  143. data/vendor/kreuzberg/src/extractors/html.rs +419 -393
  144. data/vendor/kreuzberg/src/extractors/image.rs +219 -198
  145. data/vendor/kreuzberg/src/extractors/jats.rs +1054 -1051
  146. data/vendor/kreuzberg/src/extractors/jupyter.rs +368 -367
  147. data/vendor/kreuzberg/src/extractors/latex.rs +653 -652
  148. data/vendor/kreuzberg/src/extractors/markdown.rs +701 -700
  149. data/vendor/kreuzberg/src/extractors/mod.rs +429 -365
  150. data/vendor/kreuzberg/src/extractors/odt.rs +628 -628
  151. data/vendor/kreuzberg/src/extractors/opml.rs +635 -634
  152. data/vendor/kreuzberg/src/extractors/orgmode.rs +529 -528
  153. data/vendor/kreuzberg/src/extractors/pdf.rs +761 -493
  154. data/vendor/kreuzberg/src/extractors/pptx.rs +279 -248
  155. data/vendor/kreuzberg/src/extractors/rst.rs +577 -576
  156. data/vendor/kreuzberg/src/extractors/rtf.rs +809 -810
  157. data/vendor/kreuzberg/src/extractors/security.rs +484 -484
  158. data/vendor/kreuzberg/src/extractors/security_tests.rs +367 -367
  159. data/vendor/kreuzberg/src/extractors/structured.rs +142 -140
  160. data/vendor/kreuzberg/src/extractors/text.rs +265 -260
  161. data/vendor/kreuzberg/src/extractors/typst.rs +651 -650
  162. data/vendor/kreuzberg/src/extractors/xml.rs +147 -135
  163. data/vendor/kreuzberg/src/image/dpi.rs +164 -164
  164. data/vendor/kreuzberg/src/image/mod.rs +6 -6
  165. data/vendor/kreuzberg/src/image/preprocessing.rs +417 -417
  166. data/vendor/kreuzberg/src/image/resize.rs +89 -89
  167. data/vendor/kreuzberg/src/keywords/config.rs +154 -154
  168. data/vendor/kreuzberg/src/keywords/mod.rs +237 -237
  169. data/vendor/kreuzberg/src/keywords/processor.rs +275 -267
  170. data/vendor/kreuzberg/src/keywords/rake.rs +293 -293
  171. data/vendor/kreuzberg/src/keywords/types.rs +68 -68
  172. data/vendor/kreuzberg/src/keywords/yake.rs +163 -163
  173. data/vendor/kreuzberg/src/language_detection/mod.rs +985 -942
  174. data/vendor/kreuzberg/src/language_detection/processor.rs +218 -0
  175. data/vendor/kreuzberg/src/lib.rs +114 -105
  176. data/vendor/kreuzberg/src/mcp/mod.rs +35 -32
  177. data/vendor/kreuzberg/src/mcp/server.rs +2090 -1968
  178. data/vendor/kreuzberg/src/ocr/cache.rs +469 -469
  179. data/vendor/kreuzberg/src/ocr/error.rs +37 -37
  180. data/vendor/kreuzberg/src/ocr/hocr.rs +216 -216
  181. data/vendor/kreuzberg/src/ocr/language_registry.rs +520 -0
  182. data/vendor/kreuzberg/src/ocr/mod.rs +60 -58
  183. data/vendor/kreuzberg/src/ocr/processor.rs +858 -863
  184. data/vendor/kreuzberg/src/ocr/table/mod.rs +4 -4
  185. data/vendor/kreuzberg/src/ocr/table/tsv_parser.rs +144 -144
  186. data/vendor/kreuzberg/src/ocr/tesseract_backend.rs +456 -450
  187. data/vendor/kreuzberg/src/ocr/types.rs +393 -393
  188. data/vendor/kreuzberg/src/ocr/utils.rs +47 -47
  189. data/vendor/kreuzberg/src/ocr/validation.rs +206 -206
  190. data/vendor/kreuzberg/src/panic_context.rs +154 -154
  191. data/vendor/kreuzberg/src/pdf/bindings.rs +306 -0
  192. data/vendor/kreuzberg/src/pdf/bundled.rs +408 -0
  193. data/vendor/kreuzberg/src/pdf/error.rs +214 -122
  194. data/vendor/kreuzberg/src/pdf/fonts.rs +358 -0
  195. data/vendor/kreuzberg/src/pdf/hierarchy.rs +903 -0
  196. data/vendor/kreuzberg/src/pdf/images.rs +139 -139
  197. data/vendor/kreuzberg/src/pdf/metadata.rs +509 -346
  198. data/vendor/kreuzberg/src/pdf/mod.rs +81 -50
  199. data/vendor/kreuzberg/src/pdf/rendering.rs +369 -369
  200. data/vendor/kreuzberg/src/pdf/table.rs +417 -393
  201. data/vendor/kreuzberg/src/pdf/text.rs +553 -158
  202. data/vendor/kreuzberg/src/plugins/extractor.rs +1042 -1013
  203. data/vendor/kreuzberg/src/plugins/mod.rs +212 -209
  204. data/vendor/kreuzberg/src/plugins/ocr.rs +637 -620
  205. data/vendor/kreuzberg/src/plugins/processor.rs +650 -642
  206. data/vendor/kreuzberg/src/plugins/registry.rs +1339 -1337
  207. data/vendor/kreuzberg/src/plugins/traits.rs +258 -258
  208. data/vendor/kreuzberg/src/plugins/validator.rs +967 -956
  209. data/vendor/kreuzberg/src/stopwords/mod.rs +1470 -1470
  210. data/vendor/kreuzberg/src/text/mod.rs +27 -19
  211. data/vendor/kreuzberg/src/text/quality.rs +710 -697
  212. data/vendor/kreuzberg/src/text/quality_processor.rs +231 -0
  213. data/vendor/kreuzberg/src/text/string_utils.rs +229 -217
  214. data/vendor/kreuzberg/src/text/token_reduction/cjk_utils.rs +164 -164
  215. data/vendor/kreuzberg/src/text/token_reduction/config.rs +100 -100
  216. data/vendor/kreuzberg/src/text/token_reduction/core.rs +832 -796
  217. data/vendor/kreuzberg/src/text/token_reduction/filters.rs +923 -902
  218. data/vendor/kreuzberg/src/text/token_reduction/mod.rs +160 -160
  219. data/vendor/kreuzberg/src/text/token_reduction/semantic.rs +619 -619
  220. data/vendor/kreuzberg/src/text/token_reduction/simd_text.rs +148 -147
  221. data/vendor/kreuzberg/src/text/utf8_validation.rs +193 -0
  222. data/vendor/kreuzberg/src/types.rs +1713 -903
  223. data/vendor/kreuzberg/src/utils/mod.rs +31 -17
  224. data/vendor/kreuzberg/src/utils/pool.rs +503 -0
  225. data/vendor/kreuzberg/src/utils/pool_sizing.rs +364 -0
  226. data/vendor/kreuzberg/src/utils/quality.rs +968 -959
  227. data/vendor/kreuzberg/src/utils/string_pool.rs +761 -0
  228. data/vendor/kreuzberg/src/utils/string_utils.rs +381 -381
  229. data/vendor/kreuzberg/stopwords/af_stopwords.json +53 -53
  230. data/vendor/kreuzberg/stopwords/ar_stopwords.json +482 -482
  231. data/vendor/kreuzberg/stopwords/bg_stopwords.json +261 -261
  232. data/vendor/kreuzberg/stopwords/bn_stopwords.json +400 -400
  233. data/vendor/kreuzberg/stopwords/br_stopwords.json +1205 -1205
  234. data/vendor/kreuzberg/stopwords/ca_stopwords.json +280 -280
  235. data/vendor/kreuzberg/stopwords/cs_stopwords.json +425 -425
  236. data/vendor/kreuzberg/stopwords/da_stopwords.json +172 -172
  237. data/vendor/kreuzberg/stopwords/de_stopwords.json +622 -622
  238. data/vendor/kreuzberg/stopwords/el_stopwords.json +849 -849
  239. data/vendor/kreuzberg/stopwords/en_stopwords.json +1300 -1300
  240. data/vendor/kreuzberg/stopwords/eo_stopwords.json +175 -175
  241. data/vendor/kreuzberg/stopwords/es_stopwords.json +734 -734
  242. data/vendor/kreuzberg/stopwords/et_stopwords.json +37 -37
  243. data/vendor/kreuzberg/stopwords/eu_stopwords.json +100 -100
  244. data/vendor/kreuzberg/stopwords/fa_stopwords.json +801 -801
  245. data/vendor/kreuzberg/stopwords/fi_stopwords.json +849 -849
  246. data/vendor/kreuzberg/stopwords/fr_stopwords.json +693 -693
  247. data/vendor/kreuzberg/stopwords/ga_stopwords.json +111 -111
  248. data/vendor/kreuzberg/stopwords/gl_stopwords.json +162 -162
  249. data/vendor/kreuzberg/stopwords/gu_stopwords.json +226 -226
  250. data/vendor/kreuzberg/stopwords/ha_stopwords.json +41 -41
  251. data/vendor/kreuzberg/stopwords/he_stopwords.json +196 -196
  252. data/vendor/kreuzberg/stopwords/hi_stopwords.json +227 -227
  253. data/vendor/kreuzberg/stopwords/hr_stopwords.json +181 -181
  254. data/vendor/kreuzberg/stopwords/hu_stopwords.json +791 -791
  255. data/vendor/kreuzberg/stopwords/hy_stopwords.json +47 -47
  256. data/vendor/kreuzberg/stopwords/id_stopwords.json +760 -760
  257. data/vendor/kreuzberg/stopwords/it_stopwords.json +634 -634
  258. data/vendor/kreuzberg/stopwords/ja_stopwords.json +136 -136
  259. data/vendor/kreuzberg/stopwords/kn_stopwords.json +84 -84
  260. data/vendor/kreuzberg/stopwords/ko_stopwords.json +681 -681
  261. data/vendor/kreuzberg/stopwords/ku_stopwords.json +64 -64
  262. data/vendor/kreuzberg/stopwords/la_stopwords.json +51 -51
  263. data/vendor/kreuzberg/stopwords/lt_stopwords.json +476 -476
  264. data/vendor/kreuzberg/stopwords/lv_stopwords.json +163 -163
  265. data/vendor/kreuzberg/stopwords/ml_stopwords.json +1 -1
  266. data/vendor/kreuzberg/stopwords/mr_stopwords.json +101 -101
  267. data/vendor/kreuzberg/stopwords/ms_stopwords.json +477 -477
  268. data/vendor/kreuzberg/stopwords/ne_stopwords.json +490 -490
  269. data/vendor/kreuzberg/stopwords/nl_stopwords.json +415 -415
  270. data/vendor/kreuzberg/stopwords/no_stopwords.json +223 -223
  271. data/vendor/kreuzberg/stopwords/pl_stopwords.json +331 -331
  272. data/vendor/kreuzberg/stopwords/pt_stopwords.json +562 -562
  273. data/vendor/kreuzberg/stopwords/ro_stopwords.json +436 -436
  274. data/vendor/kreuzberg/stopwords/ru_stopwords.json +561 -561
  275. data/vendor/kreuzberg/stopwords/si_stopwords.json +193 -193
  276. data/vendor/kreuzberg/stopwords/sk_stopwords.json +420 -420
  277. data/vendor/kreuzberg/stopwords/sl_stopwords.json +448 -448
  278. data/vendor/kreuzberg/stopwords/so_stopwords.json +32 -32
  279. data/vendor/kreuzberg/stopwords/st_stopwords.json +33 -33
  280. data/vendor/kreuzberg/stopwords/sv_stopwords.json +420 -420
  281. data/vendor/kreuzberg/stopwords/sw_stopwords.json +76 -76
  282. data/vendor/kreuzberg/stopwords/ta_stopwords.json +129 -129
  283. data/vendor/kreuzberg/stopwords/te_stopwords.json +54 -54
  284. data/vendor/kreuzberg/stopwords/th_stopwords.json +118 -118
  285. data/vendor/kreuzberg/stopwords/tl_stopwords.json +149 -149
  286. data/vendor/kreuzberg/stopwords/tr_stopwords.json +506 -506
  287. data/vendor/kreuzberg/stopwords/uk_stopwords.json +75 -75
  288. data/vendor/kreuzberg/stopwords/ur_stopwords.json +519 -519
  289. data/vendor/kreuzberg/stopwords/vi_stopwords.json +647 -647
  290. data/vendor/kreuzberg/stopwords/yo_stopwords.json +62 -62
  291. data/vendor/kreuzberg/stopwords/zh_stopwords.json +796 -796
  292. data/vendor/kreuzberg/stopwords/zu_stopwords.json +31 -31
  293. data/vendor/kreuzberg/tests/api_embed.rs +360 -0
  294. data/vendor/kreuzberg/tests/api_extract_multipart.rs +52 -52
  295. data/vendor/kreuzberg/tests/api_large_pdf_extraction.rs +471 -0
  296. data/vendor/kreuzberg/tests/api_large_pdf_extraction_diagnostics.rs +289 -0
  297. data/vendor/kreuzberg/tests/api_tests.rs +1472 -966
  298. data/vendor/kreuzberg/tests/archive_integration.rs +545 -543
  299. data/vendor/kreuzberg/tests/batch_orchestration.rs +587 -556
  300. data/vendor/kreuzberg/tests/batch_pooling_benchmark.rs +154 -0
  301. data/vendor/kreuzberg/tests/batch_processing.rs +328 -316
  302. data/vendor/kreuzberg/tests/bibtex_parity_test.rs +421 -421
  303. data/vendor/kreuzberg/tests/concurrency_stress.rs +541 -525
  304. data/vendor/kreuzberg/tests/config_features.rs +612 -598
  305. data/vendor/kreuzberg/tests/config_integration_test.rs +753 -0
  306. data/vendor/kreuzberg/tests/config_loading_tests.rs +416 -415
  307. data/vendor/kreuzberg/tests/core_integration.rs +519 -510
  308. data/vendor/kreuzberg/tests/csv_integration.rs +414 -414
  309. data/vendor/kreuzberg/tests/data/hierarchy_ground_truth.json +294 -0
  310. data/vendor/kreuzberg/tests/docbook_extractor_tests.rs +500 -498
  311. data/vendor/kreuzberg/tests/docx_metadata_extraction_test.rs +122 -122
  312. data/vendor/kreuzberg/tests/docx_vs_pandoc_comparison.rs +370 -370
  313. data/vendor/kreuzberg/tests/email_integration.rs +327 -325
  314. data/vendor/kreuzberg/tests/epub_native_extractor_tests.rs +275 -275
  315. data/vendor/kreuzberg/tests/error_handling.rs +402 -393
  316. data/vendor/kreuzberg/tests/fictionbook_extractor_tests.rs +228 -228
  317. data/vendor/kreuzberg/tests/format_integration.rs +165 -159
  318. data/vendor/kreuzberg/tests/helpers/mod.rs +202 -142
  319. data/vendor/kreuzberg/tests/html_table_test.rs +551 -551
  320. data/vendor/kreuzberg/tests/image_integration.rs +255 -253
  321. data/vendor/kreuzberg/tests/instrumentation_test.rs +139 -139
  322. data/vendor/kreuzberg/tests/jats_extractor_tests.rs +639 -639
  323. data/vendor/kreuzberg/tests/jupyter_extractor_tests.rs +704 -704
  324. data/vendor/kreuzberg/tests/keywords_integration.rs +479 -479
  325. data/vendor/kreuzberg/tests/keywords_quality.rs +509 -509
  326. data/vendor/kreuzberg/tests/latex_extractor_tests.rs +496 -496
  327. data/vendor/kreuzberg/tests/markdown_extractor_tests.rs +490 -490
  328. data/vendor/kreuzberg/tests/mime_detection.rs +429 -428
  329. data/vendor/kreuzberg/tests/ocr_configuration.rs +514 -510
  330. data/vendor/kreuzberg/tests/ocr_errors.rs +698 -676
  331. data/vendor/kreuzberg/tests/ocr_language_registry.rs +191 -0
  332. data/vendor/kreuzberg/tests/ocr_quality.rs +629 -627
  333. data/vendor/kreuzberg/tests/ocr_stress.rs +469 -469
  334. data/vendor/kreuzberg/tests/odt_extractor_tests.rs +674 -695
  335. data/vendor/kreuzberg/tests/opml_extractor_tests.rs +616 -616
  336. data/vendor/kreuzberg/tests/orgmode_extractor_tests.rs +822 -822
  337. data/vendor/kreuzberg/tests/page_markers.rs +297 -0
  338. data/vendor/kreuzberg/tests/pdf_hierarchy_detection.rs +301 -0
  339. data/vendor/kreuzberg/tests/pdf_hierarchy_quality.rs +589 -0
  340. data/vendor/kreuzberg/tests/pdf_integration.rs +45 -43
  341. data/vendor/kreuzberg/tests/pdf_ocr_triggering.rs +301 -0
  342. data/vendor/kreuzberg/tests/pdf_text_merging.rs +475 -0
  343. data/vendor/kreuzberg/tests/pdfium_linking.rs +340 -0
  344. data/vendor/kreuzberg/tests/pipeline_integration.rs +1446 -1411
  345. data/vendor/kreuzberg/tests/plugin_ocr_backend_test.rs +776 -771
  346. data/vendor/kreuzberg/tests/plugin_postprocessor_test.rs +577 -560
  347. data/vendor/kreuzberg/tests/plugin_system.rs +927 -921
  348. data/vendor/kreuzberg/tests/plugin_validator_test.rs +783 -783
  349. data/vendor/kreuzberg/tests/registry_integration_tests.rs +587 -586
  350. data/vendor/kreuzberg/tests/rst_extractor_tests.rs +694 -692
  351. data/vendor/kreuzberg/tests/rtf_extractor_tests.rs +775 -776
  352. data/vendor/kreuzberg/tests/security_validation.rs +416 -415
  353. data/vendor/kreuzberg/tests/stopwords_integration_test.rs +888 -888
  354. data/vendor/kreuzberg/tests/test_fastembed.rs +631 -609
  355. data/vendor/kreuzberg/tests/typst_behavioral_tests.rs +1260 -1259
  356. data/vendor/kreuzberg/tests/typst_extractor_tests.rs +648 -647
  357. data/vendor/kreuzberg/tests/xlsx_metadata_extraction_test.rs +87 -87
  358. data/vendor/kreuzberg-ffi/Cargo.toml +67 -0
  359. data/vendor/kreuzberg-ffi/README.md +851 -0
  360. data/vendor/kreuzberg-ffi/benches/result_view_benchmark.rs +227 -0
  361. data/vendor/kreuzberg-ffi/build.rs +168 -0
  362. data/vendor/kreuzberg-ffi/cbindgen.toml +37 -0
  363. data/vendor/kreuzberg-ffi/kreuzberg-ffi.pc.in +12 -0
  364. data/vendor/kreuzberg-ffi/kreuzberg.h +3012 -0
  365. data/vendor/kreuzberg-ffi/src/batch_streaming.rs +588 -0
  366. data/vendor/kreuzberg-ffi/src/config.rs +1341 -0
  367. data/vendor/kreuzberg-ffi/src/error.rs +901 -0
  368. data/vendor/kreuzberg-ffi/src/extraction.rs +555 -0
  369. data/vendor/kreuzberg-ffi/src/helpers.rs +879 -0
  370. data/vendor/kreuzberg-ffi/src/lib.rs +977 -0
  371. data/vendor/kreuzberg-ffi/src/memory.rs +493 -0
  372. data/vendor/kreuzberg-ffi/src/mime.rs +329 -0
  373. data/vendor/kreuzberg-ffi/src/panic_shield.rs +265 -0
  374. data/vendor/kreuzberg-ffi/src/plugins/document_extractor.rs +442 -0
  375. data/vendor/kreuzberg-ffi/src/plugins/mod.rs +14 -0
  376. data/vendor/kreuzberg-ffi/src/plugins/ocr_backend.rs +628 -0
  377. data/vendor/kreuzberg-ffi/src/plugins/post_processor.rs +438 -0
  378. data/vendor/kreuzberg-ffi/src/plugins/validator.rs +329 -0
  379. data/vendor/kreuzberg-ffi/src/result.rs +510 -0
  380. data/vendor/kreuzberg-ffi/src/result_pool.rs +639 -0
  381. data/vendor/kreuzberg-ffi/src/result_view.rs +773 -0
  382. data/vendor/kreuzberg-ffi/src/string_intern.rs +568 -0
  383. data/vendor/kreuzberg-ffi/src/types.rs +363 -0
  384. data/vendor/kreuzberg-ffi/src/util.rs +210 -0
  385. data/vendor/kreuzberg-ffi/src/validation.rs +848 -0
  386. data/vendor/kreuzberg-ffi/tests.disabled/README.md +48 -0
  387. data/vendor/kreuzberg-ffi/tests.disabled/config_loading_tests.rs +299 -0
  388. data/vendor/kreuzberg-ffi/tests.disabled/config_tests.rs +346 -0
  389. data/vendor/kreuzberg-ffi/tests.disabled/extractor_tests.rs +232 -0
  390. data/vendor/kreuzberg-ffi/tests.disabled/plugin_registration_tests.rs +470 -0
  391. data/vendor/kreuzberg-tesseract/.commitlintrc.json +13 -0
  392. data/vendor/kreuzberg-tesseract/.crate-ignore +2 -0
  393. data/vendor/kreuzberg-tesseract/Cargo.lock +2933 -0
  394. data/vendor/kreuzberg-tesseract/Cargo.toml +57 -0
  395. data/vendor/{rb-sys/LICENSE-MIT → kreuzberg-tesseract/LICENSE} +22 -21
  396. data/vendor/kreuzberg-tesseract/README.md +399 -0
  397. data/vendor/kreuzberg-tesseract/build.rs +1127 -0
  398. data/vendor/kreuzberg-tesseract/patches/README.md +71 -0
  399. data/vendor/kreuzberg-tesseract/patches/tesseract.diff +199 -0
  400. data/vendor/kreuzberg-tesseract/src/api.rs +1371 -0
  401. data/vendor/kreuzberg-tesseract/src/choice_iterator.rs +77 -0
  402. data/vendor/kreuzberg-tesseract/src/enums.rs +297 -0
  403. data/vendor/kreuzberg-tesseract/src/error.rs +81 -0
  404. data/vendor/kreuzberg-tesseract/src/lib.rs +145 -0
  405. data/vendor/kreuzberg-tesseract/src/monitor.rs +57 -0
  406. data/vendor/kreuzberg-tesseract/src/mutable_iterator.rs +197 -0
  407. data/vendor/kreuzberg-tesseract/src/page_iterator.rs +253 -0
  408. data/vendor/kreuzberg-tesseract/src/result_iterator.rs +286 -0
  409. data/vendor/kreuzberg-tesseract/src/result_renderer.rs +183 -0
  410. data/vendor/kreuzberg-tesseract/tests/integration_test.rs +211 -0
  411. metadata +196 -45
  412. data/vendor/kreuzberg/benches/otel_overhead.rs +0 -48
  413. data/vendor/kreuzberg/src/extractors/fictionbook.rs.backup2 +0 -738
  414. data/vendor/rb-sys/.cargo-ok +0 -1
  415. data/vendor/rb-sys/.cargo_vcs_info.json +0 -6
  416. data/vendor/rb-sys/Cargo.lock +0 -393
  417. data/vendor/rb-sys/Cargo.toml +0 -70
  418. data/vendor/rb-sys/Cargo.toml.orig +0 -57
  419. data/vendor/rb-sys/LICENSE-APACHE +0 -190
  420. data/vendor/rb-sys/bin/release.sh +0 -21
  421. data/vendor/rb-sys/build/features.rs +0 -108
  422. data/vendor/rb-sys/build/main.rs +0 -246
  423. data/vendor/rb-sys/build/stable_api_config.rs +0 -153
  424. data/vendor/rb-sys/build/version.rs +0 -48
  425. data/vendor/rb-sys/readme.md +0 -36
  426. data/vendor/rb-sys/src/bindings.rs +0 -21
  427. data/vendor/rb-sys/src/hidden.rs +0 -11
  428. data/vendor/rb-sys/src/lib.rs +0 -34
  429. data/vendor/rb-sys/src/macros.rs +0 -371
  430. data/vendor/rb-sys/src/memory.rs +0 -53
  431. data/vendor/rb-sys/src/ruby_abi_version.rs +0 -38
  432. data/vendor/rb-sys/src/special_consts.rs +0 -31
  433. data/vendor/rb-sys/src/stable_api/compiled.c +0 -179
  434. data/vendor/rb-sys/src/stable_api/compiled.rs +0 -257
  435. data/vendor/rb-sys/src/stable_api/ruby_2_6.rs +0 -316
  436. data/vendor/rb-sys/src/stable_api/ruby_2_7.rs +0 -316
  437. data/vendor/rb-sys/src/stable_api/ruby_3_0.rs +0 -324
  438. data/vendor/rb-sys/src/stable_api/ruby_3_1.rs +0 -317
  439. data/vendor/rb-sys/src/stable_api/ruby_3_2.rs +0 -315
  440. data/vendor/rb-sys/src/stable_api/ruby_3_3.rs +0 -326
  441. data/vendor/rb-sys/src/stable_api/ruby_3_4.rs +0 -327
  442. data/vendor/rb-sys/src/stable_api.rs +0 -261
  443. data/vendor/rb-sys/src/symbol.rs +0 -31
  444. data/vendor/rb-sys/src/tracking_allocator.rs +0 -332
  445. data/vendor/rb-sys/src/utils.rs +0 -89
  446. data/vendor/rb-sys/src/value_type.rs +0 -7
@@ -1,1337 +1,1339 @@
1
- //! Plugin registration and discovery.
2
- //!
3
- //! This module provides registries for managing plugins of different types.
4
- //! Each plugin type (OcrBackend, DocumentExtractor, etc.) has its own registry
5
- //! with type-safe registration and lookup.
6
-
7
- use crate::plugins::{DocumentExtractor, OcrBackend, PostProcessor, ProcessingStage, Validator};
8
- use crate::{KreuzbergError, Result};
9
- use indexmap::IndexMap;
10
- use once_cell::sync::Lazy;
11
- use std::collections::{BTreeMap, HashMap};
12
- use std::sync::{Arc, RwLock};
13
-
14
- /// Validate a plugin name before registration.
15
- ///
16
- /// # Rules
17
- ///
18
- /// - Name cannot be empty
19
- /// - Name cannot contain whitespace
20
- /// - Name should follow kebab-case convention (lowercase with hyphens)
21
- ///
22
- /// # Errors
23
- ///
24
- /// Returns `ValidationError` if the name is invalid.
25
- fn validate_plugin_name(name: &str) -> Result<()> {
26
- if name.is_empty() {
27
- return Err(KreuzbergError::Validation {
28
- message: "Plugin name cannot be empty".to_string(),
29
- source: None,
30
- });
31
- }
32
-
33
- if name.contains(char::is_whitespace) {
34
- return Err(KreuzbergError::Validation {
35
- message: format!("Plugin name '{}' cannot contain whitespace", name),
36
- source: None,
37
- });
38
- }
39
-
40
- Ok(())
41
- }
42
-
43
- /// Registry for OCR backend plugins.
44
- ///
45
- /// Manages OCR backends with backend type and language-based selection.
46
- ///
47
- /// # Thread Safety
48
- ///
49
- /// The registry is thread-safe and can be accessed concurrently from multiple threads.
50
- ///
51
- /// # Example
52
- ///
53
- /// ```rust,no_run
54
- /// use kreuzberg::plugins::registry::OcrBackendRegistry;
55
- /// use std::sync::Arc;
56
- ///
57
- /// let registry = OcrBackendRegistry::new();
58
- /// // Register OCR backends
59
- /// // registry.register(Arc::new(TesseractBackend::new()));
60
- /// ```
61
- pub struct OcrBackendRegistry {
62
- backends: HashMap<String, Arc<dyn OcrBackend>>,
63
- }
64
-
65
- impl OcrBackendRegistry {
66
- /// Create a new OCR backend registry with default backends.
67
- ///
68
- /// Registers the Tesseract backend by default if the "ocr" feature is enabled.
69
- pub fn new() -> Self {
70
- #[cfg(feature = "ocr")]
71
- let mut registry = Self {
72
- backends: HashMap::new(),
73
- };
74
-
75
- #[cfg(not(feature = "ocr"))]
76
- let registry = Self {
77
- backends: HashMap::new(),
78
- };
79
-
80
- #[cfg(feature = "ocr")]
81
- {
82
- use crate::ocr::tesseract_backend::TesseractBackend;
83
- if let Ok(backend) = TesseractBackend::new() {
84
- let _ = registry.register(Arc::new(backend));
85
- }
86
- }
87
-
88
- registry
89
- }
90
-
91
- /// Create a new empty OCR backend registry without default backends.
92
- ///
93
- /// This is useful for testing or when you want full control over backend registration.
94
- pub fn new_empty() -> Self {
95
- Self {
96
- backends: HashMap::new(),
97
- }
98
- }
99
-
100
- /// Register an OCR backend.
101
- ///
102
- /// # Arguments
103
- ///
104
- /// * `backend` - The OCR backend to register
105
- ///
106
- /// # Returns
107
- ///
108
- /// - `Ok(())` if registration succeeded
109
- /// - `Err(...)` if initialization failed
110
- ///
111
- /// # Example
112
- ///
113
- /// ```rust,no_run
114
- /// # use kreuzberg::plugins::registry::OcrBackendRegistry;
115
- /// # use std::sync::Arc;
116
- /// let mut registry = OcrBackendRegistry::new();
117
- /// // let backend = Arc::new(MyOcrBackend::new());
118
- /// // registry.register(backend)?;
119
- /// # Ok::<(), kreuzberg::KreuzbergError>(())
120
- /// ```
121
- pub fn register(&mut self, backend: Arc<dyn OcrBackend>) -> Result<()> {
122
- let name = backend.name().to_string();
123
-
124
- validate_plugin_name(&name)?;
125
-
126
- backend.initialize()?;
127
-
128
- self.backends.insert(name, backend);
129
- Ok(())
130
- }
131
-
132
- /// Get an OCR backend by name.
133
- ///
134
- /// # Arguments
135
- ///
136
- /// * `name` - Backend name
137
- ///
138
- /// # Returns
139
- ///
140
- /// The backend if found, or an error if not registered.
141
- pub fn get(&self, name: &str) -> Result<Arc<dyn OcrBackend>> {
142
- self.backends.get(name).cloned().ok_or_else(|| KreuzbergError::Plugin {
143
- message: format!("OCR backend '{}' not registered", name),
144
- plugin_name: name.to_string(),
145
- })
146
- }
147
-
148
- /// Get an OCR backend that supports a specific language.
149
- ///
150
- /// Returns the first backend that supports the language.
151
- ///
152
- /// # Arguments
153
- ///
154
- /// * `language` - Language code (e.g., "eng", "deu")
155
- ///
156
- /// # Returns
157
- ///
158
- /// The first backend supporting the language, or an error if none found.
159
- pub fn get_for_language(&self, language: &str) -> Result<Arc<dyn OcrBackend>> {
160
- self.backends
161
- .values()
162
- .find(|backend| backend.supports_language(language))
163
- .cloned()
164
- .ok_or_else(|| KreuzbergError::Plugin {
165
- message: format!("No OCR backend supports language '{}'", language),
166
- plugin_name: language.to_string(),
167
- })
168
- }
169
-
170
- /// List all registered backend names.
171
- pub fn list(&self) -> Vec<String> {
172
- self.backends.keys().cloned().collect()
173
- }
174
-
175
- /// Remove a backend from the registry.
176
- ///
177
- /// Calls `shutdown()` on the backend before removing.
178
- pub fn remove(&mut self, name: &str) -> Result<()> {
179
- if let Some(backend) = self.backends.remove(name) {
180
- backend.shutdown()?;
181
- }
182
- Ok(())
183
- }
184
-
185
- /// Shutdown all backends and clear the registry.
186
- pub fn shutdown_all(&mut self) -> Result<()> {
187
- let names: Vec<_> = self.backends.keys().cloned().collect();
188
- for name in names {
189
- self.remove(&name)?;
190
- }
191
- Ok(())
192
- }
193
- }
194
-
195
- impl Default for OcrBackendRegistry {
196
- fn default() -> Self {
197
- Self::new()
198
- }
199
- }
200
-
201
- /// Registry for document extractor plugins.
202
- ///
203
- /// Manages extractors with MIME type and priority-based selection.
204
- ///
205
- /// # Thread Safety
206
- ///
207
- /// The registry is thread-safe and can be accessed concurrently from multiple threads.
208
- pub struct DocumentExtractorRegistry {
209
- extractors: HashMap<String, BTreeMap<i32, Arc<dyn DocumentExtractor>>>,
210
- name_index: HashMap<String, Vec<(String, i32)>>,
211
- }
212
-
213
- impl DocumentExtractorRegistry {
214
- /// Create a new empty extractor registry.
215
- pub fn new() -> Self {
216
- Self {
217
- extractors: HashMap::new(),
218
- name_index: HashMap::new(),
219
- }
220
- }
221
-
222
- /// Register a document extractor.
223
- ///
224
- /// The extractor is registered for all MIME types it supports.
225
- ///
226
- /// # Arguments
227
- ///
228
- /// * `extractor` - The extractor to register
229
- ///
230
- /// # Returns
231
- ///
232
- /// - `Ok(())` if registration succeeded
233
- /// - `Err(...)` if initialization failed
234
- pub fn register(&mut self, extractor: Arc<dyn DocumentExtractor>) -> Result<()> {
235
- let name = extractor.name().to_string();
236
- let priority = extractor.priority();
237
- let mime_types: Vec<String> = extractor.supported_mime_types().iter().map(|s| s.to_string()).collect();
238
-
239
- validate_plugin_name(&name)?;
240
-
241
- extractor.initialize()?;
242
-
243
- let mut index_entries = Vec::new();
244
-
245
- for mime_type in &mime_types {
246
- self.extractors
247
- .entry(mime_type.clone())
248
- .or_default()
249
- .insert(priority, Arc::clone(&extractor));
250
- index_entries.push((mime_type.clone(), priority));
251
- }
252
-
253
- self.name_index.insert(name, index_entries);
254
-
255
- Ok(())
256
- }
257
-
258
- /// Get the highest priority extractor for a MIME type.
259
- ///
260
- /// # Arguments
261
- ///
262
- /// * `mime_type` - MIME type to look up
263
- ///
264
- /// # Returns
265
- ///
266
- /// The highest priority extractor, or an error if none found.
267
- #[cfg_attr(feature = "otel", tracing::instrument(
268
- skip(self),
269
- fields(
270
- registry.mime_type = %mime_type,
271
- registry.found = tracing::field::Empty,
272
- )
273
- ))]
274
- pub fn get(&self, mime_type: &str) -> Result<Arc<dyn DocumentExtractor>> {
275
- if let Some(priority_map) = self.extractors.get(mime_type)
276
- && let Some((_priority, extractor)) = priority_map.iter().next_back()
277
- {
278
- #[cfg(feature = "otel")]
279
- tracing::Span::current().record("registry.found", true);
280
- return Ok(Arc::clone(extractor));
281
- }
282
-
283
- let mut best_match: Option<(i32, Arc<dyn DocumentExtractor>)> = None;
284
-
285
- for (registered_mime, priority_map) in &self.extractors {
286
- if registered_mime.ends_with("/*") {
287
- let prefix = &registered_mime[..registered_mime.len() - 1];
288
- if mime_type.starts_with(prefix)
289
- && let Some((_priority, extractor)) = priority_map.iter().next_back()
290
- {
291
- let priority = extractor.priority();
292
- match &best_match {
293
- None => best_match = Some((priority, Arc::clone(extractor))),
294
- Some((current_priority, _)) => {
295
- if priority > *current_priority {
296
- best_match = Some((priority, Arc::clone(extractor)));
297
- }
298
- }
299
- }
300
- }
301
- }
302
- }
303
-
304
- if let Some((_priority, extractor)) = best_match {
305
- #[cfg(feature = "otel")]
306
- tracing::Span::current().record("registry.found", true);
307
- return Ok(extractor);
308
- }
309
-
310
- #[cfg(feature = "otel")]
311
- tracing::Span::current().record("registry.found", false);
312
- Err(KreuzbergError::UnsupportedFormat(mime_type.to_string()))
313
- }
314
-
315
- /// List all registered extractors.
316
- pub fn list(&self) -> Vec<String> {
317
- self.name_index.keys().cloned().collect()
318
- }
319
-
320
- /// Remove an extractor from the registry.
321
- pub fn remove(&mut self, name: &str) -> Result<()> {
322
- let index_entries = match self.name_index.remove(name) {
323
- Some(entries) => entries,
324
- None => return Ok(()),
325
- };
326
-
327
- let mut extractor_to_shutdown: Option<Arc<dyn DocumentExtractor>> = None;
328
-
329
- for (mime_type, priority) in index_entries {
330
- if let Some(priority_map) = self.extractors.get_mut(&mime_type) {
331
- if let Some(extractor) = priority_map.remove(&priority)
332
- && extractor_to_shutdown.is_none()
333
- {
334
- extractor_to_shutdown = Some(extractor);
335
- }
336
-
337
- if priority_map.is_empty() {
338
- self.extractors.remove(&mime_type);
339
- }
340
- }
341
- }
342
-
343
- if let Some(extractor) = extractor_to_shutdown {
344
- extractor.shutdown()?;
345
- }
346
-
347
- Ok(())
348
- }
349
-
350
- /// Shutdown all extractors and clear the registry.
351
- pub fn shutdown_all(&mut self) -> Result<()> {
352
- let names = self.list();
353
- for name in names {
354
- self.remove(&name)?;
355
- }
356
- Ok(())
357
- }
358
- }
359
-
360
- impl Default for DocumentExtractorRegistry {
361
- fn default() -> Self {
362
- Self::new()
363
- }
364
- }
365
-
366
- /// Registry for post-processor plugins.
367
- ///
368
- /// Manages post-processors organized by processing stage.
369
- pub struct PostProcessorRegistry {
370
- processors: HashMap<ProcessingStage, BTreeMap<i32, Vec<Arc<dyn PostProcessor>>>>,
371
- name_index: HashMap<String, (ProcessingStage, i32)>,
372
- }
373
-
374
- impl PostProcessorRegistry {
375
- /// Create a new empty post-processor registry.
376
- pub fn new() -> Self {
377
- Self {
378
- processors: HashMap::new(),
379
- name_index: HashMap::new(),
380
- }
381
- }
382
-
383
- /// Register a post-processor.
384
- ///
385
- /// # Arguments
386
- ///
387
- /// * `processor` - The post-processor to register
388
- /// * `priority` - Execution priority (higher = runs first within stage)
389
- pub fn register(&mut self, processor: Arc<dyn PostProcessor>, priority: i32) -> Result<()> {
390
- let name = processor.name().to_string();
391
- let stage = processor.processing_stage();
392
-
393
- validate_plugin_name(&name)?;
394
-
395
- processor.initialize()?;
396
-
397
- if self.name_index.contains_key(&name) {
398
- self.remove(&name)?;
399
- }
400
-
401
- self.processors
402
- .entry(stage)
403
- .or_default()
404
- .entry(priority)
405
- .or_default()
406
- .push(Arc::clone(&processor));
407
-
408
- self.name_index.insert(name, (stage, priority));
409
-
410
- Ok(())
411
- }
412
-
413
- /// Get all processors for a specific stage, in priority order.
414
- ///
415
- /// # Arguments
416
- ///
417
- /// * `stage` - The processing stage
418
- ///
419
- /// # Returns
420
- ///
421
- /// Vector of processors in priority order (highest first).
422
- pub fn get_for_stage(&self, stage: ProcessingStage) -> Vec<Arc<dyn PostProcessor>> {
423
- let mut result = Vec::new();
424
-
425
- if let Some(priority_map) = self.processors.get(&stage) {
426
- for (_priority, processors) in priority_map.iter().rev() {
427
- for processor in processors {
428
- result.push(Arc::clone(processor));
429
- }
430
- }
431
- }
432
-
433
- result
434
- }
435
-
436
- /// List all registered processor names.
437
- pub fn list(&self) -> Vec<String> {
438
- self.name_index.keys().cloned().collect()
439
- }
440
-
441
- /// Remove a processor from the registry.
442
- pub fn remove(&mut self, name: &str) -> Result<()> {
443
- let (stage, priority) = match self.name_index.remove(name) {
444
- Some(location) => location,
445
- None => return Ok(()),
446
- };
447
-
448
- let processor_to_shutdown = if let Some(priority_map) = self.processors.get_mut(&stage) {
449
- let processor = priority_map.get_mut(&priority).and_then(|processors| {
450
- processors
451
- .iter()
452
- .position(|p| p.name() == name)
453
- .map(|pos| processors.remove(pos))
454
- });
455
-
456
- if let Some(processors) = priority_map.get(&priority)
457
- && processors.is_empty()
458
- {
459
- priority_map.remove(&priority);
460
- }
461
-
462
- if priority_map.is_empty() {
463
- self.processors.remove(&stage);
464
- }
465
- processor
466
- } else {
467
- None
468
- };
469
-
470
- if let Some(processor) = processor_to_shutdown {
471
- processor.shutdown()?;
472
- }
473
-
474
- Ok(())
475
- }
476
-
477
- /// Shutdown all processors and clear the registry.
478
- pub fn shutdown_all(&mut self) -> Result<()> {
479
- let names = self.list();
480
- for name in names {
481
- self.remove(&name)?;
482
- }
483
- Ok(())
484
- }
485
- }
486
-
487
- impl Default for PostProcessorRegistry {
488
- fn default() -> Self {
489
- Self::new()
490
- }
491
- }
492
-
493
- /// Registry for validator plugins.
494
- ///
495
- /// Manages validators with priority-based execution order.
496
- pub struct ValidatorRegistry {
497
- validators: BTreeMap<i32, IndexMap<String, Arc<dyn Validator>>>,
498
- }
499
-
500
- impl ValidatorRegistry {
501
- /// Create a new empty validator registry.
502
- pub fn new() -> Self {
503
- Self {
504
- validators: BTreeMap::new(),
505
- }
506
- }
507
-
508
- /// Register a validator.
509
- ///
510
- /// # Arguments
511
- ///
512
- /// * `validator` - The validator to register
513
- pub fn register(&mut self, validator: Arc<dyn Validator>) -> Result<()> {
514
- let name = validator.name().to_string();
515
- let priority = validator.priority();
516
-
517
- validate_plugin_name(&name)?;
518
-
519
- validator.initialize()?;
520
-
521
- self.validators.entry(priority).or_default().insert(name, validator);
522
-
523
- Ok(())
524
- }
525
-
526
- /// Get all validators in priority order.
527
- ///
528
- /// # Returns
529
- ///
530
- /// Vector of validators in priority order (highest first).
531
- pub fn get_all(&self) -> Vec<Arc<dyn Validator>> {
532
- let mut result = Vec::new();
533
-
534
- for (_priority, validators) in self.validators.iter().rev() {
535
- for validator in validators.values() {
536
- result.push(Arc::clone(validator));
537
- }
538
- }
539
-
540
- result
541
- }
542
-
543
- /// List all registered validator names.
544
- pub fn list(&self) -> Vec<String> {
545
- let mut names = std::collections::HashSet::new();
546
- for validators in self.validators.values() {
547
- names.extend(validators.keys().cloned());
548
- }
549
- names.into_iter().collect()
550
- }
551
-
552
- /// Remove a validator from the registry.
553
- pub fn remove(&mut self, name: &str) -> Result<()> {
554
- let mut validator_to_shutdown: Option<Arc<dyn Validator>> = None;
555
-
556
- for validators in self.validators.values_mut() {
557
- if let Some(validator) = validators.shift_remove(name)
558
- && validator_to_shutdown.is_none()
559
- {
560
- validator_to_shutdown = Some(validator);
561
- }
562
- }
563
-
564
- if let Some(validator) = validator_to_shutdown {
565
- validator.shutdown()?;
566
- }
567
-
568
- self.validators.retain(|_, validators| !validators.is_empty());
569
-
570
- Ok(())
571
- }
572
-
573
- /// Shutdown all validators and clear the registry.
574
- pub fn shutdown_all(&mut self) -> Result<()> {
575
- let names = self.list();
576
- for name in names {
577
- self.remove(&name)?;
578
- }
579
- Ok(())
580
- }
581
- }
582
-
583
- impl Default for ValidatorRegistry {
584
- fn default() -> Self {
585
- Self::new()
586
- }
587
- }
588
-
589
- /// Global OCR backend registry singleton.
590
- pub static OCR_BACKEND_REGISTRY: Lazy<Arc<RwLock<OcrBackendRegistry>>> =
591
- Lazy::new(|| Arc::new(RwLock::new(OcrBackendRegistry::new())));
592
-
593
- /// Global document extractor registry singleton.
594
- pub static DOCUMENT_EXTRACTOR_REGISTRY: Lazy<Arc<RwLock<DocumentExtractorRegistry>>> =
595
- Lazy::new(|| Arc::new(RwLock::new(DocumentExtractorRegistry::new())));
596
-
597
- /// Global post-processor registry singleton.
598
- pub static POST_PROCESSOR_REGISTRY: Lazy<Arc<RwLock<PostProcessorRegistry>>> =
599
- Lazy::new(|| Arc::new(RwLock::new(PostProcessorRegistry::new())));
600
-
601
- /// Global validator registry singleton.
602
- pub static VALIDATOR_REGISTRY: Lazy<Arc<RwLock<ValidatorRegistry>>> =
603
- Lazy::new(|| Arc::new(RwLock::new(ValidatorRegistry::new())));
604
-
605
- /// Get the global OCR backend registry.
606
- pub fn get_ocr_backend_registry() -> Arc<RwLock<OcrBackendRegistry>> {
607
- OCR_BACKEND_REGISTRY.clone()
608
- }
609
-
610
- /// Get the global document extractor registry.
611
- pub fn get_document_extractor_registry() -> Arc<RwLock<DocumentExtractorRegistry>> {
612
- DOCUMENT_EXTRACTOR_REGISTRY.clone()
613
- }
614
-
615
- /// Get the global post-processor registry.
616
- pub fn get_post_processor_registry() -> Arc<RwLock<PostProcessorRegistry>> {
617
- POST_PROCESSOR_REGISTRY.clone()
618
- }
619
-
620
- /// Get the global validator registry.
621
- pub fn get_validator_registry() -> Arc<RwLock<ValidatorRegistry>> {
622
- VALIDATOR_REGISTRY.clone()
623
- }
624
-
625
- #[cfg(test)]
626
- mod tests {
627
- use super::*;
628
- use crate::core::config::{ExtractionConfig, OcrConfig};
629
- use crate::plugins::{Plugin, PostProcessor, ProcessingStage, Validator};
630
- use crate::types::ExtractionResult;
631
- use async_trait::async_trait;
632
-
633
- struct MockOcrBackend {
634
- name: String,
635
- languages: Vec<String>,
636
- }
637
-
638
- impl Plugin for MockOcrBackend {
639
- fn name(&self) -> &str {
640
- &self.name
641
- }
642
- fn version(&self) -> String {
643
- "1.0.0".to_string()
644
- }
645
- fn initialize(&self) -> Result<()> {
646
- Ok(())
647
- }
648
- fn shutdown(&self) -> Result<()> {
649
- Ok(())
650
- }
651
- }
652
-
653
- #[async_trait]
654
- impl OcrBackend for MockOcrBackend {
655
- async fn process_image(&self, _: &[u8], _: &OcrConfig) -> Result<ExtractionResult> {
656
- Ok(ExtractionResult {
657
- content: "test".to_string(),
658
- mime_type: "text/plain".to_string(),
659
- metadata: crate::types::Metadata::default(),
660
- tables: vec![],
661
- detected_languages: None,
662
- chunks: None,
663
- images: None,
664
- })
665
- }
666
-
667
- fn supports_language(&self, lang: &str) -> bool {
668
- self.languages.iter().any(|l| l == lang)
669
- }
670
-
671
- fn backend_type(&self) -> crate::plugins::ocr::OcrBackendType {
672
- crate::plugins::ocr::OcrBackendType::Custom
673
- }
674
- }
675
-
676
- struct MockExtractor {
677
- name: String,
678
- mime_types: &'static [&'static str],
679
- priority: i32,
680
- }
681
-
682
- impl Plugin for MockExtractor {
683
- fn name(&self) -> &str {
684
- &self.name
685
- }
686
- fn version(&self) -> String {
687
- "1.0.0".to_string()
688
- }
689
- fn initialize(&self) -> Result<()> {
690
- Ok(())
691
- }
692
- fn shutdown(&self) -> Result<()> {
693
- Ok(())
694
- }
695
- }
696
-
697
- #[async_trait]
698
- impl DocumentExtractor for MockExtractor {
699
- async fn extract_bytes(&self, _: &[u8], _: &str, _: &ExtractionConfig) -> Result<ExtractionResult> {
700
- Ok(ExtractionResult {
701
- content: "test".to_string(),
702
- mime_type: "text/plain".to_string(),
703
- metadata: crate::types::Metadata::default(),
704
- tables: vec![],
705
- detected_languages: None,
706
- chunks: None,
707
- images: None,
708
- })
709
- }
710
-
711
- fn supported_mime_types(&self) -> &[&str] {
712
- self.mime_types
713
- }
714
-
715
- fn priority(&self) -> i32 {
716
- self.priority
717
- }
718
- }
719
-
720
- struct MockPostProcessor {
721
- name: String,
722
- stage: ProcessingStage,
723
- }
724
-
725
- impl Plugin for MockPostProcessor {
726
- fn name(&self) -> &str {
727
- &self.name
728
- }
729
- fn version(&self) -> String {
730
- "1.0.0".to_string()
731
- }
732
- fn initialize(&self) -> Result<()> {
733
- Ok(())
734
- }
735
- fn shutdown(&self) -> Result<()> {
736
- Ok(())
737
- }
738
- }
739
-
740
- #[async_trait]
741
- impl PostProcessor for MockPostProcessor {
742
- async fn process(&self, _result: &mut ExtractionResult, _: &ExtractionConfig) -> Result<()> {
743
- Ok(())
744
- }
745
-
746
- fn processing_stage(&self) -> ProcessingStage {
747
- self.stage
748
- }
749
- }
750
-
751
- struct MockValidator {
752
- name: String,
753
- priority: i32,
754
- }
755
-
756
- impl Plugin for MockValidator {
757
- fn name(&self) -> &str {
758
- &self.name
759
- }
760
- fn version(&self) -> String {
761
- "1.0.0".to_string()
762
- }
763
- fn initialize(&self) -> Result<()> {
764
- Ok(())
765
- }
766
- fn shutdown(&self) -> Result<()> {
767
- Ok(())
768
- }
769
- }
770
-
771
- #[async_trait]
772
- impl Validator for MockValidator {
773
- async fn validate(&self, _: &ExtractionResult, _: &ExtractionConfig) -> Result<()> {
774
- Ok(())
775
- }
776
-
777
- fn priority(&self) -> i32 {
778
- self.priority
779
- }
780
- }
781
-
782
- #[test]
783
- fn test_ocr_backend_registry() {
784
- let mut registry = OcrBackendRegistry::new_empty();
785
-
786
- let backend = Arc::new(MockOcrBackend {
787
- name: "test-ocr".to_string(),
788
- languages: vec!["eng".to_string(), "deu".to_string()],
789
- });
790
-
791
- registry.register(backend).unwrap();
792
-
793
- let retrieved = registry.get("test-ocr").unwrap();
794
- assert_eq!(retrieved.name(), "test-ocr");
795
-
796
- let eng_backend = registry.get_for_language("eng").unwrap();
797
- assert_eq!(eng_backend.name(), "test-ocr");
798
-
799
- let names = registry.list();
800
- assert_eq!(names.len(), 1);
801
- assert!(names.contains(&"test-ocr".to_string()));
802
- }
803
-
804
- #[test]
805
- fn test_post_processor_registry() {
806
- let mut registry = PostProcessorRegistry::new();
807
-
808
- let early = Arc::new(MockPostProcessor {
809
- name: "early-processor".to_string(),
810
- stage: ProcessingStage::Early,
811
- });
812
-
813
- let middle = Arc::new(MockPostProcessor {
814
- name: "middle-processor".to_string(),
815
- stage: ProcessingStage::Middle,
816
- });
817
-
818
- registry.register(early, 100).unwrap();
819
- registry.register(middle, 50).unwrap();
820
-
821
- let early_processors = registry.get_for_stage(ProcessingStage::Early);
822
- assert_eq!(early_processors.len(), 1);
823
- assert_eq!(early_processors[0].name(), "early-processor");
824
-
825
- let middle_processors = registry.get_for_stage(ProcessingStage::Middle);
826
- assert_eq!(middle_processors.len(), 1);
827
-
828
- let names = registry.list();
829
- assert_eq!(names.len(), 2);
830
- }
831
-
832
- #[test]
833
- fn test_validator_registry() {
834
- let mut registry = ValidatorRegistry::new();
835
-
836
- let high_priority = Arc::new(MockValidator {
837
- name: "high-priority".to_string(),
838
- priority: 100,
839
- });
840
-
841
- let low_priority = Arc::new(MockValidator {
842
- name: "low-priority".to_string(),
843
- priority: 10,
844
- });
845
-
846
- registry.register(high_priority).unwrap();
847
- registry.register(low_priority).unwrap();
848
-
849
- let validators = registry.get_all();
850
- assert_eq!(validators.len(), 2);
851
- assert_eq!(validators[0].name(), "high-priority");
852
- assert_eq!(validators[1].name(), "low-priority");
853
- }
854
-
855
- #[test]
856
- fn test_document_extractor_registry_exact_match() {
857
- let mut registry = DocumentExtractorRegistry::new();
858
-
859
- let extractor = Arc::new(MockExtractor {
860
- name: "pdf-extractor".to_string(),
861
- mime_types: &["application/pdf"],
862
- priority: 100,
863
- });
864
-
865
- registry.register(extractor).unwrap();
866
-
867
- let retrieved = registry.get("application/pdf").unwrap();
868
- assert_eq!(retrieved.name(), "pdf-extractor");
869
-
870
- let names = registry.list();
871
- assert_eq!(names.len(), 1);
872
- assert!(names.contains(&"pdf-extractor".to_string()));
873
- }
874
-
875
- #[test]
876
- fn test_document_extractor_registry_prefix_match() {
877
- let mut registry = DocumentExtractorRegistry::new();
878
-
879
- let image_extractor = Arc::new(MockExtractor {
880
- name: "image-extractor".to_string(),
881
- mime_types: &["image/*"],
882
- priority: 50,
883
- });
884
-
885
- registry.register(image_extractor).unwrap();
886
-
887
- let retrieved = registry.get("image/png").unwrap();
888
- assert_eq!(retrieved.name(), "image-extractor");
889
-
890
- let retrieved_jpg = registry.get("image/jpeg").unwrap();
891
- assert_eq!(retrieved_jpg.name(), "image-extractor");
892
- }
893
-
894
- #[test]
895
- fn test_document_extractor_registry_priority() {
896
- let mut registry = DocumentExtractorRegistry::new();
897
-
898
- let low_priority = Arc::new(MockExtractor {
899
- name: "low-priority-pdf".to_string(),
900
- mime_types: &["application/pdf"],
901
- priority: 10,
902
- });
903
-
904
- let high_priority = Arc::new(MockExtractor {
905
- name: "high-priority-pdf".to_string(),
906
- mime_types: &["application/pdf"],
907
- priority: 100,
908
- });
909
-
910
- registry.register(low_priority).unwrap();
911
- registry.register(high_priority).unwrap();
912
-
913
- let retrieved = registry.get("application/pdf").unwrap();
914
- assert_eq!(retrieved.name(), "high-priority-pdf");
915
- }
916
-
917
- #[test]
918
- fn test_document_extractor_registry_not_found() {
919
- let registry = DocumentExtractorRegistry::new();
920
-
921
- let result = registry.get("application/unknown");
922
- assert!(matches!(result, Err(KreuzbergError::UnsupportedFormat(_))));
923
- }
924
-
925
- #[test]
926
- fn test_document_extractor_registry_remove() {
927
- let mut registry = DocumentExtractorRegistry::new();
928
-
929
- let extractor = Arc::new(MockExtractor {
930
- name: "test-extractor".to_string(),
931
- mime_types: &["text/plain"],
932
- priority: 50,
933
- });
934
-
935
- registry.register(extractor).unwrap();
936
- assert!(registry.get("text/plain").is_ok());
937
-
938
- registry.remove("test-extractor").unwrap();
939
- assert!(registry.get("text/plain").is_err());
940
- }
941
-
942
- #[test]
943
- fn test_document_extractor_registry_shutdown_all() {
944
- let mut registry = DocumentExtractorRegistry::new();
945
-
946
- let extractor1 = Arc::new(MockExtractor {
947
- name: "extractor1".to_string(),
948
- mime_types: &["text/plain"],
949
- priority: 50,
950
- });
951
-
952
- let extractor2 = Arc::new(MockExtractor {
953
- name: "extractor2".to_string(),
954
- mime_types: &["application/pdf"],
955
- priority: 50,
956
- });
957
-
958
- registry.register(extractor1).unwrap();
959
- registry.register(extractor2).unwrap();
960
-
961
- assert_eq!(registry.list().len(), 2);
962
-
963
- registry.shutdown_all().unwrap();
964
- assert_eq!(registry.list().len(), 0);
965
- }
966
-
967
- #[test]
968
- fn test_ocr_backend_registry_not_found() {
969
- let registry = OcrBackendRegistry::new();
970
-
971
- let result = registry.get("nonexistent");
972
- assert!(matches!(result, Err(KreuzbergError::Plugin { .. })));
973
- }
974
-
975
- #[test]
976
- fn test_ocr_backend_registry_language_not_found() {
977
- let registry = OcrBackendRegistry::new();
978
-
979
- let result = registry.get_for_language("xyz");
980
- assert!(matches!(result, Err(KreuzbergError::Plugin { .. })));
981
- }
982
-
983
- #[test]
984
- fn test_ocr_backend_registry_remove() {
985
- let mut registry = OcrBackendRegistry::new();
986
-
987
- let backend = Arc::new(MockOcrBackend {
988
- name: "test-ocr".to_string(),
989
- languages: vec!["eng".to_string()],
990
- });
991
-
992
- registry.register(backend).unwrap();
993
- assert!(registry.get("test-ocr").is_ok());
994
-
995
- registry.remove("test-ocr").unwrap();
996
- assert!(registry.get("test-ocr").is_err());
997
- }
998
-
999
- #[test]
1000
- fn test_post_processor_registry_remove() {
1001
- let mut registry = PostProcessorRegistry::new();
1002
-
1003
- let processor = Arc::new(MockPostProcessor {
1004
- name: "test-processor".to_string(),
1005
- stage: ProcessingStage::Early,
1006
- });
1007
-
1008
- registry.register(processor, 50).unwrap();
1009
- assert_eq!(registry.get_for_stage(ProcessingStage::Early).len(), 1);
1010
-
1011
- registry.remove("test-processor").unwrap();
1012
- assert_eq!(registry.get_for_stage(ProcessingStage::Early).len(), 0);
1013
- }
1014
-
1015
- #[test]
1016
- fn test_validator_registry_remove() {
1017
- let mut registry = ValidatorRegistry::new();
1018
-
1019
- let validator = Arc::new(MockValidator {
1020
- name: "test-validator".to_string(),
1021
- priority: 50,
1022
- });
1023
-
1024
- registry.register(validator).unwrap();
1025
- assert_eq!(registry.get_all().len(), 1);
1026
-
1027
- registry.remove("test-validator").unwrap();
1028
- assert_eq!(registry.get_all().len(), 0);
1029
- }
1030
-
1031
- #[test]
1032
- fn test_global_registry_access() {
1033
- let ocr_registry = get_ocr_backend_registry();
1034
- let _ = ocr_registry
1035
- .read()
1036
- .expect("Failed to acquire read lock on OCR registry in test")
1037
- .list();
1038
-
1039
- let extractor_registry = get_document_extractor_registry();
1040
- let _ = extractor_registry
1041
- .read()
1042
- .expect("Failed to acquire read lock on extractor registry in test")
1043
- .list();
1044
-
1045
- let processor_registry = get_post_processor_registry();
1046
- let _ = processor_registry
1047
- .read()
1048
- .expect("Failed to acquire read lock on processor registry in test")
1049
- .list();
1050
-
1051
- let validator_registry = get_validator_registry();
1052
- let _ = validator_registry
1053
- .read()
1054
- .expect("Failed to acquire read lock on validator registry in test")
1055
- .list();
1056
- }
1057
-
1058
- #[test]
1059
- fn test_ocr_backend_registry_shutdown_all() {
1060
- let mut registry = OcrBackendRegistry::new();
1061
- let baseline = registry.list().len();
1062
-
1063
- let backend1 = Arc::new(MockOcrBackend {
1064
- name: "backend1".to_string(),
1065
- languages: vec!["eng".to_string()],
1066
- });
1067
-
1068
- let backend2 = Arc::new(MockOcrBackend {
1069
- name: "backend2".to_string(),
1070
- languages: vec!["deu".to_string()],
1071
- });
1072
-
1073
- registry.register(backend1).unwrap();
1074
- registry.register(backend2).unwrap();
1075
-
1076
- assert_eq!(registry.list().len(), baseline + 2);
1077
-
1078
- registry.shutdown_all().unwrap();
1079
- assert_eq!(registry.list().len(), 0);
1080
- }
1081
-
1082
- #[test]
1083
- fn test_post_processor_registry_shutdown_all() {
1084
- let mut registry = PostProcessorRegistry::new();
1085
-
1086
- let early = Arc::new(MockPostProcessor {
1087
- name: "early".to_string(),
1088
- stage: ProcessingStage::Early,
1089
- });
1090
-
1091
- let late = Arc::new(MockPostProcessor {
1092
- name: "late".to_string(),
1093
- stage: ProcessingStage::Late,
1094
- });
1095
-
1096
- registry.register(early, 100).unwrap();
1097
- registry.register(late, 50).unwrap();
1098
-
1099
- assert_eq!(registry.list().len(), 2);
1100
-
1101
- registry.shutdown_all().unwrap();
1102
- assert_eq!(registry.list().len(), 0);
1103
- }
1104
-
1105
- #[test]
1106
- fn test_validator_registry_shutdown_all() {
1107
- let mut registry = ValidatorRegistry::new();
1108
-
1109
- let validator1 = Arc::new(MockValidator {
1110
- name: "validator1".to_string(),
1111
- priority: 100,
1112
- });
1113
-
1114
- let validator2 = Arc::new(MockValidator {
1115
- name: "validator2".to_string(),
1116
- priority: 50,
1117
- });
1118
-
1119
- registry.register(validator1).unwrap();
1120
- registry.register(validator2).unwrap();
1121
-
1122
- assert_eq!(registry.get_all().len(), 2);
1123
-
1124
- registry.shutdown_all().unwrap();
1125
- assert_eq!(registry.get_all().len(), 0);
1126
- }
1127
-
1128
- #[test]
1129
- fn test_document_extractor_registry_multiple_mime_types() {
1130
- let mut registry = DocumentExtractorRegistry::new();
1131
-
1132
- let multi_extractor = Arc::new(MockExtractor {
1133
- name: "multi-extractor".to_string(),
1134
- mime_types: &["text/plain", "text/markdown", "text/html"],
1135
- priority: 50,
1136
- });
1137
-
1138
- registry.register(multi_extractor).unwrap();
1139
-
1140
- assert_eq!(registry.get("text/plain").unwrap().name(), "multi-extractor");
1141
- assert_eq!(registry.get("text/markdown").unwrap().name(), "multi-extractor");
1142
- assert_eq!(registry.get("text/html").unwrap().name(), "multi-extractor");
1143
- }
1144
-
1145
- #[test]
1146
- fn test_post_processor_registry_priority_order() {
1147
- let mut registry = PostProcessorRegistry::new();
1148
-
1149
- let low = Arc::new(MockPostProcessor {
1150
- name: "low-priority".to_string(),
1151
- stage: ProcessingStage::Early,
1152
- });
1153
-
1154
- let high = Arc::new(MockPostProcessor {
1155
- name: "high-priority".to_string(),
1156
- stage: ProcessingStage::Early,
1157
- });
1158
-
1159
- registry.register(low, 10).unwrap();
1160
- registry.register(high, 100).unwrap();
1161
-
1162
- let processors = registry.get_for_stage(ProcessingStage::Early);
1163
- assert_eq!(processors.len(), 2);
1164
- assert_eq!(processors[0].name(), "high-priority");
1165
- assert_eq!(processors[1].name(), "low-priority");
1166
- }
1167
-
1168
- #[test]
1169
- fn test_post_processor_registry_empty_stage() {
1170
- let registry = PostProcessorRegistry::new();
1171
-
1172
- let processors = registry.get_for_stage(ProcessingStage::Late);
1173
- assert_eq!(processors.len(), 0);
1174
- }
1175
-
1176
- #[test]
1177
- fn test_ocr_backend_registry_default() {
1178
- let registry = OcrBackendRegistry::default();
1179
- #[cfg(feature = "ocr")]
1180
- assert!(
1181
- !registry.list().is_empty(),
1182
- "expected at least one default OCR backend when the 'ocr' feature is enabled"
1183
- );
1184
- #[cfg(not(feature = "ocr"))]
1185
- assert_eq!(registry.list().len(), 0);
1186
- }
1187
-
1188
- #[test]
1189
- fn test_document_extractor_registry_default() {
1190
- let registry = DocumentExtractorRegistry::default();
1191
- assert_eq!(registry.list().len(), 0);
1192
- }
1193
-
1194
- #[test]
1195
- fn test_post_processor_registry_default() {
1196
- let registry = PostProcessorRegistry::default();
1197
- assert_eq!(registry.list().len(), 0);
1198
- }
1199
-
1200
- #[test]
1201
- fn test_validator_registry_default() {
1202
- let registry = ValidatorRegistry::default();
1203
- assert_eq!(registry.get_all().len(), 0);
1204
- }
1205
-
1206
- #[test]
1207
- fn test_document_extractor_registry_exact_over_prefix() {
1208
- let mut registry = DocumentExtractorRegistry::new();
1209
-
1210
- let prefix_extractor = Arc::new(MockExtractor {
1211
- name: "prefix-extractor".to_string(),
1212
- mime_types: &["image/*"],
1213
- priority: 100,
1214
- });
1215
-
1216
- let exact_extractor = Arc::new(MockExtractor {
1217
- name: "exact-extractor".to_string(),
1218
- mime_types: &["image/png"],
1219
- priority: 50,
1220
- });
1221
-
1222
- registry.register(prefix_extractor).unwrap();
1223
- registry.register(exact_extractor).unwrap();
1224
-
1225
- let retrieved = registry.get("image/png").unwrap();
1226
- assert_eq!(retrieved.name(), "exact-extractor");
1227
-
1228
- let retrieved_jpg = registry.get("image/jpeg").unwrap();
1229
- assert_eq!(retrieved_jpg.name(), "prefix-extractor");
1230
- }
1231
-
1232
- #[test]
1233
- fn test_ocr_backend_registry_invalid_name_empty() {
1234
- let mut registry = OcrBackendRegistry::new();
1235
-
1236
- let backend = Arc::new(MockOcrBackend {
1237
- name: "".to_string(),
1238
- languages: vec!["eng".to_string()],
1239
- });
1240
-
1241
- let result = registry.register(backend);
1242
- assert!(matches!(result, Err(KreuzbergError::Validation { .. })));
1243
- }
1244
-
1245
- #[test]
1246
- fn test_ocr_backend_registry_invalid_name_whitespace() {
1247
- let mut registry = OcrBackendRegistry::new();
1248
-
1249
- let backend = Arc::new(MockOcrBackend {
1250
- name: "my ocr backend".to_string(),
1251
- languages: vec!["eng".to_string()],
1252
- });
1253
-
1254
- let result = registry.register(backend);
1255
- assert!(matches!(result, Err(KreuzbergError::Validation { .. })));
1256
- }
1257
-
1258
- #[test]
1259
- fn test_document_extractor_registry_invalid_name_empty() {
1260
- let mut registry = DocumentExtractorRegistry::new();
1261
-
1262
- let extractor = Arc::new(MockExtractor {
1263
- name: "".to_string(),
1264
- mime_types: &["text/plain"],
1265
- priority: 50,
1266
- });
1267
-
1268
- let result = registry.register(extractor);
1269
- assert!(matches!(result, Err(KreuzbergError::Validation { .. })));
1270
- }
1271
-
1272
- #[test]
1273
- fn test_document_extractor_registry_invalid_name_whitespace() {
1274
- let mut registry = DocumentExtractorRegistry::new();
1275
-
1276
- let extractor = Arc::new(MockExtractor {
1277
- name: "my extractor".to_string(),
1278
- mime_types: &["text/plain"],
1279
- priority: 50,
1280
- });
1281
-
1282
- let result = registry.register(extractor);
1283
- assert!(matches!(result, Err(KreuzbergError::Validation { .. })));
1284
- }
1285
-
1286
- #[test]
1287
- fn test_post_processor_registry_invalid_name_empty() {
1288
- let mut registry = PostProcessorRegistry::new();
1289
-
1290
- let processor = Arc::new(MockPostProcessor {
1291
- name: "".to_string(),
1292
- stage: ProcessingStage::Early,
1293
- });
1294
-
1295
- let result = registry.register(processor, 50);
1296
- assert!(matches!(result, Err(KreuzbergError::Validation { .. })));
1297
- }
1298
-
1299
- #[test]
1300
- fn test_post_processor_registry_invalid_name_whitespace() {
1301
- let mut registry = PostProcessorRegistry::new();
1302
-
1303
- let processor = Arc::new(MockPostProcessor {
1304
- name: "my processor".to_string(),
1305
- stage: ProcessingStage::Early,
1306
- });
1307
-
1308
- let result = registry.register(processor, 50);
1309
- assert!(matches!(result, Err(KreuzbergError::Validation { .. })));
1310
- }
1311
-
1312
- #[test]
1313
- fn test_validator_registry_invalid_name_empty() {
1314
- let mut registry = ValidatorRegistry::new();
1315
-
1316
- let validator = Arc::new(MockValidator {
1317
- name: "".to_string(),
1318
- priority: 50,
1319
- });
1320
-
1321
- let result = registry.register(validator);
1322
- assert!(matches!(result, Err(KreuzbergError::Validation { .. })));
1323
- }
1324
-
1325
- #[test]
1326
- fn test_validator_registry_invalid_name_whitespace() {
1327
- let mut registry = ValidatorRegistry::new();
1328
-
1329
- let validator = Arc::new(MockValidator {
1330
- name: "my validator".to_string(),
1331
- priority: 50,
1332
- });
1333
-
1334
- let result = registry.register(validator);
1335
- assert!(matches!(result, Err(KreuzbergError::Validation { .. })));
1336
- }
1337
- }
1
+ //! Plugin registration and discovery.
2
+ //!
3
+ //! This module provides registries for managing plugins of different types.
4
+ //! Each plugin type (OcrBackend, DocumentExtractor, etc.) has its own registry
5
+ //! with type-safe registration and lookup.
6
+
7
+ use crate::plugins::{DocumentExtractor, OcrBackend, PostProcessor, ProcessingStage, Validator};
8
+ use crate::{KreuzbergError, Result};
9
+ use indexmap::IndexMap;
10
+ use once_cell::sync::Lazy;
11
+ use std::collections::{BTreeMap, HashMap};
12
+ use std::sync::{Arc, RwLock};
13
+
14
+ /// Validate a plugin name before registration.
15
+ ///
16
+ /// # Rules
17
+ ///
18
+ /// - Name cannot be empty
19
+ /// - Name cannot contain whitespace
20
+ /// - Name should follow kebab-case convention (lowercase with hyphens)
21
+ ///
22
+ /// # Errors
23
+ ///
24
+ /// Returns `ValidationError` if the name is invalid.
25
+ fn validate_plugin_name(name: &str) -> Result<()> {
26
+ if name.is_empty() {
27
+ return Err(KreuzbergError::Validation {
28
+ message: "Plugin name cannot be empty".to_string(),
29
+ source: None,
30
+ });
31
+ }
32
+
33
+ if name.contains(char::is_whitespace) {
34
+ return Err(KreuzbergError::Validation {
35
+ message: format!("Plugin name '{}' cannot contain whitespace", name),
36
+ source: None,
37
+ });
38
+ }
39
+
40
+ Ok(())
41
+ }
42
+
43
+ /// Registry for OCR backend plugins.
44
+ ///
45
+ /// Manages OCR backends with backend type and language-based selection.
46
+ ///
47
+ /// # Thread Safety
48
+ ///
49
+ /// The registry is thread-safe and can be accessed concurrently from multiple threads.
50
+ ///
51
+ /// # Example
52
+ ///
53
+ /// ```rust,no_run
54
+ /// use kreuzberg::plugins::registry::OcrBackendRegistry;
55
+ /// use std::sync::Arc;
56
+ ///
57
+ /// let registry = OcrBackendRegistry::new();
58
+ /// // Register OCR backends
59
+ /// // registry.register(Arc::new(TesseractBackend::new()));
60
+ /// ```
61
+ pub struct OcrBackendRegistry {
62
+ backends: HashMap<String, Arc<dyn OcrBackend>>,
63
+ }
64
+
65
+ impl OcrBackendRegistry {
66
+ /// Create a new OCR backend registry with default backends.
67
+ ///
68
+ /// Registers the Tesseract backend by default if the "ocr" feature is enabled.
69
+ pub fn new() -> Self {
70
+ #[cfg(feature = "ocr")]
71
+ let mut registry = Self {
72
+ backends: HashMap::new(),
73
+ };
74
+
75
+ #[cfg(not(feature = "ocr"))]
76
+ let registry = Self {
77
+ backends: HashMap::new(),
78
+ };
79
+
80
+ #[cfg(feature = "ocr")]
81
+ {
82
+ use crate::ocr::tesseract_backend::TesseractBackend;
83
+ if let Ok(backend) = TesseractBackend::new() {
84
+ let _ = registry.register(Arc::new(backend));
85
+ }
86
+ }
87
+
88
+ registry
89
+ }
90
+
91
+ /// Create a new empty OCR backend registry without default backends.
92
+ ///
93
+ /// This is useful for testing or when you want full control over backend registration.
94
+ pub fn new_empty() -> Self {
95
+ Self {
96
+ backends: HashMap::new(),
97
+ }
98
+ }
99
+
100
+ /// Register an OCR backend.
101
+ ///
102
+ /// # Arguments
103
+ ///
104
+ /// * `backend` - The OCR backend to register
105
+ ///
106
+ /// # Returns
107
+ ///
108
+ /// - `Ok(())` if registration succeeded
109
+ /// - `Err(...)` if initialization failed
110
+ ///
111
+ /// # Example
112
+ ///
113
+ /// ```rust,no_run
114
+ /// # use kreuzberg::plugins::registry::OcrBackendRegistry;
115
+ /// # use std::sync::Arc;
116
+ /// let mut registry = OcrBackendRegistry::new();
117
+ /// // let backend = Arc::new(MyOcrBackend::new());
118
+ /// // registry.register(backend)?;
119
+ /// # Ok::<(), kreuzberg::KreuzbergError>(())
120
+ /// ```
121
+ pub fn register(&mut self, backend: Arc<dyn OcrBackend>) -> Result<()> {
122
+ let name = backend.name().to_string();
123
+
124
+ validate_plugin_name(&name)?;
125
+
126
+ backend.initialize()?;
127
+
128
+ self.backends.insert(name, backend);
129
+ Ok(())
130
+ }
131
+
132
+ /// Get an OCR backend by name.
133
+ ///
134
+ /// # Arguments
135
+ ///
136
+ /// * `name` - Backend name
137
+ ///
138
+ /// # Returns
139
+ ///
140
+ /// The backend if found, or an error if not registered.
141
+ pub fn get(&self, name: &str) -> Result<Arc<dyn OcrBackend>> {
142
+ self.backends.get(name).cloned().ok_or_else(|| KreuzbergError::Plugin {
143
+ message: format!("OCR backend '{}' not registered", name),
144
+ plugin_name: name.to_string(),
145
+ })
146
+ }
147
+
148
+ /// Get an OCR backend that supports a specific language.
149
+ ///
150
+ /// Returns the first backend that supports the language.
151
+ ///
152
+ /// # Arguments
153
+ ///
154
+ /// * `language` - Language code (e.g., "eng", "deu")
155
+ ///
156
+ /// # Returns
157
+ ///
158
+ /// The first backend supporting the language, or an error if none found.
159
+ pub fn get_for_language(&self, language: &str) -> Result<Arc<dyn OcrBackend>> {
160
+ self.backends
161
+ .values()
162
+ .find(|backend| backend.supports_language(language))
163
+ .cloned()
164
+ .ok_or_else(|| KreuzbergError::Plugin {
165
+ message: format!("No OCR backend supports language '{}'", language),
166
+ plugin_name: language.to_string(),
167
+ })
168
+ }
169
+
170
+ /// List all registered backend names.
171
+ pub fn list(&self) -> Vec<String> {
172
+ self.backends.keys().cloned().collect()
173
+ }
174
+
175
+ /// Remove a backend from the registry.
176
+ ///
177
+ /// Calls `shutdown()` on the backend before removing.
178
+ pub fn remove(&mut self, name: &str) -> Result<()> {
179
+ if let Some(backend) = self.backends.remove(name) {
180
+ backend.shutdown()?;
181
+ }
182
+ Ok(())
183
+ }
184
+
185
+ /// Shutdown all backends and clear the registry.
186
+ pub fn shutdown_all(&mut self) -> Result<()> {
187
+ let names: Vec<_> = self.backends.keys().cloned().collect();
188
+ for name in names {
189
+ self.remove(&name)?;
190
+ }
191
+ Ok(())
192
+ }
193
+ }
194
+
195
+ impl Default for OcrBackendRegistry {
196
+ fn default() -> Self {
197
+ Self::new()
198
+ }
199
+ }
200
+
201
+ /// Registry for document extractor plugins.
202
+ ///
203
+ /// Manages extractors with MIME type and priority-based selection.
204
+ ///
205
+ /// # Thread Safety
206
+ ///
207
+ /// The registry is thread-safe and can be accessed concurrently from multiple threads.
208
+ pub struct DocumentExtractorRegistry {
209
+ extractors: HashMap<String, BTreeMap<i32, Arc<dyn DocumentExtractor>>>,
210
+ name_index: HashMap<String, Vec<(String, i32)>>,
211
+ }
212
+
213
+ impl DocumentExtractorRegistry {
214
+ /// Create a new empty extractor registry.
215
+ pub fn new() -> Self {
216
+ Self {
217
+ extractors: HashMap::new(),
218
+ name_index: HashMap::new(),
219
+ }
220
+ }
221
+
222
+ /// Register a document extractor.
223
+ ///
224
+ /// The extractor is registered for all MIME types it supports.
225
+ ///
226
+ /// # Arguments
227
+ ///
228
+ /// * `extractor` - The extractor to register
229
+ ///
230
+ /// # Returns
231
+ ///
232
+ /// - `Ok(())` if registration succeeded
233
+ /// - `Err(...)` if initialization failed
234
+ pub fn register(&mut self, extractor: Arc<dyn DocumentExtractor>) -> Result<()> {
235
+ let name = extractor.name().to_string();
236
+ let priority = extractor.priority();
237
+ let mime_types: Vec<String> = extractor.supported_mime_types().iter().map(|s| s.to_string()).collect();
238
+
239
+ validate_plugin_name(&name)?;
240
+
241
+ extractor.initialize()?;
242
+
243
+ let mut index_entries = Vec::new();
244
+
245
+ for mime_type in &mime_types {
246
+ self.extractors
247
+ .entry(mime_type.clone())
248
+ .or_default()
249
+ .insert(priority, Arc::clone(&extractor));
250
+ index_entries.push((mime_type.clone(), priority));
251
+ }
252
+
253
+ self.name_index.insert(name, index_entries);
254
+
255
+ Ok(())
256
+ }
257
+
258
+ /// Get the highest priority extractor for a MIME type.
259
+ ///
260
+ /// # Arguments
261
+ ///
262
+ /// * `mime_type` - MIME type to look up
263
+ ///
264
+ /// # Returns
265
+ ///
266
+ /// The highest priority extractor, or an error if none found.
267
+ #[cfg_attr(feature = "otel", tracing::instrument(
268
+ skip(self),
269
+ fields(
270
+ registry.mime_type = %mime_type,
271
+ registry.found = tracing::field::Empty,
272
+ )
273
+ ))]
274
+ pub fn get(&self, mime_type: &str) -> Result<Arc<dyn DocumentExtractor>> {
275
+ if let Some(priority_map) = self.extractors.get(mime_type)
276
+ && let Some((_priority, extractor)) = priority_map.iter().next_back()
277
+ {
278
+ #[cfg(feature = "otel")]
279
+ tracing::Span::current().record("registry.found", true);
280
+ return Ok(Arc::clone(extractor));
281
+ }
282
+
283
+ let mut best_match: Option<(i32, Arc<dyn DocumentExtractor>)> = None;
284
+
285
+ for (registered_mime, priority_map) in &self.extractors {
286
+ if registered_mime.ends_with("/*") {
287
+ let prefix = &registered_mime[..registered_mime.len() - 1];
288
+ if mime_type.starts_with(prefix)
289
+ && let Some((_priority, extractor)) = priority_map.iter().next_back()
290
+ {
291
+ let priority = extractor.priority();
292
+ match &best_match {
293
+ None => best_match = Some((priority, Arc::clone(extractor))),
294
+ Some((current_priority, _)) => {
295
+ if priority > *current_priority {
296
+ best_match = Some((priority, Arc::clone(extractor)));
297
+ }
298
+ }
299
+ }
300
+ }
301
+ }
302
+ }
303
+
304
+ if let Some((_priority, extractor)) = best_match {
305
+ #[cfg(feature = "otel")]
306
+ tracing::Span::current().record("registry.found", true);
307
+ return Ok(extractor);
308
+ }
309
+
310
+ #[cfg(feature = "otel")]
311
+ tracing::Span::current().record("registry.found", false);
312
+ Err(KreuzbergError::UnsupportedFormat(mime_type.to_string()))
313
+ }
314
+
315
+ /// List all registered extractors.
316
+ pub fn list(&self) -> Vec<String> {
317
+ self.name_index.keys().cloned().collect()
318
+ }
319
+
320
+ /// Remove an extractor from the registry.
321
+ pub fn remove(&mut self, name: &str) -> Result<()> {
322
+ let index_entries = match self.name_index.remove(name) {
323
+ Some(entries) => entries,
324
+ None => return Ok(()),
325
+ };
326
+
327
+ let mut extractor_to_shutdown: Option<Arc<dyn DocumentExtractor>> = None;
328
+
329
+ for (mime_type, priority) in index_entries {
330
+ if let Some(priority_map) = self.extractors.get_mut(&mime_type) {
331
+ if let Some(extractor) = priority_map.remove(&priority)
332
+ && extractor_to_shutdown.is_none()
333
+ {
334
+ extractor_to_shutdown = Some(extractor);
335
+ }
336
+
337
+ if priority_map.is_empty() {
338
+ self.extractors.remove(&mime_type);
339
+ }
340
+ }
341
+ }
342
+
343
+ if let Some(extractor) = extractor_to_shutdown {
344
+ extractor.shutdown()?;
345
+ }
346
+
347
+ Ok(())
348
+ }
349
+
350
+ /// Shutdown all extractors and clear the registry.
351
+ pub fn shutdown_all(&mut self) -> Result<()> {
352
+ let names = self.list();
353
+ for name in names {
354
+ self.remove(&name)?;
355
+ }
356
+ Ok(())
357
+ }
358
+ }
359
+
360
+ impl Default for DocumentExtractorRegistry {
361
+ fn default() -> Self {
362
+ Self::new()
363
+ }
364
+ }
365
+
366
+ /// Registry for post-processor plugins.
367
+ ///
368
+ /// Manages post-processors organized by processing stage.
369
+ pub struct PostProcessorRegistry {
370
+ processors: HashMap<ProcessingStage, BTreeMap<i32, Vec<Arc<dyn PostProcessor>>>>,
371
+ name_index: HashMap<String, (ProcessingStage, i32)>,
372
+ }
373
+
374
+ impl PostProcessorRegistry {
375
+ /// Create a new empty post-processor registry.
376
+ pub fn new() -> Self {
377
+ Self {
378
+ processors: HashMap::new(),
379
+ name_index: HashMap::new(),
380
+ }
381
+ }
382
+
383
+ /// Register a post-processor.
384
+ ///
385
+ /// # Arguments
386
+ ///
387
+ /// * `processor` - The post-processor to register
388
+ /// * `priority` - Execution priority (higher = runs first within stage)
389
+ pub fn register(&mut self, processor: Arc<dyn PostProcessor>, priority: i32) -> Result<()> {
390
+ let name = processor.name().to_string();
391
+ let stage = processor.processing_stage();
392
+
393
+ validate_plugin_name(&name)?;
394
+
395
+ processor.initialize()?;
396
+
397
+ if self.name_index.contains_key(&name) {
398
+ self.remove(&name)?;
399
+ }
400
+
401
+ self.processors
402
+ .entry(stage)
403
+ .or_default()
404
+ .entry(priority)
405
+ .or_default()
406
+ .push(Arc::clone(&processor));
407
+
408
+ self.name_index.insert(name, (stage, priority));
409
+
410
+ Ok(())
411
+ }
412
+
413
+ /// Get all processors for a specific stage, in priority order.
414
+ ///
415
+ /// # Arguments
416
+ ///
417
+ /// * `stage` - The processing stage
418
+ ///
419
+ /// # Returns
420
+ ///
421
+ /// Vector of processors in priority order (highest first).
422
+ pub fn get_for_stage(&self, stage: ProcessingStage) -> Vec<Arc<dyn PostProcessor>> {
423
+ let mut result = Vec::new();
424
+
425
+ if let Some(priority_map) = self.processors.get(&stage) {
426
+ for (_priority, processors) in priority_map.iter().rev() {
427
+ for processor in processors {
428
+ result.push(Arc::clone(processor));
429
+ }
430
+ }
431
+ }
432
+
433
+ result
434
+ }
435
+
436
+ /// List all registered processor names.
437
+ pub fn list(&self) -> Vec<String> {
438
+ self.name_index.keys().cloned().collect()
439
+ }
440
+
441
+ /// Remove a processor from the registry.
442
+ pub fn remove(&mut self, name: &str) -> Result<()> {
443
+ let (stage, priority) = match self.name_index.remove(name) {
444
+ Some(location) => location,
445
+ None => return Ok(()),
446
+ };
447
+
448
+ let processor_to_shutdown = if let Some(priority_map) = self.processors.get_mut(&stage) {
449
+ let processor = priority_map.get_mut(&priority).and_then(|processors| {
450
+ processors
451
+ .iter()
452
+ .position(|p| p.name() == name)
453
+ .map(|pos| processors.remove(pos))
454
+ });
455
+
456
+ if let Some(processors) = priority_map.get(&priority)
457
+ && processors.is_empty()
458
+ {
459
+ priority_map.remove(&priority);
460
+ }
461
+
462
+ if priority_map.is_empty() {
463
+ self.processors.remove(&stage);
464
+ }
465
+ processor
466
+ } else {
467
+ None
468
+ };
469
+
470
+ if let Some(processor) = processor_to_shutdown {
471
+ processor.shutdown()?;
472
+ }
473
+
474
+ Ok(())
475
+ }
476
+
477
+ /// Shutdown all processors and clear the registry.
478
+ pub fn shutdown_all(&mut self) -> Result<()> {
479
+ let names = self.list();
480
+ for name in names {
481
+ self.remove(&name)?;
482
+ }
483
+ Ok(())
484
+ }
485
+ }
486
+
487
+ impl Default for PostProcessorRegistry {
488
+ fn default() -> Self {
489
+ Self::new()
490
+ }
491
+ }
492
+
493
+ /// Registry for validator plugins.
494
+ ///
495
+ /// Manages validators with priority-based execution order.
496
+ pub struct ValidatorRegistry {
497
+ validators: BTreeMap<i32, IndexMap<String, Arc<dyn Validator>>>,
498
+ }
499
+
500
+ impl ValidatorRegistry {
501
+ /// Create a new empty validator registry.
502
+ pub fn new() -> Self {
503
+ Self {
504
+ validators: BTreeMap::new(),
505
+ }
506
+ }
507
+
508
+ /// Register a validator.
509
+ ///
510
+ /// # Arguments
511
+ ///
512
+ /// * `validator` - The validator to register
513
+ pub fn register(&mut self, validator: Arc<dyn Validator>) -> Result<()> {
514
+ let name = validator.name().to_string();
515
+ let priority = validator.priority();
516
+
517
+ validate_plugin_name(&name)?;
518
+
519
+ validator.initialize()?;
520
+
521
+ self.validators.entry(priority).or_default().insert(name, validator);
522
+
523
+ Ok(())
524
+ }
525
+
526
+ /// Get all validators in priority order.
527
+ ///
528
+ /// # Returns
529
+ ///
530
+ /// Vector of validators in priority order (highest first).
531
+ pub fn get_all(&self) -> Vec<Arc<dyn Validator>> {
532
+ let mut result = Vec::new();
533
+
534
+ for (_priority, validators) in self.validators.iter().rev() {
535
+ for validator in validators.values() {
536
+ result.push(Arc::clone(validator));
537
+ }
538
+ }
539
+
540
+ result
541
+ }
542
+
543
+ /// List all registered validator names.
544
+ pub fn list(&self) -> Vec<String> {
545
+ let mut names = std::collections::HashSet::new();
546
+ for validators in self.validators.values() {
547
+ names.extend(validators.keys().cloned());
548
+ }
549
+ names.into_iter().collect()
550
+ }
551
+
552
+ /// Remove a validator from the registry.
553
+ pub fn remove(&mut self, name: &str) -> Result<()> {
554
+ let mut validator_to_shutdown: Option<Arc<dyn Validator>> = None;
555
+
556
+ for validators in self.validators.values_mut() {
557
+ if let Some(validator) = validators.shift_remove(name)
558
+ && validator_to_shutdown.is_none()
559
+ {
560
+ validator_to_shutdown = Some(validator);
561
+ }
562
+ }
563
+
564
+ if let Some(validator) = validator_to_shutdown {
565
+ validator.shutdown()?;
566
+ }
567
+
568
+ self.validators.retain(|_, validators| !validators.is_empty());
569
+
570
+ Ok(())
571
+ }
572
+
573
+ /// Shutdown all validators and clear the registry.
574
+ pub fn shutdown_all(&mut self) -> Result<()> {
575
+ let names = self.list();
576
+ for name in names {
577
+ self.remove(&name)?;
578
+ }
579
+ Ok(())
580
+ }
581
+ }
582
+
583
+ impl Default for ValidatorRegistry {
584
+ fn default() -> Self {
585
+ Self::new()
586
+ }
587
+ }
588
+
589
+ /// Global OCR backend registry singleton.
590
+ pub static OCR_BACKEND_REGISTRY: Lazy<Arc<RwLock<OcrBackendRegistry>>> =
591
+ Lazy::new(|| Arc::new(RwLock::new(OcrBackendRegistry::new())));
592
+
593
+ /// Global document extractor registry singleton.
594
+ pub static DOCUMENT_EXTRACTOR_REGISTRY: Lazy<Arc<RwLock<DocumentExtractorRegistry>>> =
595
+ Lazy::new(|| Arc::new(RwLock::new(DocumentExtractorRegistry::new())));
596
+
597
+ /// Global post-processor registry singleton.
598
+ pub static POST_PROCESSOR_REGISTRY: Lazy<Arc<RwLock<PostProcessorRegistry>>> =
599
+ Lazy::new(|| Arc::new(RwLock::new(PostProcessorRegistry::new())));
600
+
601
+ /// Global validator registry singleton.
602
+ pub static VALIDATOR_REGISTRY: Lazy<Arc<RwLock<ValidatorRegistry>>> =
603
+ Lazy::new(|| Arc::new(RwLock::new(ValidatorRegistry::new())));
604
+
605
+ /// Get the global OCR backend registry.
606
+ pub fn get_ocr_backend_registry() -> Arc<RwLock<OcrBackendRegistry>> {
607
+ OCR_BACKEND_REGISTRY.clone()
608
+ }
609
+
610
+ /// Get the global document extractor registry.
611
+ pub fn get_document_extractor_registry() -> Arc<RwLock<DocumentExtractorRegistry>> {
612
+ DOCUMENT_EXTRACTOR_REGISTRY.clone()
613
+ }
614
+
615
+ /// Get the global post-processor registry.
616
+ pub fn get_post_processor_registry() -> Arc<RwLock<PostProcessorRegistry>> {
617
+ POST_PROCESSOR_REGISTRY.clone()
618
+ }
619
+
620
+ /// Get the global validator registry.
621
+ pub fn get_validator_registry() -> Arc<RwLock<ValidatorRegistry>> {
622
+ VALIDATOR_REGISTRY.clone()
623
+ }
624
+
625
+ #[cfg(test)]
626
+ mod tests {
627
+ use super::*;
628
+ use crate::core::config::{ExtractionConfig, OcrConfig};
629
+ use crate::plugins::{Plugin, PostProcessor, ProcessingStage, Validator};
630
+ use crate::types::ExtractionResult;
631
+ use async_trait::async_trait;
632
+
633
+ struct MockOcrBackend {
634
+ name: String,
635
+ languages: Vec<String>,
636
+ }
637
+
638
+ impl Plugin for MockOcrBackend {
639
+ fn name(&self) -> &str {
640
+ &self.name
641
+ }
642
+ fn version(&self) -> String {
643
+ "1.0.0".to_string()
644
+ }
645
+ fn initialize(&self) -> Result<()> {
646
+ Ok(())
647
+ }
648
+ fn shutdown(&self) -> Result<()> {
649
+ Ok(())
650
+ }
651
+ }
652
+
653
+ #[async_trait]
654
+ impl OcrBackend for MockOcrBackend {
655
+ async fn process_image(&self, _: &[u8], _: &OcrConfig) -> Result<ExtractionResult> {
656
+ Ok(ExtractionResult {
657
+ content: "test".to_string(),
658
+ mime_type: "text/plain".to_string(),
659
+ metadata: crate::types::Metadata::default(),
660
+ tables: vec![],
661
+ detected_languages: None,
662
+ chunks: None,
663
+ images: None,
664
+ pages: None,
665
+ })
666
+ }
667
+
668
+ fn supports_language(&self, lang: &str) -> bool {
669
+ self.languages.iter().any(|l| l == lang)
670
+ }
671
+
672
+ fn backend_type(&self) -> crate::plugins::ocr::OcrBackendType {
673
+ crate::plugins::ocr::OcrBackendType::Custom
674
+ }
675
+ }
676
+
677
+ struct MockExtractor {
678
+ name: String,
679
+ mime_types: &'static [&'static str],
680
+ priority: i32,
681
+ }
682
+
683
+ impl Plugin for MockExtractor {
684
+ fn name(&self) -> &str {
685
+ &self.name
686
+ }
687
+ fn version(&self) -> String {
688
+ "1.0.0".to_string()
689
+ }
690
+ fn initialize(&self) -> Result<()> {
691
+ Ok(())
692
+ }
693
+ fn shutdown(&self) -> Result<()> {
694
+ Ok(())
695
+ }
696
+ }
697
+
698
+ #[async_trait]
699
+ impl DocumentExtractor for MockExtractor {
700
+ async fn extract_bytes(&self, _: &[u8], _: &str, _: &ExtractionConfig) -> Result<ExtractionResult> {
701
+ Ok(ExtractionResult {
702
+ content: "test".to_string(),
703
+ mime_type: "text/plain".to_string(),
704
+ metadata: crate::types::Metadata::default(),
705
+ tables: vec![],
706
+ detected_languages: None,
707
+ chunks: None,
708
+ images: None,
709
+ pages: None,
710
+ })
711
+ }
712
+
713
+ fn supported_mime_types(&self) -> &[&str] {
714
+ self.mime_types
715
+ }
716
+
717
+ fn priority(&self) -> i32 {
718
+ self.priority
719
+ }
720
+ }
721
+
722
+ struct MockPostProcessor {
723
+ name: String,
724
+ stage: ProcessingStage,
725
+ }
726
+
727
+ impl Plugin for MockPostProcessor {
728
+ fn name(&self) -> &str {
729
+ &self.name
730
+ }
731
+ fn version(&self) -> String {
732
+ "1.0.0".to_string()
733
+ }
734
+ fn initialize(&self) -> Result<()> {
735
+ Ok(())
736
+ }
737
+ fn shutdown(&self) -> Result<()> {
738
+ Ok(())
739
+ }
740
+ }
741
+
742
+ #[async_trait]
743
+ impl PostProcessor for MockPostProcessor {
744
+ async fn process(&self, _result: &mut ExtractionResult, _: &ExtractionConfig) -> Result<()> {
745
+ Ok(())
746
+ }
747
+
748
+ fn processing_stage(&self) -> ProcessingStage {
749
+ self.stage
750
+ }
751
+ }
752
+
753
+ struct MockValidator {
754
+ name: String,
755
+ priority: i32,
756
+ }
757
+
758
+ impl Plugin for MockValidator {
759
+ fn name(&self) -> &str {
760
+ &self.name
761
+ }
762
+ fn version(&self) -> String {
763
+ "1.0.0".to_string()
764
+ }
765
+ fn initialize(&self) -> Result<()> {
766
+ Ok(())
767
+ }
768
+ fn shutdown(&self) -> Result<()> {
769
+ Ok(())
770
+ }
771
+ }
772
+
773
+ #[async_trait]
774
+ impl Validator for MockValidator {
775
+ async fn validate(&self, _: &ExtractionResult, _: &ExtractionConfig) -> Result<()> {
776
+ Ok(())
777
+ }
778
+
779
+ fn priority(&self) -> i32 {
780
+ self.priority
781
+ }
782
+ }
783
+
784
+ #[test]
785
+ fn test_ocr_backend_registry() {
786
+ let mut registry = OcrBackendRegistry::new_empty();
787
+
788
+ let backend = Arc::new(MockOcrBackend {
789
+ name: "test-ocr".to_string(),
790
+ languages: vec!["eng".to_string(), "deu".to_string()],
791
+ });
792
+
793
+ registry.register(backend).unwrap();
794
+
795
+ let retrieved = registry.get("test-ocr").unwrap();
796
+ assert_eq!(retrieved.name(), "test-ocr");
797
+
798
+ let eng_backend = registry.get_for_language("eng").unwrap();
799
+ assert_eq!(eng_backend.name(), "test-ocr");
800
+
801
+ let names = registry.list();
802
+ assert_eq!(names.len(), 1);
803
+ assert!(names.contains(&"test-ocr".to_string()));
804
+ }
805
+
806
+ #[test]
807
+ fn test_post_processor_registry() {
808
+ let mut registry = PostProcessorRegistry::new();
809
+
810
+ let early = Arc::new(MockPostProcessor {
811
+ name: "early-processor".to_string(),
812
+ stage: ProcessingStage::Early,
813
+ });
814
+
815
+ let middle = Arc::new(MockPostProcessor {
816
+ name: "middle-processor".to_string(),
817
+ stage: ProcessingStage::Middle,
818
+ });
819
+
820
+ registry.register(early, 100).unwrap();
821
+ registry.register(middle, 50).unwrap();
822
+
823
+ let early_processors = registry.get_for_stage(ProcessingStage::Early);
824
+ assert_eq!(early_processors.len(), 1);
825
+ assert_eq!(early_processors[0].name(), "early-processor");
826
+
827
+ let middle_processors = registry.get_for_stage(ProcessingStage::Middle);
828
+ assert_eq!(middle_processors.len(), 1);
829
+
830
+ let names = registry.list();
831
+ assert_eq!(names.len(), 2);
832
+ }
833
+
834
+ #[test]
835
+ fn test_validator_registry() {
836
+ let mut registry = ValidatorRegistry::new();
837
+
838
+ let high_priority = Arc::new(MockValidator {
839
+ name: "high-priority".to_string(),
840
+ priority: 100,
841
+ });
842
+
843
+ let low_priority = Arc::new(MockValidator {
844
+ name: "low-priority".to_string(),
845
+ priority: 10,
846
+ });
847
+
848
+ registry.register(high_priority).unwrap();
849
+ registry.register(low_priority).unwrap();
850
+
851
+ let validators = registry.get_all();
852
+ assert_eq!(validators.len(), 2);
853
+ assert_eq!(validators[0].name(), "high-priority");
854
+ assert_eq!(validators[1].name(), "low-priority");
855
+ }
856
+
857
+ #[test]
858
+ fn test_document_extractor_registry_exact_match() {
859
+ let mut registry = DocumentExtractorRegistry::new();
860
+
861
+ let extractor = Arc::new(MockExtractor {
862
+ name: "pdf-extractor".to_string(),
863
+ mime_types: &["application/pdf"],
864
+ priority: 100,
865
+ });
866
+
867
+ registry.register(extractor).unwrap();
868
+
869
+ let retrieved = registry.get("application/pdf").unwrap();
870
+ assert_eq!(retrieved.name(), "pdf-extractor");
871
+
872
+ let names = registry.list();
873
+ assert_eq!(names.len(), 1);
874
+ assert!(names.contains(&"pdf-extractor".to_string()));
875
+ }
876
+
877
+ #[test]
878
+ fn test_document_extractor_registry_prefix_match() {
879
+ let mut registry = DocumentExtractorRegistry::new();
880
+
881
+ let image_extractor = Arc::new(MockExtractor {
882
+ name: "image-extractor".to_string(),
883
+ mime_types: &["image/*"],
884
+ priority: 50,
885
+ });
886
+
887
+ registry.register(image_extractor).unwrap();
888
+
889
+ let retrieved = registry.get("image/png").unwrap();
890
+ assert_eq!(retrieved.name(), "image-extractor");
891
+
892
+ let retrieved_jpg = registry.get("image/jpeg").unwrap();
893
+ assert_eq!(retrieved_jpg.name(), "image-extractor");
894
+ }
895
+
896
+ #[test]
897
+ fn test_document_extractor_registry_priority() {
898
+ let mut registry = DocumentExtractorRegistry::new();
899
+
900
+ let low_priority = Arc::new(MockExtractor {
901
+ name: "low-priority-pdf".to_string(),
902
+ mime_types: &["application/pdf"],
903
+ priority: 10,
904
+ });
905
+
906
+ let high_priority = Arc::new(MockExtractor {
907
+ name: "high-priority-pdf".to_string(),
908
+ mime_types: &["application/pdf"],
909
+ priority: 100,
910
+ });
911
+
912
+ registry.register(low_priority).unwrap();
913
+ registry.register(high_priority).unwrap();
914
+
915
+ let retrieved = registry.get("application/pdf").unwrap();
916
+ assert_eq!(retrieved.name(), "high-priority-pdf");
917
+ }
918
+
919
+ #[test]
920
+ fn test_document_extractor_registry_not_found() {
921
+ let registry = DocumentExtractorRegistry::new();
922
+
923
+ let result = registry.get("application/unknown");
924
+ assert!(matches!(result, Err(KreuzbergError::UnsupportedFormat(_))));
925
+ }
926
+
927
+ #[test]
928
+ fn test_document_extractor_registry_remove() {
929
+ let mut registry = DocumentExtractorRegistry::new();
930
+
931
+ let extractor = Arc::new(MockExtractor {
932
+ name: "test-extractor".to_string(),
933
+ mime_types: &["text/plain"],
934
+ priority: 50,
935
+ });
936
+
937
+ registry.register(extractor).unwrap();
938
+ assert!(registry.get("text/plain").is_ok());
939
+
940
+ registry.remove("test-extractor").unwrap();
941
+ assert!(registry.get("text/plain").is_err());
942
+ }
943
+
944
+ #[test]
945
+ fn test_document_extractor_registry_shutdown_all() {
946
+ let mut registry = DocumentExtractorRegistry::new();
947
+
948
+ let extractor1 = Arc::new(MockExtractor {
949
+ name: "extractor1".to_string(),
950
+ mime_types: &["text/plain"],
951
+ priority: 50,
952
+ });
953
+
954
+ let extractor2 = Arc::new(MockExtractor {
955
+ name: "extractor2".to_string(),
956
+ mime_types: &["application/pdf"],
957
+ priority: 50,
958
+ });
959
+
960
+ registry.register(extractor1).unwrap();
961
+ registry.register(extractor2).unwrap();
962
+
963
+ assert_eq!(registry.list().len(), 2);
964
+
965
+ registry.shutdown_all().unwrap();
966
+ assert_eq!(registry.list().len(), 0);
967
+ }
968
+
969
+ #[test]
970
+ fn test_ocr_backend_registry_not_found() {
971
+ let registry = OcrBackendRegistry::new();
972
+
973
+ let result = registry.get("nonexistent");
974
+ assert!(matches!(result, Err(KreuzbergError::Plugin { .. })));
975
+ }
976
+
977
+ #[test]
978
+ fn test_ocr_backend_registry_language_not_found() {
979
+ let registry = OcrBackendRegistry::new();
980
+
981
+ let result = registry.get_for_language("xyz");
982
+ assert!(matches!(result, Err(KreuzbergError::Plugin { .. })));
983
+ }
984
+
985
+ #[test]
986
+ fn test_ocr_backend_registry_remove() {
987
+ let mut registry = OcrBackendRegistry::new();
988
+
989
+ let backend = Arc::new(MockOcrBackend {
990
+ name: "test-ocr".to_string(),
991
+ languages: vec!["eng".to_string()],
992
+ });
993
+
994
+ registry.register(backend).unwrap();
995
+ assert!(registry.get("test-ocr").is_ok());
996
+
997
+ registry.remove("test-ocr").unwrap();
998
+ assert!(registry.get("test-ocr").is_err());
999
+ }
1000
+
1001
+ #[test]
1002
+ fn test_post_processor_registry_remove() {
1003
+ let mut registry = PostProcessorRegistry::new();
1004
+
1005
+ let processor = Arc::new(MockPostProcessor {
1006
+ name: "test-processor".to_string(),
1007
+ stage: ProcessingStage::Early,
1008
+ });
1009
+
1010
+ registry.register(processor, 50).unwrap();
1011
+ assert_eq!(registry.get_for_stage(ProcessingStage::Early).len(), 1);
1012
+
1013
+ registry.remove("test-processor").unwrap();
1014
+ assert_eq!(registry.get_for_stage(ProcessingStage::Early).len(), 0);
1015
+ }
1016
+
1017
+ #[test]
1018
+ fn test_validator_registry_remove() {
1019
+ let mut registry = ValidatorRegistry::new();
1020
+
1021
+ let validator = Arc::new(MockValidator {
1022
+ name: "test-validator".to_string(),
1023
+ priority: 50,
1024
+ });
1025
+
1026
+ registry.register(validator).unwrap();
1027
+ assert_eq!(registry.get_all().len(), 1);
1028
+
1029
+ registry.remove("test-validator").unwrap();
1030
+ assert_eq!(registry.get_all().len(), 0);
1031
+ }
1032
+
1033
+ #[test]
1034
+ fn test_global_registry_access() {
1035
+ let ocr_registry = get_ocr_backend_registry();
1036
+ let _ = ocr_registry
1037
+ .read()
1038
+ .expect("Failed to acquire read lock on OCR registry in test")
1039
+ .list();
1040
+
1041
+ let extractor_registry = get_document_extractor_registry();
1042
+ let _ = extractor_registry
1043
+ .read()
1044
+ .expect("Failed to acquire read lock on extractor registry in test")
1045
+ .list();
1046
+
1047
+ let processor_registry = get_post_processor_registry();
1048
+ let _ = processor_registry
1049
+ .read()
1050
+ .expect("Failed to acquire read lock on processor registry in test")
1051
+ .list();
1052
+
1053
+ let validator_registry = get_validator_registry();
1054
+ let _ = validator_registry
1055
+ .read()
1056
+ .expect("Failed to acquire read lock on validator registry in test")
1057
+ .list();
1058
+ }
1059
+
1060
+ #[test]
1061
+ fn test_ocr_backend_registry_shutdown_all() {
1062
+ let mut registry = OcrBackendRegistry::new();
1063
+ let baseline = registry.list().len();
1064
+
1065
+ let backend1 = Arc::new(MockOcrBackend {
1066
+ name: "backend1".to_string(),
1067
+ languages: vec!["eng".to_string()],
1068
+ });
1069
+
1070
+ let backend2 = Arc::new(MockOcrBackend {
1071
+ name: "backend2".to_string(),
1072
+ languages: vec!["deu".to_string()],
1073
+ });
1074
+
1075
+ registry.register(backend1).unwrap();
1076
+ registry.register(backend2).unwrap();
1077
+
1078
+ assert_eq!(registry.list().len(), baseline + 2);
1079
+
1080
+ registry.shutdown_all().unwrap();
1081
+ assert_eq!(registry.list().len(), 0);
1082
+ }
1083
+
1084
+ #[test]
1085
+ fn test_post_processor_registry_shutdown_all() {
1086
+ let mut registry = PostProcessorRegistry::new();
1087
+
1088
+ let early = Arc::new(MockPostProcessor {
1089
+ name: "early".to_string(),
1090
+ stage: ProcessingStage::Early,
1091
+ });
1092
+
1093
+ let late = Arc::new(MockPostProcessor {
1094
+ name: "late".to_string(),
1095
+ stage: ProcessingStage::Late,
1096
+ });
1097
+
1098
+ registry.register(early, 100).unwrap();
1099
+ registry.register(late, 50).unwrap();
1100
+
1101
+ assert_eq!(registry.list().len(), 2);
1102
+
1103
+ registry.shutdown_all().unwrap();
1104
+ assert_eq!(registry.list().len(), 0);
1105
+ }
1106
+
1107
+ #[test]
1108
+ fn test_validator_registry_shutdown_all() {
1109
+ let mut registry = ValidatorRegistry::new();
1110
+
1111
+ let validator1 = Arc::new(MockValidator {
1112
+ name: "validator1".to_string(),
1113
+ priority: 100,
1114
+ });
1115
+
1116
+ let validator2 = Arc::new(MockValidator {
1117
+ name: "validator2".to_string(),
1118
+ priority: 50,
1119
+ });
1120
+
1121
+ registry.register(validator1).unwrap();
1122
+ registry.register(validator2).unwrap();
1123
+
1124
+ assert_eq!(registry.get_all().len(), 2);
1125
+
1126
+ registry.shutdown_all().unwrap();
1127
+ assert_eq!(registry.get_all().len(), 0);
1128
+ }
1129
+
1130
+ #[test]
1131
+ fn test_document_extractor_registry_multiple_mime_types() {
1132
+ let mut registry = DocumentExtractorRegistry::new();
1133
+
1134
+ let multi_extractor = Arc::new(MockExtractor {
1135
+ name: "multi-extractor".to_string(),
1136
+ mime_types: &["text/plain", "text/markdown", "text/html"],
1137
+ priority: 50,
1138
+ });
1139
+
1140
+ registry.register(multi_extractor).unwrap();
1141
+
1142
+ assert_eq!(registry.get("text/plain").unwrap().name(), "multi-extractor");
1143
+ assert_eq!(registry.get("text/markdown").unwrap().name(), "multi-extractor");
1144
+ assert_eq!(registry.get("text/html").unwrap().name(), "multi-extractor");
1145
+ }
1146
+
1147
+ #[test]
1148
+ fn test_post_processor_registry_priority_order() {
1149
+ let mut registry = PostProcessorRegistry::new();
1150
+
1151
+ let low = Arc::new(MockPostProcessor {
1152
+ name: "low-priority".to_string(),
1153
+ stage: ProcessingStage::Early,
1154
+ });
1155
+
1156
+ let high = Arc::new(MockPostProcessor {
1157
+ name: "high-priority".to_string(),
1158
+ stage: ProcessingStage::Early,
1159
+ });
1160
+
1161
+ registry.register(low, 10).unwrap();
1162
+ registry.register(high, 100).unwrap();
1163
+
1164
+ let processors = registry.get_for_stage(ProcessingStage::Early);
1165
+ assert_eq!(processors.len(), 2);
1166
+ assert_eq!(processors[0].name(), "high-priority");
1167
+ assert_eq!(processors[1].name(), "low-priority");
1168
+ }
1169
+
1170
+ #[test]
1171
+ fn test_post_processor_registry_empty_stage() {
1172
+ let registry = PostProcessorRegistry::new();
1173
+
1174
+ let processors = registry.get_for_stage(ProcessingStage::Late);
1175
+ assert_eq!(processors.len(), 0);
1176
+ }
1177
+
1178
+ #[test]
1179
+ fn test_ocr_backend_registry_default() {
1180
+ let registry = OcrBackendRegistry::default();
1181
+ #[cfg(feature = "ocr")]
1182
+ assert!(
1183
+ !registry.list().is_empty(),
1184
+ "expected at least one default OCR backend when the 'ocr' feature is enabled"
1185
+ );
1186
+ #[cfg(not(feature = "ocr"))]
1187
+ assert_eq!(registry.list().len(), 0);
1188
+ }
1189
+
1190
+ #[test]
1191
+ fn test_document_extractor_registry_default() {
1192
+ let registry = DocumentExtractorRegistry::default();
1193
+ assert_eq!(registry.list().len(), 0);
1194
+ }
1195
+
1196
+ #[test]
1197
+ fn test_post_processor_registry_default() {
1198
+ let registry = PostProcessorRegistry::default();
1199
+ assert_eq!(registry.list().len(), 0);
1200
+ }
1201
+
1202
+ #[test]
1203
+ fn test_validator_registry_default() {
1204
+ let registry = ValidatorRegistry::default();
1205
+ assert_eq!(registry.get_all().len(), 0);
1206
+ }
1207
+
1208
+ #[test]
1209
+ fn test_document_extractor_registry_exact_over_prefix() {
1210
+ let mut registry = DocumentExtractorRegistry::new();
1211
+
1212
+ let prefix_extractor = Arc::new(MockExtractor {
1213
+ name: "prefix-extractor".to_string(),
1214
+ mime_types: &["image/*"],
1215
+ priority: 100,
1216
+ });
1217
+
1218
+ let exact_extractor = Arc::new(MockExtractor {
1219
+ name: "exact-extractor".to_string(),
1220
+ mime_types: &["image/png"],
1221
+ priority: 50,
1222
+ });
1223
+
1224
+ registry.register(prefix_extractor).unwrap();
1225
+ registry.register(exact_extractor).unwrap();
1226
+
1227
+ let retrieved = registry.get("image/png").unwrap();
1228
+ assert_eq!(retrieved.name(), "exact-extractor");
1229
+
1230
+ let retrieved_jpg = registry.get("image/jpeg").unwrap();
1231
+ assert_eq!(retrieved_jpg.name(), "prefix-extractor");
1232
+ }
1233
+
1234
+ #[test]
1235
+ fn test_ocr_backend_registry_invalid_name_empty() {
1236
+ let mut registry = OcrBackendRegistry::new();
1237
+
1238
+ let backend = Arc::new(MockOcrBackend {
1239
+ name: "".to_string(),
1240
+ languages: vec!["eng".to_string()],
1241
+ });
1242
+
1243
+ let result = registry.register(backend);
1244
+ assert!(matches!(result, Err(KreuzbergError::Validation { .. })));
1245
+ }
1246
+
1247
+ #[test]
1248
+ fn test_ocr_backend_registry_invalid_name_whitespace() {
1249
+ let mut registry = OcrBackendRegistry::new();
1250
+
1251
+ let backend = Arc::new(MockOcrBackend {
1252
+ name: "my ocr backend".to_string(),
1253
+ languages: vec!["eng".to_string()],
1254
+ });
1255
+
1256
+ let result = registry.register(backend);
1257
+ assert!(matches!(result, Err(KreuzbergError::Validation { .. })));
1258
+ }
1259
+
1260
+ #[test]
1261
+ fn test_document_extractor_registry_invalid_name_empty() {
1262
+ let mut registry = DocumentExtractorRegistry::new();
1263
+
1264
+ let extractor = Arc::new(MockExtractor {
1265
+ name: "".to_string(),
1266
+ mime_types: &["text/plain"],
1267
+ priority: 50,
1268
+ });
1269
+
1270
+ let result = registry.register(extractor);
1271
+ assert!(matches!(result, Err(KreuzbergError::Validation { .. })));
1272
+ }
1273
+
1274
+ #[test]
1275
+ fn test_document_extractor_registry_invalid_name_whitespace() {
1276
+ let mut registry = DocumentExtractorRegistry::new();
1277
+
1278
+ let extractor = Arc::new(MockExtractor {
1279
+ name: "my extractor".to_string(),
1280
+ mime_types: &["text/plain"],
1281
+ priority: 50,
1282
+ });
1283
+
1284
+ let result = registry.register(extractor);
1285
+ assert!(matches!(result, Err(KreuzbergError::Validation { .. })));
1286
+ }
1287
+
1288
+ #[test]
1289
+ fn test_post_processor_registry_invalid_name_empty() {
1290
+ let mut registry = PostProcessorRegistry::new();
1291
+
1292
+ let processor = Arc::new(MockPostProcessor {
1293
+ name: "".to_string(),
1294
+ stage: ProcessingStage::Early,
1295
+ });
1296
+
1297
+ let result = registry.register(processor, 50);
1298
+ assert!(matches!(result, Err(KreuzbergError::Validation { .. })));
1299
+ }
1300
+
1301
+ #[test]
1302
+ fn test_post_processor_registry_invalid_name_whitespace() {
1303
+ let mut registry = PostProcessorRegistry::new();
1304
+
1305
+ let processor = Arc::new(MockPostProcessor {
1306
+ name: "my processor".to_string(),
1307
+ stage: ProcessingStage::Early,
1308
+ });
1309
+
1310
+ let result = registry.register(processor, 50);
1311
+ assert!(matches!(result, Err(KreuzbergError::Validation { .. })));
1312
+ }
1313
+
1314
+ #[test]
1315
+ fn test_validator_registry_invalid_name_empty() {
1316
+ let mut registry = ValidatorRegistry::new();
1317
+
1318
+ let validator = Arc::new(MockValidator {
1319
+ name: "".to_string(),
1320
+ priority: 50,
1321
+ });
1322
+
1323
+ let result = registry.register(validator);
1324
+ assert!(matches!(result, Err(KreuzbergError::Validation { .. })));
1325
+ }
1326
+
1327
+ #[test]
1328
+ fn test_validator_registry_invalid_name_whitespace() {
1329
+ let mut registry = ValidatorRegistry::new();
1330
+
1331
+ let validator = Arc::new(MockValidator {
1332
+ name: "my validator".to_string(),
1333
+ priority: 50,
1334
+ });
1335
+
1336
+ let result = registry.register(validator);
1337
+ assert!(matches!(result, Err(KreuzbergError::Validation { .. })));
1338
+ }
1339
+ }