kreuzberg 4.0.0.rc1

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 (265) hide show
  1. checksums.yaml +7 -0
  2. data/.gitignore +8 -0
  3. data/.rspec +3 -0
  4. data/.rubocop.yaml +534 -0
  5. data/Gemfile +9 -0
  6. data/Gemfile.lock +157 -0
  7. data/README.md +421 -0
  8. data/Rakefile +25 -0
  9. data/Steepfile +47 -0
  10. data/examples/async_patterns.rb +340 -0
  11. data/ext/kreuzberg_rb/extconf.rb +35 -0
  12. data/ext/kreuzberg_rb/native/Cargo.toml +36 -0
  13. data/ext/kreuzberg_rb/native/README.md +425 -0
  14. data/ext/kreuzberg_rb/native/build.rs +17 -0
  15. data/ext/kreuzberg_rb/native/include/ieeefp.h +11 -0
  16. data/ext/kreuzberg_rb/native/include/msvc_compat/strings.h +14 -0
  17. data/ext/kreuzberg_rb/native/include/strings.h +20 -0
  18. data/ext/kreuzberg_rb/native/include/unistd.h +47 -0
  19. data/ext/kreuzberg_rb/native/src/lib.rs +2939 -0
  20. data/extconf.rb +28 -0
  21. data/kreuzberg.gemspec +105 -0
  22. data/lib/kreuzberg/api_proxy.rb +142 -0
  23. data/lib/kreuzberg/cache_api.rb +45 -0
  24. data/lib/kreuzberg/cli.rb +55 -0
  25. data/lib/kreuzberg/cli_proxy.rb +127 -0
  26. data/lib/kreuzberg/config.rb +684 -0
  27. data/lib/kreuzberg/errors.rb +50 -0
  28. data/lib/kreuzberg/extraction_api.rb +84 -0
  29. data/lib/kreuzberg/mcp_proxy.rb +186 -0
  30. data/lib/kreuzberg/ocr_backend_protocol.rb +113 -0
  31. data/lib/kreuzberg/post_processor_protocol.rb +86 -0
  32. data/lib/kreuzberg/result.rb +216 -0
  33. data/lib/kreuzberg/setup_lib_path.rb +79 -0
  34. data/lib/kreuzberg/validator_protocol.rb +89 -0
  35. data/lib/kreuzberg/version.rb +5 -0
  36. data/lib/kreuzberg.rb +82 -0
  37. data/pkg/kreuzberg-4.0.0.rc1.gem +0 -0
  38. data/sig/kreuzberg/internal.rbs +184 -0
  39. data/sig/kreuzberg.rbs +468 -0
  40. data/spec/binding/cache_spec.rb +227 -0
  41. data/spec/binding/cli_proxy_spec.rb +87 -0
  42. data/spec/binding/cli_spec.rb +54 -0
  43. data/spec/binding/config_spec.rb +345 -0
  44. data/spec/binding/config_validation_spec.rb +283 -0
  45. data/spec/binding/error_handling_spec.rb +213 -0
  46. data/spec/binding/errors_spec.rb +66 -0
  47. data/spec/binding/plugins/ocr_backend_spec.rb +307 -0
  48. data/spec/binding/plugins/postprocessor_spec.rb +269 -0
  49. data/spec/binding/plugins/validator_spec.rb +274 -0
  50. data/spec/examples.txt +104 -0
  51. data/spec/fixtures/config.toml +39 -0
  52. data/spec/fixtures/config.yaml +42 -0
  53. data/spec/fixtures/invalid_config.toml +4 -0
  54. data/spec/smoke/package_spec.rb +178 -0
  55. data/spec/spec_helper.rb +42 -0
  56. data/vendor/kreuzberg/Cargo.toml +134 -0
  57. data/vendor/kreuzberg/README.md +175 -0
  58. data/vendor/kreuzberg/build.rs +460 -0
  59. data/vendor/kreuzberg/src/api/error.rs +81 -0
  60. data/vendor/kreuzberg/src/api/handlers.rs +199 -0
  61. data/vendor/kreuzberg/src/api/mod.rs +79 -0
  62. data/vendor/kreuzberg/src/api/server.rs +353 -0
  63. data/vendor/kreuzberg/src/api/types.rs +170 -0
  64. data/vendor/kreuzberg/src/bin/profile_extract.rs +455 -0
  65. data/vendor/kreuzberg/src/cache/mod.rs +1143 -0
  66. data/vendor/kreuzberg/src/chunking/mod.rs +677 -0
  67. data/vendor/kreuzberg/src/core/batch_mode.rs +35 -0
  68. data/vendor/kreuzberg/src/core/config.rs +1032 -0
  69. data/vendor/kreuzberg/src/core/extractor.rs +903 -0
  70. data/vendor/kreuzberg/src/core/io.rs +327 -0
  71. data/vendor/kreuzberg/src/core/mime.rs +615 -0
  72. data/vendor/kreuzberg/src/core/mod.rs +42 -0
  73. data/vendor/kreuzberg/src/core/pipeline.rs +906 -0
  74. data/vendor/kreuzberg/src/embeddings.rs +323 -0
  75. data/vendor/kreuzberg/src/error.rs +431 -0
  76. data/vendor/kreuzberg/src/extraction/archive.rs +954 -0
  77. data/vendor/kreuzberg/src/extraction/docx.rs +40 -0
  78. data/vendor/kreuzberg/src/extraction/email.rs +854 -0
  79. data/vendor/kreuzberg/src/extraction/excel.rs +688 -0
  80. data/vendor/kreuzberg/src/extraction/html.rs +553 -0
  81. data/vendor/kreuzberg/src/extraction/image.rs +368 -0
  82. data/vendor/kreuzberg/src/extraction/libreoffice.rs +564 -0
  83. data/vendor/kreuzberg/src/extraction/mod.rs +77 -0
  84. data/vendor/kreuzberg/src/extraction/office_metadata/app_properties.rs +398 -0
  85. data/vendor/kreuzberg/src/extraction/office_metadata/core_properties.rs +247 -0
  86. data/vendor/kreuzberg/src/extraction/office_metadata/custom_properties.rs +240 -0
  87. data/vendor/kreuzberg/src/extraction/office_metadata/mod.rs +128 -0
  88. data/vendor/kreuzberg/src/extraction/pandoc/batch.rs +275 -0
  89. data/vendor/kreuzberg/src/extraction/pandoc/mime_types.rs +178 -0
  90. data/vendor/kreuzberg/src/extraction/pandoc/mod.rs +491 -0
  91. data/vendor/kreuzberg/src/extraction/pandoc/server.rs +496 -0
  92. data/vendor/kreuzberg/src/extraction/pandoc/subprocess.rs +1188 -0
  93. data/vendor/kreuzberg/src/extraction/pandoc/version.rs +162 -0
  94. data/vendor/kreuzberg/src/extraction/pptx.rs +3000 -0
  95. data/vendor/kreuzberg/src/extraction/structured.rs +490 -0
  96. data/vendor/kreuzberg/src/extraction/table.rs +328 -0
  97. data/vendor/kreuzberg/src/extraction/text.rs +269 -0
  98. data/vendor/kreuzberg/src/extraction/xml.rs +333 -0
  99. data/vendor/kreuzberg/src/extractors/archive.rs +425 -0
  100. data/vendor/kreuzberg/src/extractors/docx.rs +479 -0
  101. data/vendor/kreuzberg/src/extractors/email.rs +129 -0
  102. data/vendor/kreuzberg/src/extractors/excel.rs +344 -0
  103. data/vendor/kreuzberg/src/extractors/html.rs +410 -0
  104. data/vendor/kreuzberg/src/extractors/image.rs +195 -0
  105. data/vendor/kreuzberg/src/extractors/mod.rs +268 -0
  106. data/vendor/kreuzberg/src/extractors/pandoc.rs +201 -0
  107. data/vendor/kreuzberg/src/extractors/pdf.rs +496 -0
  108. data/vendor/kreuzberg/src/extractors/pptx.rs +234 -0
  109. data/vendor/kreuzberg/src/extractors/structured.rs +126 -0
  110. data/vendor/kreuzberg/src/extractors/text.rs +242 -0
  111. data/vendor/kreuzberg/src/extractors/xml.rs +128 -0
  112. data/vendor/kreuzberg/src/image/dpi.rs +164 -0
  113. data/vendor/kreuzberg/src/image/mod.rs +6 -0
  114. data/vendor/kreuzberg/src/image/preprocessing.rs +417 -0
  115. data/vendor/kreuzberg/src/image/resize.rs +89 -0
  116. data/vendor/kreuzberg/src/keywords/config.rs +154 -0
  117. data/vendor/kreuzberg/src/keywords/mod.rs +237 -0
  118. data/vendor/kreuzberg/src/keywords/processor.rs +267 -0
  119. data/vendor/kreuzberg/src/keywords/rake.rs +294 -0
  120. data/vendor/kreuzberg/src/keywords/types.rs +68 -0
  121. data/vendor/kreuzberg/src/keywords/yake.rs +163 -0
  122. data/vendor/kreuzberg/src/language_detection/mod.rs +942 -0
  123. data/vendor/kreuzberg/src/lib.rs +102 -0
  124. data/vendor/kreuzberg/src/mcp/mod.rs +32 -0
  125. data/vendor/kreuzberg/src/mcp/server.rs +1966 -0
  126. data/vendor/kreuzberg/src/ocr/cache.rs +469 -0
  127. data/vendor/kreuzberg/src/ocr/error.rs +37 -0
  128. data/vendor/kreuzberg/src/ocr/hocr.rs +216 -0
  129. data/vendor/kreuzberg/src/ocr/mod.rs +58 -0
  130. data/vendor/kreuzberg/src/ocr/processor.rs +847 -0
  131. data/vendor/kreuzberg/src/ocr/table/mod.rs +4 -0
  132. data/vendor/kreuzberg/src/ocr/table/tsv_parser.rs +144 -0
  133. data/vendor/kreuzberg/src/ocr/tesseract_backend.rs +450 -0
  134. data/vendor/kreuzberg/src/ocr/types.rs +393 -0
  135. data/vendor/kreuzberg/src/ocr/utils.rs +47 -0
  136. data/vendor/kreuzberg/src/ocr/validation.rs +206 -0
  137. data/vendor/kreuzberg/src/pdf/error.rs +122 -0
  138. data/vendor/kreuzberg/src/pdf/images.rs +139 -0
  139. data/vendor/kreuzberg/src/pdf/metadata.rs +346 -0
  140. data/vendor/kreuzberg/src/pdf/mod.rs +50 -0
  141. data/vendor/kreuzberg/src/pdf/rendering.rs +369 -0
  142. data/vendor/kreuzberg/src/pdf/table.rs +420 -0
  143. data/vendor/kreuzberg/src/pdf/text.rs +161 -0
  144. data/vendor/kreuzberg/src/plugins/extractor.rs +1010 -0
  145. data/vendor/kreuzberg/src/plugins/mod.rs +209 -0
  146. data/vendor/kreuzberg/src/plugins/ocr.rs +629 -0
  147. data/vendor/kreuzberg/src/plugins/processor.rs +641 -0
  148. data/vendor/kreuzberg/src/plugins/registry.rs +1324 -0
  149. data/vendor/kreuzberg/src/plugins/traits.rs +258 -0
  150. data/vendor/kreuzberg/src/plugins/validator.rs +955 -0
  151. data/vendor/kreuzberg/src/stopwords/mod.rs +1470 -0
  152. data/vendor/kreuzberg/src/text/mod.rs +19 -0
  153. data/vendor/kreuzberg/src/text/quality.rs +697 -0
  154. data/vendor/kreuzberg/src/text/string_utils.rs +217 -0
  155. data/vendor/kreuzberg/src/text/token_reduction/cjk_utils.rs +164 -0
  156. data/vendor/kreuzberg/src/text/token_reduction/config.rs +100 -0
  157. data/vendor/kreuzberg/src/text/token_reduction/core.rs +796 -0
  158. data/vendor/kreuzberg/src/text/token_reduction/filters.rs +902 -0
  159. data/vendor/kreuzberg/src/text/token_reduction/mod.rs +160 -0
  160. data/vendor/kreuzberg/src/text/token_reduction/semantic.rs +619 -0
  161. data/vendor/kreuzberg/src/text/token_reduction/simd_text.rs +147 -0
  162. data/vendor/kreuzberg/src/types.rs +873 -0
  163. data/vendor/kreuzberg/src/utils/mod.rs +17 -0
  164. data/vendor/kreuzberg/src/utils/quality.rs +959 -0
  165. data/vendor/kreuzberg/src/utils/string_utils.rs +381 -0
  166. data/vendor/kreuzberg/stopwords/af_stopwords.json +53 -0
  167. data/vendor/kreuzberg/stopwords/ar_stopwords.json +482 -0
  168. data/vendor/kreuzberg/stopwords/bg_stopwords.json +261 -0
  169. data/vendor/kreuzberg/stopwords/bn_stopwords.json +400 -0
  170. data/vendor/kreuzberg/stopwords/br_stopwords.json +1205 -0
  171. data/vendor/kreuzberg/stopwords/ca_stopwords.json +280 -0
  172. data/vendor/kreuzberg/stopwords/cs_stopwords.json +425 -0
  173. data/vendor/kreuzberg/stopwords/da_stopwords.json +172 -0
  174. data/vendor/kreuzberg/stopwords/de_stopwords.json +622 -0
  175. data/vendor/kreuzberg/stopwords/el_stopwords.json +849 -0
  176. data/vendor/kreuzberg/stopwords/en_stopwords.json +1300 -0
  177. data/vendor/kreuzberg/stopwords/eo_stopwords.json +175 -0
  178. data/vendor/kreuzberg/stopwords/es_stopwords.json +734 -0
  179. data/vendor/kreuzberg/stopwords/et_stopwords.json +37 -0
  180. data/vendor/kreuzberg/stopwords/eu_stopwords.json +100 -0
  181. data/vendor/kreuzberg/stopwords/fa_stopwords.json +801 -0
  182. data/vendor/kreuzberg/stopwords/fi_stopwords.json +849 -0
  183. data/vendor/kreuzberg/stopwords/fr_stopwords.json +693 -0
  184. data/vendor/kreuzberg/stopwords/ga_stopwords.json +111 -0
  185. data/vendor/kreuzberg/stopwords/gl_stopwords.json +162 -0
  186. data/vendor/kreuzberg/stopwords/gu_stopwords.json +226 -0
  187. data/vendor/kreuzberg/stopwords/ha_stopwords.json +41 -0
  188. data/vendor/kreuzberg/stopwords/he_stopwords.json +196 -0
  189. data/vendor/kreuzberg/stopwords/hi_stopwords.json +227 -0
  190. data/vendor/kreuzberg/stopwords/hr_stopwords.json +181 -0
  191. data/vendor/kreuzberg/stopwords/hu_stopwords.json +791 -0
  192. data/vendor/kreuzberg/stopwords/hy_stopwords.json +47 -0
  193. data/vendor/kreuzberg/stopwords/id_stopwords.json +760 -0
  194. data/vendor/kreuzberg/stopwords/it_stopwords.json +634 -0
  195. data/vendor/kreuzberg/stopwords/ja_stopwords.json +136 -0
  196. data/vendor/kreuzberg/stopwords/kn_stopwords.json +84 -0
  197. data/vendor/kreuzberg/stopwords/ko_stopwords.json +681 -0
  198. data/vendor/kreuzberg/stopwords/ku_stopwords.json +64 -0
  199. data/vendor/kreuzberg/stopwords/la_stopwords.json +51 -0
  200. data/vendor/kreuzberg/stopwords/lt_stopwords.json +476 -0
  201. data/vendor/kreuzberg/stopwords/lv_stopwords.json +163 -0
  202. data/vendor/kreuzberg/stopwords/ml_stopwords.json +1 -0
  203. data/vendor/kreuzberg/stopwords/mr_stopwords.json +101 -0
  204. data/vendor/kreuzberg/stopwords/ms_stopwords.json +477 -0
  205. data/vendor/kreuzberg/stopwords/ne_stopwords.json +490 -0
  206. data/vendor/kreuzberg/stopwords/nl_stopwords.json +415 -0
  207. data/vendor/kreuzberg/stopwords/no_stopwords.json +223 -0
  208. data/vendor/kreuzberg/stopwords/pl_stopwords.json +331 -0
  209. data/vendor/kreuzberg/stopwords/pt_stopwords.json +562 -0
  210. data/vendor/kreuzberg/stopwords/ro_stopwords.json +436 -0
  211. data/vendor/kreuzberg/stopwords/ru_stopwords.json +561 -0
  212. data/vendor/kreuzberg/stopwords/si_stopwords.json +193 -0
  213. data/vendor/kreuzberg/stopwords/sk_stopwords.json +420 -0
  214. data/vendor/kreuzberg/stopwords/sl_stopwords.json +448 -0
  215. data/vendor/kreuzberg/stopwords/so_stopwords.json +32 -0
  216. data/vendor/kreuzberg/stopwords/st_stopwords.json +33 -0
  217. data/vendor/kreuzberg/stopwords/sv_stopwords.json +420 -0
  218. data/vendor/kreuzberg/stopwords/sw_stopwords.json +76 -0
  219. data/vendor/kreuzberg/stopwords/ta_stopwords.json +129 -0
  220. data/vendor/kreuzberg/stopwords/te_stopwords.json +54 -0
  221. data/vendor/kreuzberg/stopwords/th_stopwords.json +118 -0
  222. data/vendor/kreuzberg/stopwords/tl_stopwords.json +149 -0
  223. data/vendor/kreuzberg/stopwords/tr_stopwords.json +506 -0
  224. data/vendor/kreuzberg/stopwords/uk_stopwords.json +75 -0
  225. data/vendor/kreuzberg/stopwords/ur_stopwords.json +519 -0
  226. data/vendor/kreuzberg/stopwords/vi_stopwords.json +647 -0
  227. data/vendor/kreuzberg/stopwords/yo_stopwords.json +62 -0
  228. data/vendor/kreuzberg/stopwords/zh_stopwords.json +796 -0
  229. data/vendor/kreuzberg/stopwords/zu_stopwords.json +31 -0
  230. data/vendor/kreuzberg/tests/api_tests.rs +966 -0
  231. data/vendor/kreuzberg/tests/archive_integration.rs +543 -0
  232. data/vendor/kreuzberg/tests/batch_orchestration.rs +542 -0
  233. data/vendor/kreuzberg/tests/batch_processing.rs +304 -0
  234. data/vendor/kreuzberg/tests/chunking_offset_demo.rs +92 -0
  235. data/vendor/kreuzberg/tests/concurrency_stress.rs +509 -0
  236. data/vendor/kreuzberg/tests/config_features.rs +580 -0
  237. data/vendor/kreuzberg/tests/config_loading_tests.rs +439 -0
  238. data/vendor/kreuzberg/tests/core_integration.rs +493 -0
  239. data/vendor/kreuzberg/tests/csv_integration.rs +424 -0
  240. data/vendor/kreuzberg/tests/docx_metadata_extraction_test.rs +124 -0
  241. data/vendor/kreuzberg/tests/email_integration.rs +325 -0
  242. data/vendor/kreuzberg/tests/error_handling.rs +393 -0
  243. data/vendor/kreuzberg/tests/format_integration.rs +159 -0
  244. data/vendor/kreuzberg/tests/helpers/mod.rs +142 -0
  245. data/vendor/kreuzberg/tests/image_integration.rs +253 -0
  246. data/vendor/kreuzberg/tests/keywords_integration.rs +479 -0
  247. data/vendor/kreuzberg/tests/keywords_quality.rs +509 -0
  248. data/vendor/kreuzberg/tests/mime_detection.rs +428 -0
  249. data/vendor/kreuzberg/tests/ocr_configuration.rs +510 -0
  250. data/vendor/kreuzberg/tests/ocr_errors.rs +676 -0
  251. data/vendor/kreuzberg/tests/ocr_quality.rs +627 -0
  252. data/vendor/kreuzberg/tests/ocr_stress.rs +469 -0
  253. data/vendor/kreuzberg/tests/pandoc_integration.rs +503 -0
  254. data/vendor/kreuzberg/tests/pdf_integration.rs +43 -0
  255. data/vendor/kreuzberg/tests/pipeline_integration.rs +1412 -0
  256. data/vendor/kreuzberg/tests/plugin_ocr_backend_test.rs +771 -0
  257. data/vendor/kreuzberg/tests/plugin_postprocessor_test.rs +561 -0
  258. data/vendor/kreuzberg/tests/plugin_system.rs +921 -0
  259. data/vendor/kreuzberg/tests/plugin_validator_test.rs +783 -0
  260. data/vendor/kreuzberg/tests/registry_integration_tests.rs +607 -0
  261. data/vendor/kreuzberg/tests/security_validation.rs +404 -0
  262. data/vendor/kreuzberg/tests/stopwords_integration_test.rs +888 -0
  263. data/vendor/kreuzberg/tests/test_fastembed.rs +609 -0
  264. data/vendor/kreuzberg/tests/xlsx_metadata_extraction_test.rs +87 -0
  265. metadata +471 -0
@@ -0,0 +1,460 @@
1
+ use std::env;
2
+ use std::fs;
3
+ use std::io;
4
+ use std::path::{Path, PathBuf};
5
+ use std::process::Command;
6
+ use std::thread;
7
+ use std::time::Duration;
8
+
9
+ fn main() {
10
+ let target = env::var("TARGET").unwrap();
11
+ let out_dir = PathBuf::from(env::var("OUT_DIR").unwrap());
12
+
13
+ println!("cargo::rustc-check-cfg=cfg(coverage)");
14
+
15
+ let (download_url, lib_name) = get_pdfium_url_and_lib(&target);
16
+
17
+ let pdfium_dir = out_dir.join("pdfium");
18
+
19
+ if let Some(prebuilt) = env::var_os("KREUZBERG_PDFIUM_PREBUILT") {
20
+ let prebuilt_path = PathBuf::from(prebuilt);
21
+ if prebuilt_path.exists() {
22
+ prepare_prebuilt_pdfium(&prebuilt_path, &pdfium_dir)
23
+ .unwrap_or_else(|err| panic!("Failed to copy Pdfium from {}: {}", prebuilt_path.display(), err));
24
+ } else {
25
+ panic!(
26
+ "Environment variable KREUZBERG_PDFIUM_PREBUILT points to '{}' but the directory does not exist",
27
+ prebuilt_path.display()
28
+ );
29
+ }
30
+ }
31
+
32
+ let (runtime_lib_name, runtime_subdir) = runtime_library_info(&target);
33
+ let runtime_lib_path = pdfium_dir.join(runtime_subdir).join(&runtime_lib_name);
34
+ let import_lib_exists = if target.contains("windows") {
35
+ let lib_dir = pdfium_dir.join("lib");
36
+ lib_dir.join("pdfium.lib").exists() || lib_dir.join("pdfium.dll.lib").exists()
37
+ } else {
38
+ true
39
+ };
40
+
41
+ if !runtime_lib_path.exists() || !import_lib_exists {
42
+ tracing::debug!("Pdfium library not found, downloading for target: {}", target);
43
+ tracing::debug!("Download URL: {}", download_url);
44
+ download_and_extract_pdfium(&download_url, &pdfium_dir);
45
+ } else {
46
+ tracing::debug!("Pdfium library already present at {}", runtime_lib_path.display());
47
+ }
48
+
49
+ if target.contains("windows") {
50
+ let lib_dir = pdfium_dir.join("lib");
51
+ let dll_lib = lib_dir.join("pdfium.dll.lib");
52
+ let expected_lib = lib_dir.join("pdfium.lib");
53
+
54
+ if dll_lib.exists() && !expected_lib.exists() {
55
+ tracing::debug!("Renaming cached {} to {}", dll_lib.display(), expected_lib.display());
56
+ fs::rename(&dll_lib, &expected_lib).expect("Failed to rename pdfium.dll.lib to pdfium.lib");
57
+ }
58
+ }
59
+
60
+ let lib_dir = pdfium_dir.join("lib");
61
+ println!("cargo:rustc-link-search=native={}", lib_dir.display());
62
+ println!("cargo:rustc-link-lib=dylib={}", lib_name);
63
+
64
+ if target.contains("darwin") {
65
+ println!("cargo:rustc-link-arg=-Wl,-rpath,@loader_path");
66
+ println!("cargo:rustc-link-arg=-Wl,-rpath,@loader_path/.");
67
+ } else if target.contains("linux") {
68
+ println!("cargo:rustc-link-arg=-Wl,-rpath,$ORIGIN");
69
+ println!("cargo:rustc-link-arg=-Wl,-rpath,$ORIGIN/.");
70
+ }
71
+
72
+ copy_lib_to_package(&pdfium_dir, &target);
73
+
74
+ if target.contains("darwin") {
75
+ println!("cargo:rustc-link-lib=framework=CoreFoundation");
76
+ println!("cargo:rustc-link-lib=framework=CoreGraphics");
77
+ println!("cargo:rustc-link-lib=framework=CoreText");
78
+ println!("cargo:rustc-link-lib=framework=AppKit");
79
+ println!("cargo:rustc-link-lib=dylib=c++");
80
+ } else if target.contains("linux") {
81
+ println!("cargo:rustc-link-lib=dylib=stdc++");
82
+ println!("cargo:rustc-link-lib=dylib=m");
83
+ } else if target.contains("windows") {
84
+ println!("cargo:rustc-link-lib=dylib=gdi32");
85
+ println!("cargo:rustc-link-lib=dylib=user32");
86
+ println!("cargo:rustc-link-lib=dylib=advapi32");
87
+ }
88
+
89
+ println!("cargo:rerun-if-changed=build.rs");
90
+ }
91
+
92
+ fn get_latest_version(repo: &str) -> String {
93
+ let api_url = format!("https://api.github.com/repos/{}/releases/latest", repo);
94
+
95
+ let output = Command::new("curl").args(["-s", &api_url]).output();
96
+
97
+ if let Ok(output) = output
98
+ && output.status.success()
99
+ {
100
+ let json = String::from_utf8_lossy(&output.stdout);
101
+ if let Some(start) = json.find("\"tag_name\":") {
102
+ let after_colon = &json[start + "\"tag_name\":".len()..];
103
+ if let Some(opening_quote) = after_colon.find('"')
104
+ && let Some(closing_quote) = after_colon[opening_quote + 1..].find('"')
105
+ {
106
+ let tag_start = opening_quote + 1;
107
+ let tag = &after_colon[tag_start..tag_start + closing_quote];
108
+ return tag.split('/').next_back().unwrap_or(tag).to_string();
109
+ }
110
+ }
111
+ }
112
+
113
+ "7529".to_string()
114
+ }
115
+
116
+ fn get_pdfium_url_and_lib(target: &str) -> (String, String) {
117
+ if target.contains("wasm") {
118
+ let version = env::var("PDFIUM_WASM_VERSION")
119
+ .ok()
120
+ .filter(|v| !v.is_empty())
121
+ .unwrap_or_else(|| get_latest_version("paulocoutinhox/pdfium-lib"));
122
+ tracing::debug!("Using pdfium-lib version: {}", version);
123
+
124
+ let wasm_arch = if target.contains("wasm32") { "wasm32" } else { "wasm64" };
125
+ return (
126
+ format!(
127
+ "https://github.com/paulocoutinhox/pdfium-lib/releases/download/{}/pdfium-{}.tar.gz",
128
+ version, wasm_arch
129
+ ),
130
+ "pdfium".to_string(),
131
+ );
132
+ }
133
+
134
+ let (platform, arch) = if target.contains("darwin") {
135
+ let arch = if target.contains("aarch64") { "arm64" } else { "x64" };
136
+ ("mac", arch)
137
+ } else if target.contains("linux") {
138
+ let arch = if target.contains("aarch64") {
139
+ "arm64"
140
+ } else if target.contains("arm") {
141
+ "arm"
142
+ } else {
143
+ "x64"
144
+ };
145
+ ("linux", arch)
146
+ } else if target.contains("windows") {
147
+ let arch = if target.contains("aarch64") {
148
+ "arm64"
149
+ } else if target.contains("i686") {
150
+ "x86"
151
+ } else {
152
+ "x64"
153
+ };
154
+ ("win", arch)
155
+ } else {
156
+ panic!("Unsupported target platform: {}", target);
157
+ };
158
+
159
+ let version = env::var("PDFIUM_VERSION")
160
+ .ok()
161
+ .filter(|v| !v.is_empty())
162
+ .unwrap_or_else(|| get_latest_version("bblanchon/pdfium-binaries"));
163
+ tracing::debug!("Using pdfium-binaries version: {}", version);
164
+
165
+ let url = format!(
166
+ "https://github.com/bblanchon/pdfium-binaries/releases/download/chromium/{}/pdfium-{}-{}.tgz",
167
+ version, platform, arch
168
+ );
169
+
170
+ (url, "pdfium".to_string())
171
+ }
172
+
173
+ fn download_and_extract_pdfium(url: &str, dest_dir: &Path) {
174
+ fs::create_dir_all(dest_dir).expect("Failed to create pdfium directory");
175
+
176
+ let archive_path = dest_dir.join("pdfium.tar.gz");
177
+ let retries = env::var("KREUZBERG_PDFIUM_DOWNLOAD_RETRIES")
178
+ .ok()
179
+ .and_then(|value| value.parse::<u32>().ok())
180
+ .filter(|value| *value > 0)
181
+ .unwrap_or(5);
182
+ let base_delay = env::var("KREUZBERG_PDFIUM_DOWNLOAD_BACKOFF_SECS")
183
+ .ok()
184
+ .and_then(|value| value.parse::<u64>().ok())
185
+ .filter(|value| *value > 0)
186
+ .unwrap_or(2);
187
+
188
+ let archive_path_str = archive_path
189
+ .to_str()
190
+ .unwrap_or_else(|| panic!("Non-UTF8 path for archive: {}", archive_path.display()));
191
+ let mut last_error = String::new();
192
+
193
+ for attempt in 1..=retries {
194
+ let _ = fs::remove_file(&archive_path);
195
+ tracing::debug!(
196
+ "Downloading Pdfium archive from: {} (attempt {}/{})",
197
+ url,
198
+ attempt,
199
+ retries
200
+ );
201
+
202
+ let status = Command::new("curl")
203
+ .args(["-f", "-L", "-o", archive_path_str, url])
204
+ .status();
205
+
206
+ match status {
207
+ Ok(code) if code.success() => {
208
+ last_error.clear();
209
+ break;
210
+ }
211
+ Ok(code) => {
212
+ last_error = format!("curl exited with {:?}", code.code());
213
+ }
214
+ Err(err) => {
215
+ last_error = format!("failed to spawn curl: {err}");
216
+ }
217
+ }
218
+
219
+ if attempt == retries {
220
+ panic!(
221
+ "Failed to download Pdfium from {} after {} attempts. Last error: {}",
222
+ url, retries, last_error
223
+ );
224
+ }
225
+
226
+ let exponent = u32::min(attempt, 5);
227
+ let multiplier = 1u64 << exponent;
228
+ let delay_secs = base_delay.saturating_mul(multiplier).min(30);
229
+ println!(
230
+ "cargo:warning=Pdfium download failed (attempt {}/{}) - {}. Retrying in {}s",
231
+ attempt, retries, last_error, delay_secs
232
+ );
233
+ thread::sleep(Duration::from_secs(delay_secs));
234
+ }
235
+
236
+ let file_type = Command::new("file")
237
+ .arg(archive_path.to_str().unwrap())
238
+ .output()
239
+ .expect("Failed to check file type");
240
+
241
+ let file_type_output = String::from_utf8_lossy(&file_type.stdout);
242
+ tracing::debug!("Downloaded file type: {}", file_type_output.trim());
243
+
244
+ if !file_type_output.to_lowercase().contains("gzip") && !file_type_output.to_lowercase().contains("compressed") {
245
+ fs::remove_file(&archive_path).ok();
246
+ panic!(
247
+ "Downloaded file is not a valid gzip archive. URL may be incorrect or version unavailable: {}",
248
+ url
249
+ );
250
+ }
251
+
252
+ tracing::debug!("Extracting Pdfium archive...");
253
+ let status = Command::new("tar")
254
+ .args(["-xzf", archive_path.to_str().unwrap(), "-C", dest_dir.to_str().unwrap()])
255
+ .status()
256
+ .expect("Failed to execute tar");
257
+
258
+ if !status.success() {
259
+ fs::remove_file(&archive_path).ok();
260
+ panic!("Failed to extract Pdfium archive from {}", url);
261
+ }
262
+
263
+ fs::remove_file(&archive_path).ok();
264
+
265
+ let target = env::var("TARGET").unwrap();
266
+ if target.contains("windows") {
267
+ let lib_dir = dest_dir.join("lib");
268
+ let dll_lib = lib_dir.join("pdfium.dll.lib");
269
+ let expected_lib = lib_dir.join("pdfium.lib");
270
+
271
+ if dll_lib.exists() {
272
+ tracing::debug!("Ensuring Windows import library at {}", expected_lib.display());
273
+ if let Err(err) = fs::copy(&dll_lib, &expected_lib) {
274
+ panic!("Failed to copy pdfium.dll.lib to pdfium.lib: {err}");
275
+ }
276
+ } else {
277
+ tracing::debug!("Warning: Expected {} not found after extraction", dll_lib.display());
278
+ }
279
+ }
280
+
281
+ tracing::debug!("Pdfium downloaded and extracted successfully");
282
+ }
283
+
284
+ fn copy_lib_to_package(pdfium_dir: &Path, target: &str) {
285
+ let (runtime_lib_name, runtime_subdir) = runtime_library_info(target);
286
+ let src_lib = pdfium_dir.join(runtime_subdir).join(&runtime_lib_name);
287
+
288
+ if !src_lib.exists() {
289
+ tracing::debug!("Source library not found: {}", src_lib.display());
290
+ return;
291
+ }
292
+
293
+ // Fix install_name on macOS to use @rpath
294
+ if target.contains("darwin") {
295
+ fix_macos_install_name(&src_lib, &runtime_lib_name);
296
+ codesign_if_needed(target, &src_lib);
297
+ }
298
+
299
+ let crate_dir = PathBuf::from(env::var("CARGO_MANIFEST_DIR").unwrap());
300
+ let workspace_root = crate_dir.parent().unwrap().parent().unwrap();
301
+
302
+ // Copy to target directory for CLI binary
303
+ if let Ok(profile) = env::var("PROFILE") {
304
+ let target_dir = workspace_root.join("target").join(profile);
305
+ if target_dir.exists() {
306
+ copy_lib_if_needed(
307
+ &src_lib,
308
+ &target_dir.join(&runtime_lib_name),
309
+ "CLI target directory",
310
+ target,
311
+ );
312
+ }
313
+ }
314
+
315
+ let python_dest_dir = workspace_root.join("packages").join("python").join("kreuzberg");
316
+ if python_dest_dir.exists() {
317
+ copy_lib_if_needed(
318
+ &src_lib,
319
+ &python_dest_dir.join(&runtime_lib_name),
320
+ "Python package",
321
+ target,
322
+ );
323
+ } else {
324
+ tracing::debug!("Python package directory not found, skipping Python library copy");
325
+ }
326
+
327
+ let node_dest_dir = workspace_root.join("crates").join("kreuzberg-node");
328
+ if node_dest_dir.exists() {
329
+ copy_lib_if_needed(
330
+ &src_lib,
331
+ &node_dest_dir.join(&runtime_lib_name),
332
+ "Node.js package",
333
+ target,
334
+ );
335
+ } else {
336
+ tracing::debug!("Node.js package directory not found, skipping Node library copy");
337
+ }
338
+
339
+ let ruby_dest_dir = workspace_root.join("packages").join("ruby").join("lib");
340
+ if ruby_dest_dir.exists() {
341
+ copy_lib_if_needed(&src_lib, &ruby_dest_dir.join(&runtime_lib_name), "Ruby package", target);
342
+ } else {
343
+ tracing::debug!("Ruby package directory not found, skipping Ruby library copy");
344
+ }
345
+ }
346
+
347
+ fn copy_lib_if_needed(src: &Path, dest: &Path, package_name: &str, target: &str) {
348
+ use std::fs;
349
+
350
+ let should_copy = if dest.exists() {
351
+ let src_metadata = fs::metadata(src).ok();
352
+ let dest_metadata = fs::metadata(dest).ok();
353
+ match (src_metadata, dest_metadata) {
354
+ (Some(src), Some(dest)) => src.modified().ok() > dest.modified().ok(),
355
+ _ => true,
356
+ }
357
+ } else {
358
+ true
359
+ };
360
+
361
+ if should_copy {
362
+ match fs::copy(src, dest) {
363
+ Ok(_) => {
364
+ tracing::debug!("Copied {} to {} ({})", src.display(), dest.display(), package_name);
365
+ codesign_if_needed(target, dest);
366
+ }
367
+ Err(e) => tracing::debug!("Failed to copy library to {}: {}", package_name, e),
368
+ }
369
+ }
370
+ }
371
+
372
+ fn codesign_if_needed(target: &str, binary: &Path) {
373
+ if !target.contains("apple-darwin") || !binary.exists() {
374
+ return;
375
+ }
376
+
377
+ let identity = env::var("KREUZBERG_CODESIGN_IDENTITY").unwrap_or_else(|_| "-".to_string());
378
+ let status = Command::new("codesign")
379
+ .arg("--force")
380
+ .arg("--timestamp=none")
381
+ .arg("--sign")
382
+ .arg(identity)
383
+ .arg(binary)
384
+ .status();
385
+
386
+ match status {
387
+ Ok(result) if result.success() => {
388
+ tracing::debug!("Codesigned {}", binary.display());
389
+ }
390
+ Ok(result) => {
391
+ tracing::debug!(
392
+ "codesign exited with status {} while signing {}",
393
+ result,
394
+ binary.display()
395
+ );
396
+ }
397
+ Err(err) => {
398
+ tracing::debug!("Failed to run codesign for {}: {}", binary.display(), err);
399
+ }
400
+ }
401
+ }
402
+
403
+ fn runtime_library_info(target: &str) -> (String, &'static str) {
404
+ if target.contains("windows") {
405
+ ("pdfium.dll".to_string(), "bin")
406
+ } else if target.contains("darwin") {
407
+ ("libpdfium.dylib".to_string(), "lib")
408
+ } else {
409
+ ("libpdfium.so".to_string(), "lib")
410
+ }
411
+ }
412
+
413
+ fn prepare_prebuilt_pdfium(prebuilt_src: &Path, dest_dir: &Path) -> io::Result<()> {
414
+ if dest_dir.exists() {
415
+ fs::remove_dir_all(dest_dir)?;
416
+ }
417
+ copy_dir_all(prebuilt_src, dest_dir)
418
+ }
419
+
420
+ fn copy_dir_all(src: &Path, dst: &Path) -> io::Result<()> {
421
+ fs::create_dir_all(dst)?;
422
+ for entry in fs::read_dir(src)? {
423
+ let entry = entry?;
424
+ let file_type = entry.file_type()?;
425
+ let target_path = dst.join(entry.file_name());
426
+ if file_type.is_dir() {
427
+ copy_dir_all(&entry.path(), &target_path)?;
428
+ } else {
429
+ fs::copy(entry.path(), &target_path)?;
430
+ }
431
+ }
432
+ Ok(())
433
+ }
434
+
435
+ fn fix_macos_install_name(lib_path: &Path, lib_name: &str) {
436
+ use std::process::Command;
437
+
438
+ // Change install_name from ./libpdfium.dylib to @rpath/libpdfium.dylib
439
+ let new_install_name = format!("@rpath/{}", lib_name);
440
+
441
+ tracing::debug!("Fixing install_name for {} to {}", lib_path.display(), new_install_name);
442
+
443
+ let status = Command::new("install_name_tool")
444
+ .arg("-id")
445
+ .arg(&new_install_name)
446
+ .arg(lib_path)
447
+ .status();
448
+
449
+ match status {
450
+ Ok(s) if s.success() => {
451
+ tracing::debug!("Successfully updated install_name");
452
+ }
453
+ Ok(s) => {
454
+ tracing::debug!("install_name_tool failed with status: {}", s);
455
+ }
456
+ Err(e) => {
457
+ tracing::debug!("Failed to run install_name_tool: {}", e);
458
+ }
459
+ }
460
+ }
@@ -0,0 +1,81 @@
1
+ //! API error handling.
2
+
3
+ use axum::{
4
+ Json,
5
+ http::StatusCode,
6
+ response::{IntoResponse, Response},
7
+ };
8
+
9
+ use crate::error::KreuzbergError;
10
+
11
+ use super::types::ErrorResponse;
12
+
13
+ /// API-specific error wrapper.
14
+ #[derive(Debug)]
15
+ pub struct ApiError {
16
+ /// HTTP status code
17
+ pub status: StatusCode,
18
+ /// Error response body
19
+ pub body: ErrorResponse,
20
+ }
21
+
22
+ impl ApiError {
23
+ /// Create a new API error.
24
+ pub fn new(status: StatusCode, error: KreuzbergError) -> Self {
25
+ let error_type = match &error {
26
+ KreuzbergError::Validation { .. } => "ValidationError",
27
+ KreuzbergError::Parsing { .. } => "ParsingError",
28
+ KreuzbergError::Ocr { .. } => "OCRError",
29
+ KreuzbergError::Io(_) => "IOError",
30
+ KreuzbergError::Cache { .. } => "CacheError",
31
+ KreuzbergError::ImageProcessing { .. } => "ImageProcessingError",
32
+ KreuzbergError::Serialization { .. } => "SerializationError",
33
+ KreuzbergError::MissingDependency(_) => "MissingDependencyError",
34
+ KreuzbergError::Plugin { .. } => "PluginError",
35
+ KreuzbergError::LockPoisoned(_) => "LockPoisonedError",
36
+ KreuzbergError::UnsupportedFormat(_) => "UnsupportedFormatError",
37
+ KreuzbergError::Other(_) => "Error",
38
+ };
39
+
40
+ Self {
41
+ status,
42
+ body: ErrorResponse {
43
+ error_type: error_type.to_string(),
44
+ message: error.to_string(),
45
+ traceback: None,
46
+ status_code: status.as_u16(),
47
+ },
48
+ }
49
+ }
50
+
51
+ /// Create a validation error (400).
52
+ pub fn validation(error: KreuzbergError) -> Self {
53
+ Self::new(StatusCode::BAD_REQUEST, error)
54
+ }
55
+
56
+ /// Create an unprocessable entity error (422).
57
+ pub fn unprocessable(error: KreuzbergError) -> Self {
58
+ Self::new(StatusCode::UNPROCESSABLE_ENTITY, error)
59
+ }
60
+
61
+ /// Create an internal server error (500).
62
+ pub fn internal(error: KreuzbergError) -> Self {
63
+ Self::new(StatusCode::INTERNAL_SERVER_ERROR, error)
64
+ }
65
+ }
66
+
67
+ impl IntoResponse for ApiError {
68
+ fn into_response(self) -> Response {
69
+ (self.status, Json(self.body)).into_response()
70
+ }
71
+ }
72
+
73
+ impl From<KreuzbergError> for ApiError {
74
+ fn from(error: KreuzbergError) -> Self {
75
+ match &error {
76
+ KreuzbergError::Validation { .. } => Self::validation(error),
77
+ KreuzbergError::Parsing { .. } | KreuzbergError::Ocr { .. } => Self::unprocessable(error),
78
+ _ => Self::internal(error),
79
+ }
80
+ }
81
+ }