kordoc 4.14.2 → 4.14.4

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 (151) hide show
  1. package/README.md +51 -6
  2. package/dist/{-AQKXLFIM.js → -Z4TJ3QGK.js} +13 -11
  3. package/dist/{chunk-QOMNEWCQ.cjs → chunk-5LAZQCUJ.cjs} +305 -31
  4. package/dist/chunk-5LAZQCUJ.cjs.map +1 -0
  5. package/dist/{chunk-7IAQ4HOK.js → chunk-6BWU6SQX.js} +1 -1
  6. package/dist/chunk-6BWU6SQX.js.map +1 -0
  7. package/dist/{chunk-FOJZNMMK.js → chunk-6CUIXJOC.js} +1 -1
  8. package/dist/chunk-6CUIXJOC.js.map +1 -0
  9. package/dist/{chunk-ZWD734EY.js → chunk-7SE4EZ45.js} +2 -2
  10. package/dist/{chunk-TSFS5VKD.js → chunk-D4U2TXDC.js} +332 -141
  11. package/dist/chunk-D4U2TXDC.js.map +1 -0
  12. package/dist/chunk-DPWWXUOG.js +2222 -0
  13. package/dist/chunk-DPWWXUOG.js.map +1 -0
  14. package/dist/{chunk-GDW4YP7K.js → chunk-GLSBEEZJ.js} +2 -2
  15. package/dist/{chunk-JQEFP45L.js → chunk-GUJ4Q73I.js} +2 -2
  16. package/dist/{chunk-ZWOFQRIB.js → chunk-HIJUKZHM.js} +3 -3
  17. package/dist/chunk-IMAAFSAX.cjs +1352 -0
  18. package/dist/chunk-IMAAFSAX.cjs.map +1 -0
  19. package/dist/{chunk-TNCKOL2O.js → chunk-LCW7SEPG.js} +2 -2
  20. package/dist/chunk-LFSQAMSM.js +1352 -0
  21. package/dist/chunk-LFSQAMSM.js.map +1 -0
  22. package/dist/chunk-MG6DJEX4.js +1299 -0
  23. package/dist/chunk-MG6DJEX4.js.map +1 -0
  24. package/dist/{chunk-QC5T6D4J.js → chunk-NLU6LBLY.js} +299 -25
  25. package/dist/chunk-NLU6LBLY.js.map +1 -0
  26. package/dist/{chunk-EQZI77PP.cjs → chunk-QEJUY6SK.cjs} +3 -3
  27. package/dist/{chunk-EQZI77PP.cjs.map → chunk-QEJUY6SK.cjs.map} +1 -1
  28. package/dist/{chunk-VQ5W4MYO.js → chunk-QIF2VXBQ.js} +298 -24
  29. package/dist/chunk-QIF2VXBQ.js.map +1 -0
  30. package/dist/{chunk-UET462HB.js → chunk-TITDJWVZ.js} +2 -2
  31. package/dist/{chunk-247ANM72.cjs → chunk-USHQCF5T.cjs} +870 -147
  32. package/dist/chunk-USHQCF5T.cjs.map +1 -0
  33. package/dist/{chunk-C7FQBSON.js → chunk-V3DX4DUF.js} +2 -2
  34. package/dist/{chunk-DOGPOY4B.js → chunk-VDN7LMPL.js} +2 -2
  35. package/dist/{chunk-EDNJLUSE.js → chunk-VJIL6PIL.js} +2 -2
  36. package/dist/{chunk-L7DMSMAG.cjs → chunk-VWUYVEFN.cjs} +1 -1
  37. package/dist/chunk-VWUYVEFN.cjs.map +1 -0
  38. package/dist/{chunk-HR2XNERB.cjs → chunk-WJHAU73B.cjs} +2 -2
  39. package/dist/{chunk-HR2XNERB.cjs.map → chunk-WJHAU73B.cjs.map} +1 -1
  40. package/dist/{chunk-NS55SM7A.js → chunk-WJJGW5PL.js} +2 -2
  41. package/dist/{chunk-EISO3POT.js → chunk-XVXHXQIO.js} +788 -194
  42. package/dist/chunk-XVXHXQIO.js.map +1 -0
  43. package/dist/{chunk-ILU76UF3.js → chunk-YAKC46HZ.js} +2 -2
  44. package/dist/chunk-YAKC46HZ.js.map +1 -0
  45. package/dist/{chunk-3DUADAY2.js → chunk-YAVHVFRY.js} +851 -128
  46. package/dist/chunk-YAVHVFRY.js.map +1 -0
  47. package/dist/{chunk-A7QC2ZVP.js → chunk-YT3GC62E.js} +851 -128
  48. package/dist/chunk-YT3GC62E.js.map +1 -0
  49. package/dist/chunks-CCMRL4MJ.js +10 -0
  50. package/dist/cli.js +228 -97
  51. package/dist/cli.js.map +1 -1
  52. package/dist/{formula-PJJ6S3Z2.cjs → formula-5LOVHAKE.cjs} +9 -9
  53. package/dist/{formula-PJJ6S3Z2.cjs.map → formula-5LOVHAKE.cjs.map} +1 -1
  54. package/dist/{formula-JGFTXWYT.js → formula-7RANCW5L.js} +3 -3
  55. package/dist/{formula-5S7WWXGH.js → formula-7UYJOVTB.js} +4 -4
  56. package/dist/{image-ocr-WCAYXLOA.js → image-ocr-I6JOCATS.js} +34 -16
  57. package/dist/image-ocr-I6JOCATS.js.map +1 -0
  58. package/dist/{image-ocr-ZTYAGV3W.cjs → image-ocr-NYBJEOH2.cjs} +36 -18
  59. package/dist/image-ocr-NYBJEOH2.cjs.map +1 -0
  60. package/dist/{image-ocr-FLIL7MGN.js → image-ocr-ZV6PDBUQ.js} +32 -14
  61. package/dist/image-ocr-ZV6PDBUQ.js.map +1 -0
  62. package/dist/index.cjs +4048 -1216
  63. package/dist/index.cjs.map +1 -1
  64. package/dist/index.d.cts +66 -28
  65. package/dist/index.d.ts +66 -28
  66. package/dist/index.js +3487 -655
  67. package/dist/index.js.map +1 -1
  68. package/dist/mcp.js +74 -53
  69. package/dist/mcp.js.map +1 -1
  70. package/dist/{model-bundle-FSPZZ2EM.js → model-bundle-2C5IYDLE.js} +5 -5
  71. package/dist/{models-CJPAQK5R.js → models-SEQZHZFT.js} +5 -5
  72. package/dist/{parser-SCUTZAIJ.js → parser-3FNXEP6T.js} +708 -250
  73. package/dist/parser-3FNXEP6T.js.map +1 -0
  74. package/dist/{parser-A4TMPJAM.js → parser-FJBJXRJK.js} +708 -250
  75. package/dist/parser-FJBJXRJK.js.map +1 -0
  76. package/dist/{parser-VCXE4I5A.cjs → parser-HVEKF73B.cjs} +733 -275
  77. package/dist/parser-HVEKF73B.cjs.map +1 -0
  78. package/dist/pdf-ocr-JRDBJ6TA.cjs +14 -0
  79. package/dist/{pdf-ocr-MBZZ57LD.cjs.map → pdf-ocr-JRDBJ6TA.cjs.map} +1 -1
  80. package/dist/pdf-ocr-UUBWBEHL.js +16 -0
  81. package/dist/pdf-ocr-VJSF2SQC.js +13 -0
  82. package/dist/{profile-io-LBM263K7.js → profile-io-RUFZOZO6.js} +3 -3
  83. package/dist/{rasterize-46RG4ZF6.js → rasterize-IOYXR5SI.js} +3 -3
  84. package/dist/redact-NMIKVHMP.js +18 -0
  85. package/dist/redact-doc-QBHNJ2BU.js +1284 -0
  86. package/dist/redact-doc-QBHNJ2BU.js.map +1 -0
  87. package/dist/{render-NGP5T24S.js → render-YZ2TEC3N.js} +7 -7
  88. package/dist/seal-ECV44TUH.js +10 -0
  89. package/dist/{setup-AU4OCHNH.js → setup-MPIEP6XT.js} +3 -3
  90. package/dist/visual-ICMOWUCR.js +19 -0
  91. package/dist/{watch-MKPUCPXM.js → watch-RYO7MN5B.js} +12 -12
  92. package/package.json +2 -2
  93. package/dist/chunk-247ANM72.cjs.map +0 -1
  94. package/dist/chunk-3DUADAY2.js.map +0 -1
  95. package/dist/chunk-422427XJ.cjs +0 -571
  96. package/dist/chunk-422427XJ.cjs.map +0 -1
  97. package/dist/chunk-7IAQ4HOK.js.map +0 -1
  98. package/dist/chunk-A7QC2ZVP.js.map +0 -1
  99. package/dist/chunk-DZIXKL7E.js +0 -145
  100. package/dist/chunk-DZIXKL7E.js.map +0 -1
  101. package/dist/chunk-EISO3POT.js.map +0 -1
  102. package/dist/chunk-EP7KOI2V.js +0 -571
  103. package/dist/chunk-EP7KOI2V.js.map +0 -1
  104. package/dist/chunk-FOJZNMMK.js.map +0 -1
  105. package/dist/chunk-ILU76UF3.js.map +0 -1
  106. package/dist/chunk-KVWV4ULP.js +0 -516
  107. package/dist/chunk-KVWV4ULP.js.map +0 -1
  108. package/dist/chunk-L7DMSMAG.cjs.map +0 -1
  109. package/dist/chunk-QC5T6D4J.js.map +0 -1
  110. package/dist/chunk-QOMNEWCQ.cjs.map +0 -1
  111. package/dist/chunk-TSFS5VKD.js.map +0 -1
  112. package/dist/chunk-VQ5W4MYO.js.map +0 -1
  113. package/dist/chunks-SX77P5CV.js +0 -10
  114. package/dist/image-ocr-FLIL7MGN.js.map +0 -1
  115. package/dist/image-ocr-WCAYXLOA.js.map +0 -1
  116. package/dist/image-ocr-ZTYAGV3W.cjs.map +0 -1
  117. package/dist/parser-A4TMPJAM.js.map +0 -1
  118. package/dist/parser-SCUTZAIJ.js.map +0 -1
  119. package/dist/parser-VCXE4I5A.cjs.map +0 -1
  120. package/dist/pdf-ocr-76SAUXMZ.js +0 -13
  121. package/dist/pdf-ocr-CPU55CVF.js +0 -16
  122. package/dist/pdf-ocr-MBZZ57LD.cjs +0 -14
  123. package/dist/redact-7ILEAUTS.js +0 -12
  124. package/dist/seal-RGUPOLXD.js +0 -10
  125. package/dist/visual-NZLYIYEM.js +0 -19
  126. /package/dist/{-AQKXLFIM.js.map → -Z4TJ3QGK.js.map} +0 -0
  127. /package/dist/{chunk-ZWD734EY.js.map → chunk-7SE4EZ45.js.map} +0 -0
  128. /package/dist/{chunk-GDW4YP7K.js.map → chunk-GLSBEEZJ.js.map} +0 -0
  129. /package/dist/{chunk-JQEFP45L.js.map → chunk-GUJ4Q73I.js.map} +0 -0
  130. /package/dist/{chunk-ZWOFQRIB.js.map → chunk-HIJUKZHM.js.map} +0 -0
  131. /package/dist/{chunk-TNCKOL2O.js.map → chunk-LCW7SEPG.js.map} +0 -0
  132. /package/dist/{chunk-UET462HB.js.map → chunk-TITDJWVZ.js.map} +0 -0
  133. /package/dist/{chunk-C7FQBSON.js.map → chunk-V3DX4DUF.js.map} +0 -0
  134. /package/dist/{chunk-DOGPOY4B.js.map → chunk-VDN7LMPL.js.map} +0 -0
  135. /package/dist/{chunk-EDNJLUSE.js.map → chunk-VJIL6PIL.js.map} +0 -0
  136. /package/dist/{chunk-NS55SM7A.js.map → chunk-WJJGW5PL.js.map} +0 -0
  137. /package/dist/{chunks-SX77P5CV.js.map → chunks-CCMRL4MJ.js.map} +0 -0
  138. /package/dist/{formula-JGFTXWYT.js.map → formula-7RANCW5L.js.map} +0 -0
  139. /package/dist/{formula-5S7WWXGH.js.map → formula-7UYJOVTB.js.map} +0 -0
  140. /package/dist/{model-bundle-FSPZZ2EM.js.map → model-bundle-2C5IYDLE.js.map} +0 -0
  141. /package/dist/{models-CJPAQK5R.js.map → models-SEQZHZFT.js.map} +0 -0
  142. /package/dist/{pdf-ocr-76SAUXMZ.js.map → pdf-ocr-UUBWBEHL.js.map} +0 -0
  143. /package/dist/{pdf-ocr-CPU55CVF.js.map → pdf-ocr-VJSF2SQC.js.map} +0 -0
  144. /package/dist/{profile-io-LBM263K7.js.map → profile-io-RUFZOZO6.js.map} +0 -0
  145. /package/dist/{rasterize-46RG4ZF6.js.map → rasterize-IOYXR5SI.js.map} +0 -0
  146. /package/dist/{redact-7ILEAUTS.js.map → redact-NMIKVHMP.js.map} +0 -0
  147. /package/dist/{render-NGP5T24S.js.map → render-YZ2TEC3N.js.map} +0 -0
  148. /package/dist/{seal-RGUPOLXD.js.map → seal-ECV44TUH.js.map} +0 -0
  149. /package/dist/{setup-AU4OCHNH.js.map → setup-MPIEP6XT.js.map} +0 -0
  150. /package/dist/{visual-NZLYIYEM.js.map → visual-ICMOWUCR.js.map} +0 -0
  151. /package/dist/{watch-MKPUCPXM.js.map → watch-RYO7MN5B.js.map} +0 -0
@@ -1 +1 @@
1
- {"version":3,"sources":["/Users/chris_gomdori/workspace/kordoc/dist/chunk-EQZI77PP.cjs","../src/pdf/formula/models.ts","../src/shared/offline.ts"],"names":[],"mappings":"AAAA;AACE;AACF,wDAA6B;AAC7B;AACA;ACMA,gCAA2B;AAC3B,wBAAiC;AACjC,uCAA4C;AAC5C;AACA,wBAAwB;AACxB,4BAA8B;AAC9B,4CAAyB;AACzB,gCAAyB;ADJzB;AACA;AEKA,IAAM,OAAA,kBAAS,IAAI,GAAA,CAAI,CAAC,GAAA,EAAK,MAAA,EAAQ,KAAA,EAAO,IAAI,CAAC,CAAA;AAEjD,SAAS,OAAA,CAAQ,IAAA,EAAuB;AACtC,EAAA,MAAM,EAAA,EAAI,OAAA,CAAQ,GAAA,CAAI,IAAI,CAAA;AAC1B,EAAA,OAAO,CAAC,CAAC,EAAA,GAAK,MAAA,CAAO,GAAA,CAAI,CAAA,CAAE,IAAA,CAAK,CAAA,CAAE,WAAA,CAAY,CAAC,CAAA;AACjD;AAGO,SAAS,aAAA,CAAA,EAAyB;AACvC,EAAA,OAAO,OAAA,CAAQ,gBAAgB,CAAA;AACjC;AAOO,SAAS,oBAAA,CAAqB,IAAA,EAAc,IAAA,EAAqB;AACtE,EAAA,GAAA,CAAI,CAAC,aAAA,CAAc,CAAA,EAAG,MAAA;AACtB,EAAA,MAAM,IAAI,kCAAA;AAAA,IACR,CAAA,gFAAA,EAAiC,IAAI,CAAA,EAAA;AACvC,EAAA;AACF;AFZ4F;AACA;ACDxD;AAC5B,EAAA;AACI,EAAA;AACL,EAAA;AACG,EAAA;AACA,EAAA;AACV;AAE4C;AACpC,EAAA;AACI,EAAA;AACL,EAAA;AACG,EAAA;AACA,EAAA;AACV;AAE4C;AACpC,EAAA;AACI,EAAA;AACL,EAAA;AACG,EAAA;AACA,EAAA;AACV;AAEwC;AAChC,EAAA;AACI,EAAA;AACL,EAAA;AACG,EAAA;AACA,EAAA;AACV;AAE4D;AAC1D,EAAA;AACA,EAAA;AACA,EAAA;AACA,EAAA;AACF;AAOqD;AACtB,EAAA;AACI,EAAA;AACH,IAAA;AAC9B,EAAA;AAC2D,EAAA;AAC7D;AAE8C;AACd,EAAA;AAChC;AAa6G;AAC5E,EAAA;AACL,EAAA;AACiB,IAAA;AAC5B,IAAA;AACT,IAAA;AAC4B,MAAA;AACE,MAAA;AAC1B,IAAA;AACG,MAAA;AACX,IAAA;AACa,IAAA;AACoD,MAAA;AAC/D,MAAA;AACF,IAAA;AACI,IAAA;AACyC,MAAA;AACf,MAAA;AACmC,QAAA;AACxD,MAAA;AACO,QAAA;AACV,UAAA;AACA,UAAA;AACQ,UAAA;AACE,UAAA;AAC4D,UAAA;AACvE,QAAA;AACH,MAAA;AACU,IAAA;AACE,MAAA;AACV,QAAA;AACA,QAAA;AACQ,QAAA;AACE,QAAA;AACgD,QAAA;AAC3D,MAAA;AACH,IAAA;AACF,EAAA;AACO,EAAA;AACT;AAGsE;AACH,EAAA;AACnE;AAoBgI;AAC1F,EAAA;AAEV,EAAA;AACiB,IAAA;AAEU,IAAA;AACpC,sBAAA;AACX,QAAA;AACY,QAAA;AACL,QAAA;AACA,QAAA;AACE,QAAA;AACV,MAAA;AACD,MAAA;AACF,IAAA;AAGI,IAAA;AACoB,MAAA;AAChB,IAAA;AAER,IAAA;AAEgD,IAAA;AAClD,EAAA;AACF;AAGuF;AACV,EAAA;AAC7E;AAGsG;AACpE,EAAA;AACI,EAAA;AACK,EAAA;AACU,EAAA;AACe,oBAAA;AAChE,IAAA;AACF,EAAA;AACI,EAAA;AACoB,IAAA;AAChB,EAAA;AAER,EAAA;AACgD,EAAA;AAClD;AAE4F;AACtF,EAAA;AAC4B,IAAA;AACU,IAAA;AAClC,EAAA;AACC,IAAA;AACT,EAAA;AACI,EAAA;AACyC,IAAA;AACzB,IAAA;AACZ,EAAA;AACC,IAAA;AACT,EAAA;AACF;AAMiB;AAEf,EAAA;AACc,IAAA;AACZ,IAAA;AACF,EAAA;AAG6B,EAAA;AACsB,EAAA;AAEhB,EAAA;AACxB,IAAA;AAAA;AAEO,MAAA;AAChB,IAAA;AACD,EAAA;AAC2B,EAAA;AAChB,IAAA;AAC0C,MAAA;AACpD,IAAA;AACF,EAAA;AAEmD,EAAA;AACQ,EAAA;AAC1C,EAAA;AAEoB,EAAA;AACjC,EAAA;AACyF,IAAA;AACzC,IAAA;AAC5B,MAAA;AACP,sBAAA;AACX,QAAA;AACA,QAAA;AACA,QAAA;AACO,QAAA;AACR,MAAA;AACF,IAAA;AACwB,IAAA;AACf,EAAA;AACN,IAAA;AACmB,MAAA;AACf,IAAA;AAER,IAAA;AAC+D,IAAA;AACjE,EAAA;AAEa,kBAAA;AACX,IAAA;AACA,IAAA;AACA,IAAA;AACO,IAAA;AACR,EAAA;AAGG,EAAA;AACA,EAAA;AACkC,IAAA;AAC1B,EAAA;AACN,IAAA;AACmB,MAAA;AACf,IAAA;AAER,IAAA;AACiE,IAAA;AACnE,EAAA;AAE4B,EAAA;AACtB,IAAA;AACmB,MAAA;AACf,IAAA;AAER,IAAA;AACU,IAAA;AAC4D,MAAA;AACtE,IAAA;AACF,EAAA;AAEgC,EAAA;AACnB,kBAAA;AACX,IAAA;AACA,IAAA;AACA,IAAA;AACO,IAAA;AACR,EAAA;AACH;AAEwD;AACzB,EAAA;AACI,EAAA;AACY,EAAA;AACZ,IAAA;AACf,MAAA;AAEhB,IAAA;AACD,EAAA;AACoB,EAAA;AACvB;AD1E4F;AACA;AACA;AACA;AACA;AACA;AACA;AACA;AACA;AACA;AACA;AACA;AACA;AACA","file":"/Users/chris_gomdori/workspace/kordoc/dist/chunk-EQZI77PP.cjs","sourcesContent":[null,"/**\n * 수식 OCR 모델 자동 다운로드 + SHA-256 검증\n *\n * 캐시 위치: `~/.cache/kordoc/models/pix2text/`\n * 총 ~155MB (MFD 44MB + MFR encoder 87MB + MFR decoder 30MB + tokenizer 40KB)\n *\n * HF URL + SHA 는 Docufinder archive/rust-formula-ocr-wip 에서 검증 완료됨.\n * 모델 버전 갱신 시 URL + SHA 를 함께 갱신해야 한다.\n */\n\nimport { createHash } from \"crypto\"\nimport { createReadStream } from \"fs\"\nimport { mkdir, stat, unlink, rename } from \"fs/promises\"\nimport { createWriteStream } from \"fs\"\nimport { homedir } from \"os\"\nimport { join, dirname } from \"path\"\nimport { pipeline } from \"stream/promises\"\nimport { Readable } from \"stream\"\nimport { assertNetworkAllowed } from \"../../shared/offline.js\"\n\nexport interface ModelSpec {\n name: string\n filename: string\n url: string\n sha256: string\n sizeMb: number\n}\n\n/** 모델 스펙 (SHA-256 은 실제 다운로드로 검증됨 — 변경 금지) */\nexport const MFD_MODEL: ModelSpec = {\n name: \"Pix2Text MFD\",\n filename: \"mfd.onnx\",\n url: \"https://huggingface.co/breezedeus/pix2text-mfd/resolve/main/mfd-v20240618.onnx\",\n sha256: \"51a8854743b17ae654729af8db82a630c1ccfa06debf4856c8b28055f87d02c1\",\n sizeMb: 42,\n}\n\nexport const MFR_ENCODER_MODEL: ModelSpec = {\n name: \"Pix2Text MFR encoder\",\n filename: \"encoder_model.onnx\",\n url: \"https://huggingface.co/breezedeus/pix2text-mfr/resolve/main/encoder_model.onnx\",\n sha256: \"bd8d5c322792e9ec45793af5569e9748f82a3d728a9e00213dbfc56c1486f37d\",\n sizeMb: 87,\n}\n\nexport const MFR_DECODER_MODEL: ModelSpec = {\n name: \"Pix2Text MFR decoder\",\n filename: \"decoder_model.onnx\",\n url: \"https://huggingface.co/breezedeus/pix2text-mfr/resolve/main/decoder_model.onnx\",\n sha256: \"fd0f92d7a012f3dae41e1ac79421aea0ea888b5a66cb3f9a004e424f82f3daed\",\n sizeMb: 30,\n}\n\nexport const MFR_TOKENIZER: ModelSpec = {\n name: \"Pix2Text MFR tokenizer\",\n filename: \"tokenizer.json\",\n url: \"https://huggingface.co/breezedeus/pix2text-mfr/resolve/main/tokenizer.json\",\n sha256: \"3e2ab757277d22639bec28c9d7972e352d3d1dba223051fa674002dc5ab64df3\",\n sizeMb: 1,\n}\n\nexport const ALL_FORMULA_MODELS: ReadonlyArray<ModelSpec> = [\n MFD_MODEL,\n MFR_ENCODER_MODEL,\n MFR_DECODER_MODEL,\n MFR_TOKENIZER,\n]\n\n/**\n * 모델 캐시 하위 디렉토리 경로. 환경 변수 `KORDOC_MODEL_CACHE` 가 설정되면\n * 해당 경로를 우선 사용, 없으면 `~/.cache/kordoc/models/<subdir>/`.\n * (수식 OCR 외 텍스트 OCR 등 다른 모델군도 공용 — subdir 로 격리)\n */\nexport function getModelsDir(subdir: string): string {\n const override = process.env.KORDOC_MODEL_CACHE\n if (override && override.trim()) {\n return join(override, subdir)\n }\n return join(homedir(), \".cache\", \"kordoc\", \"models\", subdir)\n}\n\nexport function getFormulaModelsDir(): string {\n return getModelsDir(\"pix2text\")\n}\n\nexport interface ModelStatus {\n spec: ModelSpec\n localPath: string\n exists: boolean\n /** SHA 검증까지 끝났는가 (exists && valid) */\n verified: boolean\n /** 검증 실패 시 이유 */\n invalidReason?: string\n}\n\n/** 지정 모델군의 존재/유효성 상태 반환 (다운로드 없이 확인만) */\nexport async function getModelStatusIn(dir: string, specs: ReadonlyArray<ModelSpec>): Promise<ModelStatus[]> {\n const result: ModelStatus[] = []\n for (const spec of specs) {\n const localPath = join(dir, spec.filename)\n let exists = false\n try {\n const s = await stat(localPath)\n exists = s.isFile() && s.size > 0\n } catch {\n exists = false\n }\n if (!exists) {\n result.push({ spec, localPath, exists: false, verified: false })\n continue\n }\n try {\n const actual = await sha256OfFile(localPath)\n if (actual === spec.sha256) {\n result.push({ spec, localPath, exists: true, verified: true })\n } else {\n result.push({\n spec,\n localPath,\n exists: true,\n verified: false,\n invalidReason: `SHA256 mismatch: expected ${spec.sha256}, got ${actual}`,\n })\n }\n } catch (e) {\n result.push({\n spec,\n localPath,\n exists: true,\n verified: false,\n invalidReason: `SHA compute failed: ${(e as Error).message}`,\n })\n }\n }\n return result\n}\n\n/** 모든 수식 모델의 존재/유효성 상태 반환 (다운로드 없이 확인만) */\nexport async function getFormulaModelStatus(): Promise<ModelStatus[]> {\n return getModelStatusIn(getFormulaModelsDir(), ALL_FORMULA_MODELS)\n}\n\nexport interface DownloadProgress {\n spec: ModelSpec\n /** 현재까지 다운로드한 바이트 */\n downloaded: number\n /** 전체 바이트 (알 수 있으면) */\n total: number | null\n /** \"download\" | \"verify\" | \"done\" | \"skip\" */\n phase: \"download\" | \"verify\" | \"done\" | \"skip\" | \"error\"\n message?: string\n}\n\nexport type ProgressHandler = (p: DownloadProgress) => void\n\n/**\n * 지정 모델군을 다운로드/검증한다.\n * - 이미 있고 SHA 일치 → skip\n * - 없음 → 다운로드 후 SHA 검증. 실패 시 파일 삭제.\n */\nexport async function ensureModelsIn(dir: string, specs: ReadonlyArray<ModelSpec>, onProgress?: ProgressHandler): Promise<void> {\n await mkdir(dir, { recursive: true })\n\n for (const spec of specs) {\n const localPath = join(dir, spec.filename)\n\n if (await isExistingValid(localPath, spec.sha256)) {\n onProgress?.({\n spec,\n downloaded: 0,\n total: null,\n phase: \"skip\",\n message: \"이미 존재 + SHA 일치\",\n })\n continue\n }\n\n // 기존 파일 있지만 SHA 불일치 → 삭제\n try {\n await unlink(localPath)\n } catch {\n // 없을 수 있음\n }\n\n await downloadToFile(spec, localPath, onProgress)\n }\n}\n\n/** 필요한 모든 수식 모델을 다운로드/검증한다. */\nexport async function ensureFormulaModels(onProgress?: ProgressHandler): Promise<void> {\n return ensureModelsIn(getFormulaModelsDir(), ALL_FORMULA_MODELS, onProgress)\n}\n\n/** 단일 모델만 확인/다운로드 (진단 UI 에서 개별 상태 새로고침용) */\nexport async function ensureSingleModel(spec: ModelSpec, onProgress?: ProgressHandler): Promise<void> {\n const dir = getFormulaModelsDir()\n await mkdir(dir, { recursive: true })\n const localPath = join(dir, spec.filename)\n if (await isExistingValid(localPath, spec.sha256)) {\n onProgress?.({ spec, downloaded: 0, total: null, phase: \"skip\" })\n return\n }\n try {\n await unlink(localPath)\n } catch {\n // 재다운로드 전 손상/잔존 파일 정리 best-effort — 애초에 없으면 실패해도 무방\n }\n await downloadToFile(spec, localPath, onProgress)\n}\n\nasync function isExistingValid(localPath: string, sha256Expected: string): Promise<boolean> {\n try {\n const s = await stat(localPath)\n if (!s.isFile() || s.size === 0) return false\n } catch {\n return false\n }\n try {\n const actual = await sha256OfFile(localPath)\n return actual === sha256Expected\n } catch {\n return false\n }\n}\n\nasync function downloadToFile(\n spec: ModelSpec,\n localPath: string,\n onProgress?: ProgressHandler,\n): Promise<void> {\n // 폐쇄망 모드에서는 요청 자체를 보내지 않는다 (사이드로드 안내로 대체)\n assertNetworkAllowed(\n `${spec.name} 모델 다운로드`,\n \"온라인 PC에서 `kordoc models --export <디렉토리>` 로 내보낸 뒤 이 PC에서 `kordoc models --import <디렉토리>` 하거나, KORDOC_MODEL_CACHE 로 모델 캐시 경로를 지정하세요\",\n )\n\n // 먼저 .part 로 받고 검증 후 rename — 중단된 다운로드가 \"정상 파일\"로 오인되는 걸 방지\n const partPath = `${localPath}.part`\n await mkdir(dirname(localPath), { recursive: true })\n\n const resp = await fetch(spec.url, {\n headers: {\n // HF CDN 은 UA 없으면 가끔 403 을 뱉는다\n \"User-Agent\": \"kordoc-formula-ocr/1.0 (+https://github.com/chrisryugj/kordoc)\",\n },\n })\n if (!resp.ok || !resp.body) {\n throw new Error(\n `${spec.name} 다운로드 실패: HTTP ${resp.status} ${resp.statusText} (${spec.url})`,\n )\n }\n\n const lenHeader = resp.headers.get(\"content-length\")\n const total = lenHeader ? Number.parseInt(lenHeader, 10) : null\n let downloaded = 0\n\n const ws = createWriteStream(partPath)\n try {\n const reader = Readable.fromWeb(resp.body as unknown as import(\"stream/web\").ReadableStream)\n reader.on(\"data\", (chunk: Buffer | Uint8Array) => {\n downloaded += chunk.length\n onProgress?.({\n spec,\n downloaded,\n total,\n phase: \"download\",\n })\n })\n await pipeline(reader, ws)\n } catch (e) {\n try {\n await unlink(partPath)\n } catch {\n // .part 임시파일 정리 best-effort — 본 에러(스트리밍 실패)를 가리지 않도록 무시\n }\n throw new Error(`${spec.name} 스트리밍 실패: ${(e as Error).message}`)\n }\n\n onProgress?.({\n spec,\n downloaded,\n total,\n phase: \"verify\",\n })\n\n // SHA 검증\n let actual: string\n try {\n actual = await sha256OfFile(partPath)\n } catch (e) {\n try {\n await unlink(partPath)\n } catch {\n // .part 임시파일 정리 best-effort — 본 에러(SHA 계산 실패)를 가리지 않도록 무시\n }\n throw new Error(`${spec.name} SHA 계산 실패: ${(e as Error).message}`)\n }\n\n if (actual !== spec.sha256) {\n try {\n await unlink(partPath)\n } catch {\n // 오염된 .part 정리 best-effort — 본 에러(SHA mismatch)를 가리지 않도록 무시\n }\n throw new Error(\n `${spec.name} SHA256 mismatch: expected ${spec.sha256}, got ${actual} — 모델 URL 이 오염되었거나 전송 중 손상되었습니다.`,\n )\n }\n\n await rename(partPath, localPath)\n onProgress?.({\n spec,\n downloaded,\n total,\n phase: \"done\",\n })\n}\n\nasync function sha256OfFile(p: string): Promise<string> {\n const h = createHash(\"sha256\")\n const stream = createReadStream(p)\n await pipeline(stream, async function* (src) {\n for await (const chunk of src) {\n h.update(chunk)\n // yield 안 함 (소비만)\n }\n })\n return h.digest(\"hex\")\n}\n","/**\n * 폐쇄망(내부망) 배포용 아웃바운드 통신 킬스위치 + 파일 접근 루트 제한.\n *\n * kordoc 자체는 문서 파서라 상시 외부 통신이 없지만, 두 경로만 나간다:\n * 1. OCR/수식 모델 최초 1회 다운로드 (huggingface.co)\n * 2. `kordoc watch --webhook` 사용자가 명시한 URL\n *\n * `KORDOC_OFFLINE=1` 이면 위 경로를 **시도 전에** 차단한다. 보안성 검토에서\n * \"외부 통신 없음\"을 런타임으로 증명하기 위한 단일 관문 — 새 아웃바운드 경로를\n * 추가하면 반드시 `assertNetworkAllowed()` 를 먼저 통과시킬 것.\n *\n * `KORDOC_ROOT` 는 파일 접근을 특정 디렉토리 하위로 제한한다(미설정 시 제한 없음 —\n * 기존 동작 보존). 내부망 배포에서는 두 변수를 함께 고정하는 것을 권장.\n */\n\nimport { realpathSync } from \"fs\"\nimport { isAbsolute, relative, resolve, sep } from \"path\"\nimport { KordocError } from \"../utils.js\"\n\nconst TRUTHY = new Set([\"1\", \"true\", \"yes\", \"on\"])\n\nfunction envFlag(name: string): boolean {\n const v = process.env[name]\n return !!v && TRUTHY.has(v.trim().toLowerCase())\n}\n\n/** 폐쇄망 모드 여부 (`KORDOC_OFFLINE=1|true|yes|on`) */\nexport function isOfflineMode(): boolean {\n return envFlag(\"KORDOC_OFFLINE\")\n}\n\n/**\n * 아웃바운드 통신 관문. 폐쇄망 모드면 요청을 보내기 전에 예외로 끊는다.\n * @param what 차단 대상 설명 (사용자에게 노출됨 — 호스트/URL 은 넣지 말 것)\n * @param hint 대안 안내 (오프라인 사이드로드 방법 등)\n */\nexport function assertNetworkAllowed(what: string, hint?: string): void {\n if (!isOfflineMode()) return\n throw new KordocError(\n `폐쇄망 모드(KORDOC_OFFLINE)에서 차단됨: ${what}` + (hint ? ` — ${hint}` : \"\"),\n )\n}\n\n/** 파일 접근 루트 (`KORDOC_ROOT`). 미설정이면 null = 제한 없음 */\nexport function getAccessRoot(): string | null {\n const raw = process.env.KORDOC_ROOT\n if (!raw || !raw.trim()) return null\n const resolved = resolve(raw.trim())\n try {\n return realpathSync(resolved)\n } catch {\n // 루트가 실재하지 않으면 모든 접근이 막히는 게 안전한 기본값\n return resolved\n }\n}\n\n/** 경로가 루트 하위인지 (루트 자신 포함). 경계 문자열 매칭이 아닌 세그먼트 단위 비교 */\nexport function isWithinRoot(realPath: string, root: string): boolean {\n if (realPath === root) return true\n const rel = relative(root, realPath)\n return rel !== \"\" && !rel.startsWith(\"..\") && !isAbsolute(rel) && !rel.startsWith(`..${sep}`)\n}\n\n/**\n * `KORDOC_ROOT` 가 설정된 경우 경로가 루트 하위인지 검증한다.\n * 인자는 이미 realpath 로 해석된 절대경로여야 한다 (심볼릭 링크 탈출 방지).\n */\nexport function assertWithinRoot(realPath: string): void {\n const root = getAccessRoot()\n if (!root) return\n if (!isWithinRoot(realPath, root)) {\n throw new KordocError(`허용된 작업 디렉토리(KORDOC_ROOT) 밖의 경로입니다: ${realPath}`)\n }\n}\n"]}
1
+ {"version":3,"sources":["/Users/chris_gomdori/workspace/kordoc/dist/chunk-QEJUY6SK.cjs","../src/pdf/formula/models.ts","../src/shared/offline.ts"],"names":[],"mappings":"AAAA;AACE;AACF,wDAA6B;AAC7B;AACA;ACMA,gCAA2B;AAC3B,wBAAiC;AACjC,uCAA4C;AAC5C;AACA,wBAAwB;AACxB,4BAA8B;AAC9B,4CAAyB;AACzB,gCAAyB;ADJzB;AACA;AEKA,IAAM,OAAA,kBAAS,IAAI,GAAA,CAAI,CAAC,GAAA,EAAK,MAAA,EAAQ,KAAA,EAAO,IAAI,CAAC,CAAA;AAEjD,SAAS,OAAA,CAAQ,IAAA,EAAuB;AACtC,EAAA,MAAM,EAAA,EAAI,OAAA,CAAQ,GAAA,CAAI,IAAI,CAAA;AAC1B,EAAA,OAAO,CAAC,CAAC,EAAA,GAAK,MAAA,CAAO,GAAA,CAAI,CAAA,CAAE,IAAA,CAAK,CAAA,CAAE,WAAA,CAAY,CAAC,CAAA;AACjD;AAGO,SAAS,aAAA,CAAA,EAAyB;AACvC,EAAA,OAAO,OAAA,CAAQ,gBAAgB,CAAA;AACjC;AAOO,SAAS,oBAAA,CAAqB,IAAA,EAAc,IAAA,EAAqB;AACtE,EAAA,GAAA,CAAI,CAAC,aAAA,CAAc,CAAA,EAAG,MAAA;AACtB,EAAA,MAAM,IAAI,kCAAA;AAAA,IACR,CAAA,gFAAA,EAAiC,IAAI,CAAA,EAAA;AACvC,EAAA;AACF;AFZ4F;AACA;ACDxD;AAC5B,EAAA;AACI,EAAA;AACL,EAAA;AACG,EAAA;AACA,EAAA;AACV;AAE4C;AACpC,EAAA;AACI,EAAA;AACL,EAAA;AACG,EAAA;AACA,EAAA;AACV;AAE4C;AACpC,EAAA;AACI,EAAA;AACL,EAAA;AACG,EAAA;AACA,EAAA;AACV;AAEwC;AAChC,EAAA;AACI,EAAA;AACL,EAAA;AACG,EAAA;AACA,EAAA;AACV;AAE4D;AAC1D,EAAA;AACA,EAAA;AACA,EAAA;AACA,EAAA;AACF;AAOqD;AACtB,EAAA;AACI,EAAA;AACH,IAAA;AAC9B,EAAA;AAC2D,EAAA;AAC7D;AAE8C;AACd,EAAA;AAChC;AAa6G;AAC5E,EAAA;AACL,EAAA;AACiB,IAAA;AAC5B,IAAA;AACT,IAAA;AAC4B,MAAA;AACE,MAAA;AAC1B,IAAA;AACG,MAAA;AACX,IAAA;AACa,IAAA;AACoD,MAAA;AAC/D,MAAA;AACF,IAAA;AACI,IAAA;AACyC,MAAA;AACf,MAAA;AACmC,QAAA;AACxD,MAAA;AACO,QAAA;AACV,UAAA;AACA,UAAA;AACQ,UAAA;AACE,UAAA;AAC4D,UAAA;AACvE,QAAA;AACH,MAAA;AACU,IAAA;AACE,MAAA;AACV,QAAA;AACA,QAAA;AACQ,QAAA;AACE,QAAA;AACgD,QAAA;AAC3D,MAAA;AACH,IAAA;AACF,EAAA;AACO,EAAA;AACT;AAGsE;AACH,EAAA;AACnE;AAoBgI;AAC1F,EAAA;AAEV,EAAA;AACiB,IAAA;AAEU,IAAA;AACpC,sBAAA;AACX,QAAA;AACY,QAAA;AACL,QAAA;AACA,QAAA;AACE,QAAA;AACV,MAAA;AACD,MAAA;AACF,IAAA;AAGI,IAAA;AACoB,MAAA;AAChB,IAAA;AAER,IAAA;AAEgD,IAAA;AAClD,EAAA;AACF;AAGuF;AACV,EAAA;AAC7E;AAGsG;AACpE,EAAA;AACI,EAAA;AACK,EAAA;AACU,EAAA;AACe,oBAAA;AAChE,IAAA;AACF,EAAA;AACI,EAAA;AACoB,IAAA;AAChB,EAAA;AAER,EAAA;AACgD,EAAA;AAClD;AAE4F;AACtF,EAAA;AAC4B,IAAA;AACU,IAAA;AAClC,EAAA;AACC,IAAA;AACT,EAAA;AACI,EAAA;AACyC,IAAA;AACzB,IAAA;AACZ,EAAA;AACC,IAAA;AACT,EAAA;AACF;AAMiB;AAEf,EAAA;AACc,IAAA;AACZ,IAAA;AACF,EAAA;AAG6B,EAAA;AACsB,EAAA;AAEhB,EAAA;AACxB,IAAA;AAAA;AAEO,MAAA;AAChB,IAAA;AACD,EAAA;AAC2B,EAAA;AAChB,IAAA;AAC0C,MAAA;AACpD,IAAA;AACF,EAAA;AAEmD,EAAA;AACQ,EAAA;AAC1C,EAAA;AAEoB,EAAA;AACjC,EAAA;AACyF,IAAA;AACzC,IAAA;AAC5B,MAAA;AACP,sBAAA;AACX,QAAA;AACA,QAAA;AACA,QAAA;AACO,QAAA;AACR,MAAA;AACF,IAAA;AACwB,IAAA;AACf,EAAA;AACN,IAAA;AACmB,MAAA;AACf,IAAA;AAER,IAAA;AAC+D,IAAA;AACjE,EAAA;AAEa,kBAAA;AACX,IAAA;AACA,IAAA;AACA,IAAA;AACO,IAAA;AACR,EAAA;AAGG,EAAA;AACA,EAAA;AACkC,IAAA;AAC1B,EAAA;AACN,IAAA;AACmB,MAAA;AACf,IAAA;AAER,IAAA;AACiE,IAAA;AACnE,EAAA;AAE4B,EAAA;AACtB,IAAA;AACmB,MAAA;AACf,IAAA;AAER,IAAA;AACU,IAAA;AAC4D,MAAA;AACtE,IAAA;AACF,EAAA;AAEgC,EAAA;AACnB,kBAAA;AACX,IAAA;AACA,IAAA;AACA,IAAA;AACO,IAAA;AACR,EAAA;AACH;AAEwD;AACzB,EAAA;AACI,EAAA;AACY,EAAA;AACZ,IAAA;AACf,MAAA;AAEhB,IAAA;AACD,EAAA;AACoB,EAAA;AACvB;AD1E4F;AACA;AACA;AACA;AACA;AACA;AACA;AACA;AACA;AACA;AACA;AACA;AACA;AACA","file":"/Users/chris_gomdori/workspace/kordoc/dist/chunk-QEJUY6SK.cjs","sourcesContent":[null,"/**\n * 수식 OCR 모델 자동 다운로드 + SHA-256 검증\n *\n * 캐시 위치: `~/.cache/kordoc/models/pix2text/`\n * 총 ~155MB (MFD 44MB + MFR encoder 87MB + MFR decoder 30MB + tokenizer 40KB)\n *\n * HF URL + SHA 는 Docufinder archive/rust-formula-ocr-wip 에서 검증 완료됨.\n * 모델 버전 갱신 시 URL + SHA 를 함께 갱신해야 한다.\n */\n\nimport { createHash } from \"crypto\"\nimport { createReadStream } from \"fs\"\nimport { mkdir, stat, unlink, rename } from \"fs/promises\"\nimport { createWriteStream } from \"fs\"\nimport { homedir } from \"os\"\nimport { join, dirname } from \"path\"\nimport { pipeline } from \"stream/promises\"\nimport { Readable } from \"stream\"\nimport { assertNetworkAllowed } from \"../../shared/offline.js\"\n\nexport interface ModelSpec {\n name: string\n filename: string\n url: string\n sha256: string\n sizeMb: number\n}\n\n/** 모델 스펙 (SHA-256 은 실제 다운로드로 검증됨 — 변경 금지) */\nexport const MFD_MODEL: ModelSpec = {\n name: \"Pix2Text MFD\",\n filename: \"mfd.onnx\",\n url: \"https://huggingface.co/breezedeus/pix2text-mfd/resolve/main/mfd-v20240618.onnx\",\n sha256: \"51a8854743b17ae654729af8db82a630c1ccfa06debf4856c8b28055f87d02c1\",\n sizeMb: 42,\n}\n\nexport const MFR_ENCODER_MODEL: ModelSpec = {\n name: \"Pix2Text MFR encoder\",\n filename: \"encoder_model.onnx\",\n url: \"https://huggingface.co/breezedeus/pix2text-mfr/resolve/main/encoder_model.onnx\",\n sha256: \"bd8d5c322792e9ec45793af5569e9748f82a3d728a9e00213dbfc56c1486f37d\",\n sizeMb: 87,\n}\n\nexport const MFR_DECODER_MODEL: ModelSpec = {\n name: \"Pix2Text MFR decoder\",\n filename: \"decoder_model.onnx\",\n url: \"https://huggingface.co/breezedeus/pix2text-mfr/resolve/main/decoder_model.onnx\",\n sha256: \"fd0f92d7a012f3dae41e1ac79421aea0ea888b5a66cb3f9a004e424f82f3daed\",\n sizeMb: 30,\n}\n\nexport const MFR_TOKENIZER: ModelSpec = {\n name: \"Pix2Text MFR tokenizer\",\n filename: \"tokenizer.json\",\n url: \"https://huggingface.co/breezedeus/pix2text-mfr/resolve/main/tokenizer.json\",\n sha256: \"3e2ab757277d22639bec28c9d7972e352d3d1dba223051fa674002dc5ab64df3\",\n sizeMb: 1,\n}\n\nexport const ALL_FORMULA_MODELS: ReadonlyArray<ModelSpec> = [\n MFD_MODEL,\n MFR_ENCODER_MODEL,\n MFR_DECODER_MODEL,\n MFR_TOKENIZER,\n]\n\n/**\n * 모델 캐시 하위 디렉토리 경로. 환경 변수 `KORDOC_MODEL_CACHE` 가 설정되면\n * 해당 경로를 우선 사용, 없으면 `~/.cache/kordoc/models/<subdir>/`.\n * (수식 OCR 외 텍스트 OCR 등 다른 모델군도 공용 — subdir 로 격리)\n */\nexport function getModelsDir(subdir: string): string {\n const override = process.env.KORDOC_MODEL_CACHE\n if (override && override.trim()) {\n return join(override, subdir)\n }\n return join(homedir(), \".cache\", \"kordoc\", \"models\", subdir)\n}\n\nexport function getFormulaModelsDir(): string {\n return getModelsDir(\"pix2text\")\n}\n\nexport interface ModelStatus {\n spec: ModelSpec\n localPath: string\n exists: boolean\n /** SHA 검증까지 끝났는가 (exists && valid) */\n verified: boolean\n /** 검증 실패 시 이유 */\n invalidReason?: string\n}\n\n/** 지정 모델군의 존재/유효성 상태 반환 (다운로드 없이 확인만) */\nexport async function getModelStatusIn(dir: string, specs: ReadonlyArray<ModelSpec>): Promise<ModelStatus[]> {\n const result: ModelStatus[] = []\n for (const spec of specs) {\n const localPath = join(dir, spec.filename)\n let exists = false\n try {\n const s = await stat(localPath)\n exists = s.isFile() && s.size > 0\n } catch {\n exists = false\n }\n if (!exists) {\n result.push({ spec, localPath, exists: false, verified: false })\n continue\n }\n try {\n const actual = await sha256OfFile(localPath)\n if (actual === spec.sha256) {\n result.push({ spec, localPath, exists: true, verified: true })\n } else {\n result.push({\n spec,\n localPath,\n exists: true,\n verified: false,\n invalidReason: `SHA256 mismatch: expected ${spec.sha256}, got ${actual}`,\n })\n }\n } catch (e) {\n result.push({\n spec,\n localPath,\n exists: true,\n verified: false,\n invalidReason: `SHA compute failed: ${(e as Error).message}`,\n })\n }\n }\n return result\n}\n\n/** 모든 수식 모델의 존재/유효성 상태 반환 (다운로드 없이 확인만) */\nexport async function getFormulaModelStatus(): Promise<ModelStatus[]> {\n return getModelStatusIn(getFormulaModelsDir(), ALL_FORMULA_MODELS)\n}\n\nexport interface DownloadProgress {\n spec: ModelSpec\n /** 현재까지 다운로드한 바이트 */\n downloaded: number\n /** 전체 바이트 (알 수 있으면) */\n total: number | null\n /** \"download\" | \"verify\" | \"done\" | \"skip\" */\n phase: \"download\" | \"verify\" | \"done\" | \"skip\" | \"error\"\n message?: string\n}\n\nexport type ProgressHandler = (p: DownloadProgress) => void\n\n/**\n * 지정 모델군을 다운로드/검증한다.\n * - 이미 있고 SHA 일치 → skip\n * - 없음 → 다운로드 후 SHA 검증. 실패 시 파일 삭제.\n */\nexport async function ensureModelsIn(dir: string, specs: ReadonlyArray<ModelSpec>, onProgress?: ProgressHandler): Promise<void> {\n await mkdir(dir, { recursive: true })\n\n for (const spec of specs) {\n const localPath = join(dir, spec.filename)\n\n if (await isExistingValid(localPath, spec.sha256)) {\n onProgress?.({\n spec,\n downloaded: 0,\n total: null,\n phase: \"skip\",\n message: \"이미 존재 + SHA 일치\",\n })\n continue\n }\n\n // 기존 파일 있지만 SHA 불일치 → 삭제\n try {\n await unlink(localPath)\n } catch {\n // 없을 수 있음\n }\n\n await downloadToFile(spec, localPath, onProgress)\n }\n}\n\n/** 필요한 모든 수식 모델을 다운로드/검증한다. */\nexport async function ensureFormulaModels(onProgress?: ProgressHandler): Promise<void> {\n return ensureModelsIn(getFormulaModelsDir(), ALL_FORMULA_MODELS, onProgress)\n}\n\n/** 단일 모델만 확인/다운로드 (진단 UI 에서 개별 상태 새로고침용) */\nexport async function ensureSingleModel(spec: ModelSpec, onProgress?: ProgressHandler): Promise<void> {\n const dir = getFormulaModelsDir()\n await mkdir(dir, { recursive: true })\n const localPath = join(dir, spec.filename)\n if (await isExistingValid(localPath, spec.sha256)) {\n onProgress?.({ spec, downloaded: 0, total: null, phase: \"skip\" })\n return\n }\n try {\n await unlink(localPath)\n } catch {\n // 재다운로드 전 손상/잔존 파일 정리 best-effort — 애초에 없으면 실패해도 무방\n }\n await downloadToFile(spec, localPath, onProgress)\n}\n\nasync function isExistingValid(localPath: string, sha256Expected: string): Promise<boolean> {\n try {\n const s = await stat(localPath)\n if (!s.isFile() || s.size === 0) return false\n } catch {\n return false\n }\n try {\n const actual = await sha256OfFile(localPath)\n return actual === sha256Expected\n } catch {\n return false\n }\n}\n\nasync function downloadToFile(\n spec: ModelSpec,\n localPath: string,\n onProgress?: ProgressHandler,\n): Promise<void> {\n // 폐쇄망 모드에서는 요청 자체를 보내지 않는다 (사이드로드 안내로 대체)\n assertNetworkAllowed(\n `${spec.name} 모델 다운로드`,\n \"온라인 PC에서 `kordoc models --export <디렉토리>` 로 내보낸 뒤 이 PC에서 `kordoc models --import <디렉토리>` 하거나, KORDOC_MODEL_CACHE 로 모델 캐시 경로를 지정하세요\",\n )\n\n // 먼저 .part 로 받고 검증 후 rename — 중단된 다운로드가 \"정상 파일\"로 오인되는 걸 방지\n const partPath = `${localPath}.part`\n await mkdir(dirname(localPath), { recursive: true })\n\n const resp = await fetch(spec.url, {\n headers: {\n // HF CDN 은 UA 없으면 가끔 403 을 뱉는다\n \"User-Agent\": \"kordoc-formula-ocr/1.0 (+https://github.com/chrisryugj/kordoc)\",\n },\n })\n if (!resp.ok || !resp.body) {\n throw new Error(\n `${spec.name} 다운로드 실패: HTTP ${resp.status} ${resp.statusText} (${spec.url})`,\n )\n }\n\n const lenHeader = resp.headers.get(\"content-length\")\n const total = lenHeader ? Number.parseInt(lenHeader, 10) : null\n let downloaded = 0\n\n const ws = createWriteStream(partPath)\n try {\n const reader = Readable.fromWeb(resp.body as unknown as import(\"stream/web\").ReadableStream)\n reader.on(\"data\", (chunk: Buffer | Uint8Array) => {\n downloaded += chunk.length\n onProgress?.({\n spec,\n downloaded,\n total,\n phase: \"download\",\n })\n })\n await pipeline(reader, ws)\n } catch (e) {\n try {\n await unlink(partPath)\n } catch {\n // .part 임시파일 정리 best-effort — 본 에러(스트리밍 실패)를 가리지 않도록 무시\n }\n throw new Error(`${spec.name} 스트리밍 실패: ${(e as Error).message}`)\n }\n\n onProgress?.({\n spec,\n downloaded,\n total,\n phase: \"verify\",\n })\n\n // SHA 검증\n let actual: string\n try {\n actual = await sha256OfFile(partPath)\n } catch (e) {\n try {\n await unlink(partPath)\n } catch {\n // .part 임시파일 정리 best-effort — 본 에러(SHA 계산 실패)를 가리지 않도록 무시\n }\n throw new Error(`${spec.name} SHA 계산 실패: ${(e as Error).message}`)\n }\n\n if (actual !== spec.sha256) {\n try {\n await unlink(partPath)\n } catch {\n // 오염된 .part 정리 best-effort — 본 에러(SHA mismatch)를 가리지 않도록 무시\n }\n throw new Error(\n `${spec.name} SHA256 mismatch: expected ${spec.sha256}, got ${actual} — 모델 URL 이 오염되었거나 전송 중 손상되었습니다.`,\n )\n }\n\n await rename(partPath, localPath)\n onProgress?.({\n spec,\n downloaded,\n total,\n phase: \"done\",\n })\n}\n\nasync function sha256OfFile(p: string): Promise<string> {\n const h = createHash(\"sha256\")\n const stream = createReadStream(p)\n await pipeline(stream, async function* (src) {\n for await (const chunk of src) {\n h.update(chunk)\n // yield 안 함 (소비만)\n }\n })\n return h.digest(\"hex\")\n}\n","/**\n * 폐쇄망(내부망) 배포용 아웃바운드 통신 킬스위치 + 파일 접근 루트 제한.\n *\n * kordoc 자체는 문서 파서라 상시 외부 통신이 없지만, 두 경로만 나간다:\n * 1. OCR/수식 모델 최초 1회 다운로드 (huggingface.co)\n * 2. `kordoc watch --webhook` 사용자가 명시한 URL\n *\n * `KORDOC_OFFLINE=1` 이면 위 경로를 **시도 전에** 차단한다. 보안성 검토에서\n * \"외부 통신 없음\"을 런타임으로 증명하기 위한 단일 관문 — 새 아웃바운드 경로를\n * 추가하면 반드시 `assertNetworkAllowed()` 를 먼저 통과시킬 것.\n *\n * `KORDOC_ROOT` 는 파일 접근을 특정 디렉토리 하위로 제한한다(미설정 시 제한 없음 —\n * 기존 동작 보존). 내부망 배포에서는 두 변수를 함께 고정하는 것을 권장.\n */\n\nimport { realpathSync } from \"fs\"\nimport { isAbsolute, relative, resolve, sep } from \"path\"\nimport { KordocError } from \"../utils.js\"\n\nconst TRUTHY = new Set([\"1\", \"true\", \"yes\", \"on\"])\n\nfunction envFlag(name: string): boolean {\n const v = process.env[name]\n return !!v && TRUTHY.has(v.trim().toLowerCase())\n}\n\n/** 폐쇄망 모드 여부 (`KORDOC_OFFLINE=1|true|yes|on`) */\nexport function isOfflineMode(): boolean {\n return envFlag(\"KORDOC_OFFLINE\")\n}\n\n/**\n * 아웃바운드 통신 관문. 폐쇄망 모드면 요청을 보내기 전에 예외로 끊는다.\n * @param what 차단 대상 설명 (사용자에게 노출됨 — 호스트/URL 은 넣지 말 것)\n * @param hint 대안 안내 (오프라인 사이드로드 방법 등)\n */\nexport function assertNetworkAllowed(what: string, hint?: string): void {\n if (!isOfflineMode()) return\n throw new KordocError(\n `폐쇄망 모드(KORDOC_OFFLINE)에서 차단됨: ${what}` + (hint ? ` — ${hint}` : \"\"),\n )\n}\n\n/** 파일 접근 루트 (`KORDOC_ROOT`). 미설정이면 null = 제한 없음 */\nexport function getAccessRoot(): string | null {\n const raw = process.env.KORDOC_ROOT\n if (!raw || !raw.trim()) return null\n const resolved = resolve(raw.trim())\n try {\n return realpathSync(resolved)\n } catch {\n // 루트가 실재하지 않으면 모든 접근이 막히는 게 안전한 기본값\n return resolved\n }\n}\n\n/** 경로가 루트 하위인지 (루트 자신 포함). 경계 문자열 매칭이 아닌 세그먼트 단위 비교 */\nexport function isWithinRoot(realPath: string, root: string): boolean {\n if (realPath === root) return true\n const rel = relative(root, realPath)\n return rel !== \"\" && !rel.startsWith(\"..\") && !isAbsolute(rel) && !rel.startsWith(`..${sep}`)\n}\n\n/**\n * `KORDOC_ROOT` 가 설정된 경우 경로가 루트 하위인지 검증한다.\n * 인자는 이미 realpath 로 해석된 절대경로여야 한다 (심볼릭 링크 탈출 방지).\n */\nexport function assertWithinRoot(realPath: string): void {\n const root = getAccessRoot()\n if (!root) return\n if (!isWithinRoot(realPath, root)) {\n throw new KordocError(`허용된 작업 디렉토리(KORDOC_ROOT) 밖의 경로입니다: ${realPath}`)\n }\n}\n"]}
@@ -1,6 +1,237 @@
1
1
  import {
2
2
  sanitizeHref
3
- } from "./chunk-JQEFP45L.js";
3
+ } from "./chunk-GUJ4Q73I.js";
4
+
5
+ // src/shared/symbol-fonts.ts
6
+ var WINGDINGS = [
7
+ "\u{1F589}",
8
+ "\u2702",
9
+ "\u2701",
10
+ "\u{1F453}",
11
+ "\u{1F56D}",
12
+ "\u{1F56E}",
13
+ "\u{1F56F}",
14
+ "\u{1F57F}",
15
+ "\u2706",
16
+ "\u{1F582}",
17
+ "\u{1F583}",
18
+ "\u{1F4EA}",
19
+ "\u{1F4EB}",
20
+ "\u{1F4EC}",
21
+ "\u{1F4ED}",
22
+ "\u{1F4C1}",
23
+ "\u{1F4C2}",
24
+ "\u{1F4C4}",
25
+ "\u{1F5CF}",
26
+ "\u{1F5D0}",
27
+ "\u{1F5C4}",
28
+ "\u231B",
29
+ "\u{1F5AE}",
30
+ "\u{1F5B0}",
31
+ "\u{1F5B2}",
32
+ "\u{1F5B3}",
33
+ "\u{1F5B4}",
34
+ "\u{1F5AB}",
35
+ "\u{1F5AC}",
36
+ "\u2707",
37
+ "\u270D",
38
+ "\u{1F58E}",
39
+ "\u270C",
40
+ "\u{1F44C}",
41
+ "\u{1F44D}",
42
+ "\u{1F44E}",
43
+ "\u261C",
44
+ "\u261E",
45
+ "\u261D",
46
+ "\u261F",
47
+ "\u{1F590}",
48
+ "\u263A",
49
+ "\u{1F610}",
50
+ "\u2639",
51
+ "\u{1F4A3}",
52
+ "\u2620",
53
+ "\u{1F3F3}",
54
+ "\u{1F3F1}",
55
+ "\u2708",
56
+ "\u263C",
57
+ "\u{1F4A7}",
58
+ "\u2744",
59
+ "\u{1F546}",
60
+ "\u271E",
61
+ "\u{1F548}",
62
+ "\u2720",
63
+ "\u2721",
64
+ "\u262A",
65
+ "\u262F",
66
+ "\u0950",
67
+ "\u2638",
68
+ "\u2648",
69
+ "\u2649",
70
+ "\u264A",
71
+ "\u264B",
72
+ "\u264C",
73
+ "\u264D",
74
+ "\u264E",
75
+ "\u264F",
76
+ "\u2650",
77
+ "\u2651",
78
+ "\u2652",
79
+ "\u2653",
80
+ "\u{1F670}",
81
+ "\u{1F675}",
82
+ "\u25CF",
83
+ "\u25CB",
84
+ "\u25A0",
85
+ "\u25A1",
86
+ "\u25A1",
87
+ "\u2751",
88
+ "\u2752",
89
+ "\u2B27",
90
+ "\u29EB",
91
+ "\u25C6",
92
+ "\u2756",
93
+ "\u2B25",
94
+ "\u2327",
95
+ "\u2BB9",
96
+ "\u2318",
97
+ "\u{1F3F5}",
98
+ "\u{1F3F6}",
99
+ "\u{1F676}",
100
+ "\u{1F677}",
101
+ "",
102
+ "\u24EA",
103
+ "\u2460",
104
+ "\u2461",
105
+ "\u2462",
106
+ "\u2463",
107
+ "\u2464",
108
+ "\u2465",
109
+ "\u2466",
110
+ "\u2467",
111
+ "\u2468",
112
+ "\u2469",
113
+ "\u24FF",
114
+ "\u2776",
115
+ "\u2777",
116
+ "\u2778",
117
+ "\u2779",
118
+ "\u277A",
119
+ "\u277B",
120
+ "\u277C",
121
+ "\u277D",
122
+ "\u277E",
123
+ "\u277F",
124
+ "\u{1F662}",
125
+ "\u{1F660}",
126
+ "\u{1F661}",
127
+ "\u{1F663}",
128
+ "\u{1F65E}",
129
+ "\u{1F65C}",
130
+ "\u{1F65D}",
131
+ "\u{1F65F}",
132
+ "\xB7",
133
+ "\u2022",
134
+ "\u25AA",
135
+ "\u26AA",
136
+ "\u25CB",
137
+ "\u25EF",
138
+ "\u25C9",
139
+ "\u25CE",
140
+ "\u{1F53F}",
141
+ "\u25AA",
142
+ "\u25FB",
143
+ "\u{1F7C2}",
144
+ "\u2726",
145
+ "\u2605",
146
+ "\u2736",
147
+ "\u2734",
148
+ "\u2739",
149
+ "\u2735",
150
+ "\u2BD0",
151
+ "\u2316",
152
+ "\u27E1",
153
+ "\u2311",
154
+ "\u2BD1",
155
+ "\u272A",
156
+ "\u2730",
157
+ "\u{1F550}",
158
+ "\u{1F551}",
159
+ "\u{1F552}",
160
+ "\u{1F553}",
161
+ "\u{1F554}",
162
+ "\u{1F555}",
163
+ "\u{1F556}",
164
+ "\u{1F557}",
165
+ "\u{1F558}",
166
+ "\u{1F559}",
167
+ "\u{1F55A}",
168
+ "\u{1F55B}",
169
+ "\u2BB0",
170
+ "\u2BB1",
171
+ "\u2BB2",
172
+ "\u2BB3",
173
+ "\u2BB4",
174
+ "\u2BB5",
175
+ "\u2BB6",
176
+ "\u2BB7",
177
+ "\u{1F66A}",
178
+ "\u{1F66B}",
179
+ "\u{1F655}",
180
+ "\u{1F654}",
181
+ "\u{1F657}",
182
+ "\u{1F656}",
183
+ "\u{1F650}",
184
+ "\u{1F651}",
185
+ "\u{1F652}",
186
+ "\u{1F653}",
187
+ "\u232B",
188
+ "\u2326",
189
+ "\u2B98",
190
+ "\u2B9A",
191
+ "\u2B99",
192
+ "\u2B9B",
193
+ "\u2B88",
194
+ "\u2B8A",
195
+ "\u2B89",
196
+ "\u2B8B",
197
+ "\u2190",
198
+ "\u2192",
199
+ "\u2191",
200
+ "\u2193",
201
+ "\u2196",
202
+ "\u2197",
203
+ "\u2199",
204
+ "\u2198",
205
+ "\u2B05",
206
+ "\u2794",
207
+ "\u2B06",
208
+ "\u2B07",
209
+ "\u2B09",
210
+ "\u2B08",
211
+ "\u2B0B",
212
+ "\u2B0A",
213
+ "\u21E6",
214
+ "\u21E8",
215
+ "\u21E7",
216
+ "\u21E9",
217
+ "\u2B04",
218
+ "\u21F3",
219
+ "\u2B00",
220
+ "\u2B01",
221
+ "\u2B03",
222
+ "\u2B02",
223
+ "\u{1F8AC}",
224
+ "\u{1F8AD}",
225
+ "\u2717",
226
+ "\u2714",
227
+ "\u2612",
228
+ "\u2611",
229
+ ""
230
+ ];
231
+ function wingdingsChar(code) {
232
+ if (code < 33 || code > 255) return void 0;
233
+ return WINGDINGS[code - 33] || void 0;
234
+ }
4
235
 
5
236
  // src/shared/pua.ts
6
237
  var BMP_SYMBOL_MAP = {
@@ -188,7 +419,7 @@ var BOXED_NUMBER_START = 983729;
188
419
  var BOXED_NUMBER_END = 983748;
189
420
  function mapPuaChar(code) {
190
421
  if (code >= 61472 && code <= 61695) {
191
- return BMP_SYMBOL_MAP[code - 61440];
422
+ return BMP_SYMBOL_MAP[code - 61440] ?? wingdingsChar(code - 61440);
192
423
  }
193
424
  if (code >= BOXED_NUMBER_START && code <= BOXED_NUMBER_END) {
194
425
  return String.fromCodePoint(9312 + (code - BOXED_NUMBER_START));
@@ -264,32 +495,54 @@ function buildTable(rows, options) {
264
495
  }
265
496
  function buildTableDirect(rows, numRows, options) {
266
497
  let maxCols = 0;
498
+ const rowCap = Math.min(MAX_ROWS, numRows + rows.reduce((n, r) => n + r.length, 0) + 1);
267
499
  for (const row of rows) {
268
500
  for (const cell of row) {
269
501
  const end = (cell.colAddr ?? 0) + cell.colSpan;
270
502
  if (end > maxCols) maxCols = end;
503
+ if (cell.rowAddr !== void 0 && cell.rowAddr >= numRows) numRows = Math.min(cell.rowAddr + 1, rowCap);
271
504
  }
272
505
  }
273
506
  if (maxCols > MAX_COLS) maxCols = MAX_COLS;
274
- if (maxCols === 0) return { rows: 0, cols: 0, cells: [], hasHeader: false };
507
+ if (maxCols === 0 || numRows === 0) return { rows: 0, cols: 0, cells: [], hasHeader: false };
275
508
  const grid = Array.from(
276
509
  { length: numRows },
277
510
  () => Array.from({ length: maxCols }, () => ({ text: "", colSpan: 1, rowSpan: 1 }))
278
511
  );
512
+ const owner = Array.from({ length: numRows }, () => new Array(maxCols));
279
513
  const anchorCols = /* @__PURE__ */ new Set();
280
- for (const row of rows) {
281
- for (const cell of row) {
282
- const r = cell.rowAddr ?? 0;
283
- const c = cell.colAddr ?? 0;
284
- if (r >= numRows || c >= maxCols || r < 0 || c < 0) continue;
514
+ for (let ri = 0; ri < rows.length; ri++) {
515
+ for (const cell of rows[ri]) {
516
+ const text = options?.keepEmptyParagraphs ? cell.text : cell.text.trim();
517
+ const r = Math.min(cell.rowAddr ?? ri, numRows - 1);
518
+ let c = cell.colAddr ?? 0;
519
+ if (r < 0 || c < 0) continue;
520
+ if (cell.colAddr === void 0) while (c < maxCols - 1 && owner[r][c]) c++;
521
+ if (c >= maxCols) c = maxCols - 1;
522
+ const own = owner[r][c];
523
+ if (own) {
524
+ if (text.trim()) own.text = own.text ? `${own.text}
525
+ ${text}` : text;
526
+ continue;
527
+ }
528
+ let colSpan = Math.max(1, Math.min(cell.colSpan, maxCols - c));
529
+ for (let dc = 1; dc < colSpan; dc++) if (owner[r][c + dc]) {
530
+ colSpan = dc;
531
+ break;
532
+ }
533
+ let rowSpan = Math.max(1, Math.min(cell.rowSpan, numRows - r));
534
+ for (let dr = 1; dr < rowSpan; dr++) {
535
+ if (owner[r + dr].slice(c, c + colSpan).some(Boolean)) {
536
+ rowSpan = dr;
537
+ break;
538
+ }
539
+ }
540
+ const ir = { text, colSpan, rowSpan };
285
541
  anchorCols.add(c);
286
- grid[r][c] = { text: options?.keepEmptyParagraphs ? cell.text : cell.text.trim(), colSpan: cell.colSpan, rowSpan: cell.rowSpan };
287
- for (let dr = 0; dr < cell.rowSpan; dr++) {
288
- for (let dc = 0; dc < cell.colSpan; dc++) {
289
- if (dr === 0 && dc === 0) continue;
290
- if (r + dr < numRows && c + dc < maxCols) {
291
- grid[r + dr][c + dc] = { text: "", colSpan: 1, rowSpan: 1 };
292
- }
542
+ for (let dr = 0; dr < rowSpan; dr++) {
543
+ for (let dc = 0; dc < colSpan; dc++) {
544
+ owner[r + dr][c + dc] = ir;
545
+ grid[r + dr][c + dc] = dr === 0 && dc === 0 ? ir : { text: "", colSpan: 1, rowSpan: 1 };
293
546
  }
294
547
  }
295
548
  }
@@ -306,6 +559,11 @@ function trimAndReturn(grid, numRows, maxCols, anchorCols, options) {
306
559
  }
307
560
  if (effectiveCols < maxCols && effectiveCols > 0) {
308
561
  const trimmed = grid.map((row) => row.slice(0, effectiveCols));
562
+ for (const row of trimmed) {
563
+ for (let c = 0; c < row.length; c++) {
564
+ if (c + row[c].colSpan > effectiveCols) row[c].colSpan = effectiveCols - c;
565
+ }
566
+ }
309
567
  return { rows: numRows, cols: effectiveCols, cells: trimmed, hasHeader: numRows > 1 };
310
568
  }
311
569
  return { rows: numRows, cols: maxCols, cells: grid, hasHeader: numRows > 1 };
@@ -318,13 +576,16 @@ function convertTableToText(rows) {
318
576
  function escapeGfm(text) {
319
577
  const NUL = String.fromCharCode(0);
320
578
  const spans = [];
321
- const masked = text.replace(/!\[[^\]]*\]\([^)\n]*\)|\]\((?:https?:|mailto:|tel:|#)[^)\n]*\)|\$\$[^$]*\$\$|\$[^$\n]*\$/gi, (m) => {
579
+ const masked = text.replace(/!\[[^\]]*\]\([^)\n]*\)|\]\((?:https?:|mailto:|tel:|#)[^)\n]*\)|(?<!\\)\$\$(?:\\[\s\S]|[^\\$])*\$\$|(?<!\\)\$(?:\\[^\n]|[^\\$\n])*\$/gi, (m) => {
322
580
  spans.push(m);
323
581
  return NUL + (spans.length - 1) + NUL;
324
582
  });
325
- const escaped = masked.replace(/([~*_`])/g, "\\$1");
583
+ const escaped = masked.replace(/\\(?=[!-#%-\/:-@\[-\x60{}~])/g, "\\\\").replace(/([~*_`])/g, "\\$1").replace(/(?<!\\)\|/g, "\\|").replace(/^([ \t]*)(?=#{1,6}(?:[ \t]|$))/gm, "$1\\").replace(/<(?!\/?u>)(?=[A-Za-z/!?])/g, "\\<");
326
584
  return escaped.replace(new RegExp(NUL + "(\\d+)" + NUL, "g"), (_, n) => spans[Number(n)]);
327
585
  }
586
+ function escapeLiteralDollar(text) {
587
+ return text.includes("$") ? text.replace(/\$/g, "\\$") : text;
588
+ }
328
589
  var HWP_SHAPE_ALT_TEXT_RE = /^(?:모서리가 둥근 |둥근 )?(?:사각형|직사각형|정사각형|원|타원|삼각형|이등변 삼각형|직각 삼각형|선|직선|곡선|화살표|굵은 화살표|이중 화살표|오각형|육각형|팔각형|별|[4-8]점별|십자|십자형|구름|구름형|마름모|도넛|평행사변형|사다리꼴|부채꼴|호|반원|물결|번개|하트|빗금|블록 화살표|수식|표|그림|개체|그리기\s?개체|묶음\s?개체|글상자|수식\s?개체|OLE\s?개체)\s?입니다\.?$/gm;
329
590
  function sanitizePua(text) {
330
591
  return mapPuaText(text).replace(/[\u{F0000}-\u{FFFFD}]/gu, "");
@@ -423,6 +684,7 @@ function dedupeRunningHeaders(blocks) {
423
684
  function spansToMarkdown(spans) {
424
685
  let out = "";
425
686
  for (const s of spans) {
687
+ if (s.placeholder) continue;
426
688
  const text = sanitizePua(s.text ?? "");
427
689
  if (!text) continue;
428
690
  let marker = s.code ? "`" : s.bold && s.italic ? "***" : s.bold ? "**" : s.italic ? "*" : "";
@@ -455,7 +717,8 @@ function blocksToMarkdown(blocks) {
455
717
  if (block.type === "heading" && block.text) {
456
718
  const prefix = "#".repeat(Math.min(block.level || 2, 6));
457
719
  const headingText = sanitizeText(block.text);
458
- if (headingText) lines.push("", `${prefix} ${escapeGfm(headingText)}`, "");
720
+ const note = block.footnoteText ? ` (\uC8FC: ${block.footnoteText})` : "";
721
+ if (headingText) lines.push("", `${prefix} ${escapeGfm(headingText + note)}`, "");
459
722
  continue;
460
723
  }
461
724
  if (block.type === "image" && block.text) {
@@ -549,6 +812,12 @@ function hasStructuredCellContent(table) {
549
812
  }
550
813
  return false;
551
814
  }
815
+ function noteSuffix(b) {
816
+ return b.footnoteText && b.text ? ` (\uC8FC: ${b.footnoteText})` : "";
817
+ }
818
+ function visibleText(b) {
819
+ return b.spans?.some((s) => s.placeholder) ? b.spans.filter((s) => !s.placeholder).map((s) => s.text).join("") : b.text ?? "";
820
+ }
552
821
  function cellInnerHtml(cell) {
553
822
  if (cell.blocks?.length) {
554
823
  return cell.blocks.map((b) => {
@@ -557,8 +826,8 @@ function cellInnerHtml(cell) {
557
826
  return (cap ? cap + "<br>" : "") + tableToHtml(b.table);
558
827
  }
559
828
  if (b.type === "image" && b.text) return `<img src="${b.text}" alt="image">`;
560
- const t = sanitizeText(b.text ?? "");
561
- return t ? t.replace(/\n/g, "<br>") : "";
829
+ const t = sanitizeText(visibleText(b));
830
+ return t ? (t + noteSuffix(b)).replace(/\n/g, "<br>") : "";
562
831
  }).filter(Boolean).join("<br>");
563
832
  }
564
833
  return sanitizeText(cell.text).replace(/\n/g, "<br>");
@@ -598,7 +867,7 @@ function tableToHtml(table) {
598
867
  const attrStr = attrs.length ? " " + attrs.join(" ") : "";
599
868
  rowHtml.push(`<${tag}${attrStr}>${text}</${tag}>`);
600
869
  }
601
- if (rowHtml.length) lines.push(`<tr>${rowHtml.join("")}</tr>`);
870
+ lines.push(`<tr>${rowHtml.join("")}</tr>`);
602
871
  }
603
872
  lines.push("</table>");
604
873
  return lines.join("\n");
@@ -606,7 +875,8 @@ function tableToHtml(table) {
606
875
  function tableToMarkdown(table) {
607
876
  if (table.rows === 0 || table.cols === 0) return "";
608
877
  const { cells, rows: numRows, cols: numCols } = table;
609
- if ((hasMergedCells(table) || hasStructuredCellContent(table)) && !tableContainsInlineMath(table)) {
878
+ if (hasStructuredCellContent(table)) return tableToHtml(table);
879
+ if (hasMergedCells(table) && !tableContainsInlineMath(table)) {
610
880
  return tableToHtml(table);
611
881
  }
612
882
  if (numRows === 1 && numCols === 1) {
@@ -630,7 +900,7 @@ function tableToMarkdown(table) {
630
900
  if (skip.has(`${r},${c}`)) continue;
631
901
  const cell = cells[r]?.[c];
632
902
  if (!cell) continue;
633
- display[r][c] = (cell.blocks?.some((b) => b.spans || b.type === "image" && b.text) ? cell.blocks.map((b) => b.type === "image" && b.text ? `![image](${b.text})` : b.spans ? spansToMarkdown(b.spans) : escapeGfm(sanitizeText(b.text ?? ""))).filter(Boolean).join("<br>") : escapeGfm(sanitizeText(cell.text)).replace(/\n/g, "<br>")).replace(/\|/g, "\\|");
903
+ display[r][c] = (cell.blocks?.some((b) => b.spans || b.type === "image" && b.text) ? cell.blocks.map((b) => b.type === "image" && b.text ? `![image](${b.text})` : b.spans ? spansToMarkdown(b.spans) + escapeGfm(noteSuffix(b)) : escapeGfm(sanitizeText(b.text ?? "") + noteSuffix(b))).filter(Boolean).join("<br>") : escapeGfm(sanitizeText(cell.text)).replace(/\n/g, "<br>")).replace(/(?<!\\)\|/g, "\\|");
634
904
  for (let dr = 0; dr < cell.rowSpan; dr++) {
635
905
  for (let dc = 0; dc < cell.colSpan; dc++) {
636
906
  if (dr === 0 && dc === 0) continue;
@@ -818,16 +1088,20 @@ function inlineImagesIntoMarkdown(markdown, images, opts) {
818
1088
  }
819
1089
 
820
1090
  export {
1091
+ WINGDINGS,
821
1092
  mapPuaText,
822
1093
  MAX_COLS,
823
1094
  MAX_ROWS,
824
1095
  buildTable,
825
1096
  convertTableToText,
1097
+ escapeGfm,
1098
+ escapeLiteralDollar,
826
1099
  flattenLayoutTables,
827
1100
  dedupeRunningHeaders,
828
1101
  blocksToMarkdown,
829
1102
  hasStructuredCellContent,
1103
+ noteSuffix,
830
1104
  encodePng,
831
1105
  inlineImagesIntoMarkdown
832
1106
  };
833
- //# sourceMappingURL=chunk-VQ5W4MYO.js.map
1107
+ //# sourceMappingURL=chunk-QIF2VXBQ.js.map