chore: pin document recognition engine

app/ingestion/ocr.py

@@ -1,87 +1,22 @@
-"""OCRmyPDF integration with Tesseract.
-
-We treat OCR as best-effort: if the input PDF already has a text layer (or OCR is
-disabled by config), we skip OCR and use the original PDF. On failure, the
-caller is expected to mark the document ``OCR_FAILED`` and continue without it.
-"""
+"""LegacyHUB adapter for the shared document recognition engine."""
 
 from __future__ import annotations
 
-from dataclasses import dataclass
 from pathlib import Path
 
-import ocrmypdf
+from teamhub_document_recognition_engine import OcrResult
+from teamhub_document_recognition_engine import run_ocr as _engine_run_ocr
 
 from app.config import settings
-from app.logging_config import get_logger
-from app.utils.pdf import has_searchable_text
-
-logger = get_logger(__name__)
-
-
-@dataclass
-class OcrResult:
-    output_path: Path
-    skipped: bool
-    reason: str
-    languages: str
 
 
 def run_ocr(input_pdf: Path, output_pdf: Path, languages: str | None = None) -> OcrResult:
-    """Run OCRmyPDF.
-
-    - If ``OCR_ENABLED`` is false: copy the input as the output and skip.
-    - If the input already has searchable text: skip OCR but still produce
-      ``output_pdf`` (a hard-link / copy to keep downstream code simple).
-    - On unexpected exceptions: re-raise (caller handles status update).
-    """
-    langs = languages or settings.ocr_languages
-
-    if not settings.ocr_enabled:
-        return _skip(input_pdf, output_pdf, langs, "ocr_disabled")
-
-    if has_searchable_text(input_pdf):
-        return _skip(input_pdf, output_pdf, langs, "already_searchable")
-
-    output_pdf.parent.mkdir(parents=True, exist_ok=True)
-    logger.info("ocr.start", input=str(input_pdf), output=str(output_pdf), languages=langs)
-
-    try:
-        ocrmypdf.ocr(
-            input_file=str(input_pdf),
-            output_file=str(output_pdf),
-            language=langs,
-            skip_text=False,
-            redo_ocr=False,
-            force_ocr=False,
-            deskew=settings.ocr_deskew,
-            clean=settings.ocr_clean,
-            optimize=settings.ocr_optimize,
-            progress_bar=False,
-            jobs=1,
-            output_type="pdf",
-            # tolerate already-OCR pages where present
-            skip_big=200.0,
-        )
-    except ocrmypdf.exceptions.PriorOcrFoundError:
-        logger.info("ocr.skip.prior_ocr", input=str(input_pdf))
-        return _skip(input_pdf, output_pdf, langs, "prior_ocr_found")
-    except ocrmypdf.exceptions.DigitalSignatureError:
-        logger.warning("ocr.skip.signed_pdf", input=str(input_pdf))
-        return _skip(input_pdf, output_pdf, langs, "digitally_signed")
-    except ocrmypdf.exceptions.EncryptedPdfError as exc:
-        logger.warning("ocr.encrypted", input=str(input_pdf), error=str(exc))
-        raise
-    except ocrmypdf.exceptions.MissingDependencyError as exc:
-        logger.error("ocr.missing_dependency", error=str(exc))
-        raise
-
-    logger.info("ocr.done", output=str(output_pdf))
-    return OcrResult(output_path=output_pdf, skipped=False, reason="ocr_completed", languages=langs)
-
-
-def _skip(input_pdf: Path, output_pdf: Path, langs: str, reason: str) -> OcrResult:
-    output_pdf.parent.mkdir(parents=True, exist_ok=True)
-    if not output_pdf.exists() or output_pdf.resolve() != input_pdf.resolve():
-        output_pdf.write_bytes(input_pdf.read_bytes())
-    return OcrResult(output_path=output_pdf, skipped=True, reason=reason, languages=langs)
+    return _engine_run_ocr(
+        input_pdf,
+        output_pdf,
+        languages=languages or settings.ocr_languages,
+        enabled=settings.ocr_enabled,
+        deskew=settings.ocr_deskew,
+        clean=settings.ocr_clean,
+        optimize=settings.ocr_optimize,
+    )