feat: add reviewed documents JSONL export endpoint

app/routes/documents.py

@@ -4,10 +4,12 @@ from decimal import Decimal, InvalidOperation
 import re
 import os
 import hashlib
+import json
+from decimal import Decimal
 from pathlib import Path
 
 from fastapi import APIRouter, Depends, Form, Query, Request
-from fastapi.responses import HTMLResponse, RedirectResponse
+from fastapi.responses import FileResponse, HTMLResponse, RedirectResponse
 from fastapi.templating import Jinja2Templates
 from sqlalchemy import distinct
 from sqlalchemy.orm import Session, selectinload
@@ -81,6 +83,95 @@ def _version_file_available(version, expected_document_id: str) -> bool:
     except Exception:
         return False
 
+
+def _json_safe(value):
+    if isinstance(value, Decimal):
+        return float(value)
+    if hasattr(value, "isoformat"):
+        return value.isoformat()
+    return value
+
+
+def _serialize_model_row(row, fields: list[str]) -> dict:
+    if not row:
+        return {}
+    data = {}
+    for field in fields:
+        value = getattr(row, field, None)
+        data[field] = _json_safe(value)
+    return data
+
+
+def _document_export_payload(document) -> dict:
+    raw_ocr, reviewed_ocr = _get_current_text_versions(document)
+    extracted = get_current_extracted_fields(document)
+    additional = _get_current_additional_fields(document)
+
+    versions = []
+    for version in sorted(getattr(document, "versions", []), key=lambda v: v.version_number):
+        created_at = getattr(version, "created_at", None)
+        versions.append({
+            "version_number": _json_safe(version.version_number),
+            "version_type": _json_safe(version.version_type),
+            "file_path": _json_safe(version.file_path),
+            "sha256": _json_safe(version.sha256),
+            "created_by": _json_safe(version.created_by),
+            "notes": _json_safe(version.notes),
+            "created_at": _json_safe(created_at),
+        })
+
+    return {
+        "document_id": document.document_id,
+        "document_type": document.document_type,
+        "review_status": document.review_status,
+        "source_path": document.source_path,
+        "original_path": document.original_path,
+        "current_path": document.current_path,
+        "share_path": document.share_path,
+        "original_filename": document.original_filename,
+        "canonical_filename": document.canonical_filename,
+        "mime_type": document.mime_type,
+        "file_size": _json_safe(document.file_size),
+        "page_count": _json_safe(document.page_count),
+        "sha256_original": _json_safe(document.sha256_original),
+        "sha256_current": _json_safe(document.sha256_current),
+        "raw_ocr_text": _json_safe(raw_ocr.text_content if raw_ocr else None),
+        "reviewed_ocr_text": _json_safe(reviewed_ocr.text_content if reviewed_ocr else None),
+        "ocr_quality_score": _json_safe(raw_ocr.quality_score if raw_ocr else None),
+        "quality_flags": raw_ocr.quality_flags if raw_ocr and raw_ocr.quality_flags else [],
+        "quality_note": _json_safe(raw_ocr.quality_note if raw_ocr else None),
+        "extracted_fields": _serialize_model_row(extracted, [
+            "merchant_raw",
+            "merchant_normalized",
+            "transaction_date",
+            "transaction_time",
+            "subtotal",
+            "tax",
+            "total",
+            "currency",
+            "payment_method",
+            "receipt_number",
+            "location",
+            "counterparty",
+        ]),
+        "additional_fields": _serialize_model_row(additional, [
+            "owner_primary",
+            "owner_secondary",
+            "paid_by_person",
+            "occasion_note",
+            "is_shared_expense",
+            "covered_people",
+            "attendees",
+            "reimbursement_expected_from",
+            "reimbursement_paid_by",
+            "reimbursement_paid_to",
+            "reimbursement_paid_amount",
+            "reimbursement_paid_date",
+            "reimbursement_note",
+        ]),
+        "versions": versions,
+    }
+
 BASE_DIR = Path(__file__).resolve().parent.parent
 templates = Jinja2Templates(directory=str(BASE_DIR / "templates"))
 
@@ -996,3 +1087,37 @@ def document_detail(document_id: str, request: Request, queue: str | None = None
             "active_page": "documents",
         },
     )
+
+
+
+@router.get("/export/reviewed.jsonl")
+def export_reviewed_jsonl(db: Session = Depends(get_db)):
+    docs = (
+        db.query(Document)
+        .options(
+            selectinload(Document.text_versions),
+            selectinload(Document.naming_fields),
+            selectinload(Document.extracted_fields),
+            selectinload(Document.additional_fields),
+            selectinload(Document.versions),
+        )
+        .filter(Document.review_status == "reviewed")
+        .order_by(Document.updated_at.asc())
+        .all()
+    )
+
+    export_dir = Path("/mnt/storage/document-processor/exports")
+    export_dir.mkdir(parents=True, exist_ok=True)
+    out_path = export_dir / "reviewed_documents.jsonl"
+
+    with out_path.open("w", encoding="utf-8") as f:
+        for document in docs:
+            payload = _document_export_payload(document)
+            f.write(json.dumps(payload, ensure_ascii=False) + "\n")
+
+    return FileResponse(
+        path=str(out_path),
+        media_type="application/json",
+        filename=out_path.name,
+    )
+