fix ocr

backend/alembic/env.py

@@ -1,8 +1,16 @@
 from logging.config import fileConfig
+import sys
+from pathlib import Path
 
 from sqlalchemy import engine_from_config, pool
 from alembic import context
 
+# Ensure `backend/` is on sys.path so `import app...` works
+# no matter where `alembic` is executed from.
+BACKEND_ROOT = Path(__file__).resolve().parents[1]
+if str(BACKEND_ROOT) not in sys.path:
+    sys.path.insert(0, str(BACKEND_ROOT))
+
 from app.core.config import settings
 from app.core.database import Base
 import app.models  # noqa: F401 – ensure all models are imported

backend/alembic/script.py.mako

@@ -8,6 +8,7 @@ from typing import Sequence, Union
 
 from alembic import op
 import sqlalchemy as sa
+from sqlalchemy import Text
 ${imports if imports else ""}
 
 revision: str = ${repr(up_revision)}

backend/alembic/versions/be562b8079e3_init_sqlite_test.py

@@ -0,0 +1,169 @@
+"""init_sqlite_test
+
+Revision ID: be562b8079e3
+Revises: 
+Create Date: 2026-03-11 17:33:30.695730
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy import Text
+from sqlalchemy.dialects import postgresql
+
+revision: str = 'be562b8079e3'
+down_revision: Union[str, None] = None
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.create_table('cases',
+    sa.Column('id', sa.UUID(), nullable=False),
+    sa.Column('case_no', sa.String(length=64), nullable=False),
+    sa.Column('title', sa.String(length=256), nullable=False),
+    sa.Column('victim_name', sa.String(length=128), nullable=False),
+    sa.Column('handler', sa.String(length=128), nullable=False),
+    sa.Column('status', sa.Enum('pending', 'uploading', 'analyzing', 'reviewing', 'completed', name='casestatus'), nullable=False),
+    sa.Column('image_count', sa.Integer(), nullable=False),
+    sa.Column('total_amount', sa.Numeric(precision=14, scale=2), nullable=False),
+    sa.Column('created_by', sa.String(length=128), nullable=False),
+    sa.Column('created_at', sa.DateTime(timezone=True), server_default=sa.text('(CURRENT_TIMESTAMP)'), nullable=False),
+    sa.Column('updated_at', sa.DateTime(timezone=True), server_default=sa.text('(CURRENT_TIMESTAMP)'), nullable=False),
+    sa.Column('deleted_at', sa.DateTime(timezone=True), nullable=True),
+    sa.PrimaryKeyConstraint('id')
+    )
+    op.create_index(op.f('ix_cases_case_no'), 'cases', ['case_no'], unique=True)
+    op.create_table('evidence_images',
+    sa.Column('id', sa.UUID(), nullable=False),
+    sa.Column('case_id', sa.UUID(), nullable=False),
+    sa.Column('file_path', sa.String(length=512), nullable=False),
+    sa.Column('thumb_path', sa.String(length=512), nullable=False),
+    sa.Column('source_app', sa.Enum('wechat', 'alipay', 'bank', 'digital_wallet', 'other', name='sourceapp'), nullable=False),
+    sa.Column('page_type', sa.Enum('bill_list', 'bill_detail', 'transfer_receipt', 'sms_notice', 'balance', 'unknown', name='pagetype'), nullable=False),
+    sa.Column('ocr_status', sa.Enum('pending', 'processing', 'done', 'failed', name='ocrstatus'), nullable=False),
+    sa.Column('file_hash', sa.String(length=128), nullable=False),
+    sa.Column('file_size', sa.Integer(), nullable=False),
+    sa.Column('uploaded_at', sa.DateTime(timezone=True), server_default=sa.text('(CURRENT_TIMESTAMP)'), nullable=False),
+    sa.ForeignKeyConstraint(['case_id'], ['cases.id'], ),
+    sa.PrimaryKeyConstraint('id')
+    )
+    op.create_index(op.f('ix_evidence_images_case_id'), 'evidence_images', ['case_id'], unique=False)
+    op.create_index(op.f('ix_evidence_images_file_hash'), 'evidence_images', ['file_hash'], unique=True)
+    op.create_table('export_reports',
+    sa.Column('id', sa.UUID(), nullable=False),
+    sa.Column('case_id', sa.UUID(), nullable=False),
+    sa.Column('report_type', sa.Enum('pdf', 'excel', 'word', name='reporttype'), nullable=False),
+    sa.Column('file_path', sa.String(length=512), nullable=False),
+    sa.Column('version', sa.Integer(), nullable=False),
+    sa.Column('content_snapshot', postgresql.JSONB(astext_type=Text()), nullable=False),
+    sa.Column('created_at', sa.DateTime(timezone=True), server_default=sa.text('(CURRENT_TIMESTAMP)'), nullable=False),
+    sa.ForeignKeyConstraint(['case_id'], ['cases.id'], ),
+    sa.PrimaryKeyConstraint('id')
+    )
+    op.create_index(op.f('ix_export_reports_case_id'), 'export_reports', ['case_id'], unique=False)
+    op.create_table('fund_flow_edges',
+    sa.Column('id', sa.UUID(), nullable=False),
+    sa.Column('case_id', sa.UUID(), nullable=False),
+    sa.Column('source_node', sa.String(length=256), nullable=False),
+    sa.Column('target_node', sa.String(length=256), nullable=False),
+    sa.Column('source_type', sa.String(length=32), nullable=False),
+    sa.Column('target_type', sa.String(length=32), nullable=False),
+    sa.Column('amount', sa.Numeric(precision=14, scale=2), nullable=False),
+    sa.Column('tx_count', sa.Integer(), nullable=False),
+    sa.Column('earliest_time', sa.DateTime(timezone=True), nullable=True),
+    sa.Column('created_at', sa.DateTime(timezone=True), server_default=sa.text('(CURRENT_TIMESTAMP)'), nullable=False),
+    sa.ForeignKeyConstraint(['case_id'], ['cases.id'], ),
+    sa.PrimaryKeyConstraint('id')
+    )
+    op.create_index(op.f('ix_fund_flow_edges_case_id'), 'fund_flow_edges', ['case_id'], unique=False)
+    op.create_table('transaction_clusters',
+    sa.Column('id', sa.UUID(), nullable=False),
+    sa.Column('case_id', sa.UUID(), nullable=False),
+    sa.Column('primary_tx_id', sa.UUID(), nullable=True),
+    sa.Column('match_reason', sa.String(length=512), nullable=False),
+    sa.ForeignKeyConstraint(['case_id'], ['cases.id'], ),
+    sa.PrimaryKeyConstraint('id')
+    )
+    op.create_index(op.f('ix_transaction_clusters_case_id'), 'transaction_clusters', ['case_id'], unique=False)
+    op.create_table('ocr_blocks',
+    sa.Column('id', sa.UUID(), nullable=False),
+    sa.Column('image_id', sa.UUID(), nullable=False),
+    sa.Column('content', sa.Text(), nullable=False),
+    sa.Column('bbox', postgresql.JSONB(astext_type=Text()), nullable=False),
+    sa.Column('seq_order', sa.Integer(), nullable=False),
+    sa.Column('confidence', sa.Float(), nullable=False),
+    sa.ForeignKeyConstraint(['image_id'], ['evidence_images.id'], ),
+    sa.PrimaryKeyConstraint('id')
+    )
+    op.create_index(op.f('ix_ocr_blocks_image_id'), 'ocr_blocks', ['image_id'], unique=False)
+    op.create_table('transaction_records',
+    sa.Column('id', sa.UUID(), nullable=False),
+    sa.Column('case_id', sa.UUID(), nullable=False),
+    sa.Column('evidence_image_id', sa.UUID(), nullable=True),
+    sa.Column('cluster_id', sa.UUID(), nullable=True),
+    sa.Column('source_app', sa.Enum('wechat', 'alipay', 'bank', 'digital_wallet', 'other', name='sourceapp'), nullable=False),
+    sa.Column('trade_time', sa.DateTime(timezone=True), nullable=False),
+    sa.Column('amount', sa.Numeric(precision=14, scale=2), nullable=False),
+    sa.Column('direction', sa.Enum('in_', 'out', name='direction'), nullable=False),
+    sa.Column('counterparty_name', sa.String(length=256), nullable=False),
+    sa.Column('counterparty_account', sa.String(length=256), nullable=False),
+    sa.Column('self_account_tail_no', sa.String(length=32), nullable=False),
+    sa.Column('order_no', sa.String(length=128), nullable=False),
+    sa.Column('remark', sa.Text(), nullable=False),
+    sa.Column('confidence', sa.Float(), nullable=False),
+    sa.Column('is_duplicate', sa.Boolean(), nullable=False),
+    sa.Column('is_transit', sa.Boolean(), nullable=False),
+    sa.Column('created_at', sa.DateTime(timezone=True), server_default=sa.text('(CURRENT_TIMESTAMP)'), nullable=False),
+    sa.ForeignKeyConstraint(['case_id'], ['cases.id'], ),
+    sa.ForeignKeyConstraint(['cluster_id'], ['transaction_clusters.id'], ),
+    sa.ForeignKeyConstraint(['evidence_image_id'], ['evidence_images.id'], ),
+    sa.PrimaryKeyConstraint('id')
+    )
+    op.create_index(op.f('ix_transaction_records_case_id'), 'transaction_records', ['case_id'], unique=False)
+    op.create_index(op.f('ix_transaction_records_order_no'), 'transaction_records', ['order_no'], unique=False)
+    op.create_index(op.f('ix_transaction_records_trade_time'), 'transaction_records', ['trade_time'], unique=False)
+    op.create_table('fraud_assessments',
+    sa.Column('id', sa.UUID(), nullable=False),
+    sa.Column('case_id', sa.UUID(), nullable=False),
+    sa.Column('transaction_id', sa.UUID(), nullable=False),
+    sa.Column('confidence_level', sa.Enum('high', 'medium', 'low', name='confidencelevel'), nullable=False),
+    sa.Column('assessed_amount', sa.Numeric(precision=14, scale=2), nullable=False),
+    sa.Column('reason', sa.Text(), nullable=False),
+    sa.Column('exclude_reason', sa.Text(), nullable=False),
+    sa.Column('review_status', sa.Enum('pending', 'confirmed', 'rejected', 'needs_info', name='reviewstatus'), nullable=False),
+    sa.Column('review_note', sa.Text(), nullable=False),
+    sa.Column('reviewed_by', sa.String(length=128), nullable=False),
+    sa.Column('reviewed_at', sa.DateTime(timezone=True), nullable=True),
+    sa.Column('created_at', sa.DateTime(timezone=True), server_default=sa.text('(CURRENT_TIMESTAMP)'), nullable=False),
+    sa.ForeignKeyConstraint(['case_id'], ['cases.id'], ),
+    sa.ForeignKeyConstraint(['transaction_id'], ['transaction_records.id'], ),
+    sa.PrimaryKeyConstraint('id')
+    )
+    op.create_index(op.f('ix_fraud_assessments_case_id'), 'fraud_assessments', ['case_id'], unique=False)
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.drop_index(op.f('ix_fraud_assessments_case_id'), table_name='fraud_assessments')
+    op.drop_table('fraud_assessments')
+    op.drop_index(op.f('ix_transaction_records_trade_time'), table_name='transaction_records')
+    op.drop_index(op.f('ix_transaction_records_order_no'), table_name='transaction_records')
+    op.drop_index(op.f('ix_transaction_records_case_id'), table_name='transaction_records')
+    op.drop_table('transaction_records')
+    op.drop_index(op.f('ix_ocr_blocks_image_id'), table_name='ocr_blocks')
+    op.drop_table('ocr_blocks')
+    op.drop_index(op.f('ix_transaction_clusters_case_id'), table_name='transaction_clusters')
+    op.drop_table('transaction_clusters')
+    op.drop_index(op.f('ix_fund_flow_edges_case_id'), table_name='fund_flow_edges')
+    op.drop_table('fund_flow_edges')
+    op.drop_index(op.f('ix_export_reports_case_id'), table_name='export_reports')
+    op.drop_table('export_reports')
+    op.drop_index(op.f('ix_evidence_images_file_hash'), table_name='evidence_images')
+    op.drop_index(op.f('ix_evidence_images_case_id'), table_name='evidence_images')
+    op.drop_table('evidence_images')
+    op.drop_index(op.f('ix_cases_case_no'), table_name='cases')
+    op.drop_table('cases')
+    # ### end Alembic commands ###

backend/app/api/v1/assessments.py

@@ -2,10 +2,12 @@ from uuid import UUID
 from datetime import datetime, timezone
 
 from fastapi import APIRouter, Depends, Query, HTTPException
+from sqlalchemy import select
 from sqlalchemy.ext.asyncio import AsyncSession
+from sqlalchemy.orm import selectinload
 
 from app.core.database import get_db
-from app.models.assessment import ConfidenceLevel
+from app.models.assessment import FraudAssessment, ConfidenceLevel
 from app.repositories.assessment_repo import AssessmentRepository
 from app.schemas.assessment import (
     AssessmentOut,
@@ -46,6 +48,14 @@ async def review_assessment(
         "reviewed_by": body.reviewed_by,
         "reviewed_at": datetime.now(timezone.utc),
     })
+
+    # eager-load the transaction relationship to avoid lazy-load in async context
+    result = await db.execute(
+        select(FraudAssessment)
+        .options(selectinload(FraudAssessment.transaction))
+        .where(FraudAssessment.id == assessment_id)
+    )
+    assessment = result.scalar_one()
     return assessment
 
 

backend/app/api/v1/images.py

@@ -1,15 +1,16 @@
 from uuid import UUID
+import asyncio
 
-from fastapi import APIRouter, Depends, UploadFile, File, HTTPException, Query
+from fastapi import APIRouter, Depends, UploadFile, File, HTTPException
 from fastapi.responses import FileResponse
 from sqlalchemy.ext.asyncio import AsyncSession
 
 from app.core.config import settings
 from app.core.database import get_db
-from app.models.evidence_image import EvidenceImage, SourceApp, PageType
+from app.models.evidence_image import EvidenceImage, SourceApp, PageType, OcrStatus
 from app.repositories.image_repo import ImageRepository
 from app.repositories.case_repo import CaseRepository
-from app.schemas.image import ImageOut, ImageDetailOut, OcrFieldCorrection
+from app.schemas.image import ImageOut, ImageDetailOut, OcrFieldCorrection, CaseOcrStartIn
 from app.utils.hash import sha256_file
 from app.utils.file_storage import save_upload
 
@@ -32,9 +33,11 @@ async def upload_images(
 
     for f in files:
         data = await f.read()
-        file_hash = sha256_file(data)
+        raw_hash = sha256_file(data)
+        # Scope hash by case to avoid cross-case unique conflicts while still deduplicating inside one case.
+        scoped_hash = f"{raw_hash}:{case_id}"
 
-        existing = await img_repo.find_by_hash(file_hash)
+        existing = await img_repo.find_by_hash_in_case(case_id, [raw_hash, scoped_hash])
         if existing:
             results.append(existing)
             continue
@@ -44,7 +47,7 @@ async def upload_images(
             case_id=case_id,
             file_path=file_path,
             thumb_path=thumb_path,
-            file_hash=file_hash,
+            file_hash=scoped_hash,
             file_size=len(data),
         )
         image = await img_repo.create(image)
@@ -53,14 +56,11 @@ async def upload_images(
     case.image_count = await img_repo.count_by_case(case_id)
     await db.flush()
 
-    # trigger OCR tasks (non-blocking)
+    # trigger OCR tasks in-process background (non-blocking for API response)
-    from app.workers.ocr_tasks import process_image_ocr
+    from app.workers.ocr_tasks import process_image_ocr_async
     for img in results:
         if img.ocr_status.value == "pending":
-            try:
+            asyncio.create_task(process_image_ocr_async(str(img.id)))
-                process_image_ocr.delay(str(img.id))
-            except Exception:
-                pass
 
     return results
 
@@ -73,7 +73,21 @@ async def list_images(
     db: AsyncSession = Depends(get_db),
 ):
     repo = ImageRepository(db)
-    return await repo.list_by_case(case_id, source_app=source_app, page_type=page_type)
+    images = await repo.list_by_case(case_id, source_app=source_app, page_type=page_type)
+    return [
+        ImageOut(
+            id=img.id,
+            case_id=img.case_id,
+            url=f"/api/v1/images/{img.id}/file",
+            thumb_url=f"/api/v1/images/{img.id}/file",
+            source_app=img.source_app,
+            page_type=img.page_type,
+            ocr_status=img.ocr_status,
+            file_hash=img.file_hash,
+            uploaded_at=img.uploaded_at,
+        )
+        for img in images
+    ]
 
 
 @router.get("/images/{image_id}", response_model=ImageDetailOut)
@@ -128,3 +142,43 @@ async def get_image_file(image_id: UUID, db: AsyncSession = Depends(get_db)):
     if not full_path.exists():
         raise HTTPException(404, "文件不存在")
     return FileResponse(full_path)
+
+
+@router.post("/cases/{case_id}/ocr/start")
+async def start_case_ocr(
+    case_id: UUID,
+    payload: CaseOcrStartIn | None = None,
+    db: AsyncSession = Depends(get_db),
+):
+    case_repo = CaseRepository(db)
+    case = await case_repo.get(case_id)
+    if not case:
+        raise HTTPException(404, "案件不存在")
+
+    repo = ImageRepository(db)
+    include_done = payload.include_done if payload else False
+    image_ids = payload.image_ids if payload else []
+    if image_ids:
+        images = await repo.list_by_ids_in_case(case_id, image_ids)
+        # For explicit re-run, mark selected images as processing immediately
+        # so frontend can reflect state transition without full page refresh.
+        for img in images:
+            img.ocr_status = OcrStatus.processing
+        await db.flush()
+        await db.commit()
+    else:
+        images = await repo.list_for_ocr(case_id, include_done=include_done)
+
+    from app.workers.ocr_tasks import process_image_ocr_async
+
+    submitted = 0
+    for img in images:
+        asyncio.create_task(process_image_ocr_async(str(img.id)))
+        submitted += 1
+
+    return {
+        "caseId": str(case_id),
+        "submitted": submitted,
+        "totalCandidates": len(images),
+        "message": f"已提交 {submitted} 张截图的 OCR 任务",
+    }

backend/app/core/config.py

@@ -17,6 +17,7 @@ class Settings(BaseSettings):
 
     OCR_API_KEY: str = ""
     OCR_API_URL: str = ""
+    OCR_MODEL: str = ""
     LLM_API_KEY: str = ""
     LLM_API_URL: str = ""
     LLM_MODEL: str = ""

backend/app/repositories/image_repo.py

@@ -3,7 +3,7 @@ from uuid import UUID
 from sqlalchemy import select, func
 from sqlalchemy.ext.asyncio import AsyncSession
 
-from app.models.evidence_image import EvidenceImage, SourceApp, PageType
+from app.models.evidence_image import EvidenceImage, SourceApp, PageType, OcrStatus
 from app.repositories.base import BaseRepository
 
 
@@ -17,6 +17,17 @@ class ImageRepository(BaseRepository[EvidenceImage]):
         )
         return result.scalar_one_or_none()
 
+    async def find_by_hash_in_case(self, case_id: UUID, file_hashes: list[str]) -> EvidenceImage | None:
+        if not file_hashes:
+            return None
+        result = await self.session.execute(
+            select(EvidenceImage).where(
+                EvidenceImage.case_id == case_id,
+                EvidenceImage.file_hash.in_(file_hashes),
+            )
+        )
+        return result.scalar_one_or_none()
+
     async def list_by_case(
         self,
         case_id: UUID,
@@ -37,3 +48,20 @@ class ImageRepository(BaseRepository[EvidenceImage]):
             select(func.count()).select_from(EvidenceImage).where(EvidenceImage.case_id == case_id)
         )
         return result.scalar() or 0
+
+    async def list_for_ocr(self, case_id: UUID, include_done: bool = False) -> list[EvidenceImage]:
+        query = select(EvidenceImage).where(EvidenceImage.case_id == case_id)
+        if not include_done:
+            query = query.where(EvidenceImage.ocr_status != OcrStatus.done)
+        result = await self.session.execute(query.order_by(EvidenceImage.uploaded_at.desc()))
+        return list(result.scalars().all())
+
+    async def list_by_ids_in_case(self, case_id: UUID, image_ids: list[UUID]) -> list[EvidenceImage]:
+        if not image_ids:
+            return []
+        result = await self.session.execute(
+            select(EvidenceImage)
+            .where(EvidenceImage.case_id == case_id, EvidenceImage.id.in_(image_ids))
+            .order_by(EvidenceImage.uploaded_at.desc())
+        )
+        return list(result.scalars().all())

backend/app/schemas/analysis.py

@@ -1,7 +1,7 @@
-from pydantic import BaseModel
+from app.schemas.base import CamelModel
 
 
-class AnalysisStatusOut(BaseModel):
+class AnalysisStatusOut(CamelModel):
     case_id: str
     status: str
     progress: int = 0
@@ -9,6 +9,6 @@ class AnalysisStatusOut(BaseModel):
     message: str = ""
 
 
-class AnalysisTriggerOut(BaseModel):
+class AnalysisTriggerOut(CamelModel):
     task_id: str
     message: str

backend/app/schemas/assessment.py

@@ -1,13 +1,12 @@
 from datetime import datetime
 from uuid import UUID
 
-from pydantic import BaseModel
-
 from app.models.assessment import ConfidenceLevel, ReviewStatus
+from app.schemas.base import CamelModel
 from app.schemas.transaction import TransactionOut
 
 
-class AssessmentOut(BaseModel):
+class AssessmentOut(CamelModel):
     id: UUID
     case_id: UUID
     transaction_id: UUID
@@ -21,19 +20,17 @@ class AssessmentOut(BaseModel):
     reviewed_by: str
     reviewed_at: datetime | None = None
 
-    model_config = {"from_attributes": True}
 
-
+class AssessmentListOut(CamelModel):
-class AssessmentListOut(BaseModel):
     items: list[AssessmentOut]
     total: int
 
 
-class ReviewSubmit(BaseModel):
+class ReviewSubmit(CamelModel):
     review_status: ReviewStatus
     review_note: str = ""
     reviewed_by: str = "demo_user"
 
 
-class InquirySuggestionOut(BaseModel):
+class InquirySuggestionOut(CamelModel):
     suggestions: list[str]

backend/app/schemas/base.py

@@ -0,0 +1,15 @@
+"""Base schema with camelCase JSON serialization."""
+from pydantic import BaseModel, ConfigDict
+
+
+def to_camel(s: str) -> str:
+    parts = s.split("_")
+    return parts[0] + "".join(p.capitalize() for p in parts[1:])
+
+
+class CamelModel(BaseModel):
+    model_config = ConfigDict(
+        from_attributes=True,
+        alias_generator=to_camel,
+        populate_by_name=True,
+    )

backend/app/schemas/case.py

@@ -1,26 +1,25 @@
 from datetime import datetime
 from uuid import UUID
 
-from pydantic import BaseModel
-
 from app.models.case import CaseStatus
+from app.schemas.base import CamelModel
 
 
-class CaseCreate(BaseModel):
+class CaseCreate(CamelModel):
     case_no: str
     title: str
     victim_name: str
     handler: str = ""
 
 
-class CaseUpdate(BaseModel):
+class CaseUpdate(CamelModel):
     title: str | None = None
     victim_name: str | None = None
     handler: str | None = None
     status: CaseStatus | None = None
 
 
-class CaseOut(BaseModel):
+class CaseOut(CamelModel):
     id: UUID
     case_no: str
     title: str
@@ -32,9 +31,7 @@ class CaseOut(BaseModel):
     created_at: datetime
     updated_at: datetime
 
-    model_config = {"from_attributes": True}
 
-
+class CaseListOut(CamelModel):
-class CaseListOut(BaseModel):
     items: list[CaseOut]
     total: int

backend/app/schemas/image.py

@@ -1,12 +1,11 @@
 from datetime import datetime
 from uuid import UUID
 
-from pydantic import BaseModel
-
 from app.models.evidence_image import SourceApp, PageType, OcrStatus
+from app.schemas.base import CamelModel
 
 
-class ImageOut(BaseModel):
+class ImageOut(CamelModel):
     id: UUID
     case_id: UUID
     url: str = ""
@@ -17,24 +16,25 @@ class ImageOut(BaseModel):
     file_hash: str
     uploaded_at: datetime
 
-    model_config = {"from_attributes": True}
 
-
+class OcrBlockOut(CamelModel):
-class OcrBlockOut(BaseModel):
     id: UUID
     content: str
     bbox: dict
     seq_order: int
     confidence: float
 
-    model_config = {"from_attributes": True}
-
 
 class ImageDetailOut(ImageOut):
     ocr_blocks: list[OcrBlockOut] = []
 
 
-class OcrFieldCorrection(BaseModel):
+class OcrFieldCorrection(CamelModel):
     field_name: str
     old_value: str
     new_value: str
+
+
+class CaseOcrStartIn(CamelModel):
+    include_done: bool = False
+    image_ids: list[UUID] = []

backend/app/schemas/report.py

@@ -1,12 +1,11 @@
 from datetime import datetime
 from uuid import UUID
 
-from pydantic import BaseModel
-
 from app.models.report import ReportType
+from app.schemas.base import CamelModel
 
 
-class ReportCreate(BaseModel):
+class ReportCreate(CamelModel):
     report_type: ReportType
     include_summary: bool = True
     include_transactions: bool = True
@@ -17,7 +16,7 @@ class ReportCreate(BaseModel):
     include_screenshots: bool = False
 
 
-class ReportOut(BaseModel):
+class ReportOut(CamelModel):
     id: UUID
     case_id: UUID
     report_type: ReportType
@@ -25,9 +24,7 @@ class ReportOut(BaseModel):
     version: int
     created_at: datetime
 
-    model_config = {"from_attributes": True}
 
-
+class ReportListOut(CamelModel):
-class ReportListOut(BaseModel):
     items: list[ReportOut]
     total: int

backend/app/schemas/transaction.py

@@ -1,13 +1,12 @@
 from datetime import datetime
 from uuid import UUID
 
-from pydantic import BaseModel
-
 from app.models.evidence_image import SourceApp
 from app.models.transaction import Direction
+from app.schemas.base import CamelModel
 
 
-class TransactionOut(BaseModel):
+class TransactionOut(CamelModel):
     id: UUID
     case_id: UUID
     source_app: SourceApp
@@ -25,21 +24,19 @@ class TransactionOut(BaseModel):
     is_duplicate: bool
     is_transit: bool
 
-    model_config = {"from_attributes": True}
 
-
+class TransactionListOut(CamelModel):
-class TransactionListOut(BaseModel):
     items: list[TransactionOut]
     total: int
 
 
-class FlowNodeOut(BaseModel):
+class FlowNodeOut(CamelModel):
     id: str
     label: str
     type: str
 
 
-class FlowEdgeOut(BaseModel):
+class FlowEdgeOut(CamelModel):
     source: str
     target: str
     amount: float
@@ -47,6 +44,6 @@ class FlowEdgeOut(BaseModel):
     trade_time: str
 
 
-class FlowGraphOut(BaseModel):
+class FlowGraphOut(CamelModel):
     nodes: list[FlowNodeOut]
     edges: list[FlowEdgeOut]

backend/app/services/ocr_service.py

@@ -1,12 +1,20 @@
 """OCR and multimodal extraction service.
 
-Wraps calls to cloud OCR / multimodal APIs with a provider-agnostic interface.
+Both classify_page and extract_transaction_fields use OpenAI-compatible
-When API keys are not configured, falls back to a mock implementation that
+multimodal chat completion APIs. OCR and LLM can point to different
-returns placeholder data (sufficient for demo / competition).
+providers / models via separate env vars:
+
+  OCR_API_URL / OCR_API_KEY / OCR_MODEL  — for page classification & field extraction
+  LLM_API_URL / LLM_API_KEY / LLM_MODEL — for reasoning tasks (assessment, suggestions)
+
+When OCR keys are not set, falls back to LLM keys.
+When neither is set, returns mock data (sufficient for demo).
 """
+import base64
 import json
 import logging
-from pathlib import Path
+import ast
+import re
 
 import httpx
 
@@ -14,28 +22,163 @@ from app.core.config import settings
 from app.models.evidence_image import SourceApp, PageType
 
 logger = logging.getLogger(__name__)
+ENABLE_LLM_REPAIR = False  # temporary: disabled per debugging request
+
+
+def _ocr_config() -> tuple[str, str, str]:
+    """Return (api_url, api_key, model) for OCR, falling back to LLM config."""
+    url = settings.OCR_API_URL or settings.LLM_API_URL
+    key = settings.OCR_API_KEY or settings.LLM_API_KEY
+    model = settings.OCR_MODEL or settings.LLM_MODEL
+    return url, key, model
+
+
+def _llm_config() -> tuple[str, str, str]:
+    """Return (api_url, api_key, model) for text-only LLM repair."""
+    return settings.LLM_API_URL, settings.LLM_API_KEY, settings.LLM_MODEL
+
+
+def _ocr_available() -> bool:
+    url, key, model = _ocr_config()
+    return bool(url and key and model)
+
+
+def _llm_available() -> bool:
+    url, key, model = _llm_config()
+    return bool(url and key and model)
 
 
 # ── provider-agnostic interface ──────────────────────────────────────────
 
 async def classify_page(image_path: str) -> tuple[SourceApp, PageType]:
     """Identify the source app and page type of a screenshot."""
-    if settings.LLM_API_KEY and settings.LLM_API_URL:
+    if _ocr_available():
         return await _classify_via_api(image_path)
     return _classify_mock(image_path)
 
 
-async def extract_transaction_fields(image_path: str, source_app: SourceApp, page_type: PageType) -> dict:
+async def extract_transaction_fields(
+    image_path: str, source_app: SourceApp, page_type: PageType
+) -> tuple[dict | list, str]:
     """Extract structured transaction fields from a screenshot."""
-    if settings.LLM_API_KEY and settings.LLM_API_URL:
+    if _ocr_available():
         return await _extract_via_api(image_path, source_app, page_type)
-    return _extract_mock(image_path, source_app, page_type)
+    mock_data = _extract_mock(image_path, source_app, page_type)
+    return mock_data, json.dumps(mock_data, ensure_ascii=False)
 
 
-# ── real API implementation ──────────────────────────────────────────────
+# ── OpenAI-compatible API implementation ─────────────────────────────────
+
+async def _call_vision(prompt: str, image_b64: str, max_tokens: int = 2000) -> str:
+    """Send a vision request to the OCR model endpoint."""
+    url, key, model = _ocr_config()
+    async with httpx.AsyncClient(timeout=60) as client:
+        resp = await client.post(
+            url,
+            headers={"Authorization": f"Bearer {key}"},
+            json={
+                "model": model,
+                "messages": [
+                    {
+                        "role": "user",
+                        "content": [
+                            {"type": "text", "text": prompt},
+                            {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{image_b64}"}},
+                        ],
+                    }
+                ],
+                "max_tokens": max_tokens,
+                "temperature": 0,
+            },
+        )
+        resp.raise_for_status()
+        body = resp.json()
+        choice0 = (body.get("choices") or [{}])[0]
+        msg = choice0.get("message") or {}
+        content = msg.get("content", "")
+        finish_reason = choice0.get("finish_reason")
+        usage = body.get("usage") or {}
+        return content if isinstance(content, str) else str(content)
+
+
+async def _call_text_llm(prompt: str, max_tokens: int = 1200) -> str:
+    """Send a text-only request to the reasoning LLM endpoint."""
+    url, key, model = _llm_config()
+    async with httpx.AsyncClient(timeout=60) as client:
+        resp = await client.post(
+            url,
+            headers={"Authorization": f"Bearer {key}"},
+            json={
+                "model": model,
+                "messages": [{"role": "user", "content": prompt}],
+                "max_tokens": max_tokens,
+            },
+        )
+        resp.raise_for_status()
+        return resp.json()["choices"][0]["message"]["content"]
+
+
+def _parse_json_response(text: str):
+    """Strip markdown fences and parse JSON from LLM output."""
+    cleaned = text.strip()
+    if cleaned.startswith("```"):
+        cleaned = cleaned.split("\n", 1)[-1]
+    if cleaned.endswith("```"):
+        cleaned = cleaned.rsplit("```", 1)[0]
+    cleaned = cleaned.strip().removeprefix("json").strip()
+    # Try parsing the full body first.
+    try:
+        return json.loads(cleaned)
+    except Exception:
+        pass
+
+    # Extract likely JSON block when model wraps with extra text.
+    match = re.search(r"(\{[\s\S]*\}|\[[\s\S]*\])", cleaned)
+    candidate = match.group(1) if match else cleaned
+
+    # Normalize common model output issues:
+    # - smart quotes
+    # - trailing commas
+    normalized = (
+        candidate
+        .replace("“", "\"")
+        .replace("”", "\"")
+        .replace("‘", "'")
+        .replace("’", "'")
+    )
+    normalized = re.sub(r",\s*([}\]])", r"\1", normalized)
+    try:
+        return json.loads(normalized)
+    except Exception:
+        pass
+
+    # Last fallback: Python-like literal payloads with single quotes / True/False/None.
+    py_like = re.sub(r"\btrue\b", "True", normalized, flags=re.IGNORECASE)
+    py_like = re.sub(r"\bfalse\b", "False", py_like, flags=re.IGNORECASE)
+    py_like = re.sub(r"\bnull\b", "None", py_like, flags=re.IGNORECASE)
+    parsed = ast.literal_eval(py_like)
+    if isinstance(parsed, (dict, list)):
+        return parsed
+    raise ValueError("Parsed payload is neither dict nor list")
+
+
+async def _repair_broken_json_with_llm(broken_text: str) -> str:
+    """Use LLM_MODEL to repair malformed OCR JSON text without adding new semantics."""
+    prompt = (
+        "你是JSON修复器。下面是OCR模型返回的可能损坏JSON文本。\n"
+        "任务：仅修复语法并输出一个可被 json.loads 直接解析的JSON。\n"
+        "硬性要求：\n"
+        "1) 只能输出JSON，不要Markdown，不要解释。\n"
+        "2) 保留原有字段和语义，不新增未出现的信息。\n"
+        "3) 若出现截断，允许最小闭合修复（补齐缺失括号/引号）。\n"
+        "4) 结果必须是对象或数组。\n\n"
+        "损坏文本如下：\n"
+        f"{broken_text}"
+    )
+    return await _call_text_llm(prompt, max_tokens=1200)
+
 
 async def _classify_via_api(image_path: str) -> tuple[SourceApp, PageType]:
-    import base64
     full_path = settings.upload_path / image_path
     if not full_path.exists():
         return SourceApp.other, PageType.unknown
@@ -43,79 +186,64 @@ async def _classify_via_api(image_path: str) -> tuple[SourceApp, PageType]:
     image_b64 = base64.b64encode(full_path.read_bytes()).decode()
     prompt = (
         "请分析这张手机截图，判断它来自哪个APP（wechat/alipay/bank/digital_wallet/other）"
-        "以及页面类型（bill_list/bill_detail/transfer_receipt/sms_notice/balance/unknown）。"
+        "以及页面类型（bill_list/bill_detail/transfer_receipt/sms_notice/balance/unknown）。\n"
-        "只返回JSON: {\"source_app\": \"...\", \"page_type\": \"...\"}"
+        '只返回JSON，格式: {"source_app": "...", "page_type": "..."}'
     )
 
     try:
-        async with httpx.AsyncClient(timeout=30) as client:
+        text = await _call_vision(prompt, image_b64, max_tokens=600)
-            resp = await client.post(
+        data = _parse_json_response(text)
-                settings.LLM_API_URL,
-                headers={"Authorization": f"Bearer {settings.LLM_API_KEY}"},
-                json={
-                    "model": settings.LLM_MODEL,
-                    "messages": [
-                        {
-                            "role": "user",
-                            "content": [
-                                {"type": "text", "text": prompt},
-                                {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{image_b64}"}},
-                            ],
-                        }
-                    ],
-                    "max_tokens": 200,
-                },
-            )
-            resp.raise_for_status()
-            text = resp.json()["choices"][0]["message"]["content"]
-            data = json.loads(text.strip().strip("`").removeprefix("json").strip())
         return SourceApp(data.get("source_app", "other")), PageType(data.get("page_type", "unknown"))
     except Exception as e:
         logger.warning("classify_page API failed: %s", e)
         return SourceApp.other, PageType.unknown
 
 
-async def _extract_via_api(image_path: str, source_app: SourceApp, page_type: PageType) -> dict:
+async def _extract_via_api(
-    import base64
+    image_path: str, source_app: SourceApp, page_type: PageType
+) -> tuple[dict | list, str]:
     full_path = settings.upload_path / image_path
     if not full_path.exists():
-        return {}
+        return {}, ""
 
     image_b64 = base64.b64encode(full_path.read_bytes()).decode()
     prompt = (
-        f"这是一张来自{source_app.value}的{page_type.value}截图。"
+        "你是账单OCR结构化引擎。你的输出会被程序直接 json.loads 解析，任何非JSON字符都会导致任务失败。\n"
-        "请提取其中的交易信息，返回JSON格式，字段包括："
+        f"输入图片来自 {source_app.value} / {page_type.value}。\n\n"
-        "trade_time(交易时间,格式YYYY-MM-DD HH:MM:SS), amount(金额,数字), "
+        "【硬性输出规则】\n"
-        "direction(in或out), counterparty_name(对方名称), counterparty_account(对方账号), "
+        "1) 只能输出一个合法JSON值：对象或数组；禁止Markdown代码块、禁止解释文字、禁止注释。\n"
-        "self_account_tail_no(本方账户尾号), order_no(订单号), remark(备注), confidence(0-1)。"
+        "2) 键名必须使用英文双引号；字符串值必须使用英文双引号；禁止单引号、禁止中文引号。\n"
-        "如果截图包含多笔交易，返回JSON数组。否则返回单个JSON对象。"
+        "3) 禁止尾逗号，禁止 NaN/Infinity/undefined。\n"
+        "4) 若只有1笔交易，输出对象；若>=2笔交易，输出数组。\n"
+        "5) 每笔交易字段固定为：\n"
+        '   {"trade_time":"YYYY-MM-DD HH:MM:SS","amount":123.45,"direction":"in|out",'
+        '"counterparty_name":"","counterparty_account":"","self_account_tail_no":"","order_no":"","remark":"","confidence":0.0}\n'
+        "6) 字段约束：\n"
+        "- trade_time: 必须为 YYYY-MM-DD HH:MM:SS；无法识别时填空字符串\"\"。\n"
+        "- amount: 必须是数字（不要货币符号）；无法识别时填 0。\n"
+        "- direction: 仅允许 in 或 out；无法判断默认 out。\n"
+        "- confidence: 0~1 的数字；无法判断默认 0.5。\n"
+        "- 其他文本字段无法识别时填空字符串\"\"。\n\n"
+        "【输出前自检】\n"
+        "- 检查是否是严格JSON（可被 json.loads 直接解析）。\n"
+        "- 检查每条记录都包含全部9个字段。\n"
+        "- 检查 amount/confidence 为数字类型，direction 仅为 in/out。\n"
+        "现在只输出最终JSON，不要输出任何额外文本。"
     )
 
     try:
-        async with httpx.AsyncClient(timeout=60) as client:
+        text = await _call_vision(prompt, image_b64, max_tokens=6000)
-            resp = await client.post(
+        try:
-                settings.LLM_API_URL,
+            parsed = _parse_json_response(text)
-                headers={"Authorization": f"Bearer {settings.LLM_API_KEY}"},
+        except Exception as parse_err:
-                json={
+            if not ENABLE_LLM_REPAIR or not _llm_available():
-                    "model": settings.LLM_MODEL,
+                raise
-                    "messages": [
+            repaired_text = await _repair_broken_json_with_llm(text)
-                        {
+            parsed = _parse_json_response(repaired_text)
-                            "role": "user",
+        return parsed, text
-                            "content": [
-                                {"type": "text", "text": prompt},
-                                {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{image_b64}"}},
-                            ],
-                        }
-                    ],
-                    "max_tokens": 2000,
-                },
-            )
-            resp.raise_for_status()
-            text = resp.json()["choices"][0]["message"]["content"]
-            return json.loads(text.strip().strip("`").removeprefix("json").strip())
     except Exception as e:
         logger.warning("extract_transaction_fields API failed: %s", e)
-        return {}
+        return {}, text if "text" in locals() else ""
 
 
 # ── mock fallback ────────────────────────────────────────────────────────

backend/app/services/parser_service.py

@@ -16,31 +16,42 @@ def parse_extracted_fields(
     items = raw if isinstance(raw, list) else [raw]
     records: list[TransactionRecord] = []
 
-    for item in items:
+    for idx, item in enumerate(items):
         if not item or not item.get("amount"):
             continue
 
+        raw_trade_time = item.get("trade_time")
         try:
-            trade_time = datetime.fromisoformat(item["trade_time"])
+            if isinstance(raw_trade_time, datetime):
-        except (ValueError, KeyError):
+                trade_time = raw_trade_time
+            elif isinstance(raw_trade_time, str):
+                trade_time = datetime.fromisoformat(raw_trade_time)
+            else:
+                raise TypeError("trade_time is not str/datetime")
+        except (ValueError, KeyError, TypeError):
             trade_time = datetime.now()
 
         direction_str = item.get("direction", "out")
         direction = Direction.in_ if direction_str == "in" else Direction.out
+        try:
+            amount = float(item.get("amount", 0))
+            confidence = float(item.get("confidence", 0.5))
+        except Exception:
+            raise
 
         record = TransactionRecord(
             case_id=case_id,
             evidence_image_id=image_id,
             source_app=source_app,
             trade_time=trade_time,
-            amount=float(item.get("amount", 0)),
+            amount=amount,
             direction=direction,
             counterparty_name=str(item.get("counterparty_name", "")),
             counterparty_account=str(item.get("counterparty_account", "")),
             self_account_tail_no=str(item.get("self_account_tail_no", "")),
             order_no=str(item.get("order_no", "")),
             remark=str(item.get("remark", "")),
-            confidence=float(item.get("confidence", 0.5)),
+            confidence=confidence,
         )
         records.append(record)
 

backend/app/workers/ocr_tasks.py

@@ -1,6 +1,7 @@
 """Celery tasks for OCR processing of uploaded screenshots."""
 import asyncio
 import logging
+import json
 from uuid import UUID
 
 from app.workers.celery_app import celery_app
@@ -20,18 +21,19 @@ def _run_async(coro):
 @celery_app.task(name="app.workers.ocr_tasks.process_image_ocr", bind=True, max_retries=3)
 def process_image_ocr(self, image_id: str):
     """Process a single image: classify page, extract fields, save to DB."""
-    _run_async(_process(image_id))
+    _run_async(process_image_ocr_async(image_id))
 
 
-async def _process(image_id_str: str):
+async def process_image_ocr_async(image_id_str: str):
     from app.core.database import async_session_factory
+    from sqlalchemy import delete
     from app.models.evidence_image import EvidenceImage, OcrStatus
     from app.models.ocr_block import OcrBlock
+    from app.models.transaction import TransactionRecord
     from app.services.ocr_service import classify_page, extract_transaction_fields
     from app.services.parser_service import parse_extracted_fields
 
     image_id = UUID(image_id_str)
-
     async with async_session_factory() as db:
         image = await db.get(EvidenceImage, image_id)
         if not image:
@@ -40,18 +42,34 @@ async def _process(image_id_str: str):
 
         image.ocr_status = OcrStatus.processing
         await db.flush()
+        # Re-run OCR for this image should replace old OCR blocks/records.
+        await db.execute(delete(OcrBlock).where(OcrBlock.image_id == image.id))
+        await db.execute(delete(TransactionRecord).where(TransactionRecord.evidence_image_id == image.id))
 
         try:
             source_app, page_type = await classify_page(image.file_path)
             image.source_app = source_app
             image.page_type = page_type
 
-            raw_fields = await extract_transaction_fields(image.file_path, source_app, page_type)
+            raw_fields, raw_ocr_text = await extract_transaction_fields(image.file_path, source_app, page_type)
 
-            # save raw OCR block
+            is_empty_extract = raw_fields is None or raw_fields == {} or raw_fields == []
+
+            # save raw OCR block (direct model output for debugging)
+            raw_block_content = (raw_ocr_text or "").strip()
+            if raw_block_content:
                 block = OcrBlock(
                     image_id=image.id,
-                content=str(raw_fields),
+                    content=raw_block_content,
+                    bbox={},
+                    seq_order=0,
+                    confidence=raw_fields.get("confidence", 0.5) if isinstance(raw_fields, dict) else 0.5,
+                )
+                db.add(block)
+            elif not is_empty_extract:
+                block = OcrBlock(
+                    image_id=image.id,
+                    content=json.dumps(raw_fields, ensure_ascii=False),
                     bbox={},
                     seq_order=0,
                     confidence=raw_fields.get("confidence", 0.5) if isinstance(raw_fields, dict) else 0.5,

backend/requirements.txt

@@ -0,0 +1,18 @@
+fastapi==0.135.1
+uvicorn[standard]==0.41.0
+sqlalchemy[asyncio]==2.0.48
+asyncpg==0.31.0
+alembic==1.18.4
+celery[redis]==5.6.2
+redis==6.4.0
+pydantic-settings>=2.0.0
+python-multipart==0.0.22
+Pillow==12.1.0
+httpx==0.28.1
+openpyxl==3.1.5
+python-docx==1.2.0
+psycopg2-binary==2.9.11
+
+# dev dependencies
+pytest==9.0.2
+pytest-asyncio>=0.24.0