feat(klausur-service): Add Tesseract OCR, DSFA RAG, TrOCR, grid detection and vocab session store

New modules:
- tesseract_vocab_extractor.py: Bounding-box OCR with multi-PSM pipeline
- grid_detection_service.py: CV-based grid/table detection for worksheets
- vocab_session_store.py: PostgreSQL persistence for vocab sessions
- trocr_api.py: TrOCR handwriting recognition endpoint
- dsfa_rag_api.py + dsfa_corpus_ingestion.py: DSFA RAG corpus search

Changes:
- Dockerfile: Install tesseract-ocr + deu/eng language packs
- requirements.txt: Add PyMuPDF, pytesseract, Pillow
- main.py: Register new routers, init DB pools + Qdrant collections

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

klausur-service/Dockerfile

@@ -13,9 +13,12 @@ FROM python:3.11-slim
 
 WORKDIR /app
 
-# Install system dependencies
+# Install system dependencies (incl. Tesseract OCR for bounding-box extraction)
 RUN apt-get update && apt-get install -y --no-install-recommends \
     curl \
+    tesseract-ocr \
+    tesseract-ocr-deu \
+    tesseract-ocr-eng \
     && rm -rf /var/lib/apt/lists/*
 
 # Install Python dependencies

klausur-service/backend/dsfa_rag_api.py

@@ -0,0 +1,715 @@
+"""
+DSFA RAG API Endpoints.
+
+Provides REST API for searching DSFA corpus with full source attribution.
+
+Endpoints:
+- GET  /api/v1/dsfa-rag/search         - Semantic search with attribution
+- GET  /api/v1/dsfa-rag/sources        - List all registered sources
+- POST /api/v1/dsfa-rag/sources/{code}/ingest - Trigger source ingestion
+- GET  /api/v1/dsfa-rag/chunks/{id}    - Get single chunk with attribution
+- GET  /api/v1/dsfa-rag/stats          - Get corpus statistics
+"""
+
+import os
+import uuid
+import logging
+from typing import List, Optional
+from dataclasses import dataclass, asdict
+
+import httpx
+from fastapi import APIRouter, HTTPException, Query, Depends
+from pydantic import BaseModel, Field
+
+logger = logging.getLogger(__name__)
+
+# Embedding service configuration
+EMBEDDING_SERVICE_URL = os.getenv("EMBEDDING_SERVICE_URL", "http://172.18.0.13:8087")
+
+# Import from ingestion module
+from dsfa_corpus_ingestion import (
+    DSFACorpusStore,
+    DSFAQdrantService,
+    DSFASearchResult,
+    LICENSE_REGISTRY,
+    DSFA_SOURCES,
+    generate_attribution_notice,
+    get_license_label,
+    DSFA_COLLECTION,
+    chunk_document
+)
+
+router = APIRouter(prefix="/api/v1/dsfa-rag", tags=["DSFA RAG"])
+
+
+# =============================================================================
+# Pydantic Models
+# =============================================================================
+
+class DSFASourceResponse(BaseModel):
+    """Response model for DSFA source."""
+    id: str
+    source_code: str
+    name: str
+    full_name: Optional[str] = None
+    organization: Optional[str] = None
+    source_url: Optional[str] = None
+    license_code: str
+    license_name: str
+    license_url: Optional[str] = None
+    attribution_required: bool
+    attribution_text: str
+    document_type: Optional[str] = None
+    language: str = "de"
+
+
+class DSFAChunkResponse(BaseModel):
+    """Response model for a single chunk with attribution."""
+    chunk_id: str
+    content: str
+    section_title: Optional[str] = None
+    page_number: Optional[int] = None
+    category: Optional[str] = None
+
+    # Document info
+    document_id: str
+    document_title: Optional[str] = None
+
+    # Attribution (always included)
+    source_id: str
+    source_code: str
+    source_name: str
+    attribution_text: str
+    license_code: str
+    license_name: str
+    license_url: Optional[str] = None
+    attribution_required: bool
+    source_url: Optional[str] = None
+    document_type: Optional[str] = None
+
+
+class DSFASearchResultResponse(BaseModel):
+    """Response model for search result."""
+    chunk_id: str
+    content: str
+    score: float
+
+    # Attribution
+    source_code: str
+    source_name: str
+    attribution_text: str
+    license_code: str
+    license_name: str
+    license_url: Optional[str] = None
+    attribution_required: bool
+    source_url: Optional[str] = None
+
+    # Metadata
+    document_type: Optional[str] = None
+    category: Optional[str] = None
+    section_title: Optional[str] = None
+    page_number: Optional[int] = None
+
+
+class DSFASearchResponse(BaseModel):
+    """Response model for search endpoint."""
+    query: str
+    results: List[DSFASearchResultResponse]
+    total_results: int
+
+    # Aggregated licenses for footer
+    licenses_used: List[str]
+    attribution_notice: str
+
+
+class DSFASourceStatsResponse(BaseModel):
+    """Response model for source statistics."""
+    source_id: str
+    source_code: str
+    name: str
+    organization: Optional[str] = None
+    license_code: str
+    document_type: Optional[str] = None
+    document_count: int
+    chunk_count: int
+    last_indexed_at: Optional[str] = None
+
+
+class DSFACorpusStatsResponse(BaseModel):
+    """Response model for corpus statistics."""
+    sources: List[DSFASourceStatsResponse]
+    total_sources: int
+    total_documents: int
+    total_chunks: int
+    qdrant_collection: str
+    qdrant_points_count: int
+    qdrant_status: str
+
+
+class IngestRequest(BaseModel):
+    """Request model for ingestion."""
+    document_url: Optional[str] = None
+    document_text: Optional[str] = None
+    title: Optional[str] = None
+
+
+class IngestResponse(BaseModel):
+    """Response model for ingestion."""
+    source_code: str
+    document_id: Optional[str] = None
+    chunks_created: int
+    message: str
+
+
+class LicenseInfo(BaseModel):
+    """License information."""
+    code: str
+    name: str
+    url: Optional[str] = None
+    attribution_required: bool
+    modification_allowed: bool
+    commercial_use: bool
+
+
+# =============================================================================
+# Dependency Injection
+# =============================================================================
+
+# Database pool (will be set from main.py)
+_db_pool = None
+
+
+def set_db_pool(pool):
+    """Set the database pool for API endpoints."""
+    global _db_pool
+    _db_pool = pool
+
+
+async def get_store() -> DSFACorpusStore:
+    """Get DSFA corpus store."""
+    if _db_pool is None:
+        raise HTTPException(status_code=503, detail="Database not initialized")
+    return DSFACorpusStore(_db_pool)
+
+
+async def get_qdrant() -> DSFAQdrantService:
+    """Get Qdrant service."""
+    return DSFAQdrantService()
+
+
+# =============================================================================
+# Embedding Service Integration
+# =============================================================================
+
+async def get_embedding(text: str) -> List[float]:
+    """
+    Get embedding for text using the embedding-service.
+
+    Uses BGE-M3 model which produces 1024-dimensional vectors.
+    """
+    async with httpx.AsyncClient(timeout=60.0) as client:
+        try:
+            response = await client.post(
+                f"{EMBEDDING_SERVICE_URL}/embed-single",
+                json={"text": text}
+            )
+            response.raise_for_status()
+            data = response.json()
+            return data.get("embedding", [])
+        except httpx.HTTPError as e:
+            logger.error(f"Embedding service error: {e}")
+            # Fallback to hash-based pseudo-embedding for development
+            return _generate_fallback_embedding(text)
+
+
+async def get_embeddings_batch(texts: List[str]) -> List[List[float]]:
+    """
+    Get embeddings for multiple texts in batch.
+    """
+    async with httpx.AsyncClient(timeout=120.0) as client:
+        try:
+            response = await client.post(
+                f"{EMBEDDING_SERVICE_URL}/embed",
+                json={"texts": texts}
+            )
+            response.raise_for_status()
+            data = response.json()
+            return data.get("embeddings", [])
+        except httpx.HTTPError as e:
+            logger.error(f"Embedding service batch error: {e}")
+            # Fallback
+            return [_generate_fallback_embedding(t) for t in texts]
+
+
+async def extract_text_from_url(url: str) -> str:
+    """
+    Extract text from a document URL (PDF, HTML, etc.).
+    """
+    async with httpx.AsyncClient(timeout=120.0) as client:
+        try:
+            # First try to use the embedding-service's extract-pdf endpoint
+            response = await client.post(
+                f"{EMBEDDING_SERVICE_URL}/extract-pdf",
+                json={"url": url}
+            )
+            response.raise_for_status()
+            data = response.json()
+            return data.get("text", "")
+        except httpx.HTTPError as e:
+            logger.error(f"PDF extraction error for {url}: {e}")
+            # Fallback: try to fetch HTML content directly
+            try:
+                response = await client.get(url, follow_redirects=True)
+                response.raise_for_status()
+                content_type = response.headers.get("content-type", "")
+                if "html" in content_type:
+                    # Simple HTML text extraction
+                    import re
+                    html = response.text
+                    # Remove scripts and styles
+                    html = re.sub(r'<script[^>]*>.*?</script>', '', html, flags=re.DOTALL | re.IGNORECASE)
+                    html = re.sub(r'<style[^>]*>.*?</style>', '', html, flags=re.DOTALL | re.IGNORECASE)
+                    # Remove tags
+                    text = re.sub(r'<[^>]+>', ' ', html)
+                    # Clean whitespace
+                    text = re.sub(r'\s+', ' ', text).strip()
+                    return text
+                else:
+                    return ""
+            except Exception as fetch_err:
+                logger.error(f"Fallback fetch error for {url}: {fetch_err}")
+                return ""
+
+
+def _generate_fallback_embedding(text: str) -> List[float]:
+    """
+    Generate deterministic pseudo-embedding from text hash.
+    Used as fallback when embedding service is unavailable.
+    """
+    import hashlib
+    import struct
+
+    hash_bytes = hashlib.sha256(text.encode()).digest()
+    embedding = []
+    for i in range(0, min(len(hash_bytes), 128), 4):
+        val = struct.unpack('f', hash_bytes[i:i+4])[0]
+        embedding.append(val % 1.0)
+
+    # Pad to 1024 dimensions
+    while len(embedding) < 1024:
+        embedding.extend(embedding[:min(len(embedding), 1024 - len(embedding))])
+
+    return embedding[:1024]
+
+
+# =============================================================================
+# API Endpoints
+# =============================================================================
+
+@router.get("/search", response_model=DSFASearchResponse)
+async def search_dsfa_corpus(
+    query: str = Query(..., min_length=3, description="Search query"),
+    source_codes: Optional[List[str]] = Query(None, description="Filter by source codes"),
+    document_types: Optional[List[str]] = Query(None, description="Filter by document types (guideline, checklist, regulation)"),
+    categories: Optional[List[str]] = Query(None, description="Filter by categories (threshold_analysis, risk_assessment, mitigation)"),
+    limit: int = Query(10, ge=1, le=50, description="Maximum results"),
+    include_attribution: bool = Query(True, description="Include attribution in results"),
+    store: DSFACorpusStore = Depends(get_store),
+    qdrant: DSFAQdrantService = Depends(get_qdrant)
+):
+    """
+    Search DSFA corpus with full attribution.
+
+    Returns matching chunks with source/license information for compliance.
+    """
+    # Get query embedding
+    query_embedding = await get_embedding(query)
+
+    # Search Qdrant
+    raw_results = await qdrant.search(
+        query_embedding=query_embedding,
+        source_codes=source_codes,
+        document_types=document_types,
+        categories=categories,
+        limit=limit
+    )
+
+    # Transform results
+    results = []
+    licenses_used = set()
+
+    for r in raw_results:
+        license_code = r.get("license_code", "")
+        license_info = LICENSE_REGISTRY.get(license_code, {})
+
+        result = DSFASearchResultResponse(
+            chunk_id=r.get("chunk_id", ""),
+            content=r.get("content", ""),
+            score=r.get("score", 0.0),
+            source_code=r.get("source_code", ""),
+            source_name=r.get("source_name", ""),
+            attribution_text=r.get("attribution_text", ""),
+            license_code=license_code,
+            license_name=license_info.get("name", license_code),
+            license_url=license_info.get("url"),
+            attribution_required=r.get("attribution_required", True),
+            source_url=r.get("source_url"),
+            document_type=r.get("document_type"),
+            category=r.get("category"),
+            section_title=r.get("section_title"),
+            page_number=r.get("page_number")
+        )
+        results.append(result)
+        licenses_used.add(license_code)
+
+    # Generate attribution notice
+    search_results = [
+        DSFASearchResult(
+            chunk_id=r.chunk_id,
+            content=r.content,
+            score=r.score,
+            source_code=r.source_code,
+            source_name=r.source_name,
+            attribution_text=r.attribution_text,
+            license_code=r.license_code,
+            license_url=r.license_url,
+            attribution_required=r.attribution_required,
+            source_url=r.source_url,
+            document_type=r.document_type or "",
+            category=r.category or "",
+            section_title=r.section_title,
+            page_number=r.page_number
+        )
+        for r in results
+    ]
+    attribution_notice = generate_attribution_notice(search_results) if include_attribution else ""
+
+    return DSFASearchResponse(
+        query=query,
+        results=results,
+        total_results=len(results),
+        licenses_used=list(licenses_used),
+        attribution_notice=attribution_notice
+    )
+
+
+@router.get("/sources", response_model=List[DSFASourceResponse])
+async def list_dsfa_sources(
+    document_type: Optional[str] = Query(None, description="Filter by document type"),
+    license_code: Optional[str] = Query(None, description="Filter by license"),
+    store: DSFACorpusStore = Depends(get_store)
+):
+    """List all registered DSFA sources with license info."""
+    sources = await store.list_sources()
+
+    result = []
+    for s in sources:
+        # Apply filters
+        if document_type and s.get("document_type") != document_type:
+            continue
+        if license_code and s.get("license_code") != license_code:
+            continue
+
+        license_info = LICENSE_REGISTRY.get(s.get("license_code", ""), {})
+
+        result.append(DSFASourceResponse(
+            id=str(s["id"]),
+            source_code=s["source_code"],
+            name=s["name"],
+            full_name=s.get("full_name"),
+            organization=s.get("organization"),
+            source_url=s.get("source_url"),
+            license_code=s.get("license_code", ""),
+            license_name=license_info.get("name", s.get("license_code", "")),
+            license_url=license_info.get("url"),
+            attribution_required=s.get("attribution_required", True),
+            attribution_text=s.get("attribution_text", ""),
+            document_type=s.get("document_type"),
+            language=s.get("language", "de")
+        ))
+
+    return result
+
+
+@router.get("/sources/available")
+async def list_available_sources():
+    """List all available source definitions (from DSFA_SOURCES constant)."""
+    return [
+        {
+            "source_code": s["source_code"],
+            "name": s["name"],
+            "organization": s.get("organization"),
+            "license_code": s["license_code"],
+            "document_type": s.get("document_type")
+        }
+        for s in DSFA_SOURCES
+    ]
+
+
+@router.get("/sources/{source_code}", response_model=DSFASourceResponse)
+async def get_dsfa_source(
+    source_code: str,
+    store: DSFACorpusStore = Depends(get_store)
+):
+    """Get details for a specific source."""
+    source = await store.get_source_by_code(source_code)
+
+    if not source:
+        raise HTTPException(status_code=404, detail=f"Source not found: {source_code}")
+
+    license_info = LICENSE_REGISTRY.get(source.get("license_code", ""), {})
+
+    return DSFASourceResponse(
+        id=str(source["id"]),
+        source_code=source["source_code"],
+        name=source["name"],
+        full_name=source.get("full_name"),
+        organization=source.get("organization"),
+        source_url=source.get("source_url"),
+        license_code=source.get("license_code", ""),
+        license_name=license_info.get("name", source.get("license_code", "")),
+        license_url=license_info.get("url"),
+        attribution_required=source.get("attribution_required", True),
+        attribution_text=source.get("attribution_text", ""),
+        document_type=source.get("document_type"),
+        language=source.get("language", "de")
+    )
+
+
+@router.post("/sources/{source_code}/ingest", response_model=IngestResponse)
+async def ingest_dsfa_source(
+    source_code: str,
+    request: IngestRequest,
+    store: DSFACorpusStore = Depends(get_store),
+    qdrant: DSFAQdrantService = Depends(get_qdrant)
+):
+    """
+    Trigger ingestion for a specific source.
+
+    Can provide document via URL or direct text.
+    """
+    # Get source
+    source = await store.get_source_by_code(source_code)
+    if not source:
+        raise HTTPException(status_code=404, detail=f"Source not found: {source_code}")
+
+    # Need either URL or text
+    if not request.document_text and not request.document_url:
+        raise HTTPException(
+            status_code=400,
+            detail="Either document_text or document_url must be provided"
+        )
+
+    # Ensure Qdrant collection exists
+    await qdrant.ensure_collection()
+
+    # Get text content
+    text_content = request.document_text
+    if request.document_url and not text_content:
+        # Download and extract text from URL
+        logger.info(f"Extracting text from URL: {request.document_url}")
+        text_content = await extract_text_from_url(request.document_url)
+        if not text_content:
+            raise HTTPException(
+                status_code=400,
+                detail=f"Could not extract text from URL: {request.document_url}"
+            )
+
+    if not text_content or len(text_content.strip()) < 50:
+        raise HTTPException(status_code=400, detail="Document text too short (min 50 chars)")
+
+    # Create document record
+    doc_title = request.title or f"Document for {source_code}"
+    document_id = await store.create_document(
+        source_id=str(source["id"]),
+        title=doc_title,
+        file_type="text",
+        metadata={"ingested_via": "api", "source_code": source_code}
+    )
+
+    # Chunk the document
+    chunks = chunk_document(text_content, source_code)
+
+    if not chunks:
+        return IngestResponse(
+            source_code=source_code,
+            document_id=document_id,
+            chunks_created=0,
+            message="Document created but no chunks generated"
+        )
+
+    # Generate embeddings in batch for efficiency
+    chunk_texts = [chunk["content"] for chunk in chunks]
+    logger.info(f"Generating embeddings for {len(chunk_texts)} chunks...")
+    embeddings = await get_embeddings_batch(chunk_texts)
+
+    # Create chunk records in PostgreSQL and prepare for Qdrant
+    chunk_records = []
+    for i, (chunk, embedding) in enumerate(zip(chunks, embeddings)):
+        # Create chunk in PostgreSQL
+        chunk_id = await store.create_chunk(
+            document_id=document_id,
+            source_id=str(source["id"]),
+            content=chunk["content"],
+            chunk_index=i,
+            section_title=chunk.get("section_title"),
+            page_number=chunk.get("page_number"),
+            category=chunk.get("category")
+        )
+
+        chunk_records.append({
+            "chunk_id": chunk_id,
+            "document_id": document_id,
+            "source_id": str(source["id"]),
+            "content": chunk["content"],
+            "section_title": chunk.get("section_title"),
+            "source_code": source_code,
+            "source_name": source["name"],
+            "attribution_text": source["attribution_text"],
+            "license_code": source["license_code"],
+            "attribution_required": source.get("attribution_required", True),
+            "document_type": source.get("document_type", ""),
+            "category": chunk.get("category", ""),
+            "language": source.get("language", "de"),
+            "page_number": chunk.get("page_number")
+        })
+
+    # Index in Qdrant
+    indexed_count = await qdrant.index_chunks(chunk_records, embeddings)
+
+    # Update document record
+    await store.update_document_indexed(document_id, len(chunks))
+
+    return IngestResponse(
+        source_code=source_code,
+        document_id=document_id,
+        chunks_created=indexed_count,
+        message=f"Successfully ingested {indexed_count} chunks from document"
+    )
+
+
+@router.get("/chunks/{chunk_id}", response_model=DSFAChunkResponse)
+async def get_chunk_with_attribution(
+    chunk_id: str,
+    store: DSFACorpusStore = Depends(get_store)
+):
+    """Get single chunk with full source attribution."""
+    chunk = await store.get_chunk_with_attribution(chunk_id)
+
+    if not chunk:
+        raise HTTPException(status_code=404, detail=f"Chunk not found: {chunk_id}")
+
+    license_info = LICENSE_REGISTRY.get(chunk.get("license_code", ""), {})
+
+    return DSFAChunkResponse(
+        chunk_id=str(chunk["chunk_id"]),
+        content=chunk.get("content", ""),
+        section_title=chunk.get("section_title"),
+        page_number=chunk.get("page_number"),
+        category=chunk.get("category"),
+        document_id=str(chunk.get("document_id", "")),
+        document_title=chunk.get("document_title"),
+        source_id=str(chunk.get("source_id", "")),
+        source_code=chunk.get("source_code", ""),
+        source_name=chunk.get("source_name", ""),
+        attribution_text=chunk.get("attribution_text", ""),
+        license_code=chunk.get("license_code", ""),
+        license_name=license_info.get("name", chunk.get("license_code", "")),
+        license_url=license_info.get("url"),
+        attribution_required=chunk.get("attribution_required", True),
+        source_url=chunk.get("source_url"),
+        document_type=chunk.get("document_type")
+    )
+
+
+@router.get("/stats", response_model=DSFACorpusStatsResponse)
+async def get_corpus_stats(
+    store: DSFACorpusStore = Depends(get_store),
+    qdrant: DSFAQdrantService = Depends(get_qdrant)
+):
+    """Get corpus statistics for dashboard."""
+    # Get PostgreSQL stats
+    source_stats = await store.get_source_stats()
+
+    total_docs = 0
+    total_chunks = 0
+    stats_response = []
+
+    for s in source_stats:
+        doc_count = s.get("document_count", 0) or 0
+        chunk_count = s.get("chunk_count", 0) or 0
+        total_docs += doc_count
+        total_chunks += chunk_count
+
+        last_indexed = s.get("last_indexed_at")
+
+        stats_response.append(DSFASourceStatsResponse(
+            source_id=str(s.get("source_id", "")),
+            source_code=s.get("source_code", ""),
+            name=s.get("name", ""),
+            organization=s.get("organization"),
+            license_code=s.get("license_code", ""),
+            document_type=s.get("document_type"),
+            document_count=doc_count,
+            chunk_count=chunk_count,
+            last_indexed_at=last_indexed.isoformat() if last_indexed else None
+        ))
+
+    # Get Qdrant stats
+    qdrant_stats = await qdrant.get_stats()
+
+    return DSFACorpusStatsResponse(
+        sources=stats_response,
+        total_sources=len(source_stats),
+        total_documents=total_docs,
+        total_chunks=total_chunks,
+        qdrant_collection=DSFA_COLLECTION,
+        qdrant_points_count=qdrant_stats.get("points_count", 0),
+        qdrant_status=qdrant_stats.get("status", "unknown")
+    )
+
+
+@router.get("/licenses")
+async def list_licenses():
+    """List all supported licenses with their terms."""
+    return [
+        LicenseInfo(
+            code=code,
+            name=info.get("name", code),
+            url=info.get("url"),
+            attribution_required=info.get("attribution_required", True),
+            modification_allowed=info.get("modification_allowed", True),
+            commercial_use=info.get("commercial_use", True)
+        )
+        for code, info in LICENSE_REGISTRY.items()
+    ]
+
+
+@router.post("/init")
+async def initialize_dsfa_corpus(
+    store: DSFACorpusStore = Depends(get_store),
+    qdrant: DSFAQdrantService = Depends(get_qdrant)
+):
+    """
+    Initialize DSFA corpus.
+
+    - Creates Qdrant collection
+    - Registers all predefined sources
+    """
+    # Ensure Qdrant collection exists
+    qdrant_ok = await qdrant.ensure_collection()
+
+    # Register all sources
+    registered = 0
+    for source in DSFA_SOURCES:
+        try:
+            await store.register_source(source)
+            registered += 1
+        except Exception as e:
+            print(f"Error registering source {source['source_code']}: {e}")
+
+    return {
+        "qdrant_collection_created": qdrant_ok,
+        "sources_registered": registered,
+        "total_sources": len(DSFA_SOURCES)
+    }

klausur-service/backend/main.py

@@ -20,6 +20,7 @@ This is the main entry point. All functionality is organized in modular packages
 import os
 from contextlib import asynccontextmanager
 
+import asyncpg
 from fastapi import FastAPI
 from fastapi.middleware.cors import CORSMiddleware
 from fastapi.staticfiles import StaticFiles
@@ -36,7 +37,19 @@ from admin_api import router as admin_router
 from zeugnis_api import router as zeugnis_router
 from training_api import router as training_router
 from mail.api import router as mail_router
+try:
     from trocr_api import router as trocr_router
+except ImportError:
+    trocr_router = None
+from vocab_worksheet_api import router as vocab_router, set_db_pool as set_vocab_db_pool, _init_vocab_table, _load_all_sessions, DATABASE_URL as VOCAB_DATABASE_URL
+try:
+    from dsfa_rag_api import router as dsfa_rag_router, set_db_pool as set_dsfa_db_pool
+    from dsfa_corpus_ingestion import DSFAQdrantService, DATABASE_URL as DSFA_DATABASE_URL
+except ImportError:
+    dsfa_rag_router = None
+    set_dsfa_db_pool = None
+    DSFAQdrantService = None
+    DSFA_DATABASE_URL = None
 
 # BYOEH Qdrant initialization
 from qdrant_service import init_qdrant_collection
@@ -51,12 +64,42 @@ async def lifespan(app: FastAPI):
     """Application lifespan manager for startup and shutdown events."""
     print("Klausur-Service starting...")
 
+    # Initialize database pool for Vocab Sessions
+    vocab_db_pool = None
+    try:
+        vocab_db_pool = await asyncpg.create_pool(VOCAB_DATABASE_URL, min_size=2, max_size=5)
+        set_vocab_db_pool(vocab_db_pool)
+        await _init_vocab_table()
+        await _load_all_sessions()
+        print(f"Vocab sessions database initialized")
+    except Exception as e:
+        print(f"Warning: Vocab sessions database initialization failed: {e}")
+
+    # Initialize database pool for DSFA RAG
+    dsfa_db_pool = None
+    if DSFA_DATABASE_URL and set_dsfa_db_pool:
+        try:
+            dsfa_db_pool = await asyncpg.create_pool(DSFA_DATABASE_URL, min_size=2, max_size=10)
+            set_dsfa_db_pool(dsfa_db_pool)
+            print(f"DSFA database pool initialized: {DSFA_DATABASE_URL}")
+        except Exception as e:
+            print(f"Warning: DSFA database pool initialization failed: {e}")
+
     # Initialize Qdrant collection for BYOEH
     try:
         await init_qdrant_collection()
         print("Qdrant BYOEH collection initialized")
     except Exception as e:
-        print(f"Warning: Qdrant initialization failed: {e}")
+        print(f"Warning: Qdrant BYOEH initialization failed: {e}")
+
+    # Initialize Qdrant collection for DSFA RAG
+    if DSFAQdrantService:
+        try:
+            dsfa_qdrant = DSFAQdrantService()
+            await dsfa_qdrant.ensure_collection()
+            print("Qdrant DSFA corpus collection initialized")
+        except Exception as e:
+            print(f"Warning: Qdrant DSFA initialization failed: {e}")
 
     # Ensure EH upload directory exists
     os.makedirs(EH_UPLOAD_DIR, exist_ok=True)
@@ -65,6 +108,16 @@ async def lifespan(app: FastAPI):
 
     print("Klausur-Service shutting down...")
 
+    # Close Vocab sessions database pool
+    if vocab_db_pool:
+        await vocab_db_pool.close()
+        print("Vocab sessions database pool closed")
+
+    # Close DSFA database pool
+    if dsfa_db_pool:
+        await dsfa_db_pool.close()
+        print("DSFA database pool closed")
+
 
 app = FastAPI(
     title="Klausur-Service",
@@ -94,7 +147,11 @@ app.include_router(admin_router)      # NiBiS Ingestion
 app.include_router(zeugnis_router)    # Zeugnis Rights-Aware Crawler
 app.include_router(training_router)   # Training Management
 app.include_router(mail_router)       # Unified Inbox Mail
+if trocr_router:
     app.include_router(trocr_router)      # TrOCR Handwriting OCR
+app.include_router(vocab_router)      # Vocabulary Worksheet Generator
+if dsfa_rag_router:
+    app.include_router(dsfa_rag_router)   # DSFA RAG Corpus Search
 
 
 # =============================================

klausur-service/backend/requirements.txt

@@ -9,6 +9,7 @@ python-dotenv>=1.0.0
 qdrant-client>=1.7.0
 cryptography>=41.0.0
 PyPDF2>=3.0.0
+PyMuPDF>=1.24.0
 
 # PyTorch CPU-only (smaller, no CUDA needed for Docker on Mac)
 --extra-index-url https://download.pytorch.org/whl/cpu
@@ -23,6 +24,10 @@ minio>=7.2.0
 # OpenCV for handwriting detection (headless = no GUI, smaller for CI)
 opencv-python-headless>=4.8.0
 
+# Tesseract OCR Python binding (requires system tesseract-ocr package)
+pytesseract>=0.3.10
+Pillow>=10.0.0
+
 # PostgreSQL (for metrics storage)
 psycopg2-binary>=2.9.0
 asyncpg>=0.29.0

klausur-service/backend/services/grid_detection_service.py

@@ -0,0 +1,509 @@
+"""
+Grid Detection Service v4
+
+Detects table/grid structure from OCR bounding-box data.
+Converts pixel coordinates to percentage and mm coordinates (A4 format).
+Supports deskew correction, column detection, and multi-line cell handling.
+
+Lizenz: Apache 2.0 (kommerziell nutzbar)
+"""
+
+import math
+import logging
+from enum import Enum
+from dataclasses import dataclass, field
+from typing import List, Optional, Dict, Any, Tuple
+
+logger = logging.getLogger(__name__)
+
+# A4 dimensions
+A4_WIDTH_MM = 210.0
+A4_HEIGHT_MM = 297.0
+
+# Column margin (1mm)
+COLUMN_MARGIN_MM = 1.0
+COLUMN_MARGIN_PCT = (COLUMN_MARGIN_MM / A4_WIDTH_MM) * 100
+
+
+class CellStatus(str, Enum):
+    EMPTY = "empty"
+    RECOGNIZED = "recognized"
+    PROBLEMATIC = "problematic"
+    MANUAL = "manual"
+
+
+class ColumnType(str, Enum):
+    ENGLISH = "english"
+    GERMAN = "german"
+    EXAMPLE = "example"
+    UNKNOWN = "unknown"
+
+
+@dataclass
+class OCRRegion:
+    """A word/phrase detected by OCR with bounding box coordinates in percentage (0-100)."""
+    text: str
+    confidence: float
+    x: float       # X position as percentage of page width
+    y: float       # Y position as percentage of page height
+    width: float   # Width as percentage of page width
+    height: float  # Height as percentage of page height
+
+    @property
+    def x_mm(self) -> float:
+        return round(self.x / 100 * A4_WIDTH_MM, 1)
+
+    @property
+    def y_mm(self) -> float:
+        return round(self.y / 100 * A4_HEIGHT_MM, 1)
+
+    @property
+    def width_mm(self) -> float:
+        return round(self.width / 100 * A4_WIDTH_MM, 1)
+
+    @property
+    def height_mm(self) -> float:
+        return round(self.height / 100 * A4_HEIGHT_MM, 2)
+
+    @property
+    def center_x(self) -> float:
+        return self.x + self.width / 2
+
+    @property
+    def center_y(self) -> float:
+        return self.y + self.height / 2
+
+    @property
+    def right(self) -> float:
+        return self.x + self.width
+
+    @property
+    def bottom(self) -> float:
+        return self.y + self.height
+
+
+@dataclass
+class GridCell:
+    """A cell in the detected grid with coordinates in percentage (0-100)."""
+    row: int
+    col: int
+    x: float
+    y: float
+    width: float
+    height: float
+    text: str = ""
+    confidence: float = 0.0
+    status: CellStatus = CellStatus.EMPTY
+    column_type: ColumnType = ColumnType.UNKNOWN
+    logical_row: int = 0
+    logical_col: int = 0
+    is_continuation: bool = False
+
+    @property
+    def x_mm(self) -> float:
+        return round(self.x / 100 * A4_WIDTH_MM, 1)
+
+    @property
+    def y_mm(self) -> float:
+        return round(self.y / 100 * A4_HEIGHT_MM, 1)
+
+    @property
+    def width_mm(self) -> float:
+        return round(self.width / 100 * A4_WIDTH_MM, 1)
+
+    @property
+    def height_mm(self) -> float:
+        return round(self.height / 100 * A4_HEIGHT_MM, 2)
+
+    def to_dict(self) -> dict:
+        return {
+            "row": self.row,
+            "col": self.col,
+            "x": round(self.x, 2),
+            "y": round(self.y, 2),
+            "width": round(self.width, 2),
+            "height": round(self.height, 2),
+            "x_mm": self.x_mm,
+            "y_mm": self.y_mm,
+            "width_mm": self.width_mm,
+            "height_mm": self.height_mm,
+            "text": self.text,
+            "confidence": self.confidence,
+            "status": self.status.value,
+            "column_type": self.column_type.value,
+            "logical_row": self.logical_row,
+            "logical_col": self.logical_col,
+            "is_continuation": self.is_continuation,
+        }
+
+
+@dataclass
+class GridResult:
+    """Result of grid detection."""
+    rows: int = 0
+    columns: int = 0
+    cells: List[List[GridCell]] = field(default_factory=list)
+    column_types: List[str] = field(default_factory=list)
+    column_boundaries: List[float] = field(default_factory=list)
+    row_boundaries: List[float] = field(default_factory=list)
+    deskew_angle: float = 0.0
+    stats: Dict[str, Any] = field(default_factory=dict)
+
+    def to_dict(self) -> dict:
+        cells_dicts = []
+        for row_cells in self.cells:
+            cells_dicts.append([c.to_dict() for c in row_cells])
+
+        return {
+            "rows": self.rows,
+            "columns": self.columns,
+            "cells": cells_dicts,
+            "column_types": self.column_types,
+            "column_boundaries": [round(b, 2) for b in self.column_boundaries],
+            "row_boundaries": [round(b, 2) for b in self.row_boundaries],
+            "deskew_angle": round(self.deskew_angle, 2),
+            "stats": self.stats,
+            "page_dimensions": {
+                "width_mm": A4_WIDTH_MM,
+                "height_mm": A4_HEIGHT_MM,
+                "format": "A4",
+            },
+        }
+
+
+class GridDetectionService:
+    """Detect grid/table structure from OCR bounding-box regions."""
+
+    def __init__(self, y_tolerance_pct: float = 1.5, padding_pct: float = 0.3,
+                 column_margin_mm: float = COLUMN_MARGIN_MM):
+        self.y_tolerance_pct = y_tolerance_pct
+        self.padding_pct = padding_pct
+        self.column_margin_mm = column_margin_mm
+
+    def calculate_deskew_angle(self, regions: List[OCRRegion]) -> float:
+        """Calculate page skew angle from OCR region positions.
+
+        Uses left-edge alignment of regions to detect consistent tilt.
+        Returns angle in degrees, clamped to ±5°.
+        """
+        if len(regions) < 3:
+            return 0.0
+
+        # Group by similar X position (same column)
+        sorted_by_x = sorted(regions, key=lambda r: r.x)
+
+        # Find regions that are vertically aligned (similar X)
+        x_tolerance = 3.0  # percent
+        aligned_groups: List[List[OCRRegion]] = []
+        current_group = [sorted_by_x[0]]
+
+        for r in sorted_by_x[1:]:
+            if abs(r.x - current_group[0].x) <= x_tolerance:
+                current_group.append(r)
+            else:
+                if len(current_group) >= 3:
+                    aligned_groups.append(current_group)
+                current_group = [r]
+
+        if len(current_group) >= 3:
+            aligned_groups.append(current_group)
+
+        if not aligned_groups:
+            return 0.0
+
+        # Use the largest aligned group to calculate skew
+        best_group = max(aligned_groups, key=len)
+        best_group.sort(key=lambda r: r.y)
+
+        # Linear regression: X as function of Y
+        n = len(best_group)
+        sum_y = sum(r.y for r in best_group)
+        sum_x = sum(r.x for r in best_group)
+        sum_xy = sum(r.x * r.y for r in best_group)
+        sum_y2 = sum(r.y ** 2 for r in best_group)
+
+        denom = n * sum_y2 - sum_y ** 2
+        if denom == 0:
+            return 0.0
+
+        slope = (n * sum_xy - sum_y * sum_x) / denom
+
+        # Convert slope to angle (slope is dx/dy in percent space)
+        # Adjust for aspect ratio: A4 is 210/297 ≈ 0.707
+        aspect = A4_WIDTH_MM / A4_HEIGHT_MM
+        angle_rad = math.atan(slope * aspect)
+        angle_deg = math.degrees(angle_rad)
+
+        # Clamp to ±5°
+        return max(-5.0, min(5.0, round(angle_deg, 2)))
+
+    def apply_deskew_to_regions(self, regions: List[OCRRegion], angle: float) -> List[OCRRegion]:
+        """Apply deskew correction to region coordinates.
+
+        Rotates all coordinates around the page center by -angle.
+        """
+        if abs(angle) < 0.01:
+            return regions
+
+        angle_rad = math.radians(-angle)
+        cos_a = math.cos(angle_rad)
+        sin_a = math.sin(angle_rad)
+
+        # Page center
+        cx, cy = 50.0, 50.0
+
+        result = []
+        for r in regions:
+            # Rotate center of region around page center
+            rx = r.center_x - cx
+            ry = r.center_y - cy
+            new_cx = rx * cos_a - ry * sin_a + cx
+            new_cy = rx * sin_a + ry * cos_a + cy
+            new_x = new_cx - r.width / 2
+            new_y = new_cy - r.height / 2
+
+            result.append(OCRRegion(
+                text=r.text,
+                confidence=r.confidence,
+                x=round(new_x, 2),
+                y=round(new_y, 2),
+                width=r.width,
+                height=r.height,
+            ))
+
+        return result
+
+    def _group_regions_into_rows(self, regions: List[OCRRegion]) -> List[List[OCRRegion]]:
+        """Group regions by Y position into rows."""
+        if not regions:
+            return []
+
+        sorted_regions = sorted(regions, key=lambda r: r.y)
+        rows: List[List[OCRRegion]] = []
+        current_row = [sorted_regions[0]]
+        current_y = sorted_regions[0].center_y
+
+        for r in sorted_regions[1:]:
+            if abs(r.center_y - current_y) <= self.y_tolerance_pct:
+                current_row.append(r)
+            else:
+                current_row.sort(key=lambda r: r.x)
+                rows.append(current_row)
+                current_row = [r]
+                current_y = r.center_y
+
+        if current_row:
+            current_row.sort(key=lambda r: r.x)
+            rows.append(current_row)
+
+        return rows
+
+    def _detect_column_boundaries(self, rows: List[List[OCRRegion]]) -> List[float]:
+        """Detect column boundaries from row data."""
+        if not rows:
+            return []
+
+        # Collect all X starting positions
+        all_x = []
+        for row in rows:
+            for r in row:
+                all_x.append(r.x)
+
+        if not all_x:
+            return []
+
+        all_x.sort()
+
+        # Gap-based clustering
+        min_gap = 5.0  # percent
+        clusters: List[List[float]] = []
+        current = [all_x[0]]
+
+        for x in all_x[1:]:
+            if x - current[-1] > min_gap:
+                clusters.append(current)
+                current = [x]
+            else:
+                current.append(x)
+
+        if current:
+            clusters.append(current)
+
+        # Column boundaries: start of each cluster
+        boundaries = [min(c) - self.padding_pct for c in clusters]
+        # Add right boundary
+        boundaries.append(100.0)
+
+        return boundaries
+
+    def _assign_column_types(self, boundaries: List[float]) -> List[str]:
+        """Assign column types based on position."""
+        num_cols = max(0, len(boundaries) - 1)
+        type_map = [ColumnType.ENGLISH, ColumnType.GERMAN, ColumnType.EXAMPLE]
+        result = []
+        for i in range(num_cols):
+            if i < len(type_map):
+                result.append(type_map[i].value)
+            else:
+                result.append(ColumnType.UNKNOWN.value)
+        return result
+
+    def detect_grid(self, regions: List[OCRRegion]) -> GridResult:
+        """Detect grid structure from OCR regions.
+
+        Args:
+            regions: List of OCR regions with percentage-based coordinates.
+
+        Returns:
+            GridResult with detected rows, columns, and cells.
+        """
+        if not regions:
+            return GridResult(stats={"recognized": 0, "problematic": 0, "empty": 0, "manual": 0, "total": 0, "coverage": 0.0})
+
+        # Step 1: Calculate and apply deskew
+        deskew_angle = self.calculate_deskew_angle(regions)
+        corrected_regions = self.apply_deskew_to_regions(regions, deskew_angle)
+
+        # Step 2: Group into rows
+        rows = self._group_regions_into_rows(corrected_regions)
+
+        # Step 3: Detect column boundaries
+        col_boundaries = self._detect_column_boundaries(rows)
+        column_types = self._assign_column_types(col_boundaries)
+        num_cols = max(1, len(col_boundaries) - 1)
+
+        # Step 4: Build cell grid
+        num_rows = len(rows)
+        row_boundaries = []
+        cells = []
+
+        recognized = 0
+        problematic = 0
+        empty = 0
+
+        for row_idx, row_regions in enumerate(rows):
+            # Row Y boundary
+            if row_regions:
+                row_y = min(r.y for r in row_regions) - self.padding_pct
+                row_bottom = max(r.bottom for r in row_regions) + self.padding_pct
+            else:
+                row_y = row_idx / num_rows * 100
+                row_bottom = (row_idx + 1) / num_rows * 100
+
+            row_boundaries.append(row_y)
+            row_height = row_bottom - row_y
+
+            row_cells = []
+            for col_idx in range(num_cols):
+                col_x = col_boundaries[col_idx]
+                col_right = col_boundaries[col_idx + 1] if col_idx + 1 < len(col_boundaries) else 100.0
+                col_width = col_right - col_x
+
+                # Find regions in this cell
+                cell_regions = []
+                for r in row_regions:
+                    r_center = r.center_x
+                    if col_x <= r_center < col_right:
+                        cell_regions.append(r)
+
+                if cell_regions:
+                    text = " ".join(r.text for r in cell_regions)
+                    avg_conf = sum(r.confidence for r in cell_regions) / len(cell_regions)
+                    status = CellStatus.RECOGNIZED if avg_conf >= 0.5 else CellStatus.PROBLEMATIC
+                    # Use actual bounding box from regions
+                    actual_x = min(r.x for r in cell_regions)
+                    actual_y = min(r.y for r in cell_regions)
+                    actual_right = max(r.right for r in cell_regions)
+                    actual_bottom = max(r.bottom for r in cell_regions)
+
+                    cell = GridCell(
+                        row=row_idx,
+                        col=col_idx,
+                        x=actual_x,
+                        y=actual_y,
+                        width=actual_right - actual_x,
+                        height=actual_bottom - actual_y,
+                        text=text,
+                        confidence=round(avg_conf, 3),
+                        status=status,
+                        column_type=ColumnType(column_types[col_idx]) if col_idx < len(column_types) else ColumnType.UNKNOWN,
+                        logical_row=row_idx,
+                        logical_col=col_idx,
+                    )
+
+                    if status == CellStatus.RECOGNIZED:
+                        recognized += 1
+                    else:
+                        problematic += 1
+                else:
+                    cell = GridCell(
+                        row=row_idx,
+                        col=col_idx,
+                        x=col_x,
+                        y=row_y,
+                        width=col_width,
+                        height=row_height,
+                        status=CellStatus.EMPTY,
+                        column_type=ColumnType(column_types[col_idx]) if col_idx < len(column_types) else ColumnType.UNKNOWN,
+                        logical_row=row_idx,
+                        logical_col=col_idx,
+                    )
+                    empty += 1
+
+                row_cells.append(cell)
+            cells.append(row_cells)
+
+        # Add final row boundary
+        if rows and rows[-1]:
+            row_boundaries.append(max(r.bottom for r in rows[-1]) + self.padding_pct)
+        else:
+            row_boundaries.append(100.0)
+
+        total = num_rows * num_cols
+        coverage = (recognized + problematic) / max(total, 1)
+
+        return GridResult(
+            rows=num_rows,
+            columns=num_cols,
+            cells=cells,
+            column_types=column_types,
+            column_boundaries=col_boundaries,
+            row_boundaries=row_boundaries,
+            deskew_angle=deskew_angle,
+            stats={
+                "recognized": recognized,
+                "problematic": problematic,
+                "empty": empty,
+                "manual": 0,
+                "total": total,
+                "coverage": round(coverage, 3),
+            },
+        )
+
+    def convert_tesseract_regions(self, tess_words: List[dict],
+                                   image_width: int, image_height: int) -> List[OCRRegion]:
+        """Convert Tesseract word data (pixels) to OCRRegions (percentages).
+
+        Args:
+            tess_words: Word list from tesseract_vocab_extractor.extract_bounding_boxes.
+            image_width: Image width in pixels.
+            image_height: Image height in pixels.
+
+        Returns:
+            List of OCRRegion with percentage-based coordinates.
+        """
+        if not tess_words or image_width == 0 or image_height == 0:
+            return []
+
+        regions = []
+        for w in tess_words:
+            regions.append(OCRRegion(
+                text=w["text"],
+                confidence=w.get("conf", 50) / 100.0,
+                x=w["left"] / image_width * 100,
+                y=w["top"] / image_height * 100,
+                width=w["width"] / image_width * 100,
+                height=w["height"] / image_height * 100,
+            ))
+
+        return regions

klausur-service/backend/tesseract_vocab_extractor.py

@@ -0,0 +1,346 @@
+"""
+Tesseract-based OCR extraction with word-level bounding boxes.
+
+Uses Tesseract for spatial information (WHERE text is) while
+the Vision LLM handles semantic understanding (WHAT the text means).
+
+Tesseract runs natively on ARM64 via Debian's apt package.
+
+Lizenz: Apache 2.0 (kommerziell nutzbar)
+"""
+
+import io
+import logging
+from typing import List, Dict, Any, Optional
+from difflib import SequenceMatcher
+
+logger = logging.getLogger(__name__)
+
+try:
+    import pytesseract
+    from PIL import Image
+    TESSERACT_AVAILABLE = True
+except ImportError:
+    TESSERACT_AVAILABLE = False
+    logger.warning("pytesseract or Pillow not installed - Tesseract OCR unavailable")
+
+
+async def extract_bounding_boxes(image_bytes: bytes, lang: str = "eng+deu") -> dict:
+    """Run Tesseract OCR and return word-level bounding boxes.
+
+    Args:
+        image_bytes: PNG/JPEG image as bytes.
+        lang: Tesseract language string (e.g. "eng+deu").
+
+    Returns:
+        Dict with 'words' list and 'image_width'/'image_height'.
+    """
+    if not TESSERACT_AVAILABLE:
+        return {"words": [], "image_width": 0, "image_height": 0, "error": "Tesseract not available"}
+
+    image = Image.open(io.BytesIO(image_bytes))
+    data = pytesseract.image_to_data(image, lang=lang, output_type=pytesseract.Output.DICT)
+
+    words = []
+    for i in range(len(data['text'])):
+        text = data['text'][i].strip()
+        conf = int(data['conf'][i])
+        if not text or conf < 20:
+            continue
+        words.append({
+            "text": text,
+            "left": data['left'][i],
+            "top": data['top'][i],
+            "width": data['width'][i],
+            "height": data['height'][i],
+            "conf": conf,
+            "block_num": data['block_num'][i],
+            "par_num": data['par_num'][i],
+            "line_num": data['line_num'][i],
+            "word_num": data['word_num'][i],
+        })
+
+    return {
+        "words": words,
+        "image_width": image.width,
+        "image_height": image.height,
+    }
+
+
+def group_words_into_lines(words: List[dict], y_tolerance_px: int = 15) -> List[List[dict]]:
+    """Group words by their Y position into lines.
+
+    Args:
+        words: List of word dicts from extract_bounding_boxes.
+        y_tolerance_px: Max pixel distance to consider words on the same line.
+
+    Returns:
+        List of lines, each line is a list of words sorted by X position.
+    """
+    if not words:
+        return []
+
+    # Sort by Y then X
+    sorted_words = sorted(words, key=lambda w: (w['top'], w['left']))
+
+    lines: List[List[dict]] = []
+    current_line: List[dict] = [sorted_words[0]]
+    current_y = sorted_words[0]['top']
+
+    for word in sorted_words[1:]:
+        if abs(word['top'] - current_y) <= y_tolerance_px:
+            current_line.append(word)
+        else:
+            current_line.sort(key=lambda w: w['left'])
+            lines.append(current_line)
+            current_line = [word]
+            current_y = word['top']
+
+    if current_line:
+        current_line.sort(key=lambda w: w['left'])
+        lines.append(current_line)
+
+    return lines
+
+
+def detect_columns(lines: List[List[dict]], image_width: int) -> Dict[str, Any]:
+    """Detect column boundaries from word positions.
+
+    Typical vocab table: Left=English, Middle=German, Right=Example sentences.
+
+    Returns:
+        Dict with column boundaries and type assignments.
+    """
+    if not lines or image_width == 0:
+        return {"columns": [], "column_types": []}
+
+    # Collect all word X positions
+    all_x_positions = []
+    for line in lines:
+        for word in line:
+            all_x_positions.append(word['left'])
+
+    if not all_x_positions:
+        return {"columns": [], "column_types": []}
+
+    # Find X-position clusters (column starts)
+    all_x_positions.sort()
+
+    # Simple gap-based column detection
+    min_gap = image_width * 0.08  # 8% of page width = column gap
+    clusters = []
+    current_cluster = [all_x_positions[0]]
+
+    for x in all_x_positions[1:]:
+        if x - current_cluster[-1] > min_gap:
+            clusters.append(current_cluster)
+            current_cluster = [x]
+        else:
+            current_cluster.append(x)
+
+    if current_cluster:
+        clusters.append(current_cluster)
+
+    # Each cluster represents a column start
+    columns = []
+    for cluster in clusters:
+        col_start = min(cluster)
+        columns.append({
+            "x_start": col_start,
+            "x_start_pct": col_start / image_width * 100,
+            "word_count": len(cluster),
+        })
+
+    # Assign column types based on position (left→right: EN, DE, Example)
+    type_map = ["english", "german", "example"]
+    column_types = []
+    for i, col in enumerate(columns):
+        if i < len(type_map):
+            column_types.append(type_map[i])
+        else:
+            column_types.append("unknown")
+
+    return {
+        "columns": columns,
+        "column_types": column_types,
+    }
+
+
+def words_to_vocab_entries(lines: List[List[dict]], columns: List[dict],
+                           column_types: List[str], image_width: int,
+                           image_height: int) -> List[dict]:
+    """Convert grouped words into vocabulary entries using column positions.
+
+    Args:
+        lines: Grouped word lines from group_words_into_lines.
+        columns: Column boundaries from detect_columns.
+        column_types: Column type assignments.
+        image_width: Image width in pixels.
+        image_height: Image height in pixels.
+
+    Returns:
+        List of vocabulary entry dicts with english/german/example fields.
+    """
+    if not columns or not lines:
+        return []
+
+    # Build column boundaries for word assignment
+    col_boundaries = []
+    for i, col in enumerate(columns):
+        start = col['x_start']
+        if i + 1 < len(columns):
+            end = columns[i + 1]['x_start']
+        else:
+            end = image_width
+        col_boundaries.append((start, end, column_types[i] if i < len(column_types) else "unknown"))
+
+    entries = []
+    for line in lines:
+        entry = {"english": "", "german": "", "example": ""}
+        line_words_by_col: Dict[str, List[str]] = {"english": [], "german": [], "example": []}
+        line_bbox: Dict[str, Optional[dict]] = {}
+
+        for word in line:
+            word_center_x = word['left'] + word['width'] / 2
+            assigned_type = "unknown"
+            for start, end, col_type in col_boundaries:
+                if start <= word_center_x < end:
+                    assigned_type = col_type
+                    break
+
+            if assigned_type in line_words_by_col:
+                line_words_by_col[assigned_type].append(word['text'])
+                # Track bounding box for the column
+                if assigned_type not in line_bbox or line_bbox[assigned_type] is None:
+                    line_bbox[assigned_type] = {
+                        "left": word['left'],
+                        "top": word['top'],
+                        "right": word['left'] + word['width'],
+                        "bottom": word['top'] + word['height'],
+                    }
+                else:
+                    bb = line_bbox[assigned_type]
+                    bb['left'] = min(bb['left'], word['left'])
+                    bb['top'] = min(bb['top'], word['top'])
+                    bb['right'] = max(bb['right'], word['left'] + word['width'])
+                    bb['bottom'] = max(bb['bottom'], word['top'] + word['height'])
+
+        for col_type in ["english", "german", "example"]:
+            if line_words_by_col[col_type]:
+                entry[col_type] = " ".join(line_words_by_col[col_type])
+                if line_bbox.get(col_type):
+                    bb = line_bbox[col_type]
+                    entry[f"{col_type}_bbox"] = {
+                        "x_pct": bb['left'] / image_width * 100,
+                        "y_pct": bb['top'] / image_height * 100,
+                        "w_pct": (bb['right'] - bb['left']) / image_width * 100,
+                        "h_pct": (bb['bottom'] - bb['top']) / image_height * 100,
+                    }
+
+        # Only add if at least one column has content
+        if entry["english"] or entry["german"]:
+            entries.append(entry)
+
+    return entries
+
+
+def match_positions_to_vocab(tess_words: List[dict], llm_vocab: List[dict],
+                             image_w: int, image_h: int,
+                             threshold: float = 0.6) -> List[dict]:
+    """Match Tesseract bounding boxes to LLM vocabulary entries.
+
+    For each LLM vocab entry, find the best-matching Tesseract word
+    and attach its bounding box coordinates.
+
+    Args:
+        tess_words: Word list from Tesseract with pixel coordinates.
+        llm_vocab: Vocabulary list from Vision LLM.
+        image_w: Image width in pixels.
+        image_h: Image height in pixels.
+        threshold: Minimum similarity ratio for a match.
+
+    Returns:
+        llm_vocab list with bbox_x_pct/bbox_y_pct/bbox_w_pct/bbox_h_pct added.
+    """
+    if not tess_words or not llm_vocab or image_w == 0 or image_h == 0:
+        return llm_vocab
+
+    for entry in llm_vocab:
+        english = entry.get("english", "").lower().strip()
+        german = entry.get("german", "").lower().strip()
+
+        if not english and not german:
+            continue
+
+        # Try to match English word first, then German
+        for field in ["english", "german"]:
+            search_text = entry.get(field, "").lower().strip()
+            if not search_text:
+                continue
+
+            best_word = None
+            best_ratio = 0.0
+
+            for word in tess_words:
+                ratio = SequenceMatcher(None, search_text, word['text'].lower()).ratio()
+                if ratio > best_ratio:
+                    best_ratio = ratio
+                    best_word = word
+
+            if best_word and best_ratio >= threshold:
+                entry[f"bbox_x_pct"] = best_word['left'] / image_w * 100
+                entry[f"bbox_y_pct"] = best_word['top'] / image_h * 100
+                entry[f"bbox_w_pct"] = best_word['width'] / image_w * 100
+                entry[f"bbox_h_pct"] = best_word['height'] / image_h * 100
+                entry["bbox_match_field"] = field
+                entry["bbox_match_ratio"] = round(best_ratio, 3)
+                break  # Found a match, no need to try the other field
+
+    return llm_vocab
+
+
+async def run_tesseract_pipeline(image_bytes: bytes, lang: str = "eng+deu") -> dict:
+    """Full Tesseract pipeline: extract words, group lines, detect columns, build vocab.
+
+    Args:
+        image_bytes: PNG/JPEG image as bytes.
+        lang: Tesseract language string.
+
+    Returns:
+        Dict with 'vocabulary', 'words', 'lines', 'columns', 'image_width', 'image_height'.
+    """
+    # Step 1: Extract bounding boxes
+    bbox_data = await extract_bounding_boxes(image_bytes, lang=lang)
+
+    if bbox_data.get("error"):
+        return bbox_data
+
+    words = bbox_data["words"]
+    image_w = bbox_data["image_width"]
+    image_h = bbox_data["image_height"]
+
+    # Step 2: Group into lines
+    lines = group_words_into_lines(words)
+
+    # Step 3: Detect columns
+    col_info = detect_columns(lines, image_w)
+
+    # Step 4: Build vocabulary entries
+    vocab = words_to_vocab_entries(
+        lines,
+        col_info["columns"],
+        col_info["column_types"],
+        image_w,
+        image_h,
+    )
+
+    return {
+        "vocabulary": vocab,
+        "words": words,
+        "lines_count": len(lines),
+        "columns": col_info["columns"],
+        "column_types": col_info["column_types"],
+        "image_width": image_w,
+        "image_height": image_h,
+        "word_count": len(words),
+    }

klausur-service/backend/trocr_api.py

@@ -0,0 +1,261 @@
+"""
+TrOCR API - REST endpoints for TrOCR handwriting OCR.
+
+Provides:
+- /ocr/trocr - Single image OCR
+- /ocr/trocr/batch - Batch image processing
+- /ocr/trocr/status - Model status
+- /ocr/trocr/cache - Cache statistics
+"""
+
+from fastapi import APIRouter, UploadFile, File, HTTPException, Query
+from fastapi.responses import StreamingResponse
+from pydantic import BaseModel, Field
+from typing import List, Optional
+import json
+import logging
+
+from services.trocr_service import (
+    run_trocr_ocr_enhanced,
+    run_trocr_batch,
+    run_trocr_batch_stream,
+    get_model_status,
+    get_cache_stats,
+    preload_trocr_model,
+    OCRResult,
+    BatchOCRResult
+)
+
+logger = logging.getLogger(__name__)
+
+router = APIRouter(prefix="/api/v1/ocr/trocr", tags=["TrOCR"])
+
+
+# =============================================================================
+# MODELS
+# =============================================================================
+
+class TrOCRResponse(BaseModel):
+    """Response model for single image OCR."""
+    text: str = Field(..., description="Extracted text")
+    confidence: float = Field(..., ge=0.0, le=1.0, description="Overall confidence")
+    processing_time_ms: int = Field(..., ge=0, description="Processing time in milliseconds")
+    model: str = Field(..., description="Model used for OCR")
+    has_lora_adapter: bool = Field(False, description="Whether LoRA adapter was used")
+    from_cache: bool = Field(False, description="Whether result was from cache")
+    image_hash: str = Field("", description="SHA256 hash of image (first 16 chars)")
+    word_count: int = Field(0, description="Number of words detected")
+
+
+class BatchOCRResponse(BaseModel):
+    """Response model for batch OCR."""
+    results: List[TrOCRResponse] = Field(..., description="Individual OCR results")
+    total_time_ms: int = Field(..., ge=0, description="Total processing time")
+    processed_count: int = Field(..., ge=0, description="Number of images processed")
+    cached_count: int = Field(0, description="Number of results from cache")
+    error_count: int = Field(0, description="Number of errors")
+
+
+class ModelStatusResponse(BaseModel):
+    """Response model for model status."""
+    status: str = Field(..., description="Model status: available, not_installed")
+    is_loaded: bool = Field(..., description="Whether model is loaded in memory")
+    model_name: Optional[str] = Field(None, description="Name of loaded model")
+    device: Optional[str] = Field(None, description="Device model is running on")
+    loaded_at: Optional[str] = Field(None, description="ISO timestamp when model was loaded")
+
+
+class CacheStatsResponse(BaseModel):
+    """Response model for cache statistics."""
+    size: int = Field(..., ge=0, description="Current cache size")
+    max_size: int = Field(..., ge=0, description="Maximum cache size")
+    ttl_seconds: int = Field(..., ge=0, description="Cache TTL in seconds")
+
+
+# =============================================================================
+# ENDPOINTS
+# =============================================================================
+
+@router.get("/status", response_model=ModelStatusResponse)
+async def get_trocr_status():
+    """
+    Get TrOCR model status.
+
+    Returns information about whether the model is loaded and available.
+    """
+    return get_model_status()
+
+
+@router.get("/cache", response_model=CacheStatsResponse)
+async def get_trocr_cache_stats():
+    """
+    Get TrOCR cache statistics.
+
+    Returns information about the OCR result cache.
+    """
+    return get_cache_stats()
+
+
+@router.post("/preload")
+async def preload_model(handwritten: bool = Query(True, description="Load handwritten model")):
+    """
+    Preload TrOCR model into memory.
+
+    This speeds up the first OCR request by loading the model ahead of time.
+    """
+    success = preload_trocr_model(handwritten=handwritten)
+    if success:
+        return {"status": "success", "message": "Model preloaded successfully"}
+    else:
+        raise HTTPException(status_code=500, detail="Failed to preload model")
+
+
+@router.post("", response_model=TrOCRResponse)
+async def run_trocr(
+    file: UploadFile = File(..., description="Image file to process"),
+    handwritten: bool = Query(True, description="Use handwritten model"),
+    split_lines: bool = Query(True, description="Split image into lines"),
+    use_cache: bool = Query(True, description="Use result caching")
+):
+    """
+    Run TrOCR on a single image.
+
+    Supports PNG, JPG, and other common image formats.
+    """
+    # Validate file type
+    if not file.content_type or not file.content_type.startswith("image/"):
+        raise HTTPException(status_code=400, detail="File must be an image")
+
+    try:
+        image_data = await file.read()
+
+        result = await run_trocr_ocr_enhanced(
+            image_data,
+            handwritten=handwritten,
+            split_lines=split_lines,
+            use_cache=use_cache
+        )
+
+        return TrOCRResponse(
+            text=result.text,
+            confidence=result.confidence,
+            processing_time_ms=result.processing_time_ms,
+            model=result.model,
+            has_lora_adapter=result.has_lora_adapter,
+            from_cache=result.from_cache,
+            image_hash=result.image_hash,
+            word_count=len(result.text.split()) if result.text else 0
+        )
+
+    except Exception as e:
+        logger.error(f"TrOCR API error: {e}")
+        raise HTTPException(status_code=500, detail=str(e))
+
+
+@router.post("/batch", response_model=BatchOCRResponse)
+async def run_trocr_batch_endpoint(
+    files: List[UploadFile] = File(..., description="Image files to process"),
+    handwritten: bool = Query(True, description="Use handwritten model"),
+    split_lines: bool = Query(True, description="Split images into lines"),
+    use_cache: bool = Query(True, description="Use result caching")
+):
+    """
+    Run TrOCR on multiple images.
+
+    Processes images sequentially and returns all results.
+    """
+    if not files:
+        raise HTTPException(status_code=400, detail="No files provided")
+
+    if len(files) > 50:
+        raise HTTPException(status_code=400, detail="Maximum 50 images per batch")
+
+    try:
+        images = []
+        for file in files:
+            if not file.content_type or not file.content_type.startswith("image/"):
+                raise HTTPException(status_code=400, detail=f"File {file.filename} is not an image")
+            images.append(await file.read())
+
+        batch_result = await run_trocr_batch(
+            images,
+            handwritten=handwritten,
+            split_lines=split_lines,
+            use_cache=use_cache
+        )
+
+        return BatchOCRResponse(
+            results=[
+                TrOCRResponse(
+                    text=r.text,
+                    confidence=r.confidence,
+                    processing_time_ms=r.processing_time_ms,
+                    model=r.model,
+                    has_lora_adapter=r.has_lora_adapter,
+                    from_cache=r.from_cache,
+                    image_hash=r.image_hash,
+                    word_count=len(r.text.split()) if r.text else 0
+                )
+                for r in batch_result.results
+            ],
+            total_time_ms=batch_result.total_time_ms,
+            processed_count=batch_result.processed_count,
+            cached_count=batch_result.cached_count,
+            error_count=batch_result.error_count
+        )
+
+    except HTTPException:
+        raise
+    except Exception as e:
+        logger.error(f"TrOCR batch API error: {e}")
+        raise HTTPException(status_code=500, detail=str(e))
+
+
+@router.post("/batch/stream")
+async def run_trocr_batch_stream_endpoint(
+    files: List[UploadFile] = File(..., description="Image files to process"),
+    handwritten: bool = Query(True, description="Use handwritten model"),
+    split_lines: bool = Query(True, description="Split images into lines"),
+    use_cache: bool = Query(True, description="Use result caching")
+):
+    """
+    Run TrOCR on multiple images with Server-Sent Events (SSE) progress updates.
+
+    Returns a stream of progress events as images are processed.
+    """
+    if not files:
+        raise HTTPException(status_code=400, detail="No files provided")
+
+    if len(files) > 50:
+        raise HTTPException(status_code=400, detail="Maximum 50 images per batch")
+
+    try:
+        images = []
+        for file in files:
+            if not file.content_type or not file.content_type.startswith("image/"):
+                raise HTTPException(status_code=400, detail=f"File {file.filename} is not an image")
+            images.append(await file.read())
+
+        async def event_generator():
+            async for update in run_trocr_batch_stream(
+                images,
+                handwritten=handwritten,
+                split_lines=split_lines,
+                use_cache=use_cache
+            ):
+                yield f"data: {json.dumps(update)}\n\n"
+
+        return StreamingResponse(
+            event_generator(),
+            media_type="text/event-stream",
+            headers={
+                "Cache-Control": "no-cache",
+                "Connection": "keep-alive"
+            }
+        )
+
+    except HTTPException:
+        raise
+    except Exception as e:
+        logger.error(f"TrOCR stream API error: {e}")
+        raise HTTPException(status_code=500, detail=str(e))

klausur-service/backend/vocab_session_store.py

@@ -0,0 +1,428 @@
+"""
+Vocabulary Session Store - PostgreSQL persistence for vocab extraction sessions.
+
+Replaces in-memory storage with database persistence.
+See migrations/001_vocab_sessions.sql for schema.
+"""
+
+import os
+import uuid
+import logging
+import json
+from typing import Optional, List, Dict, Any
+from datetime import datetime
+
+import asyncpg
+
+logger = logging.getLogger(__name__)
+
+# Database configuration
+DATABASE_URL = os.getenv(
+    "DATABASE_URL",
+    "postgresql://breakpilot:breakpilot@postgres:5432/breakpilot_db"
+)
+
+# Connection pool (initialized lazily)
+_pool: Optional[asyncpg.Pool] = None
+
+
+async def get_pool() -> asyncpg.Pool:
+    """Get or create the database connection pool."""
+    global _pool
+    if _pool is None:
+        _pool = await asyncpg.create_pool(DATABASE_URL, min_size=2, max_size=10)
+    return _pool
+
+
+async def init_vocab_tables():
+    """
+    Initialize vocab tables if they don't exist.
+    This is called at startup.
+    """
+    pool = await get_pool()
+    async with pool.acquire() as conn:
+        # Check if tables exist
+        tables_exist = await conn.fetchval("""
+            SELECT EXISTS (
+                SELECT FROM information_schema.tables
+                WHERE table_name = 'vocab_sessions'
+            )
+        """)
+
+        if not tables_exist:
+            logger.info("Creating vocab tables...")
+            # Read and execute migration
+            migration_path = os.path.join(
+                os.path.dirname(__file__),
+                "migrations/001_vocab_sessions.sql"
+            )
+            if os.path.exists(migration_path):
+                with open(migration_path, "r") as f:
+                    sql = f.read()
+                await conn.execute(sql)
+                logger.info("Vocab tables created successfully")
+            else:
+                logger.warning(f"Migration file not found: {migration_path}")
+        else:
+            logger.debug("Vocab tables already exist")
+
+
+# =============================================================================
+# SESSION OPERATIONS
+# =============================================================================
+
+async def create_session_db(
+    session_id: str,
+    name: str,
+    description: str = "",
+    source_language: str = "en",
+    target_language: str = "de"
+) -> Dict[str, Any]:
+    """Create a new vocabulary session in the database."""
+    pool = await get_pool()
+    async with pool.acquire() as conn:
+        row = await conn.fetchrow("""
+            INSERT INTO vocab_sessions (
+                id, name, description, source_language, target_language,
+                status, vocabulary_count
+            ) VALUES ($1, $2, $3, $4, $5, 'pending', 0)
+            RETURNING *
+        """, uuid.UUID(session_id), name, description, source_language, target_language)
+
+        return _row_to_dict(row)
+
+
+async def get_session_db(session_id: str) -> Optional[Dict[str, Any]]:
+    """Get a session by ID."""
+    pool = await get_pool()
+    async with pool.acquire() as conn:
+        row = await conn.fetchrow("""
+            SELECT * FROM vocab_sessions WHERE id = $1
+        """, uuid.UUID(session_id))
+
+        if row:
+            return _row_to_dict(row)
+        return None
+
+
+async def list_sessions_db(
+    limit: int = 50,
+    offset: int = 0,
+    status: Optional[str] = None
+) -> List[Dict[str, Any]]:
+    """List all sessions with optional filtering."""
+    pool = await get_pool()
+    async with pool.acquire() as conn:
+        if status:
+            rows = await conn.fetch("""
+                SELECT * FROM vocab_sessions
+                WHERE status = $1
+                ORDER BY created_at DESC
+                LIMIT $2 OFFSET $3
+            """, status, limit, offset)
+        else:
+            rows = await conn.fetch("""
+                SELECT * FROM vocab_sessions
+                ORDER BY created_at DESC
+                LIMIT $1 OFFSET $2
+            """, limit, offset)
+
+        return [_row_to_dict(row) for row in rows]
+
+
+async def update_session_db(
+    session_id: str,
+    **kwargs
+) -> Optional[Dict[str, Any]]:
+    """Update a session with given fields."""
+    pool = await get_pool()
+
+    # Build dynamic UPDATE query
+    fields = []
+    values = []
+    param_idx = 1
+
+    allowed_fields = [
+        'name', 'description', 'status', 'vocabulary_count',
+        'extraction_confidence', 'image_path', 'pdf_path', 'pdf_page_count',
+        'ocr_prompts', 'processed_pages', 'successful_pages', 'failed_pages'
+    ]
+
+    for key, value in kwargs.items():
+        if key in allowed_fields:
+            fields.append(f"{key} = ${param_idx}")
+            # Convert dicts/lists to JSON for JSONB columns
+            if key in ['ocr_prompts', 'processed_pages', 'successful_pages', 'failed_pages']:
+                value = json.dumps(value) if value else None
+            values.append(value)
+            param_idx += 1
+
+    if not fields:
+        return await get_session_db(session_id)
+
+    values.append(uuid.UUID(session_id))
+
+    async with pool.acquire() as conn:
+        row = await conn.fetchrow(f"""
+            UPDATE vocab_sessions
+            SET {', '.join(fields)}
+            WHERE id = ${param_idx}
+            RETURNING *
+        """, *values)
+
+        if row:
+            return _row_to_dict(row)
+        return None
+
+
+async def delete_session_db(session_id: str) -> bool:
+    """Delete a session and all related data (cascades)."""
+    pool = await get_pool()
+    async with pool.acquire() as conn:
+        result = await conn.execute("""
+            DELETE FROM vocab_sessions WHERE id = $1
+        """, uuid.UUID(session_id))
+        return result == "DELETE 1"
+
+
+# =============================================================================
+# VOCABULARY OPERATIONS
+# =============================================================================
+
+async def add_vocabulary_db(
+    session_id: str,
+    vocab_list: List[Dict[str, Any]]
+) -> List[Dict[str, Any]]:
+    """Add vocabulary entries to a session."""
+    if not vocab_list:
+        return []
+
+    pool = await get_pool()
+    results = []
+
+    async with pool.acquire() as conn:
+        for vocab in vocab_list:
+            vocab_id = str(uuid.uuid4())
+            row = await conn.fetchrow("""
+                INSERT INTO vocab_entries (
+                    id, session_id, english, german, example_sentence,
+                    example_sentence_gap, word_type, source_page
+                ) VALUES ($1, $2, $3, $4, $5, $6, $7, $8)
+                RETURNING *
+            """,
+                uuid.UUID(vocab_id),
+                uuid.UUID(session_id),
+                vocab.get('english', ''),
+                vocab.get('german', ''),
+                vocab.get('example_sentence'),
+                vocab.get('example_sentence_gap'),
+                vocab.get('word_type'),
+                vocab.get('source_page')
+            )
+            results.append(_row_to_dict(row))
+
+        # Update vocabulary count
+        await conn.execute("""
+            UPDATE vocab_sessions
+            SET vocabulary_count = (
+                SELECT COUNT(*) FROM vocab_entries WHERE session_id = $1
+            )
+            WHERE id = $1
+        """, uuid.UUID(session_id))
+
+    return results
+
+
+async def get_vocabulary_db(
+    session_id: str,
+    source_page: Optional[int] = None
+) -> List[Dict[str, Any]]:
+    """Get vocabulary entries for a session."""
+    pool = await get_pool()
+    async with pool.acquire() as conn:
+        if source_page is not None:
+            rows = await conn.fetch("""
+                SELECT * FROM vocab_entries
+                WHERE session_id = $1 AND source_page = $2
+                ORDER BY created_at
+            """, uuid.UUID(session_id), source_page)
+        else:
+            rows = await conn.fetch("""
+                SELECT * FROM vocab_entries
+                WHERE session_id = $1
+                ORDER BY source_page NULLS LAST, created_at
+            """, uuid.UUID(session_id))
+
+        return [_row_to_dict(row) for row in rows]
+
+
+async def update_vocabulary_db(
+    entry_id: str,
+    **kwargs
+) -> Optional[Dict[str, Any]]:
+    """Update a single vocabulary entry."""
+    pool = await get_pool()
+
+    fields = []
+    values = []
+    param_idx = 1
+
+    allowed_fields = [
+        'english', 'german', 'example_sentence', 'example_sentence_gap',
+        'word_type', 'source_page'
+    ]
+
+    for key, value in kwargs.items():
+        if key in allowed_fields:
+            fields.append(f"{key} = ${param_idx}")
+            values.append(value)
+            param_idx += 1
+
+    if not fields:
+        return None
+
+    values.append(uuid.UUID(entry_id))
+
+    async with pool.acquire() as conn:
+        row = await conn.fetchrow(f"""
+            UPDATE vocab_entries
+            SET {', '.join(fields)}
+            WHERE id = ${param_idx}
+            RETURNING *
+        """, *values)
+
+        if row:
+            return _row_to_dict(row)
+        return None
+
+
+async def clear_page_vocabulary_db(session_id: str, page: int) -> int:
+    """Clear all vocabulary for a specific page."""
+    pool = await get_pool()
+    async with pool.acquire() as conn:
+        result = await conn.execute("""
+            DELETE FROM vocab_entries
+            WHERE session_id = $1 AND source_page = $2
+        """, uuid.UUID(session_id), page)
+
+        # Update vocabulary count
+        await conn.execute("""
+            UPDATE vocab_sessions
+            SET vocabulary_count = (
+                SELECT COUNT(*) FROM vocab_entries WHERE session_id = $1
+            )
+            WHERE id = $1
+        """, uuid.UUID(session_id))
+
+        # Return count of deleted rows
+        count = int(result.split()[-1]) if result else 0
+        return count
+
+
+# =============================================================================
+# WORKSHEET OPERATIONS
+# =============================================================================
+
+async def create_worksheet_db(
+    session_id: str,
+    worksheet_types: List[str],
+    pdf_path: Optional[str] = None,
+    solution_path: Optional[str] = None
+) -> Dict[str, Any]:
+    """Create a worksheet record."""
+    pool = await get_pool()
+    worksheet_id = str(uuid.uuid4())
+
+    async with pool.acquire() as conn:
+        row = await conn.fetchrow("""
+            INSERT INTO vocab_worksheets (
+                id, session_id, worksheet_types, pdf_path, solution_path
+            ) VALUES ($1, $2, $3, $4, $5)
+            RETURNING *
+        """,
+            uuid.UUID(worksheet_id),
+            uuid.UUID(session_id),
+            json.dumps(worksheet_types),
+            pdf_path,
+            solution_path
+        )
+
+        return _row_to_dict(row)
+
+
+async def get_worksheet_db(worksheet_id: str) -> Optional[Dict[str, Any]]:
+    """Get a worksheet by ID."""
+    pool = await get_pool()
+    async with pool.acquire() as conn:
+        row = await conn.fetchrow("""
+            SELECT * FROM vocab_worksheets WHERE id = $1
+        """, uuid.UUID(worksheet_id))
+
+        if row:
+            return _row_to_dict(row)
+        return None
+
+
+async def delete_worksheets_for_session_db(session_id: str) -> int:
+    """Delete all worksheets for a session."""
+    pool = await get_pool()
+    async with pool.acquire() as conn:
+        result = await conn.execute("""
+            DELETE FROM vocab_worksheets WHERE session_id = $1
+        """, uuid.UUID(session_id))
+
+        count = int(result.split()[-1]) if result else 0
+        return count
+
+
+# =============================================================================
+# PDF CACHE OPERATIONS
+# =============================================================================
+
+# Simple in-memory cache for PDF data (temporary until served)
+_pdf_cache: Dict[str, bytes] = {}
+
+
+def cache_pdf_data(worksheet_id: str, pdf_data: bytes) -> None:
+    """Cache PDF data temporarily for download."""
+    _pdf_cache[worksheet_id] = pdf_data
+
+
+def get_cached_pdf_data(worksheet_id: str) -> Optional[bytes]:
+    """Get cached PDF data."""
+    return _pdf_cache.get(worksheet_id)
+
+
+def clear_cached_pdf_data(worksheet_id: str) -> None:
+    """Clear cached PDF data."""
+    _pdf_cache.pop(worksheet_id, None)
+
+
+# =============================================================================
+# HELPER FUNCTIONS
+# =============================================================================
+
+def _row_to_dict(row: asyncpg.Record) -> Dict[str, Any]:
+    """Convert asyncpg Record to dict with proper type handling."""
+    if row is None:
+        return {}
+
+    result = dict(row)
+
+    # Convert UUIDs to strings
+    for key in ['id', 'session_id']:
+        if key in result and result[key] is not None:
+            result[key] = str(result[key])
+
+    # Convert datetimes to ISO strings
+    for key in ['created_at', 'updated_at', 'generated_at']:
+        if key in result and result[key] is not None:
+            result[key] = result[key].isoformat()
+
+    # Parse JSONB fields back to dicts/lists
+    for key in ['ocr_prompts', 'processed_pages', 'successful_pages', 'failed_pages', 'worksheet_types']:
+        if key in result and result[key] is not None:
+            if isinstance(result[key], str):
+                result[key] = json.loads(result[key])
+
+    return result