Upload backend/core/views.py with huggingface_hub

backend/core/views.py

@@ -0,0 +1,320 @@
+import json
+from django.conf import settings
+from django.db.models.functions import Lower
+from django.db.models import Q
+from django.http import FileResponse, Http404
+from django.shortcuts import get_object_or_404
+from pathlib import Path
+from rest_framework.decorators import api_view, parser_classes
+from rest_framework.parsers import MultiPartParser, FormParser
+from rest_framework.response import Response
+from .models import Procedure, Fine, Office, Advisory, LegalSection, LegalDocument, Synonym, IngestionJob
+from .serializers import (
+    ProcedureSerializer,
+    FineSerializer,
+    OfficeSerializer,
+    AdvisorySerializer,
+    LegalSectionSerializer,
+    LegalDocumentSerializer,
+    IngestionJobSerializer,
+)
+from .services import enqueue_ingestion_job
+from .search_ml import search_with_ml
+# Chatbot moved to hue_portal.chatbot app
+# Keeping import for backward compatibility
+try:
+    from hue_portal.chatbot.chatbot import get_chatbot
+except ImportError:
+    from .chatbot import get_chatbot
+
+def normalize_query(q: str) -> str:
+  return (q or "").strip()
+
+@api_view(["GET"])
+def search(request):
+  """Unified search endpoint - searches across all models."""
+  q = normalize_query(request.GET.get("q", ""))
+  type_ = request.GET.get("type")  # Optional: filter by type
+  
+  if not q:
+    return Response({"error": "q parameter is required"}, status=400)
+  
+  results = []
+  
+  # Search Procedures
+  if not type_ or type_ == "procedure":
+    proc_qs = Procedure.objects.all()
+    proc_text_fields = ["title", "domain", "conditions", "dossier"]
+    proc_results = search_with_ml(proc_qs, q, proc_text_fields, top_k=10, min_score=0.1)
+    for obj in proc_results:
+      results.append({
+        "type": "procedure",
+        "data": ProcedureSerializer(obj).data,
+        "relevance": getattr(obj, '_ml_score', 0.5)
+      })
+  
+  # Search Fines
+  if not type_ or type_ == "fine":
+    fine_qs = Fine.objects.all()
+    fine_text_fields = ["name", "code", "article", "decree", "remedial"]
+    fine_results = search_with_ml(fine_qs, q, fine_text_fields, top_k=10, min_score=0.1)
+    for obj in fine_results:
+      results.append({
+        "type": "fine",
+        "data": FineSerializer(obj).data,
+        "relevance": getattr(obj, '_ml_score', 0.5)
+      })
+  
+  # Search Offices
+  if not type_ or type_ == "office":
+    office_qs = Office.objects.all()
+    office_text_fields = ["unit_name", "address", "district", "service_scope"]
+    office_results = search_with_ml(office_qs, q, office_text_fields, top_k=10, min_score=0.1)
+    for obj in office_results:
+      results.append({
+        "type": "office",
+        "data": OfficeSerializer(obj).data,
+        "relevance": getattr(obj, '_ml_score', 0.5)
+      })
+  
+  # Search Advisories
+  if not type_ or type_ == "advisory":
+    adv_qs = Advisory.objects.all()
+    adv_text_fields = ["title", "summary"]
+    adv_results = search_with_ml(adv_qs, q, adv_text_fields, top_k=10, min_score=0.1)
+    for obj in adv_results:
+      results.append({
+        "type": "advisory",
+        "data": AdvisorySerializer(obj).data,
+        "relevance": getattr(obj, '_ml_score', 0.5)
+      })
+
+  if not type_ or type_ == "legal":
+    legal_qs = LegalSection.objects.select_related("document").all()
+    legal_text_fields = ["section_title", "section_code", "content"]
+    legal_results = search_with_ml(legal_qs, q, legal_text_fields, top_k=10, min_score=0.1)
+    for obj in legal_results:
+      results.append({
+        "type": "legal",
+        "data": LegalSectionSerializer(obj, context={"request": request}).data,
+        "relevance": getattr(obj, '_ml_score', 0.5)
+      })
+  
+  # Sort by relevance score
+  results.sort(key=lambda x: x["relevance"], reverse=True)
+  
+  return Response({
+    "query": q,
+    "count": len(results),
+    "results": results[:50]  # Limit total results
+  })
+
+@api_view(["GET"])
+def procedures_list(request):
+  q = normalize_query(request.GET.get("q", ""))
+  domain = request.GET.get("domain")
+  level = request.GET.get("level")
+  qs = Procedure.objects.all()
+  if domain: qs = qs.filter(domain__iexact=domain)
+  if level: qs = qs.filter(level__iexact=level)
+  if q:
+    # Use ML-based search for better results
+    text_fields = ["title", "domain", "conditions", "dossier"]
+    qs = search_with_ml(qs, q, text_fields, top_k=100, min_score=0.1)
+  return Response(ProcedureSerializer(qs[:100], many=True).data)
+
+@api_view(["GET"])
+def procedures_detail(request, pk:int):
+  try:
+    obj = Procedure.objects.get(pk=pk)
+  except Procedure.DoesNotExist:
+    return Response(status=404)
+  return Response(ProcedureSerializer(obj).data)
+
+@api_view(["GET"])
+def fines_list(request):
+  q = normalize_query(request.GET.get("q", ""))
+  code = request.GET.get("code")
+  qs = Fine.objects.all()
+  if code: qs = qs.filter(code__iexact=code)
+  if q:
+    # Use ML-based search for better results
+    text_fields = ["name", "code", "article", "decree", "remedial"]
+    qs = search_with_ml(qs, q, text_fields, top_k=100, min_score=0.1)
+  return Response(FineSerializer(qs[:100], many=True).data)
+
+@api_view(["GET"])
+def fines_detail(request, pk:int):
+  try:
+    obj = Fine.objects.get(pk=pk)
+  except Fine.DoesNotExist:
+    return Response(status=404)
+  return Response(FineSerializer(obj).data)
+
+@api_view(["GET"])
+def offices_list(request):
+  q = normalize_query(request.GET.get("q", ""))
+  district = request.GET.get("district")
+  qs = Office.objects.all()
+  if district: qs = qs.filter(district__iexact=district)
+  if q:
+    # Use ML-based search for better results
+    text_fields = ["unit_name", "address", "district", "service_scope"]
+    qs = search_with_ml(qs, q, text_fields, top_k=100, min_score=0.1)
+  return Response(OfficeSerializer(qs[:100], many=True).data)
+
+@api_view(["GET"])
+def offices_detail(request, pk:int):
+  try:
+    obj = Office.objects.get(pk=pk)
+  except Office.DoesNotExist:
+    return Response(status=404)
+  return Response(OfficeSerializer(obj).data)
+
+@api_view(["GET"])
+def advisories_list(request):
+  q = normalize_query(request.GET.get("q", ""))
+  qs = Advisory.objects.all().order_by("-published_at")
+  if q:
+    # Use ML-based search for better results
+    text_fields = ["title", "summary"]
+    qs = search_with_ml(qs, q, text_fields, top_k=100, min_score=0.1)
+  return Response(AdvisorySerializer(qs[:100], many=True).data)
+
+@api_view(["GET"])
+def advisories_detail(request, pk:int):
+  try:
+    obj = Advisory.objects.get(pk=pk)
+  except Advisory.DoesNotExist:
+    return Response(status=404)
+  return Response(AdvisorySerializer(obj).data)
+
+@api_view(["GET"])
+def legal_sections_list(request):
+  q = normalize_query(request.GET.get("q", ""))
+  document_code = request.GET.get("document_code")
+  section_code = request.GET.get("section_code")
+  qs = LegalSection.objects.select_related("document").all()
+  if document_code:
+    qs = qs.filter(document__code__iexact=document_code)
+  if section_code:
+    qs = qs.filter(section_code__icontains=section_code)
+  if q:
+    text_fields = ["section_title", "section_code", "content"]
+    qs = search_with_ml(qs, q, text_fields, top_k=100, min_score=0.1)
+  return Response(LegalSectionSerializer(qs[:100], many=True, context={"request": request}).data)
+
+@api_view(["GET"])
+def legal_sections_detail(request, pk:int):
+  try:
+    obj = LegalSection.objects.select_related("document").get(pk=pk)
+  except LegalSection.DoesNotExist:
+    return Response(status=404)
+  return Response(LegalSectionSerializer(obj, context={"request": request}).data)
+
+@api_view(["GET"])
+def legal_document_download(request, pk:int):
+  try:
+    doc = LegalDocument.objects.get(pk=pk)
+  except LegalDocument.DoesNotExist:
+    raise Http404("Document not found")
+  if not doc.source_file:
+    raise Http404("Document missing source file")
+  file_path = Path(doc.source_file)
+  if not file_path.exists():
+    raise Http404("Source file not found on server")
+  response = FileResponse(open(file_path, "rb"), as_attachment=True, filename=file_path.name)
+  return response
+
+
+def _has_upload_access(request):
+  if getattr(request, "user", None) and request.user.is_authenticated:
+    return True
+  expected = getattr(settings, "LEGAL_UPLOAD_TOKEN", "")
+  header_token = request.headers.get("X-Upload-Token")
+  return bool(expected and header_token and header_token == expected)
+
+
+@api_view(["POST"])
+@parser_classes([MultiPartParser, FormParser])
+def legal_document_upload(request):
+  if not _has_upload_access(request):
+    return Response({"error": "unauthorized"}, status=403)
+
+  upload = request.FILES.get("file")
+  if not upload:
+    return Response({"error": "file is required"}, status=400)
+
+  code = (request.data.get("code") or "").strip()
+  if not code:
+    return Response({"error": "code is required"}, status=400)
+
+  metadata = {
+    "code": code,
+    "title": request.data.get("title") or code,
+    "doc_type": request.data.get("doc_type", "other"),
+    "summary": request.data.get("summary", ""),
+    "issued_by": request.data.get("issued_by", ""),
+    "issued_at": request.data.get("issued_at"),
+    "source_url": request.data.get("source_url", ""),
+    "mime_type": request.data.get("mime_type") or getattr(upload, "content_type", ""),
+    "metadata": {},
+  }
+  extra_meta = request.data.get("metadata")
+  if extra_meta:
+    try:
+      metadata["metadata"] = json.loads(extra_meta) if isinstance(extra_meta, str) else extra_meta
+    except Exception:
+      return Response({"error": "metadata must be valid JSON"}, status=400)
+
+  try:
+    job = enqueue_ingestion_job(
+      file_obj=upload,
+      filename=upload.name,
+      metadata=metadata,
+    )
+  except ValueError as exc:
+    return Response({"error": str(exc)}, status=400)
+  except Exception as exc:
+    return Response({"error": str(exc)}, status=500)
+
+  serialized = IngestionJobSerializer(job, context={"request": request}).data
+  return Response(serialized, status=202)
+
+
+@api_view(["GET"])
+def legal_ingestion_job_detail(request, job_id):
+  job = get_object_or_404(IngestionJob, id=job_id)
+  return Response(IngestionJobSerializer(job, context={"request": request}).data)
+
+
+@api_view(["GET"])
+def legal_ingestion_job_list(request):
+  code = request.GET.get("code")
+  qs = IngestionJob.objects.all()
+  if code:
+    qs = qs.filter(code=code)
+  qs = qs.order_by("-created_at")[:20]
+  serializer = IngestionJobSerializer(qs, many=True, context={"request": request})
+  return Response(serializer.data)
+
+@api_view(["POST"])
+def chat(request):
+  """Chatbot endpoint for natural language queries."""
+  message = request.data.get("message", "").strip()
+  if not message:
+    return Response({"error": "message is required"}, status=400)
+  
+  try:
+    chatbot = get_chatbot()
+    response = chatbot.generate_response(message)
+    return Response(response)
+  except Exception as e:
+    return Response({
+      "message": "Xin lỗi, có lỗi xảy ra. Vui lòng thử lại.",
+      "intent": "error",
+      "error": str(e),
+      "results": [],
+      "count": 0
+    }, status=500)
+