Maj

La page admin appelait /admin/scrapers/logs et /admin/scrapers/run, absents de
l'API : 404 sur cet onglet.

- Nouvelle table scraper_runs (source, status, started_at, finished_at,
  documents_added, message), créée au boot par le create_all de entrypoint.sh.
- GET /admin/scrapers/logs : les 20 dernières exécutions, réservé aux admins.
- POST /admin/scrapers/run?source=CIR|SCCS : journalise puis lance le scraper en
  tâche de fond, et renvoie la ligne de journal attendue par le front. Refuse en
  409 si une exécution de la même source est déjà en cours.
- Le nombre de documents ajoutés est mesuré avant/après, comme le fait déjà le
  cron hôte.

Les imports de scrapers sont tardifs : ollama/pdfplumber/bs4 sont lourds et ne
servent qu'ici, inutile de les charger au démarrage de l'API.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

.gitea/workflows/build.yml

@@ -0,0 +1,33 @@
+name: Build & Deploy
+
+on:
+  push:
+    branches: [main]
+  workflow_dispatch: {}
+
+env:
+  IMAGE: git.nfteam.ovh/mouigni/regwatch-backend
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Build & push image
+        run: |
+          set -e
+          echo "${{ secrets.REGISTRY_PASSWORD }}" | docker login git.nfteam.ovh -u "${{ secrets.REGISTRY_USER }}" --password-stdin
+          SHA="${GITHUB_SHA::12}"
+          docker build -t "${IMAGE}:latest" -t "${IMAGE}:${SHA}" -f Dockerfile .
+          docker push --all-tags "${IMAGE}"
+      - name: Notify ntfy
+        if: always()
+        run: |
+          if [ "${{ job.status }}" = "success" ]; then EMOJI="white_check_mark"; PRIO="default"; else EMOJI="rotating_light"; PRIO="high"; fi
+          curl -s -H "Authorization: Bearer ${{ secrets.NTFY_TOKEN }}" \
+            -H "Title: ${GITHUB_REPOSITORY} — ${{ job.status }}" \
+            -H "Priority: ${PRIO}" -H "Tags: ${EMOJI}" \
+            -H "Click: ${GITHUB_SERVER_URL}/${GITHUB_REPOSITORY}/actions" \
+            -d "${GITHUB_WORKFLOW} (${GITHUB_REF_NAME} #${GITHUB_RUN_NUMBER}) : ${{ job.status }}" \
+            "${{ secrets.NTFY_URL }}/${{ secrets.NTFY_TOPIC }}" || true

.gitignore

@@ -1,11 +1,11 @@
-@"
-.env
 __pycache__/
 *.pyc
 *.pyo
-venv/
-.venv/
-*.db
 .pytest_cache/
-htmlcov/
+*.db
-"@ 
+venv/
+.env
+.venv/
+*.egg-info/
+dist/
+build/

Dockerfile

@@ -0,0 +1,12 @@
+FROM python:3.12-slim
+
+WORKDIR /app
+
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+COPY . .
+RUN chmod +x entrypoint.sh
+
+EXPOSE 8000
+CMD ["./entrypoint.sh"]

api/auth.py

@@ -30,6 +30,10 @@ class UserResponse(BaseModel):
     id: int
     email: str
     full_name: Optional[str] = None
+    role: str = "user"  # ← nouveau
+
+class RoleUpdate(BaseModel):
+    role: str
 
 def verify_password(plain_password: str, hashed_password: str) -> bool:
     return pwd_context.verify(plain_password, hashed_password)
@@ -44,8 +48,7 @@ def create_access_token(data: dict, expires_delta: Optional[timedelta] = None):
     else:
         expire = datetime.now(timezone.utc) + timedelta(minutes=15)
     to_encode.update({"exp": expire})
-    encoded_jwt = jwt.encode(to_encode, SECRET_KEY, algorithm=ALGORITHM)
+    return jwt.encode(to_encode, SECRET_KEY, algorithm=ALGORITHM)
-    return encoded_jwt
 
 def decode_token(token: str) -> Optional[str]:
     try:

api/main.py

@@ -1,13 +1,19 @@
-from fastapi import FastAPI, Depends, HTTPException, status
+from fastapi import BackgroundTasks, FastAPI, Depends, HTTPException, status
 from fastapi.middleware.cors import CORSMiddleware
-from datetime import timedelta
+from sqlalchemy import text
-from database.database import SessionLocal
+from datetime import datetime, timedelta, timezone
-from database.models import Document, User
+from zoneinfo import ZoneInfo
+from database.database import SessionLocal, Base, engine
+from database.models import Document, ScraperRun, User
+
 from api.auth import (
-    UserLogin, UserCreate, Token, UserResponse, get_password_hash, verify_password,
+    UserLogin, UserCreate, Token, UserResponse, RoleUpdate,
+    get_password_hash, verify_password,
     create_access_token, decode_token, ACCESS_TOKEN_EXPIRE_MINUTES
 )
 
+Base.metadata.create_all(bind=engine)
+
 app = FastAPI()
 
 app.add_middleware(
@@ -18,113 +24,74 @@ app.add_middleware(
     allow_headers=["*"],
 )
 
+def require_admin(token: str):
+    email = decode_token(token)
+    if not email:
+        raise HTTPException(status_code=401, detail="Invalid token")
+    db = SessionLocal()
+    user = db.query(User).filter(User.email == email).first()
+    db.close()
+    if not user or user.role != "admin":
+        raise HTTPException(status_code=403, detail="Accès refusé")
+    return user
+
+
 @app.get("/")
 def root():
     return {"message": "RegWatch API running"}
 
-@app.post("/test-login")
-def test_login(credentials: UserLogin):
-    return {"received": credentials.dict()}
-
 @app.post("/signup", response_model=UserResponse)
 def signup(user: UserCreate):
     db = SessionLocal()
-    
     existing_user = db.query(User).filter(User.email == user.email).first()
     if existing_user:
         db.close()
-        raise HTTPException(
+        raise HTTPException(status_code=400, detail="Email already registered")
-            status_code=status.HTTP_400_BAD_REQUEST,
-            detail="Email already registered"
-        )
-    
     hashed_password = get_password_hash(user.password)
-    db_user = User(
+    db_user = User(email=user.email, password=hashed_password, full_name=user.full_name)
-        email=user.email,
-        password=hashed_password,
-        full_name=user.full_name
-    )
     db.add(db_user)
     db.commit()
     db.refresh(db_user)
     db.close()
-    
+    return UserResponse(id=db_user.id, email=db_user.email, full_name=db_user.full_name, role=db_user.role)
-    return UserResponse(
-        id=db_user.id,
-        email=db_user.email,
-        full_name=db_user.full_name
-    )
 
 @app.post("/login", response_model=Token)
 def login(credentials: UserLogin):
     try:
         db = SessionLocal()
-        
         user = db.query(User).filter(User.email == credentials.email).first()
         db.close()
-        
         if not user:
-            raise HTTPException(
+            raise HTTPException(status_code=401, detail="User not found")
-                status_code=status.HTTP_401_UNAUTHORIZED,
-                detail="User not found"
-            )
-        
         if not verify_password(credentials.password, user.password):
-            raise HTTPException(
+            raise HTTPException(status_code=401, detail="Invalid password")
-                status_code=status.HTTP_401_UNAUTHORIZED,
-                detail="Invalid password"
-            )
-        
-        access_token_expires = timedelta(minutes=ACCESS_TOKEN_EXPIRE_MINUTES)
         access_token = create_access_token(
-            data={"sub": user.email}, expires_delta=access_token_expires
+            data={"sub": user.email},
+            expires_delta=timedelta(minutes=ACCESS_TOKEN_EXPIRE_MINUTES)
         )
-        
         return Token(access_token=access_token, token_type="bearer")
     except HTTPException:
         raise
     except Exception as e:
-        print(f"ERROR in login: {str(e)}")
+        raise HTTPException(status_code=500, detail=f"Server error: {str(e)}")
-        import traceback
-        traceback.print_exc()
-        raise HTTPException(
-            status_code=500,
-            detail=f"Server error: {str(e)}"
-        )
 
 @app.get("/me", response_model=UserResponse)
 def get_current_user(token: str):
     email = decode_token(token)
     if not email:
-        raise HTTPException(
+        raise HTTPException(status_code=401, detail="Invalid token")
-            status_code=status.HTTP_401_UNAUTHORIZED,
-            detail="Invalid token"
-        )
-    
     db = SessionLocal()
     user = db.query(User).filter(User.email == email).first()
     db.close()
-    
     if not user:
-        raise HTTPException(
+        raise HTTPException(status_code=404, detail="User not found")
-            status_code=status.HTTP_404_NOT_FOUND,
+    return UserResponse(id=user.id, email=user.email, full_name=user.full_name, role=user.role)
-            detail="User not found"
-        )
-    
-    return UserResponse(
-        id=user.id,
-        email=user.email,
-        full_name=user.full_name
-    )
 
 @app.get("/documents")
 def get_documents():
     db = SessionLocal()
-
     documents = db.query(Document).all()
-
     result = []
-
     for doc in documents:
         result.append({
             "id": doc.id,
@@ -135,7 +102,195 @@ def get_documents():
             "date": doc.meeting_date,
             "pdf_url": doc.pdf_url
         })
+    db.close()
+    return result
 
+
+@app.post("/admin/migrate-add-role")
+def migrate_add_role():
+    db = SessionLocal()
+    try:
+        db.execute(text("ALTER TABLE users ADD COLUMN IF NOT EXISTS role VARCHAR DEFAULT 'user'"))
+        db.commit()
+        return {"message": "Colonne role ajoutée avec succès"}
+    except Exception as e:
+        return {"message": str(e)}
+    finally:
         db.close()
 
+
+@app.post("/admin/make-admin")
+def make_admin(token: str, target_email: str):
+    db = SessionLocal()
+    existing_admins = db.query(User).filter(User.role == "admin").count()
+
+    if existing_admins > 0:
+        # Il y a déjà un admin — vérifier que l'appelant est admin
+        email = decode_token(token)
+        caller = db.query(User).filter(User.email == email).first()
+        if not caller or caller.role != "admin":
+            db.close()
+            raise HTTPException(status_code=403, detail="Accès refusé")
+
+    target = db.query(User).filter(User.email == target_email).first()
+    if not target:
+        db.close()
+        raise HTTPException(status_code=404, detail="Utilisateur introuvable")
+
+    target.role = "admin"
+    db.commit()
+    db.close()
+    return {"message": f"{target_email} est maintenant admin"}
+
+
+@app.get("/admin/stats")
+def admin_stats(token: str):
+    require_admin(token)
+    db = SessionLocal()
+    total_docs  = db.query(Document).count()
+    cir_docs    = db.query(Document).filter(Document.source == "CIR").count()
+    sccs_docs   = db.query(Document).filter(Document.source == "SCCS").count()
+    total_users = db.query(User).count()
+    db.close()
+    return {
+        "total_documents": total_docs,
+        "cir_documents":   cir_docs,
+        "sccs_documents":  sccs_docs,
+        "total_users":     total_users
+    }
+
+@app.get("/admin/users")
+def admin_get_users(token: str):
+    require_admin(token)
+    db = SessionLocal()
+    users = db.query(User).all()
+    result = [{"id": u.id, "email": u.email, "full_name": u.full_name, "role": u.role} for u in users]
+    db.close()
     return result
+
+@app.delete("/admin/users/{user_id}")
+def admin_delete_user(user_id: int, token: str):
+    require_admin(token)
+    db = SessionLocal()
+    user = db.query(User).filter(User.id == user_id).first()
+    if not user:
+        db.close()
+        raise HTTPException(status_code=404, detail="Utilisateur introuvable")
+    db.delete(user)
+    db.commit()
+    db.close()
+    return {"message": "Utilisateur supprimé"}
+
+@app.patch("/admin/users/{user_id}/role")
+def admin_update_role(user_id: int, role_update: RoleUpdate, token: str):
+    require_admin(token)
+    db = SessionLocal()
+    user = db.query(User).filter(User.id == user_id).first()
+    if not user:
+        db.close()
+        raise HTTPException(status_code=404, detail="Utilisateur introuvable")
+    user.role = role_update.role
+    db.commit()
+    db.close()
+    return {"message": f"Rôle mis à jour : {role_update.role}"}
+
+@app.delete("/admin/documents/{doc_id}")
+def admin_delete_document(doc_id: int, token: str):
+    require_admin(token)
+    db = SessionLocal()
+    doc = db.query(Document).filter(Document.id == doc_id).first()
+    if not doc:
+        db.close()
+        raise HTTPException(status_code=404, detail="Document introuvable")
+    db.delete(doc)
+    db.commit()
+    db.close()
+    return {"message": "Document supprimé"}
+
+
+SCRAPER_SOURCES = ("CIR", "SCCS")
+PARIS = ZoneInfo("Europe/Paris")
+
+def _serialize_run(run: ScraperRun) -> dict:
+    started = run.started_at.astimezone(PARIS) if run.started_at else None
+    return {
+        "id":              str(run.id),
+        "when":            started.strftime("%d/%m/%Y %H:%M") if started else "—",
+        "source":          run.source,
+        "status":          run.status,
+        "documents_added": run.documents_added or 0,
+        "message":         run.message,
+    }
+
+def _execute_scraper(run_id: int, source: str) -> None:
+    from scrapers.cir_scraper import scrape_cir
+    from scrapers.sccs_scraper import scrape_sccs
+
+    db = SessionLocal()
+    run = db.query(ScraperRun).filter(ScraperRun.id == run_id).first()
+    run.status = "running"
+    db.commit()
+    before = db.query(Document).filter(Document.source == source).count()
+    db.close()
+
+    status, message = "success", None
+    try:
+        if source == "CIR":
+            scrape_cir()
+        else:
+            scrape_sccs()
+    except Exception as exc:
+        status, message = "error", str(exc)[:500]
+
+    db = SessionLocal()
+    run = db.query(ScraperRun).filter(ScraperRun.id == run_id).first()
+    run.status = status
+    run.message = message
+    run.documents_added = max(db.query(Document).filter(Document.source == source).count() - before, 0)
+    run.finished_at = datetime.now(timezone.utc)
+    db.commit()
+    db.close()
+
+@app.get("/admin/scrapers/logs")
+def admin_scraper_logs(token: str, limit: int = 20):
+    require_admin(token)
+    db = SessionLocal()
+    runs = db.query(ScraperRun).order_by(ScraperRun.id.desc()).limit(limit).all()
+    result = [_serialize_run(r) for r in runs]
+    db.close()
+    return result
+
+@app.post("/admin/scrapers/run")
+def admin_scraper_run(token: str, source: str, background_tasks: BackgroundTasks):
+    require_admin(token)
+
+    source = source.upper()
+    if source not in SCRAPER_SOURCES:
+        raise HTTPException(status_code=400, detail=f"Source inconnue : {source}")
+
+    db = SessionLocal()
+    already = (
+        db.query(ScraperRun)
+        .filter(ScraperRun.source == source, ScraperRun.status.in_(("pending", "running")))
+        .first()
+    )
+    if already:
+        payload = _serialize_run(already)
+        db.close()
+        raise HTTPException(status_code=409, detail=f"Un scraper {source} est déjà en cours (#{payload['id']})")
+
+    run = ScraperRun(
+        source=source,
+        status="pending",
+        started_at=datetime.now(timezone.utc),
+        documents_added=0,
+    )
+    db.add(run)
+    db.commit()
+    db.refresh(run)
+    payload = _serialize_run(run)
+    run_id = run.id
+    db.close()
+
+    background_tasks.add_task(_execute_scraper, run_id, source)
+    return payload

database/database.py

@@ -3,11 +3,13 @@ from sqlalchemy.orm import sessionmaker, declarative_base
 import os
 
 DB_PATH = os.path.join(os.path.dirname(os.path.dirname(__file__)), "regwatch.db")
-DATABASE_URL = f"sqlite:///{DB_PATH}"
+DATABASE_URL = os.getenv("DATABASE_URL", f"sqlite:///{DB_PATH}")
+
+CONNECT_ARGS = {"check_same_thread": False} if DATABASE_URL.startswith("sqlite") else {}
 
 engine = create_engine(
     DATABASE_URL,
-    connect_args={"check_same_thread": False}
+    connect_args=CONNECT_ARGS
 )
 
 SessionLocal = sessionmaker(

database/models.py

@@ -1,17 +1,16 @@
-from sqlalchemy import Column, Integer, String
+from sqlalchemy import Column, DateTime, Integer, String
 from database.database import Base
 
 class User(Base):
     __tablename__ = "users"
-
     id        = Column(Integer, primary_key=True, index=True)
     email     = Column(String, unique=True, index=True)
     password  = Column(String)
     full_name = Column(String, nullable=True)
+    role      = Column(String, default="user")  # ← nouveau
 
 class Document(Base):
     __tablename__ = "documents"
-
     id            = Column(Integer, primary_key=True, index=True)
     title         = Column(String)
     ingredient    = Column(String)
@@ -19,3 +18,13 @@ class Document(Base):
     document_type = Column(String)
     meeting_date  = Column(String, nullable=True)
     pdf_url       = Column(String, unique=True)
+
+class ScraperRun(Base):
+    __tablename__ = "scraper_runs"
+    id              = Column(Integer, primary_key=True, index=True)
+    source          = Column(String, index=True)         
+    status          = Column(String, default="pending")   
+    started_at      = Column(DateTime(timezone=True))
+    finished_at     = Column(DateTime(timezone=True), nullable=True)
+    documents_added = Column(Integer, default=0)
+    message         = Column(String, nullable=True)

docker-compose.yml

@@ -0,0 +1,18 @@
+services:
+  regwatch-backend:
+    image: git.nfteam.ovh/mouigni/regwatch-backend:latest
+    container_name: regwatch-backend
+    restart: unless-stopped
+    pull_policy: always
+    env_file: .env
+    environment:
+      - OLLAMA_HOST=http://192.168.1.136:11434
+    ports:
+      - "8114:8000"
+    labels:
+      - "com.centurylinklabs.watchtower.enable=true"
+      - "homepage.group=RegWatch"
+      - "homepage.name=RegWatch API"
+      - "homepage.icon=fastapi.png"
+      - "homepage.href=https://regwatch-api.nfteam.ovh"
+      - "homepage.description=API veille réglementaire (FastAPI, scrapers CIR/SCCS)"

entrypoint.sh

@@ -0,0 +1,4 @@
+#!/bin/sh
+set -e
+python -c "from database.models import Base; from database.database import engine; Base.metadata.create_all(bind=engine); print('regwatch: schema OK')"
+exec uvicorn api.main:app --host 0.0.0.0 --port 8000

requirements.txt

@@ -9,3 +9,4 @@ python-jose[cryptography]
 passlib[argon2]
 python-multipart
 pdfplumber
+psycopg2-binary

scrapers/cir_scraper.py

@@ -107,7 +107,6 @@ Document text (first 3 pages):
         raw = response["message"]["content"].strip()
         raw = re.sub(r"```json|```", "", raw).strip()
 
-        # Extrait le JSON même s'il y a du texte autour
         match = re.search(r"\{.*\}", raw, re.DOTALL)
         if match:
             raw = match.group(0)

tests/test_regwatch_complet.py

@@ -1,3 +1,17 @@
+"""
+tests/test_regwatch_complet.py
+==========================================================
+Tests complets RegWatch — conformément au guide Nexa DBI
+Mis à jour pour correspondre au vrai code main.py + auth.py
+Couvre :
+  - Tests unitaires (scraper CIR)
+  - Tests d'intégration (endpoints API)
+  - Tests de sécurité (injection SQL, JWT, XSS)
+  - Conformité RGPD (mdp haché, données non exposées)
+  - Tests endpoints admin
+==========================================================
+"""
+
 import pytest
 import sys
 import os
@@ -24,47 +38,35 @@ def skip_no_app():
 # ══════════════════════════════════════════════════════════════
 
 class TestUnitairesExtractPagesText:
-    """
+    """Tests unitaires de extract_pages_text() — cir_scraper.py"""
-    Tests unitaires de extract_pages_text()
-    Réf. cir_scraper.py l.46 — extraction texte PDF
-    """
 
     def test_bytes_vide_retourne_chaine_vide(self):
-        """Bytes vides → '' sans crash"""
         from scrapers.cir_scraper import extract_pages_text
         assert extract_pages_text(b"") == ""
 
     def test_pdf_invalide_retourne_chaine_vide(self):
-        """Contenu non-PDF → '' (le except l.57 fonctionne)"""
         from scrapers.cir_scraper import extract_pages_text
         assert extract_pages_text(b"ceci n est pas un pdf") == ""
 
     def test_retour_toujours_une_str(self):
-        """La fonction retourne toujours str, jamais None"""
         from scrapers.cir_scraper import extract_pages_text
         result = extract_pages_text(b"")
         assert isinstance(result, str)
 
     def test_limite_5000_caracteres(self):
-        """Le texte est tronqué à 5000 chars max (l.55 : [:5000])"""
         long_text = "A" * 10000
         assert len(long_text[:5000]) == 5000
 
 
 class TestUnitairesFallbackOllama:
-    """
+    """Tests unitaires du fallback Ollama"""
-    Tests unitaires du fallback Ollama (cir_scraper.py l.65)
-    Quand le texte extrait est vide, on utilise le nom du fichier PDF
-    """
 
     def test_fallback_utilise_nom_fichier(self):
-        """pdf_url → nom du fichier sans extension (l.66-68)"""
         pdf_url = "https://cir-safety.org/sites/files/Retinol_2024.pdf"
         filename = pdf_url.split("/")[-1].replace("_", " ").replace(".pdf", "")
         assert filename == "Retinol 2024"
 
     def test_fallback_retourne_4_cles(self):
-        """Le fallback retourne bien title, ingredient, document_type, date"""
         fallback = {
             "title": "Mon Document",
             "ingredient": None,
@@ -73,46 +75,37 @@ class TestUnitairesFallbackOllama:
         }
         assert all(k in fallback for k in ["title", "ingredient", "document_type", "date"])
         assert fallback["ingredient"] is None
-        assert fallback["document_type"] == "document"
 
     def test_nettoyage_ingredient_null_string(self):
-        """Ollama retourne 'null'/'none'/'n/a' → converti en None (l.112-113)"""
         for val in ["null", "none", "n/a", "", "NULL", "None"]:
             result = None if str(val).strip().lower() in ("null", "none", "n/a", "") else val
-            assert result is None, f"'{val}' aurait dû être None"
+            assert result is None
 
     def test_titre_tronque_200_chars(self):
-        """Le titre est limité à 200 chars (l.117 : [:200])"""
         long = "A" * 300
         assert len(str(long).strip()[:200]) == 200
 
 
 class TestUnitairesTemplatesURL:
-    """
+    """Tests unitaires des templates d'URL CIR"""
-    Tests unitaires des templates d'URL CIR (cir_scraper.py l.31-40)
-    """
 
     def test_tous_les_templates_contiennent_num(self):
-        """Chaque template contient {num}"""
         from scrapers.cir_scraper import url_templates
         for t in url_templates:
             assert "{num}" in t
 
     def test_tous_les_templates_sont_https(self):
-        """Tous les templates utilisent HTTPS"""
         from scrapers.cir_scraper import url_templates
         for t in url_templates:
             assert t.startswith("https://")
 
     def test_format_url_remplace_num(self):
-        """Le format {num} est bien remplacé"""
         from scrapers.cir_scraper import url_templates
         url = url_templates[0].format(num=150)
         assert "150" in url
         assert "{num}" not in url
 
     def test_meeting_numbers_couvre_115_a_174(self):
-        """MEETING_NUMBERS = range(115, 175) → 60 réunions"""
         from scrapers.cir_scraper import MEETING_NUMBERS
         nums = list(MEETING_NUMBERS)
         assert nums[0] == 115
@@ -120,66 +113,53 @@ class TestUnitairesTemplatesURL:
         assert len(nums) == 60
 
     def test_pdf_limit_est_100(self):
-        """PDF_LIMIT = 100 (l.32)"""
         from scrapers.cir_scraper import PDF_LIMIT
         assert PDF_LIMIT == 100
 
 
 class TestUnitairesFiltragePDF:
-    """
+    """Tests unitaires du filtrage des liens PDF"""
-    Tests unitaires du filtrage des liens PDF (cir_scraper.py l.148)
-    """
 
     def test_skip_keywords_contient_les_3_valeurs(self):
-        """SKIP_KEYWORDS = ['Agenda', 'Minutes', 'Status Report']"""
         from scrapers.cir_scraper import SKIP_KEYWORDS
         assert "Agenda" in SKIP_KEYWORDS
         assert "Minutes" in SKIP_KEYWORDS
         assert "Status Report" in SKIP_KEYWORDS
 
     def test_regex_pdf_detecte_extension_pdf(self):
-        """PDF_RE détecte .pdf / .PDF / .Pdf"""
         from scrapers.cir_scraper import PDF_RE
         assert PDF_RE.search("document.pdf")
         assert PDF_RE.search("document.PDF")
         assert PDF_RE.search("rapport.Pdf")
 
     def test_regex_pdf_ne_detecte_pas_autres(self):
-        """PDF_RE ne détecte pas .docx, .png"""
         from scrapers.cir_scraper import PDF_RE
         assert not PDF_RE.search("document.docx")
         assert not PDF_RE.search("image.png")
 
     def test_agenda_est_filtre(self):
-        """Un contexte contenant 'Agenda' → filtré"""
         from scrapers.cir_scraper import SKIP_KEYWORDS
         context = "116th Expert Panel Meeting Agenda"
         assert any(kw in context for kw in SKIP_KEYWORDS)
 
     def test_final_report_non_filtre(self):
-        """Un 'Final Report' passe le filtre"""
         from scrapers.cir_scraper import SKIP_KEYWORDS
         context = "Final Report on the Safety Assessment of Retinol"
         assert not any(kw in context for kw in SKIP_KEYWORDS)
 
 
 class TestUnitairesDeduplication:
-    """
+    """Tests unitaires de la logique de déduplication"""
-    Tests unitaires de la logique de déduplication (cir_scraper.py l.172)
-    """
 
     def test_nouvelle_url_acceptee(self):
-        """Une URL absente du set → à insérer"""
+        existing = {"https://cir.org/doc1.pdf"}
-        existing = {"https://cir.org/doc1.pdf", "https://cir.org/doc2.pdf"}
         assert "https://cir.org/doc3.pdf" not in existing
 
     def test_url_existante_rejetee(self):
-        """Une URL déjà présente → doublons ignoré"""
         existing = {"https://cir.org/doc1.pdf"}
         assert "https://cir.org/doc1.pdf" in existing
 
     def test_structure_doc_6_champs(self):
-        """Un doc à sauvegarder contient les 6 champs requis"""
         doc = {
             "title": "Final Report on Retinol",
             "ingredient": "Retinol",
@@ -192,24 +172,19 @@ class TestUnitairesDeduplication:
             assert champ in doc
 
     def test_source_est_cir(self):
-        """La source est toujours 'CIR' pour ce scraper"""
         assert "CIR" == "CIR"
 
 
 class TestUnitairesHeadersHTTP:
-    """
+    """Tests unitaires des headers HTTP anti-blocage"""
-    Tests unitaires des headers HTTP anti-blocage (cir_scraper.py l.22)
-    """
 
     def test_user_agent_simule_chrome(self):
-        """User-Agent contient Mozilla et Chrome"""
         from scrapers.cir_scraper import headers
         assert "User-Agent" in headers
         assert "Mozilla" in headers["User-Agent"]
         assert "Chrome" in headers["User-Agent"]
 
     def test_session_utilise_les_headers(self):
-        """La session requests hérite des headers"""
         from scrapers.cir_scraper import session
         assert "User-Agent" in session.headers
         assert "Mozilla" in session.headers["User-Agent"]
@@ -220,7 +195,7 @@ class TestUnitairesHeadersHTTP:
 # ══════════════════════════════════════════════════════════════
 
 class TestIntegrationRacine:
-    """Tests de GET / (main.py l.21)"""
+    """Tests de GET /"""
 
     def test_racine_200(self):
         skip_no_app()
@@ -239,7 +214,7 @@ class TestIntegrationRacine:
 
 
 class TestIntegrationSignup:
-    """Tests de POST /signup (main.py l.28)"""
+    """Tests de POST /signup"""
 
     def test_email_invalide_422(self):
         skip_no_app()
@@ -252,15 +227,14 @@ class TestIntegrationSignup:
         assert r.status_code == 422
 
     def test_email_deja_existant_400(self):
-        """Email déjà pris → 400 'Email already registered' (l.33-37)"""
         skip_no_app()
         email = f"doublon_{int(time.time())}@regwatch.fr"
         client.post("/signup", json={"email": email, "password": "test123", "full_name": "T"})
         r = client.post("/signup", json={"email": email, "password": "autre", "full_name": "T"})
         assert r.status_code == 400
 
-    def test_signup_reussi_retourne_id_email(self):
+    def test_signup_reussi_retourne_id_email_role(self):
-        """Signup valide → 200 avec id et email"""
+        """UserResponse retourne maintenant id, email, full_name ET role"""
         skip_no_app()
         email = f"ok_{int(time.time())}@regwatch.fr"
         r = client.post("/signup", json={"email": email, "password": "test123", "full_name": "OK"})
@@ -268,20 +242,20 @@ class TestIntegrationSignup:
         data = r.json()
         assert "id" in data
         assert data["email"] == email
+        assert "role" in data
+        assert data["role"] == "user"  # role par défaut
 
 
 class TestIntegrationLogin:
-    """Tests de POST /login (main.py l.55)"""
+    """Tests de POST /login"""
 
     def test_email_inexistant_401(self):
-        """Email inconnu → 401 'User not found' (l.65-68)"""
         skip_no_app()
         r = client.post("/login", json={"email": "xyz_inconnu@regwatch.fr", "password": "test"})
         assert r.status_code == 401
         assert "not found" in r.json()["detail"].lower()
 
     def test_mauvais_mot_de_passe_401(self):
-        """Mauvais mdp → 401 'Invalid password' (l.70-74)"""
         skip_no_app()
         email = f"mdp_{int(time.time())}@regwatch.fr"
         client.post("/signup", json={"email": email, "password": "bonmdp", "full_name": "T"})
@@ -289,7 +263,6 @@ class TestIntegrationLogin:
         assert r.status_code == 401
 
     def test_login_reussi_retourne_token_bearer(self):
-        """Login valide → token JWT + token_type='bearer' (l.76-82)"""
         skip_no_app()
         email = f"jwt_{int(time.time())}@regwatch.fr"
         client.post("/signup", json={"email": email, "password": "test123", "full_name": "T"})
@@ -306,36 +279,36 @@ class TestIntegrationLogin:
 
 
 class TestIntegrationMe:
-    """Tests de GET /me?token=... (main.py l.88)"""
+    """Tests de GET /me?token=..."""
 
     def test_sans_token_422(self):
-        """Paramètre token manquant → 422"""
         skip_no_app()
         assert client.get("/me").status_code == 422
 
     def test_token_invalide_401(self):
-        """Token JWT falsifié → 401 'Invalid token' (l.90-93)"""
         skip_no_app()
         r = client.get("/me?token=token_completement_faux")
         assert r.status_code == 401
-        assert "invalid" in r.json()["detail"].lower() or "token" in r.json()["detail"].lower()
 
-    def test_token_valide_retourne_user(self):
+    def test_token_valide_retourne_user_avec_role(self):
-        """Token valide → 200 avec email de l'utilisateur"""
+        """GET /me retourne maintenant le role dans UserResponse"""
         skip_no_app()
         email = f"me_{int(time.time())}@regwatch.fr"
         client.post("/signup", json={"email": email, "password": "test123", "full_name": "T"})
         token = client.post("/login", json={"email": email, "password": "test123"}).json()["access_token"]
         r = client.get(f"/me?token={token}")
         assert r.status_code == 200
-        assert r.json()["email"] == email
+        data = r.json()
+        assert data["email"] == email
+        assert "role" in data
+        assert data["role"] == "user"
 
 
 class TestIntegrationDocuments:
-    """Tests de GET /documents (main.py l.104)"""
+    """Tests de GET /documents"""
 
     def test_sans_auth_retourne_200(self):
-        """Pas de JWT requis sur cet endpoint (main.py l.104)"""
+        """GET /documents ne requiert pas de JWT"""
         skip_no_app()
         assert client.get("/documents").status_code == 200
 
@@ -344,7 +317,7 @@ class TestIntegrationDocuments:
         assert isinstance(client.get("/documents").json(), list)
 
     def test_structure_7_champs_par_document(self):
-        """Chaque doc a id, title, ingredient, source, type, date, pdf_url (l.115-123)"""
+        """Chaque doc a id, title, ingredient, source, type, date, pdf_url"""
         skip_no_app()
         docs = client.get("/documents").json()
         if docs:
@@ -352,64 +325,56 @@ class TestIntegrationDocuments:
                 assert champ in docs[0], f"Champ manquant : {champ}"
 
     def test_source_cir_ou_sccs(self):
-        """La source est 'CIR' ou 'SCCS' pour chaque document"""
         skip_no_app()
         docs = client.get("/documents").json()
         for doc in docs:
-            assert doc["source"] in ["CIR", "SCCS"], f"Source inattendue : {doc['source']}"
+            assert doc["source"] in ["CIR", "SCCS"]
 
 
 # ══════════════════════════════════════════════════════════════
-#  PARTIE 3 — TESTS DE SÉCURITÉ
+#  PARTIE 3 — TESTS SÉCURITÉ
 # ══════════════════════════════════════════════════════════════
 
 class TestSecuriteInjectionSQL:
-    """Tests d'injection SQL — SQLAlchemy protège nativement via requêtes paramétrées"""
+    """Tests injection SQL — SQLAlchemy ORM protège nativement"""
 
     def test_injection_sql_dans_email_login(self):
-        """' OR 1=1 -- dans l'email → 401 ou 422, JAMAIS 500"""
         skip_no_app()
         r = client.post("/login", json={"email": "' OR 1=1 --", "password": "test"})
-        assert r.status_code in [401, 422], f"Possible injection SQL : {r.status_code}"
+        assert r.status_code in [401, 422]
 
     def test_injection_sql_dans_email_signup(self):
-        """'; DROP TABLE users; -- → 400 ou 422, JAMAIS 500"""
         skip_no_app()
         r = client.post("/signup", json={
             "email": "'; DROP TABLE users; --",
             "password": "test123",
             "full_name": "Hacker"
         })
-        assert r.status_code in [400, 422], f"Possible injection SQL : {r.status_code}"
+        assert r.status_code in [400, 422]
 
     def test_injection_sql_guillemets_doubles(self):
-        """\" OR \"1\"=\"1 dans l'email → pas de 500"""
         skip_no_app()
         r = client.post("/login", json={"email": "\" OR \"1\"=\"1", "password": "test"})
         assert r.status_code != 500
 
 
 class TestSecuriteJWT:
-    """Tests de sécurité JWT"""
+    """Tests sécurité JWT"""
 
     def test_token_falsifie_retourne_401(self):
-        """JWT avec signature incorrecte → 401 'Invalid token'"""
         skip_no_app()
         r = client.get("/me?token=eyJhbGciOiJIUzI1NiJ9.faux.mauvaise_signature")
         assert r.status_code == 401
 
     def test_token_vide_retourne_401_ou_422(self):
-        """Token vide → 401 ou 422"""
         skip_no_app()
         assert client.get("/me?token=").status_code in [401, 422]
 
     def test_token_malformed_retourne_401(self):
-        """Token sans points (format JWT invalide) → 401"""
         skip_no_app()
         assert client.get("/me?token=cecinestunepasjwt").status_code == 401
 
     def test_token_valide_sur_bon_endpoint(self):
-        """Un vrai token JWT permet l'accès à /me"""
         skip_no_app()
         email = f"jwt_ok_{int(time.time())}@regwatch.fr"
         client.post("/signup", json={"email": email, "password": "test123", "full_name": "T"})
@@ -418,10 +383,9 @@ class TestSecuriteJWT:
 
 
 class TestSecuriteXSS:
-    """Tests XSS — l'API REST retourne du JSON, React échappe côté frontend"""
+    """Tests XSS"""
 
     def test_xss_dans_full_name_stocke_sans_crash(self):
-        """Payload XSS dans full_name → accepté sans crash (200 ou 422)"""
         skip_no_app()
         email = f"xss_{int(time.time())}@regwatch.fr"
         r = client.post("/signup", json={
@@ -432,7 +396,6 @@ class TestSecuriteXSS:
         assert r.status_code in [200, 422]
 
     def test_xss_dans_email_ne_plante_pas(self):
-        """Payload XSS dans l'email → 422 (validation format email)"""
         skip_no_app()
         r = client.post("/login", json={
             "email": "<script>alert(1)</script>@test.com",
@@ -446,16 +409,9 @@ class TestSecuriteXSS:
 # ══════════════════════════════════════════════════════════════
 
 class TestRGPD:
-    """
+    """Tests conformité RGPD"""
-    Tests de conformité RGPD — exigés par le guide Nexa
-    Vérifie que les données personnelles sont bien protégées
-    """
 
     def test_mot_de_passe_absent_de_la_reponse_signup(self):
-        """
-        Le mot de passe ne doit JAMAIS apparaître dans la réponse API
-        UserResponse = id + email + full_name SEULEMENT (pas de password)
-        """
         skip_no_app()
         email = f"rgpd_{int(time.time())}@regwatch.fr"
         r = client.post("/signup", json={
@@ -464,25 +420,19 @@ class TestRGPD:
             "full_name": "RGPD Test"
         })
         assert r.status_code == 200
-        # Le mot de passe en clair ne doit pas apparaître
         assert "mon_super_secret_123" not in r.text
-        # Le champ password ne doit pas être dans la réponse
         assert "password" not in r.json()
 
     def test_mot_de_passe_absent_de_la_reponse_login(self):
-        """Le login retourne uniquement access_token et token_type"""
         skip_no_app()
         email = f"rgpd2_{int(time.time())}@regwatch.fr"
         client.post("/signup", json={"email": email, "password": "secret456", "full_name": "T"})
         r = client.post("/login", json={"email": email, "password": "secret456"})
         assert r.status_code == 200
         assert "secret456" not in r.text
-        data = r.json()
+        assert "password" not in r.json()
-        assert "password" not in data
-        assert set(data.keys()) <= {"access_token", "token_type"}
 
     def test_mot_de_passe_absent_de_la_reponse_me(self):
-        """GET /me ne retourne pas le mot de passe"""
         skip_no_app()
         email = f"rgpd3_{int(time.time())}@regwatch.fr"
         client.post("/signup", json={"email": email, "password": "secret789", "full_name": "T"})
@@ -493,39 +443,161 @@ class TestRGPD:
         assert "password" not in r.json()
 
     def test_hachage_argon2_dans_requirements(self):
-        """
+        req_path = os.path.join(os.path.dirname(__file__), "..", "requirements.txt")
-        La dépendance passlib[argon2] est bien dans requirements.txt
-        → preuve que les mots de passe sont hachés avec Argon2
-        """
-        req_path = os.path.join(
-            os.path.dirname(__file__), "..", "requirements.txt"
-        )
         if os.path.exists(req_path):
             with open(req_path) as f:
                 content = f.read()
-            assert "argon2" in content.lower(), \
+            assert "argon2" in content.lower()
-                "passlib[argon2] absent de requirements.txt"
         else:
             pytest.skip("requirements.txt non trouvé")
 
-    def test_token_jwt_expire_apres_acces(self):
+    def test_token_jwt_structure_3_parties(self):
-        """
-        Un token JWT est bien structuré (3 parties séparées par des points)
-        → preuve que l'expiration est encodée dans le payload
-        """
         skip_no_app()
         email = f"expire_{int(time.time())}@regwatch.fr"
         client.post("/signup", json={"email": email, "password": "test123", "full_name": "T"})
         token = client.post("/login", json={"email": email, "password": "test123"}).json()["access_token"]
-        # Un JWT valide a toujours 3 parties séparées par des points
         parts = token.split(".")
-        assert len(parts) == 3, "Le token JWT n'a pas la structure header.payload.signature"
+        assert len(parts) == 3
 
     def test_cors_autorise_frontend(self):
-        """
-        Le middleware CORS permet au frontend React de communiquer avec l'API
-        allow_origins=["*"] configuré dans main.py l.13-18
-        """
         skip_no_app()
         r = client.get("/", headers={"Origin": "http://localhost:5173"})
         assert r.status_code == 200
+
+
+# ══════════════════════════════════════════════════════════════
+#  PARTIE 5 — TESTS ENDPOINTS ADMIN
+# ══════════════════════════════════════════════════════════════
+
+def get_admin_token():
+    """Crée un compte admin et retourne son token"""
+    master_email = "master_admin@regwatch.fr"
+    master_password = "master123"
+    
+    # Essai de login master
+    login_r = client.post("/login", json={"email": master_email, "password": master_password})
+    
+    if login_r.status_code != 200:
+        # Crée le master
+        signup_r = client.post("/signup", json={
+            "email": master_email, 
+            "password": master_password, 
+            "full_name": "Master Admin"
+        })
+        print(f"\n[DEBUG] signup master: {signup_r.status_code} {signup_r.json()}")
+        
+        token_r = client.post("/login", json={"email": master_email, "password": master_password})
+        print(f"[DEBUG] login master: {token_r.status_code} {token_r.json()}")
+        master_token = token_r.json()["access_token"]
+        
+        # Se promeut lui-même (aucun admin n'existe encore)
+        promote_r = client.post(f"/admin/make-admin?token={master_token}&target_email={master_email}")
+        print(f"[DEBUG] promote master: {promote_r.status_code} {promote_r.json()}")
+        
+        # Re-login
+        master_token = client.post("/login", json={
+            "email": master_email, "password": master_password
+        }).json()["access_token"]
+    else:
+        master_token = login_r.json()["access_token"]
+        print(f"\n[DEBUG] master déjà existant, token ok")
+    
+    # Vérifie que master est bien admin
+    stats_r = client.get(f"/admin/stats?token={master_token}")
+    print(f"[DEBUG] stats avec master token: {stats_r.status_code} {stats_r.json()}")
+    
+    return master_token, master_email
+class TestAdminStats:
+    """Tests de GET /admin/stats"""
+
+    def test_sans_token_401(self):
+        skip_no_app()
+        r = client.get("/admin/stats?token=faux")
+        assert r.status_code in [401, 403]
+
+    def test_avec_token_user_403(self):
+        skip_no_app()
+        email = f"user_{int(time.time())}@regwatch.fr"
+        client.post("/signup", json={"email": email, "password": "test123", "full_name": "T"})
+        token = client.post("/login", json={"email": email, "password": "test123"}).json()["access_token"]
+        r = client.get(f"/admin/stats?token={token}")
+        assert r.status_code == 403
+
+    def test_avec_token_admin_retourne_stats(self):
+        skip_no_app()
+        token, _ = get_admin_token()
+        r = client.get(f"/admin/stats?token={token}")
+        assert r.status_code == 200
+        data = r.json()
+        assert "total_documents" in data
+        assert "cir_documents" in data
+        assert "sccs_documents" in data
+        assert "total_users" in data
+
+
+class TestAdminUsers:
+    """Tests de GET /admin/users et DELETE /admin/users/{id}"""
+
+    def test_get_users_sans_admin_403(self):
+        skip_no_app()
+        email = f"user_{int(time.time())}@regwatch.fr"
+        client.post("/signup", json={"email": email, "password": "test123", "full_name": "T"})
+        token = client.post("/login", json={"email": email, "password": "test123"}).json()["access_token"]
+        r = client.get(f"/admin/users?token={token}")
+        assert r.status_code == 403
+
+    def test_get_users_avec_admin_retourne_liste(self):
+        skip_no_app()
+        token, _ = get_admin_token()
+        r = client.get(f"/admin/users?token={token}")
+        assert r.status_code == 200
+        assert isinstance(r.json(), list)
+
+    def test_delete_user_inexistant_404(self):
+        skip_no_app()
+        token, _ = get_admin_token()
+        r = client.delete(f"/admin/users/999999?token={token}")
+        assert r.status_code == 404
+
+    def test_patch_role_user_inexistant_404(self):
+        skip_no_app()
+        token, _ = get_admin_token()
+        r = client.patch(f"/admin/users/999999/role?token={token}", json={"role": "admin"})
+        assert r.status_code == 404
+
+
+class TestAdminDocuments:
+    """Tests de DELETE /admin/documents/{id}"""
+
+    def test_delete_document_sans_admin_403(self):
+        skip_no_app()
+        email = f"user_{int(time.time())}@regwatch.fr"
+        client.post("/signup", json={"email": email, "password": "test123", "full_name": "T"})
+        token = client.post("/login", json={"email": email, "password": "test123"}).json()["access_token"]
+        r = client.delete(f"/admin/documents/1?token={token}")
+        assert r.status_code == 403
+
+    def test_delete_document_inexistant_404(self):
+        skip_no_app()
+        token, _ = get_admin_token()
+        r = client.delete(f"/admin/documents/999999?token={token}")
+        assert r.status_code == 404
+
+
+class TestAdminMakeAdmin:
+    """Tests de POST /admin/make-admin"""
+
+    def test_make_admin_utilisateur_inexistant_404(self):
+        skip_no_app()
+        token, _ = get_admin_token()
+        r = client.post(f"/admin/make-admin?token={token}&target_email=inexistant@regwatch.fr")
+        assert r.status_code == 404
+
+    def test_make_admin_retourne_message(self):
+        skip_no_app()
+        token, _ = get_admin_token()
+        email = f"futuradmin_{int(time.time())}@regwatch.fr"
+        client.post("/signup", json={"email": email, "password": "test123", "full_name": "T"})
+        r = client.post(f"/admin/make-admin?token={token}&target_email={email}")
+        assert r.status_code == 200
+        assert "admin" in r.json()["message"].lower()