feat/setup SQL database and GET predict

app/main.py

@@ -5,6 +5,17 @@ from app.schemas import EmployeeInput
 import shap
 import os
 import logging
+from dotenv import load_dotenv
+from sqlalchemy import create_engine, text
+from sqlalchemy.orm import sessionmaker
+from database.create_db import PredictionLog
+from sqlalchemy.orm import Session
+
+load_dotenv(dotenv_path=os.path.join(os.path.dirname(__file__), '..', '.env'))
+DATABASE_URL = os.getenv("DATABASE_URL")
+engine = create_engine(DATABASE_URL)
+SessionLocal = sessionmaker(bind=engine)
+
 
 BASE_DIR = os.path.dirname(os.path.abspath(__file__))
 
@@ -96,19 +107,11 @@ def interpret_shap(rank: int, value: float) -> str:
     direction = "increases resignation risk" if value > 0 else "decreases resignation risk"
     return f"{intensity[rank]} — {direction}"
 
-@app.get("/") # La page d'accueil de ton API
-def read_root():
-    return {"message": "Welcome to the FUTURISYS HR predictor API"}
-
-@app.post("/predict")
-def predict(data: EmployeeInput):
-    # 1. On transforme le dictionnaire reçu en DataFrame pandas
-    df = pd.DataFrame([data.model_dump()])
-
+def run_prediction(df):
     for col, known in known_values.items():
-        val = df[col].values[0]
-        if val not in known:
-            logger.warning(f"Unknown value '{val}' for column '{col}' — prediction may be unreliable")
+            val = df[col].values[0]
+            if val not in known:
+                logger.warning(f"Unknown value '{val}' for column '{col}' — prediction may be unreliable")
 
     # Encodage binaire non inclus dans le pipeline: 
     df['genre']= df["genre"].map({"M": 1, "F": 0})
@@ -154,3 +157,55 @@ def predict(data: EmployeeInput):
             for rank, factor in enumerate(top_factors.index)
         }
     }
+
+def log_prediction(df:pd.DataFrame, result: dict,id_employee:int = None ):
+    with Session(engine) as session:
+        factors = list(result["top_5_factors"].keys())
+        log = PredictionLog(
+            id_employee=id_employee,
+            **{col: df[col].values[0].item() if hasattr(df[col].values[0], 'item') else df[col].values[0] 
+            for col in df.columns 
+            if col in PredictionLog.__table__.columns.keys()
+            and col != 'id_employee'},
+            prediction=result["statut_employe"],
+            probability_score=result["probability_score"].item() if hasattr(result["probability_score"], 'item') else result["probability_score"],
+            primary_driver=factors[0] if len(factors) > 0 else None,
+            strong_factor=factors[1] if len(factors) > 1 else None,
+            moderate_factor=factors[2] if len(factors) > 2 else None,
+            contributing_factor=factors[3] if len(factors) > 3 else None,
+            notable_factor=factors[4] if len(factors) > 4 else None,
+        )
+        session.add(log)
+        session.commit()
+
+
+
+
+
+@app.get("/") # La page d'accueil de ton API
+def read_root():
+    return {"message": "Welcome to the FUTURISYS HR predictor API"}
+
+@app.get("/predict/{id_employee}")
+def predict_by_id(id_employee: int):
+    with engine.connect() as conn:
+        result = conn.execute(
+            text("SELECT * FROM employees_full WHERE id_employee = :id"),
+            {"id": id_employee}
+        )
+        row = result.fetchone()
+    if row is None:
+        raise HTTPException(status_code=404, detail="Employee ID not found in database")
+    
+    df = pd.DataFrame([row._mapping])
+    result = run_prediction(df)
+    log_prediction(df, result, id_employee)
+    return result
+
+@app.post("/predict")
+def predict(data: EmployeeInput):
+    df = pd.DataFrame([data.model_dump()])
+    result= run_prediction(df)
+    log_prediction(df, result)
+    return result
+    

database/create_db.py

@@ -0,0 +1,114 @@
+from sqlalchemy import create_engine, Column, Integer, String, Float, DateTime
+from sqlalchemy.sql import text
+from sqlalchemy.orm import declarative_base
+from datetime import datetime
+import os
+from dotenv import load_dotenv
+
+
+Base = declarative_base()
+
+# Connection string
+load_dotenv()
+DATABASE_URL = os.getenv("DATABASE_URL")
+engine = create_engine(DATABASE_URL)
+
+class EmployeeSirh(Base):
+    __tablename__ = 'employees_sirh'
+    
+    id_employee = Column(Integer, primary_key=True)
+    age = Column(Integer)
+    genre = Column(String)
+    revenu_mensuel = Column(Integer)
+    statut_marital = Column(String)
+    departement = Column(String)
+    poste = Column(String)
+    nombre_experiences_precedentes = Column(Integer)
+    nombre_heures_travaillees = Column(Integer)
+    annee_experience_totale = Column(Integer)
+    annees_dans_l_entreprise = Column(Integer)
+    annees_dans_le_poste_actuel = Column(Integer)
+
+class EmployeeEval(Base):
+    __tablename__ = 'employees_eval'
+    
+    eval_number = Column(Integer, primary_key=True)
+    satisfaction_employee_environnement = Column(Integer)
+    note_evaluation_precedente = Column(Integer)
+    niveau_hierarchique_poste = Column(Integer)
+    satisfaction_employee_nature_travail = Column(Integer)
+    satisfaction_employee_equipe = Column(Integer)
+    satisfaction_employee_equilibre_pro_perso = Column(Integer)
+    note_evaluation_actuelle = Column(Integer)
+    heure_supplementaires = Column(String)
+    augementation_salaire_precedente = Column(String)
+
+
+class EmployeeSondage(Base):
+    __tablename__ = 'employees_sondage'
+    
+    code_sondage = Column(Integer, primary_key=True)
+    a_quitte_l_entreprise = Column(String)
+    nombre_participation_pee = Column(Integer)
+    nb_formations_suivies = Column(Integer)
+    nombre_employee_sous_responsabilite = Column(Integer)
+    distance_domicile_travail = Column(Integer)
+    niveau_education = Column(Integer)
+    domaine_etude = Column(String)
+    ayant_enfants = Column(String)
+    frequence_deplacement = Column(String)
+    annees_depuis_la_derniere_promotion = Column(Integer)
+    annes_sous_responsable_actuel = Column(Integer)
+
+
+class PredictionLog(Base):
+    __tablename__ = 'predictions_log'
+    
+    id = Column(Integer, primary_key=True, autoincrement=True)
+    id_employee = Column(Integer)
+    timestamp = Column(DateTime, default=datetime.utcnow)
+
+    # Inputs
+    genre = Column(String)
+    statut_marital = Column(String)
+    departement = Column(String)
+    poste = Column(String)
+    domaine_etude = Column(String)
+    frequence_deplacement = Column(String)
+    heure_supplementaires = Column(String)
+    age = Column(Integer)
+    revenu_mensuel = Column(Integer)
+    nombre_experiences_precedentes = Column(Integer)
+    annee_experience_totale = Column(Integer)
+    annees_dans_l_entreprise = Column(Integer)
+    annees_dans_le_poste_actuel = Column(Integer)
+    nb_formations_suivies = Column(Integer)
+    distance_domicile_travail = Column(Integer)
+    niveau_education = Column(Integer)
+    annees_depuis_la_derniere_promotion = Column(Integer)
+    annes_sous_responsable_actuel = Column(Integer)
+    satisfaction_employee_environnement = Column(Integer)
+    note_evaluation_precedente = Column(Float)
+    satisfaction_employee_nature_travail = Column(Integer)
+    satisfaction_employee_equipe = Column(Integer)
+    satisfaction_employee_equilibre_pro_perso = Column(Integer)
+    note_evaluation_actuelle = Column(Float)
+    augementation_salaire_precedente = Column(String)
+
+    # Outputs
+    prediction = Column(String)
+    probability_score = Column(Float)
+    primary_driver = Column(String)
+    strong_factor = Column(String)
+    moderate_factor = Column(String)
+    contributing_factor = Column(String)
+    notable_factor = Column(String)
+    unknown_category_warning = Column(String, nullable=True)
+    ground_truth = Column(Integer, nullable=True)
+
+
+if __name__ == "__main__":
+    Base.metadata.create_all(engine)
+    print("Tables created successfully!")
+   
+

database/insert_data.py

@@ -0,0 +1,39 @@
+import pandas as pd
+from sqlalchemy import create_engine, Column, Integer, String, Float, DateTime
+from sqlalchemy.sql import text
+import os
+from dotenv import load_dotenv
+
+# Connection string
+load_dotenv()
+DATABASE_URL = os.getenv("DATABASE_URL")
+engine = create_engine(DATABASE_URL)
+
+with engine.connect() as conn:
+    conn.execute(text("DROP VIEW IF EXISTS employees_full CASCADE"))
+    conn.commit()
+
+# Load CSVs
+df_sirh = pd.read_csv(r'C:\Users\Kevin\projects\OC P4\Projet 4\extrait_sirh.csv')
+df_eval = pd.read_csv(r'C:\Users\Kevin\projects\OC P4\Projet 4\extrait_eval.csv')
+df_eval['eval_number'] = pd.to_numeric(df_eval['eval_number'].str[2:], errors='raise')
+df_sondage = pd.read_csv(r'C:\Users\Kevin\projects\OC P4\Projet 4\extrait_sondage.csv')
+
+
+# Insert into DB
+df_sirh.to_sql('employees_sirh', engine, if_exists='replace', index=False)
+df_eval.to_sql('employees_eval', engine, if_exists='replace', index=False)
+df_sondage.to_sql('employees_sondage', engine, if_exists='replace', index=False)
+
+print("Data inserted successfully!")
+
+with engine.connect() as conn:
+    conn.execute(text("""
+        CREATE OR REPLACE VIEW employees_full AS
+        SELECT * FROM employees_sirh s
+        INNER JOIN employees_sondage so ON s.id_employee = CAST(so.code_sondage AS INTEGER)
+        INNER JOIN employees_eval e ON s.id_employee = CAST(e.eval_number AS INTEGER)
+    """))
+    conn.commit()
+
+    print("View created successfully!")

requirements.txt

@@ -7,4 +7,6 @@ shap==0.50.0
 psycopg2-binary==2.9.11
 pytest
 pytest-cov
-httpx
+httpx
+python-dotenv
+sqlalchemy