Spaces:
Running
Running
Upload folder using huggingface_hub
Browse files- .env.production +13 -0
- Dockerfile +37 -0
- README.md +207 -74
- README_HF.md +37 -21
- app.py +217 -103
- requirements.txt +103 -9
- src/__init__.py +1 -0
- src/auth.py +99 -0
- src/config.py +64 -0
- src/logger.py +223 -0
- src/models.py +153 -0
- src/preprocessing.py +243 -0
- src/rate_limit.py +40 -0
- src/schemas.py +232 -0
.env.production
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Production environment variables for HuggingFace Spaces
|
| 2 |
+
|
| 3 |
+
# Security
|
| 4 |
+
DEBUG=false
|
| 5 |
+
API_KEY=${HF_SPACE_API_KEY}
|
| 6 |
+
|
| 7 |
+
# API Configuration
|
| 8 |
+
API_VERSION=2.1.0
|
| 9 |
+
LOG_LEVEL=INFO
|
| 10 |
+
|
| 11 |
+
# HuggingFace Model
|
| 12 |
+
HF_MODEL_REPO=ASI-Engineer/employee-turnover-model
|
| 13 |
+
MODEL_FILENAME=model/model.pkl
|
Dockerfile
ADDED
|
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
FROM python:3.12-slim
|
| 2 |
+
|
| 3 |
+
WORKDIR /app
|
| 4 |
+
|
| 5 |
+
# Installer les dépendances système
|
| 6 |
+
RUN apt-get update && apt-get install -y \
|
| 7 |
+
curl \
|
| 8 |
+
&& rm -rf /var/lib/apt/lists/*
|
| 9 |
+
|
| 10 |
+
# Copier les fichiers de dépendances
|
| 11 |
+
COPY requirements.txt .
|
| 12 |
+
|
| 13 |
+
# Installer les dépendances Python
|
| 14 |
+
RUN pip install --no-cache-dir -r requirements.txt
|
| 15 |
+
|
| 16 |
+
# Copier le code de l'application
|
| 17 |
+
COPY app.py .
|
| 18 |
+
COPY src/ ./src/
|
| 19 |
+
COPY .env.example .env
|
| 20 |
+
|
| 21 |
+
# Créer le dossier logs
|
| 22 |
+
RUN mkdir -p logs
|
| 23 |
+
|
| 24 |
+
# Exposer le port
|
| 25 |
+
EXPOSE 8000
|
| 26 |
+
|
| 27 |
+
# Variables d'environnement par défaut
|
| 28 |
+
ENV DEBUG=false
|
| 29 |
+
ENV LOG_LEVEL=INFO
|
| 30 |
+
ENV API_KEY=change-me-in-production
|
| 31 |
+
|
| 32 |
+
# Healthcheck
|
| 33 |
+
HEALTHCHECK --interval=30s --timeout=10s --start-period=40s --retries=3 \
|
| 34 |
+
CMD curl -f http://localhost:8000/health || exit 1
|
| 35 |
+
|
| 36 |
+
# Commande de démarrage
|
| 37 |
+
CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "8000", "--workers", "2"]
|
README.md
CHANGED
|
@@ -1,106 +1,239 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
# Chargement du modèle depuis HF Hub
|
| 35 |
-
model_path = hf_hub_download(
|
| 36 |
-
repo_id="ASI-Engineer/employee-turnover-model",
|
| 37 |
-
filename="model/model.pkl"
|
| 38 |
-
)
|
| 39 |
-
model = mlflow.sklearn.load_model(str(Path(model_path).parent))
|
| 40 |
```
|
| 41 |
|
| 42 |
-
##
|
| 43 |
|
| 44 |
### Prérequis
|
| 45 |
- Python 3.12+
|
| 46 |
-
- Poetry
|
|
|
|
| 47 |
|
| 48 |
-
###
|
| 49 |
|
| 50 |
```bash
|
| 51 |
-
#
|
| 52 |
-
|
|
|
|
| 53 |
|
| 54 |
-
# Installer les dépendances
|
| 55 |
poetry install
|
| 56 |
|
| 57 |
-
#
|
| 58 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 59 |
|
| 60 |
-
#
|
| 61 |
-
|
| 62 |
|
| 63 |
-
#
|
| 64 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 65 |
```
|
| 66 |
|
| 67 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 68 |
|
| 69 |
-
|
| 70 |
|
| 71 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 72 |
|
| 73 |
-
|
| 74 |
```bash
|
| 75 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 76 |
```
|
| 77 |
|
| 78 |
-
|
| 79 |
|
| 80 |
```bash
|
| 81 |
-
#
|
| 82 |
-
poetry run
|
| 83 |
|
| 84 |
-
#
|
| 85 |
-
poetry run
|
| 86 |
|
| 87 |
-
#
|
| 88 |
-
|
| 89 |
```
|
| 90 |
|
| 91 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 92 |
|
| 93 |
-
|
| 94 |
-
- **Accuracy**: 79%
|
| 95 |
-
- **Données**: 1470 échantillons, 50 features
|
| 96 |
-
- **Classes**: {0: 1233, 1: 237} - Ratio 5.20:1
|
| 97 |
|
| 98 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 99 |
|
| 100 |
-
|
| 101 |
-
-
|
| 102 |
-
-
|
|
|
|
| 103 |
|
| 104 |
-
|
| 105 |
|
| 106 |
-
**
|
|
|
|
|
|
| 1 |
+
# 🚀 Employee Turnover Prediction API - v2.1.0
|
| 2 |
+
|
| 3 |
+
## 📊 Vue d'ensemble
|
| 4 |
+
|
| 5 |
+
API REST de prédiction du turnover des employés basée sur un modèle XGBoost avec SMOTE.
|
| 6 |
+
|
| 7 |
+
**✨ Nouveautés v2.1.0** :
|
| 8 |
+
- 📝 Logging structuré JSON
|
| 9 |
+
- 🛡️ Rate limiting (20 req/min par IP)
|
| 10 |
+
- ⚡ Gestion d'erreurs améliorée
|
| 11 |
+
- 📊 Monitoring des performances
|
| 12 |
+
- 🔐 Authentification API Key
|
| 13 |
+
|
| 14 |
+
## 🏗️ Architecture
|
| 15 |
+
|
| 16 |
+
```
|
| 17 |
+
OC_P5/
|
| 18 |
+
├── app.py # Point d'entrée FastAPI
|
| 19 |
+
├── src/
|
| 20 |
+
│ ├── auth.py # Authentification API Key
|
| 21 |
+
│ ├── config.py # Configuration centralisée
|
| 22 |
+
│ ├── logger.py # Logging structuré (NOUVEAU)
|
| 23 |
+
│ ├── models.py # Chargement modèle HF Hub
|
| 24 |
+
│ ├── preprocessing.py # Pipeline preprocessing
|
| 25 |
+
│ ├── rate_limit.py # Rate limiting (NOUVEAU)
|
| 26 |
+
│ └── schemas.py # Validation Pydantic
|
| 27 |
+
├── tests/ # Suite pytest (33 tests, 88% couverture)
|
| 28 |
+
├── logs/ # Logs JSON (NOUVEAU)
|
| 29 |
+
│ ├── api.log # Tous les logs
|
| 30 |
+
│ └── error.log # Erreurs uniquement
|
| 31 |
+
├── docs/ # Documentation
|
| 32 |
+
├── ml_model/ # Scripts training
|
| 33 |
+
└── data/ # Données sources
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 34 |
```
|
| 35 |
|
| 36 |
+
## 🚀 Installation
|
| 37 |
|
| 38 |
### Prérequis
|
| 39 |
- Python 3.12+
|
| 40 |
+
- Poetry 1.7+
|
| 41 |
+
- Git
|
| 42 |
|
| 43 |
+
### Setup rapide
|
| 44 |
|
| 45 |
```bash
|
| 46 |
+
# 1. Cloner le repo
|
| 47 |
+
git clone https://github.com/chaton59/OC_P5.git
|
| 48 |
+
cd OC_P5
|
| 49 |
|
| 50 |
+
# 2. Installer les dépendances
|
| 51 |
poetry install
|
| 52 |
|
| 53 |
+
# 3. Configurer l'environnement
|
| 54 |
+
cp .env.example .env
|
| 55 |
+
# Éditer .env avec vos valeurs
|
| 56 |
+
|
| 57 |
+
# 4. Lancer l'API
|
| 58 |
+
poetry run uvicorn app:app --reload
|
| 59 |
+
|
| 60 |
+
# 5. Accéder à la documentation
|
| 61 |
+
# http://localhost:8000/docs
|
| 62 |
+
```
|
| 63 |
+
|
| 64 |
+
## 📝 Configuration (.env)
|
| 65 |
+
|
| 66 |
+
```bash
|
| 67 |
+
# Mode développement (désactive auth + active logs détaillés)
|
| 68 |
+
DEBUG=true
|
| 69 |
|
| 70 |
+
# API Key (requis en production)
|
| 71 |
+
API_KEY=your-secret-key-here
|
| 72 |
|
| 73 |
+
# Logging (DEBUG, INFO, WARNING, ERROR, CRITICAL)
|
| 74 |
+
LOG_LEVEL=INFO
|
| 75 |
+
|
| 76 |
+
# HuggingFace Model
|
| 77 |
+
HF_MODEL_REPO=ASI-Engineer/employee-turnover-model
|
| 78 |
+
MODEL_FILENAME=model/model.pkl
|
| 79 |
```
|
| 80 |
|
| 81 |
+
## 🔒 Authentification
|
| 82 |
+
|
| 83 |
+
### Mode DEBUG (développement)
|
| 84 |
+
```bash
|
| 85 |
+
# L'API Key n'est PAS requise
|
| 86 |
+
curl http://localhost:8000/predict -H "Content-Type: application/json" -d '{...}'
|
| 87 |
+
```
|
| 88 |
+
|
| 89 |
+
### Mode PRODUCTION
|
| 90 |
+
```bash
|
| 91 |
+
# L'API Key est REQUISE
|
| 92 |
+
curl http://localhost:8000/predict \
|
| 93 |
+
-H "X-API-Key: your-secret-key" \
|
| 94 |
+
-H "Content-Type: application/json" \
|
| 95 |
+
-d '{...}'
|
| 96 |
+
```
|
| 97 |
|
| 98 |
+
## 📡 Endpoints
|
| 99 |
|
| 100 |
+
### 🏥 Health Check
|
| 101 |
+
```bash
|
| 102 |
+
GET /health
|
| 103 |
+
|
| 104 |
+
# Réponse
|
| 105 |
+
{
|
| 106 |
+
"status": "healthy",
|
| 107 |
+
"model_loaded": true,
|
| 108 |
+
"model_type": "Pipeline",
|
| 109 |
+
"version": "2.1.0"
|
| 110 |
+
}
|
| 111 |
+
```
|
| 112 |
|
| 113 |
+
### 🔮 Prédiction
|
| 114 |
```bash
|
| 115 |
+
POST /predict
|
| 116 |
+
Content-Type: application/json
|
| 117 |
+
X-API-Key: your-key (en production)
|
| 118 |
+
|
| 119 |
+
# Exemple payload (voir docs/API_GUIDE.md pour tous les champs)
|
| 120 |
+
{
|
| 121 |
+
"satisfaction_employee_environnement": 3,
|
| 122 |
+
"satisfaction_employee_nature_travail": 4,
|
| 123 |
+
"satisfaction_employee_equipe": 5,
|
| 124 |
+
"satisfaction_employee_equilibre_pro_perso": 3,
|
| 125 |
+
"note_evaluation_actuelle": 85,
|
| 126 |
+
"annees_depuis_la_derniere_promotion": 2,
|
| 127 |
+
"nombre_formations_realisees": 3,
|
| 128 |
+
...
|
| 129 |
+
}
|
| 130 |
+
|
| 131 |
+
# Réponse
|
| 132 |
+
{
|
| 133 |
+
"prediction": 0, # 0 = reste, 1 = part
|
| 134 |
+
"probability_0": 0.85, # Probabilité de rester
|
| 135 |
+
"probability_1": 0.15, # Probabilité de partir
|
| 136 |
+
"risk_level": "Low" # Low, Medium, High
|
| 137 |
+
}
|
| 138 |
+
```
|
| 139 |
+
|
| 140 |
+
## 📊 Logging
|
| 141 |
+
|
| 142 |
+
### Logs structurés JSON
|
| 143 |
+
|
| 144 |
+
**Fichiers** :
|
| 145 |
+
- `logs/api.log` : Tous les logs
|
| 146 |
+
- `logs/error.log` : Erreurs uniquement
|
| 147 |
+
|
| 148 |
+
**Format** :
|
| 149 |
+
```json
|
| 150 |
+
{
|
| 151 |
+
"timestamp": "2025-12-26T10:30:45",
|
| 152 |
+
"level": "INFO",
|
| 153 |
+
"logger": "employee_turnover_api",
|
| 154 |
+
"message": "Request POST /predict",
|
| 155 |
+
"method": "POST",
|
| 156 |
+
"path": "/predict",
|
| 157 |
+
"status_code": 200,
|
| 158 |
+
"duration_ms": 23.45,
|
| 159 |
+
"client_host": "127.0.0.1"
|
| 160 |
+
}
|
| 161 |
+
```
|
| 162 |
+
|
| 163 |
+
## 🛡️ Rate Limiting
|
| 164 |
+
|
| 165 |
+
**Configuration** :
|
| 166 |
+
- **Développement** : Désactivé (DEBUG=true)
|
| 167 |
+
- **Production** : 20 requêtes/minute par IP ou API Key
|
| 168 |
+
|
| 169 |
+
**En cas de dépassement** :
|
| 170 |
+
```json
|
| 171 |
+
{
|
| 172 |
+
"error": "Rate limit exceeded",
|
| 173 |
+
"message": "20 per 1 minute"
|
| 174 |
+
}
|
| 175 |
```
|
| 176 |
|
| 177 |
+
## ✅ Tests
|
| 178 |
|
| 179 |
```bash
|
| 180 |
+
# Tous les tests
|
| 181 |
+
poetry run pytest tests/ -v
|
| 182 |
|
| 183 |
+
# Avec couverture
|
| 184 |
+
poetry run pytest tests/ --cov --cov-report=html
|
| 185 |
|
| 186 |
+
# Voir rapport HTML
|
| 187 |
+
open htmlcov/index.html
|
| 188 |
```
|
| 189 |
|
| 190 |
+
**Résultats** :
|
| 191 |
+
- ✅ 33 tests passés
|
| 192 |
+
- 📊 88% de couverture globale
|
| 193 |
+
|
| 194 |
+
## 🚀 Déploiement
|
| 195 |
+
|
| 196 |
+
### Variables d'environnement requises
|
| 197 |
+
```bash
|
| 198 |
+
DEBUG=false
|
| 199 |
+
API_KEY=<votre-clé-sécurisée>
|
| 200 |
+
LOG_LEVEL=INFO
|
| 201 |
+
```
|
| 202 |
+
|
| 203 |
+
### HuggingFace Spaces
|
| 204 |
+
Prêt pour déploiement avec `app.py` et `requirements.txt`
|
| 205 |
+
|
| 206 |
+
## 📚 Documentation
|
| 207 |
+
|
| 208 |
+
- **API Interactive** : http://localhost:8000/docs
|
| 209 |
+
- **ReDoc** : http://localhost:8000/redoc
|
| 210 |
+
- **Guide complet** : [docs/API_GUIDE.md](docs/API_GUIDE.md)
|
| 211 |
+
- **Standards** : [docs/standards.md](docs/standards.md)
|
| 212 |
+
- **Couverture tests** : [docs/TEST_COVERAGE.md](docs/TEST_COVERAGE.md)
|
| 213 |
+
|
| 214 |
+
## 📦 Dépendances principales
|
| 215 |
+
|
| 216 |
+
- **FastAPI** 0.115.14 : Framework web
|
| 217 |
+
- **Pydantic** 2.12.5 : Validation données
|
| 218 |
+
- **XGBoost** 2.1.3 : Modèle ML
|
| 219 |
+
- **SlowAPI** 0.1.9 : Rate limiting
|
| 220 |
+
- **python-json-logger** 4.0.0 : Logs structurés
|
| 221 |
+
- **pytest** 9.0.2 : Tests
|
| 222 |
|
| 223 |
+
## 🔄 Changelog
|
|
|
|
|
|
|
|
|
|
| 224 |
|
| 225 |
+
### v2.1.0 (26 décembre 2025)
|
| 226 |
+
- ✨ Système de logging structuré JSON
|
| 227 |
+
- 🛡️ Rate limiting avec SlowAPI
|
| 228 |
+
- ⚡ Amélioration gestion d'erreurs
|
| 229 |
+
- 📊 Monitoring des performances
|
| 230 |
|
| 231 |
+
### v2.0.0 (26 décembre 2025)
|
| 232 |
+
- ✅ Suite de tests complète (33 tests)
|
| 233 |
+
- 🔐 Authentification API Key
|
| 234 |
+
- 📊 88% de couverture de code
|
| 235 |
|
| 236 |
+
## 👥 Auteurs
|
| 237 |
|
| 238 |
+
- **Projet** : OpenClassrooms P5
|
| 239 |
+
- **Repo** : [github.com/chaton59/OC_P5](https://github.com/chaton59/OC_P5)
|
README_HF.md
CHANGED
|
@@ -1,33 +1,49 @@
|
|
| 1 |
---
|
| 2 |
-
title: Employee Turnover Prediction
|
| 3 |
-
emoji:
|
| 4 |
colorFrom: blue
|
| 5 |
-
colorTo:
|
| 6 |
-
sdk:
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
---
|
| 11 |
|
| 12 |
-
#
|
| 13 |
|
| 14 |
-
|
| 15 |
|
| 16 |
-
##
|
| 17 |
|
| 18 |
-
-
|
| 19 |
-
-
|
| 20 |
-
-
|
| 21 |
-
-
|
|
|
|
|
|
|
| 22 |
|
| 23 |
-
##
|
| 24 |
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
| 28 |
|
| 29 |
-
##
|
| 30 |
|
| 31 |
-
|
|
|
|
|
|
|
| 32 |
|
| 33 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
+
title: Employee Turnover Prediction API
|
| 3 |
+
emoji: 👔
|
| 4 |
colorFrom: blue
|
| 5 |
+
colorTo: purple
|
| 6 |
+
sdk: docker
|
| 7 |
+
pinned: true
|
| 8 |
+
license: mit
|
| 9 |
+
app_port: 8000
|
| 10 |
---
|
| 11 |
|
| 12 |
+
# Employee Turnover Prediction API 🚀
|
| 13 |
|
| 14 |
+
API de prédiction du turnover des employés avec XGBoost + SMOTE.
|
| 15 |
|
| 16 |
+
## 🎯 Fonctionnalités
|
| 17 |
|
| 18 |
+
- ✅ Prédiction de turnover (0 = reste, 1 = part)
|
| 19 |
+
- 📊 Probabilités et niveau de risque (Low/Medium/High)
|
| 20 |
+
- 🔐 Authentification API Key
|
| 21 |
+
- 📝 Logs structurés JSON
|
| 22 |
+
- 🛡️ Rate limiting (20 req/min)
|
| 23 |
+
- 📚 Documentation OpenAPI/Swagger
|
| 24 |
|
| 25 |
+
## 🔗 Endpoints
|
| 26 |
|
| 27 |
+
- **Docs** : `/docs` - Documentation interactive
|
| 28 |
+
- **Health** : `/health` - Status de l'API
|
| 29 |
+
- **Predict** : `/predict` - Prédiction de turnover
|
| 30 |
|
| 31 |
+
## 🚀 Utilisation
|
| 32 |
|
| 33 |
+
```bash
|
| 34 |
+
# Health check
|
| 35 |
+
curl https://asi-engineer-employee-turnover-api.hf.space/health
|
| 36 |
|
| 37 |
+
# Prédiction
|
| 38 |
+
curl -X POST https://asi-engineer-employee-turnover-api.hf.space/predict \
|
| 39 |
+
-H "Content-Type: application/json" \
|
| 40 |
+
-d '{
|
| 41 |
+
"satisfaction_employee_environnement": 3,
|
| 42 |
+
"satisfaction_employee_nature_travail": 4,
|
| 43 |
+
...
|
| 44 |
+
}'
|
| 45 |
+
```
|
| 46 |
+
|
| 47 |
+
## 📚 Documentation complète
|
| 48 |
+
|
| 49 |
+
Voir [GitHub Repository](https://github.com/chaton59/OC_P5) pour la documentation complète.
|
app.py
CHANGED
|
@@ -1,138 +1,252 @@
|
|
| 1 |
#!/usr/bin/env python3
|
| 2 |
"""
|
| 3 |
-
|
| 4 |
|
| 5 |
-
|
| 6 |
-
|
|
|
|
|
|
|
|
|
|
| 7 |
"""
|
| 8 |
-
import
|
| 9 |
-
from
|
| 10 |
|
| 11 |
-
|
| 12 |
-
|
|
|
|
|
|
|
| 13 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 14 |
|
| 15 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 16 |
"""
|
| 17 |
-
|
| 18 |
|
| 19 |
-
|
| 20 |
-
Le fallback MLflow local n'est disponible qu'en développement local.
|
| 21 |
"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
try:
|
| 23 |
-
|
|
|
|
|
|
|
| 24 |
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
)
|
| 29 |
-
model = joblib.load(model_path)
|
| 30 |
-
print(f"✅ Modèle chargé depuis HF Hub: {HF_MODEL_REPO}")
|
| 31 |
-
return model, "HF Hub"
|
| 32 |
except Exception as e:
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
#
|
| 38 |
-
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
|
| 43 |
-
|
| 44 |
-
|
| 45 |
-
|
| 46 |
-
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
""
|
| 50 |
-
|
| 51 |
-
|
| 52 |
-
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 56 |
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
"status": "✅ Modèle chargé avec succès",
|
| 60 |
-
"source": model_source,
|
| 61 |
-
"model_type": type(model).__name__,
|
| 62 |
-
"features": "~50 features (après preprocessing)",
|
| 63 |
-
"algorithme": "XGBoost + SMOTE",
|
| 64 |
-
"hf_hub_repo": HF_MODEL_REPO,
|
| 65 |
-
}
|
| 66 |
-
|
| 67 |
-
info["info"] = "Interface de prédiction en développement - API FastAPI à venir"
|
| 68 |
-
return info
|
| 69 |
|
| 70 |
-
|
| 71 |
-
|
| 72 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 73 |
|
| 74 |
-
|
| 75 |
-
with gr.Blocks( # type: ignore[attr-defined]
|
| 76 |
-
title="Employee Turnover Prediction - DEV", theme=gr.themes.Soft() # type: ignore[attr-defined]
|
| 77 |
-
) as demo:
|
| 78 |
-
gr.Markdown("# 🎯 Prédiction du Turnover - Employee Attrition") # type: ignore[attr-defined]
|
| 79 |
-
gr.Markdown("## Environment DEV - Test de déploiement CI/CD") # type: ignore[attr-defined]
|
| 80 |
|
| 81 |
-
gr.Markdown( # type: ignore[attr-defined]
|
| 82 |
-
"""
|
| 83 |
-
### 📊 Statut du projet
|
| 84 |
|
| 85 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 86 |
|
| 87 |
-
**Actuellement disponible :**
|
| 88 |
-
- ✅ Pipeline d'entraînement MLflow complet (`main.py`)
|
| 89 |
-
- ✅ Déploiement automatique CI/CD (GitHub Actions → HF Spaces)
|
| 90 |
-
- ✅ Tests unitaires et linting automatisés
|
| 91 |
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
- 🚧 API FastAPI avec endpoints de prédiction
|
| 95 |
-
- 🚧 Intégration PostgreSQL pour tracking des prédictions
|
| 96 |
"""
|
| 97 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 98 |
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 103 |
|
| 104 |
-
with gr.Column(): # type: ignore[attr-defined]
|
| 105 |
-
model_output = gr.JSON(label="Statut") # type: ignore[attr-defined]
|
| 106 |
|
| 107 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 108 |
|
| 109 |
-
|
| 110 |
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
### 🛠️ Prochaines étapes (selon etapes.txt)
|
| 114 |
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
- Chargement dynamique des preprocessing artifacts (scaler, encoders)
|
| 118 |
-
- Documentation Swagger/OpenAPI automatique
|
| 119 |
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
- Traçabilité complète des requêtes
|
| 123 |
|
| 124 |
-
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
- Couverture de code avec pytest-cov
|
| 128 |
|
| 129 |
-
|
| 130 |
-
|
| 131 |
-
|
| 132 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 133 |
"""
|
| 134 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 135 |
|
| 136 |
|
| 137 |
if __name__ == "__main__":
|
| 138 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
#!/usr/bin/env python3
|
| 2 |
"""
|
| 3 |
+
API FastAPI pour le modèle Employee Turnover.
|
| 4 |
|
| 5 |
+
Cette API expose le modèle de prédiction de départ des employés avec :
|
| 6 |
+
- Validation stricte des inputs via Pydantic
|
| 7 |
+
- Preprocessing automatique
|
| 8 |
+
- Health check pour monitoring
|
| 9 |
+
- Documentation OpenAPI/Swagger automatique
|
| 10 |
"""
|
| 11 |
+
import time
|
| 12 |
+
from contextlib import asynccontextmanager
|
| 13 |
|
| 14 |
+
from fastapi import Depends, FastAPI, HTTPException, Request
|
| 15 |
+
from fastapi.middleware.cors import CORSMiddleware
|
| 16 |
+
from slowapi import _rate_limit_exceeded_handler
|
| 17 |
+
from slowapi.errors import RateLimitExceeded
|
| 18 |
|
| 19 |
+
from src.auth import verify_api_key
|
| 20 |
+
from src.config import get_settings
|
| 21 |
+
from src.logger import logger, log_model_load, log_request
|
| 22 |
+
from src.models import get_model_info, load_model
|
| 23 |
+
from src.preprocessing import preprocess_for_prediction
|
| 24 |
+
from src.rate_limit import limiter
|
| 25 |
+
from src.schemas import EmployeeInput, HealthCheck, PredictionOutput
|
| 26 |
|
| 27 |
+
# Charger la configuration
|
| 28 |
+
settings = get_settings()
|
| 29 |
+
API_VERSION = settings.API_VERSION
|
| 30 |
+
|
| 31 |
+
|
| 32 |
+
@asynccontextmanager
|
| 33 |
+
async def lifespan(app: FastAPI):
|
| 34 |
"""
|
| 35 |
+
Gestion du cycle de vie de l'application.
|
| 36 |
|
| 37 |
+
Charge le modèle au démarrage et le garde en cache.
|
|
|
|
| 38 |
"""
|
| 39 |
+
logger.info(
|
| 40 |
+
"🚀 Démarrage de l'API Employee Turnover...", extra={"version": API_VERSION}
|
| 41 |
+
)
|
| 42 |
+
|
| 43 |
+
start_time = time.time()
|
| 44 |
try:
|
| 45 |
+
# Pré-charger le modèle au démarrage
|
| 46 |
+
model = load_model()
|
| 47 |
+
duration_ms = (time.time() - start_time) * 1000
|
| 48 |
|
| 49 |
+
model_type = type(model).__name__
|
| 50 |
+
log_model_load(model_type, duration_ms, True)
|
| 51 |
+
logger.info("✅ Modèle chargé avec succès")
|
|
|
|
|
|
|
|
|
|
|
|
|
| 52 |
except Exception as e:
|
| 53 |
+
duration_ms = (time.time() - start_time) * 1000
|
| 54 |
+
log_model_load("Unknown", duration_ms, False)
|
| 55 |
+
logger.error("Le modèle n'a pas pu être chargé", extra={"error": str(e)})
|
| 56 |
+
|
| 57 |
+
yield # L'application tourne
|
| 58 |
+
|
| 59 |
+
logger.info("🛑 Arrêt de l'API")
|
| 60 |
+
|
| 61 |
+
|
| 62 |
+
# Créer l'application FastAPI
|
| 63 |
+
app = FastAPI(
|
| 64 |
+
title="Employee Turnover Prediction API",
|
| 65 |
+
description="API de prédiction du turnover des employés avec XGBoost + SMOTE",
|
| 66 |
+
version=API_VERSION,
|
| 67 |
+
lifespan=lifespan,
|
| 68 |
+
docs_url="/docs",
|
| 69 |
+
redoc_url="/redoc",
|
| 70 |
+
)
|
| 71 |
+
|
| 72 |
+
# Ajouter rate limiting
|
| 73 |
+
app.state.limiter = limiter
|
| 74 |
+
app.add_exception_handler(RateLimitExceeded, _rate_limit_exceeded_handler)
|
| 75 |
+
|
| 76 |
+
# Configurer CORS (autoriser tous les domaines en dev)
|
| 77 |
+
app.add_middleware(
|
| 78 |
+
CORSMiddleware,
|
| 79 |
+
allow_origins=["*"],
|
| 80 |
+
allow_credentials=True,
|
| 81 |
+
allow_methods=["*"],
|
| 82 |
+
allow_headers=["*"],
|
| 83 |
+
)
|
| 84 |
+
|
| 85 |
+
|
| 86 |
+
# Middleware de logging des requêtes
|
| 87 |
+
@app.middleware("http")
|
| 88 |
+
async def log_requests(request: Request, call_next):
|
| 89 |
+
"""
|
| 90 |
+
Middleware pour logger toutes les requêtes HTTP.
|
| 91 |
+
"""
|
| 92 |
+
start_time = time.time()
|
| 93 |
|
| 94 |
+
# Traiter la requête
|
| 95 |
+
response = await call_next(request)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 96 |
|
| 97 |
+
# Calculer la durée
|
| 98 |
+
duration_ms = (time.time() - start_time) * 1000
|
| 99 |
|
| 100 |
+
# Logger
|
| 101 |
+
log_request(
|
| 102 |
+
method=request.method,
|
| 103 |
+
path=request.url.path,
|
| 104 |
+
status_code=response.status_code,
|
| 105 |
+
duration_ms=duration_ms,
|
| 106 |
+
client_host=request.client.host if request.client else None,
|
| 107 |
+
)
|
| 108 |
|
| 109 |
+
return response
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 110 |
|
|
|
|
|
|
|
|
|
|
| 111 |
|
| 112 |
+
@app.get("/", tags=["Root"])
|
| 113 |
+
async def root():
|
| 114 |
+
"""
|
| 115 |
+
Endpoint racine avec informations sur l'API.
|
| 116 |
+
"""
|
| 117 |
+
return {
|
| 118 |
+
"message": "Employee Turnover Prediction API",
|
| 119 |
+
"version": API_VERSION,
|
| 120 |
+
"docs": "/docs",
|
| 121 |
+
"health": "/health",
|
| 122 |
+
"predict": "/predict (POST)",
|
| 123 |
+
}
|
| 124 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 125 |
|
| 126 |
+
@app.get("/health", response_model=HealthCheck, tags=["Monitoring"])
|
| 127 |
+
async def health_check():
|
|
|
|
|
|
|
| 128 |
"""
|
| 129 |
+
Health check endpoint pour monitoring.
|
| 130 |
+
|
| 131 |
+
Vérifie que l'API est opérationnelle et que le modèle est chargé.
|
| 132 |
+
|
| 133 |
+
Returns:
|
| 134 |
+
HealthCheck: Status de l'API et du modèle.
|
| 135 |
|
| 136 |
+
Raises:
|
| 137 |
+
HTTPException: 503 si le modèle n'est pas disponible.
|
| 138 |
+
"""
|
| 139 |
+
try:
|
| 140 |
+
model_info = get_model_info()
|
| 141 |
+
|
| 142 |
+
return HealthCheck(
|
| 143 |
+
status="healthy",
|
| 144 |
+
model_loaded=model_info.get("cached", False),
|
| 145 |
+
model_type=model_info.get("model_type", "Unknown"),
|
| 146 |
+
version=API_VERSION,
|
| 147 |
+
)
|
| 148 |
+
except Exception as e:
|
| 149 |
+
raise HTTPException(
|
| 150 |
+
status_code=503,
|
| 151 |
+
detail={
|
| 152 |
+
"status": "unhealthy",
|
| 153 |
+
"error": "Model not available",
|
| 154 |
+
"message": str(e),
|
| 155 |
+
},
|
| 156 |
+
)
|
| 157 |
|
|
|
|
|
|
|
| 158 |
|
| 159 |
+
@app.post(
|
| 160 |
+
"/predict",
|
| 161 |
+
response_model=PredictionOutput,
|
| 162 |
+
tags=["Prediction"],
|
| 163 |
+
dependencies=[Depends(verify_api_key)] if settings.is_api_key_required else [],
|
| 164 |
+
)
|
| 165 |
+
@limiter.limit("20/minute")
|
| 166 |
+
async def predict(request: Request, employee: EmployeeInput):
|
| 167 |
+
"""
|
| 168 |
+
Endpoint de prédiction du turnover d'un employé.
|
| 169 |
|
| 170 |
+
**PROTÉGÉ PAR API KEY** : Requiert le header `X-API-Key` en production.
|
| 171 |
|
| 172 |
+
Prend en entrée les données d'un employé, applique le preprocessing
|
| 173 |
+
et retourne la prédiction avec les probabilités.
|
|
|
|
| 174 |
|
| 175 |
+
Args:
|
| 176 |
+
employee: Données de l'employé validées par Pydantic.
|
|
|
|
|
|
|
| 177 |
|
| 178 |
+
Returns:
|
| 179 |
+
PredictionOutput: Prédiction et probabilités.
|
|
|
|
| 180 |
|
| 181 |
+
Raises:
|
| 182 |
+
HTTPException: 401 si API key invalide ou manquante.
|
| 183 |
+
HTTPException: 500 si erreur lors de la prédiction.
|
|
|
|
| 184 |
|
| 185 |
+
Examples:
|
| 186 |
+
```bash
|
| 187 |
+
# Avec authentification
|
| 188 |
+
curl -X POST http://localhost:8000/predict \\
|
| 189 |
+
-H "X-API-Key: your-secret-key" \\
|
| 190 |
+
-H "Content-Type: application/json" \\
|
| 191 |
+
-d '{...}'
|
| 192 |
+
```
|
| 193 |
"""
|
| 194 |
+
try:
|
| 195 |
+
# 1. Charger le modèle
|
| 196 |
+
model = load_model()
|
| 197 |
+
|
| 198 |
+
# 2. Préprocessing
|
| 199 |
+
X = preprocess_for_prediction(employee)
|
| 200 |
+
|
| 201 |
+
# 3. Prédiction
|
| 202 |
+
prediction = int(model.predict(X)[0])
|
| 203 |
+
|
| 204 |
+
# 4. Probabilités (si le modèle supporte predict_proba)
|
| 205 |
+
try:
|
| 206 |
+
probabilities = model.predict_proba(X)[0]
|
| 207 |
+
prob_0 = float(probabilities[0])
|
| 208 |
+
prob_1 = float(probabilities[1])
|
| 209 |
+
except AttributeError:
|
| 210 |
+
# Si le modèle ne supporte pas predict_proba
|
| 211 |
+
prob_0 = 1.0 if prediction == 0 else 0.0
|
| 212 |
+
prob_1 = 1.0 if prediction == 1 else 0.0
|
| 213 |
+
|
| 214 |
+
# 5. Niveau de risque
|
| 215 |
+
if prob_1 < 0.3:
|
| 216 |
+
risk_level = "Low"
|
| 217 |
+
elif prob_1 < 0.7:
|
| 218 |
+
risk_level = "Medium"
|
| 219 |
+
else:
|
| 220 |
+
risk_level = "High"
|
| 221 |
+
|
| 222 |
+
return PredictionOutput(
|
| 223 |
+
prediction=prediction,
|
| 224 |
+
probability_0=prob_0,
|
| 225 |
+
probability_1=prob_1,
|
| 226 |
+
risk_level=risk_level,
|
| 227 |
+
)
|
| 228 |
+
|
| 229 |
+
except Exception:
|
| 230 |
+
logger.exception("Unexpected error during prediction")
|
| 231 |
+
raise HTTPException(
|
| 232 |
+
status_code=500,
|
| 233 |
+
detail={
|
| 234 |
+
"error": "Prediction failed",
|
| 235 |
+
"message": "An unexpected error occurred. Please contact support.",
|
| 236 |
+
},
|
| 237 |
+
)
|
| 238 |
|
| 239 |
|
| 240 |
if __name__ == "__main__":
|
| 241 |
+
import uvicorn
|
| 242 |
+
|
| 243 |
+
print("🚀 Lancement de l'API en mode développement...")
|
| 244 |
+
print("📖 Documentation : http://localhost:8000/docs")
|
| 245 |
+
|
| 246 |
+
uvicorn.run(
|
| 247 |
+
"app:app",
|
| 248 |
+
host="0.0.0.0",
|
| 249 |
+
port=8000,
|
| 250 |
+
reload=True,
|
| 251 |
+
log_level="info",
|
| 252 |
+
)
|
requirements.txt
CHANGED
|
@@ -1,9 +1,103 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
alembic==1.17.2 ; python_version >= "3.12" and python_version < "4.0"
|
| 2 |
+
annotated-types==0.7.0 ; python_version >= "3.12" and python_version < "4.0"
|
| 3 |
+
anyio==4.12.0 ; python_version >= "3.12" and python_version < "4.0"
|
| 4 |
+
blinker==1.9.0 ; python_version >= "3.12" and python_version < "4.0"
|
| 5 |
+
cachetools==6.2.4 ; python_version >= "3.12" and python_version < "4.0"
|
| 6 |
+
certifi==2025.11.12 ; python_version >= "3.12" and python_version < "4.0"
|
| 7 |
+
cffi==2.0.0 ; python_version >= "3.12" and python_version < "4.0" and platform_python_implementation != "PyPy"
|
| 8 |
+
charset-normalizer==3.4.4 ; python_version >= "3.12" and python_version < "4.0"
|
| 9 |
+
click==8.3.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 10 |
+
cloudpickle==3.1.2 ; python_version >= "3.12" and python_version < "4.0"
|
| 11 |
+
colorama==0.4.6 ; python_version >= "3.12" and python_version < "4.0" and (platform_system == "Windows" or sys_platform == "win32")
|
| 12 |
+
contourpy==1.3.3 ; python_version >= "3.12" and python_version < "4.0"
|
| 13 |
+
cryptography==46.0.3 ; python_version >= "3.12" and python_version < "4.0"
|
| 14 |
+
cycler==0.12.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 15 |
+
databricks-sdk==0.76.0 ; python_version >= "3.12" and python_version < "4.0"
|
| 16 |
+
deprecated==1.3.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 17 |
+
docker==7.1.0 ; python_version >= "3.12" and python_version < "4.0"
|
| 18 |
+
fastapi==0.115.14 ; python_version >= "3.12" and python_version < "4.0"
|
| 19 |
+
filelock==3.20.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 20 |
+
flask-cors==6.0.2 ; python_version >= "3.12" and python_version < "4.0"
|
| 21 |
+
flask==3.1.2 ; python_version >= "3.12" and python_version < "4.0"
|
| 22 |
+
fonttools==4.61.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 23 |
+
fsspec==2025.12.0 ; python_version >= "3.12" and python_version < "4.0"
|
| 24 |
+
gitdb==4.0.12 ; python_version >= "3.12" and python_version < "4.0"
|
| 25 |
+
gitpython==3.1.45 ; python_version >= "3.12" and python_version < "4.0"
|
| 26 |
+
google-auth==2.45.0 ; python_version >= "3.12" and python_version < "4.0"
|
| 27 |
+
graphene==3.4.3 ; python_version >= "3.12" and python_version < "4.0"
|
| 28 |
+
graphql-core==3.2.7 ; python_version >= "3.12" and python_version < "4.0"
|
| 29 |
+
graphql-relay==3.2.0 ; python_version >= "3.12" and python_version < "4.0"
|
| 30 |
+
greenlet==3.3.0 ; python_version >= "3.12" and python_version < "4.0" and (platform_machine == "aarch64" or platform_machine == "ppc64le" or platform_machine == "x86_64" or platform_machine == "amd64" or platform_machine == "AMD64" or platform_machine == "win32" or platform_machine == "WIN32")
|
| 31 |
+
gunicorn==23.0.0 ; python_version >= "3.12" and python_version < "4.0" and platform_system != "Windows"
|
| 32 |
+
h11==0.16.0 ; python_version >= "3.12" and python_version < "4.0"
|
| 33 |
+
hf-xet==1.2.0 ; python_version >= "3.12" and python_version < "4.0" and (platform_machine == "x86_64" or platform_machine == "amd64" or platform_machine == "AMD64" or platform_machine == "arm64" or platform_machine == "aarch64")
|
| 34 |
+
httpcore==1.0.9 ; python_version >= "3.12" and python_version < "4.0"
|
| 35 |
+
httptools==0.7.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 36 |
+
httpx==0.28.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 37 |
+
huey==2.5.5 ; python_version >= "3.12" and python_version < "4.0"
|
| 38 |
+
huggingface-hub==1.2.3 ; python_version >= "3.12" and python_version < "4.0"
|
| 39 |
+
idna==3.11 ; python_version >= "3.12" and python_version < "4.0"
|
| 40 |
+
imbalanced-learn==0.13.0 ; python_version >= "3.12" and python_version < "4.0"
|
| 41 |
+
importlib-metadata==8.7.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 42 |
+
itsdangerous==2.2.0 ; python_version >= "3.12" and python_version < "4.0"
|
| 43 |
+
jinja2==3.1.6 ; python_version >= "3.12" and python_version < "4.0"
|
| 44 |
+
joblib==1.5.3 ; python_version >= "3.12" and python_version < "4.0"
|
| 45 |
+
kiwisolver==1.4.9 ; python_version >= "3.12" and python_version < "4.0"
|
| 46 |
+
limits==5.6.0 ; python_version >= "3.12" and python_version < "4.0"
|
| 47 |
+
mako==1.3.10 ; python_version >= "3.12" and python_version < "4.0"
|
| 48 |
+
markupsafe==3.0.3 ; python_version >= "3.12" and python_version < "4.0"
|
| 49 |
+
matplotlib==3.10.8 ; python_version >= "3.12" and python_version < "4.0"
|
| 50 |
+
mlflow-skinny==3.8.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 51 |
+
mlflow-tracing==3.8.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 52 |
+
mlflow==3.8.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 53 |
+
numpy==2.4.0 ; python_version >= "3.12" and python_version < "4.0"
|
| 54 |
+
nvidia-nccl-cu12==2.28.9 ; python_version >= "3.12" and python_version < "4.0" and platform_system == "Linux" and platform_machine != "aarch64"
|
| 55 |
+
opentelemetry-api==1.39.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 56 |
+
opentelemetry-proto==1.39.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 57 |
+
opentelemetry-sdk==1.39.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 58 |
+
opentelemetry-semantic-conventions==0.60b1 ; python_version >= "3.12" and python_version < "4.0"
|
| 59 |
+
packaging==25.0 ; python_version >= "3.12" and python_version < "4.0"
|
| 60 |
+
pandas==2.3.3 ; python_version >= "3.12" and python_version < "4.0"
|
| 61 |
+
pillow==12.0.0 ; python_version >= "3.12" and python_version < "4.0"
|
| 62 |
+
protobuf==6.33.2 ; python_version >= "3.12" and python_version < "4.0"
|
| 63 |
+
pyarrow==22.0.0 ; python_version >= "3.12" and python_version < "4.0"
|
| 64 |
+
pyasn1-modules==0.4.2 ; python_version >= "3.12" and python_version < "4.0"
|
| 65 |
+
pyasn1==0.6.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 66 |
+
pycparser==2.23 ; python_version >= "3.12" and python_version < "4.0" and platform_python_implementation != "PyPy" and implementation_name != "PyPy"
|
| 67 |
+
pydantic-core==2.41.5 ; python_version >= "3.12" and python_version < "4.0"
|
| 68 |
+
pydantic==2.12.5 ; python_version >= "3.12" and python_version < "4.0"
|
| 69 |
+
pyparsing==3.3.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 70 |
+
python-dateutil==2.9.0.post0 ; python_version >= "3.12" and python_version < "4.0"
|
| 71 |
+
python-dotenv==1.2.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 72 |
+
python-json-logger==4.0.0 ; python_version >= "3.12" and python_version < "4.0"
|
| 73 |
+
pytz==2025.2 ; python_version >= "3.12" and python_version < "4.0"
|
| 74 |
+
pywin32==311 ; python_version >= "3.12" and python_version < "4.0" and sys_platform == "win32"
|
| 75 |
+
pyyaml==6.0.3 ; python_version >= "3.12" and python_version < "4.0"
|
| 76 |
+
requests==2.32.5 ; python_version >= "3.12" and python_version < "4.0"
|
| 77 |
+
rsa==4.9.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 78 |
+
scikit-learn==1.6.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 79 |
+
scipy==1.16.3 ; python_version >= "3.12" and python_version < "4.0"
|
| 80 |
+
shellingham==1.5.4 ; python_version >= "3.12" and python_version < "4.0"
|
| 81 |
+
six==1.17.0 ; python_version >= "3.12" and python_version < "4.0"
|
| 82 |
+
sklearn-compat==0.1.5 ; python_version >= "3.12" and python_version < "4.0"
|
| 83 |
+
slowapi==0.1.9 ; python_version >= "3.12" and python_version < "4.0"
|
| 84 |
+
smmap==5.0.2 ; python_version >= "3.12" and python_version < "4.0"
|
| 85 |
+
sqlalchemy==2.0.45 ; python_version >= "3.12" and python_version < "4.0"
|
| 86 |
+
sqlparse==0.5.5 ; python_version >= "3.12" and python_version < "4.0"
|
| 87 |
+
starlette==0.46.2 ; python_version >= "3.12" and python_version < "4.0"
|
| 88 |
+
threadpoolctl==3.6.0 ; python_version >= "3.12" and python_version < "4.0"
|
| 89 |
+
tqdm==4.67.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 90 |
+
typer-slim==0.21.0 ; python_version >= "3.12" and python_version < "4.0"
|
| 91 |
+
typing-extensions==4.15.0 ; python_version >= "3.12" and python_version < "4.0"
|
| 92 |
+
typing-inspection==0.4.2 ; python_version >= "3.12" and python_version < "4.0"
|
| 93 |
+
tzdata==2025.3 ; python_version >= "3.12" and python_version < "4.0"
|
| 94 |
+
urllib3==2.6.2 ; python_version >= "3.12" and python_version < "4.0"
|
| 95 |
+
uvicorn==0.32.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 96 |
+
uvloop==0.22.1 ; python_version >= "3.12" and python_version < "4.0" and sys_platform != "win32" and sys_platform != "cygwin" and platform_python_implementation != "PyPy"
|
| 97 |
+
waitress==3.0.2 ; python_version >= "3.12" and python_version < "4.0" and platform_system == "Windows"
|
| 98 |
+
watchfiles==1.1.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 99 |
+
websockets==15.0.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 100 |
+
werkzeug==3.1.4 ; python_version >= "3.12" and python_version < "4.0"
|
| 101 |
+
wrapt==2.0.1 ; python_version >= "3.12" and python_version < "4.0"
|
| 102 |
+
xgboost==2.1.4 ; python_version >= "3.12" and python_version < "4.0"
|
| 103 |
+
zipp==3.23.0 ; python_version >= "3.12" and python_version < "4.0"
|
src/__init__.py
CHANGED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
"""Module src pour l'API FastAPI."""
|
src/auth.py
ADDED
|
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env python3
|
| 2 |
+
"""
|
| 3 |
+
Module d'authentification pour l'API.
|
| 4 |
+
|
| 5 |
+
Fournit un système de vérification de clé API via header HTTP.
|
| 6 |
+
"""
|
| 7 |
+
from fastapi import Header, HTTPException, status
|
| 8 |
+
from fastapi.security import APIKeyHeader
|
| 9 |
+
|
| 10 |
+
from src.config import get_settings
|
| 11 |
+
|
| 12 |
+
# Schéma pour la documentation Swagger
|
| 13 |
+
api_key_header = APIKeyHeader(name="X-API-Key", auto_error=False)
|
| 14 |
+
|
| 15 |
+
|
| 16 |
+
async def verify_api_key(x_api_key: str = Header(None)) -> str:
|
| 17 |
+
"""
|
| 18 |
+
Vérifie que la clé API fournie est valide.
|
| 19 |
+
|
| 20 |
+
Cette fonction est utilisée comme dépendance FastAPI (Depends).
|
| 21 |
+
Elle vérifie le header HTTP "X-API-Key" et compare avec la clé configurée.
|
| 22 |
+
|
| 23 |
+
Args:
|
| 24 |
+
x_api_key: Clé API fournie dans le header HTTP.
|
| 25 |
+
|
| 26 |
+
Returns:
|
| 27 |
+
str: La clé API validée.
|
| 28 |
+
|
| 29 |
+
Raises:
|
| 30 |
+
HTTPException: 401 si la clé est manquante ou invalide.
|
| 31 |
+
|
| 32 |
+
Comment ça marche :
|
| 33 |
+
1. FastAPI extrait automatiquement le header "X-API-Key"
|
| 34 |
+
2. La fonction compare avec la clé configurée dans .env
|
| 35 |
+
3. Si valide → continue, sinon → erreur 401
|
| 36 |
+
|
| 37 |
+
Exemple d'utilisation :
|
| 38 |
+
```python
|
| 39 |
+
@app.post("/predict", dependencies=[Depends(verify_api_key)])
|
| 40 |
+
async def predict(...):
|
| 41 |
+
# Cette route est protégée !
|
| 42 |
+
```
|
| 43 |
+
|
| 44 |
+
Exemple de requête curl :
|
| 45 |
+
```bash
|
| 46 |
+
curl -X POST http://localhost:8000/predict \\
|
| 47 |
+
-H "X-API-Key: your-secret-key" \\
|
| 48 |
+
-H "Content-Type: application/json" \\
|
| 49 |
+
-d '{...}'
|
| 50 |
+
```
|
| 51 |
+
"""
|
| 52 |
+
settings = get_settings()
|
| 53 |
+
|
| 54 |
+
# En mode DEBUG, on peut désactiver l'auth
|
| 55 |
+
if settings.DEBUG:
|
| 56 |
+
return "debug-mode-no-auth-required"
|
| 57 |
+
|
| 58 |
+
# Vérifier que la clé est fournie
|
| 59 |
+
if not x_api_key:
|
| 60 |
+
raise HTTPException(
|
| 61 |
+
status_code=status.HTTP_401_UNAUTHORIZED,
|
| 62 |
+
detail={
|
| 63 |
+
"error": "API Key missing",
|
| 64 |
+
"message": "Le header 'X-API-Key' est requis pour accéder à cette ressource",
|
| 65 |
+
"solution": "Ajoutez le header: -H 'X-API-Key: votre-cle-api'",
|
| 66 |
+
},
|
| 67 |
+
headers={"WWW-Authenticate": "ApiKey"},
|
| 68 |
+
)
|
| 69 |
+
|
| 70 |
+
# Vérifier que la clé est correcte
|
| 71 |
+
if x_api_key != settings.API_KEY:
|
| 72 |
+
raise HTTPException(
|
| 73 |
+
status_code=status.HTTP_401_UNAUTHORIZED,
|
| 74 |
+
detail={
|
| 75 |
+
"error": "Invalid API Key",
|
| 76 |
+
"message": "La clé API fournie est invalide",
|
| 77 |
+
"solution": "Vérifiez votre clé API ou contactez l'administrateur",
|
| 78 |
+
},
|
| 79 |
+
headers={"WWW-Authenticate": "ApiKey"},
|
| 80 |
+
)
|
| 81 |
+
|
| 82 |
+
return x_api_key
|
| 83 |
+
|
| 84 |
+
|
| 85 |
+
def get_api_key_dependency():
|
| 86 |
+
"""
|
| 87 |
+
Retourne la dépendance d'authentification si nécessaire.
|
| 88 |
+
|
| 89 |
+
Permet de conditionner l'authentification selon la config.
|
| 90 |
+
|
| 91 |
+
Returns:
|
| 92 |
+
Depends(verify_api_key) si auth requise, None sinon.
|
| 93 |
+
"""
|
| 94 |
+
settings = get_settings()
|
| 95 |
+
if settings.is_api_key_required:
|
| 96 |
+
from fastapi import Depends
|
| 97 |
+
|
| 98 |
+
return Depends(verify_api_key)
|
| 99 |
+
return None
|
src/config.py
ADDED
|
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env python3
|
| 2 |
+
"""
|
| 3 |
+
Module de configuration de l'application.
|
| 4 |
+
|
| 5 |
+
Charge les variables d'environnement depuis .env et fournit
|
| 6 |
+
une interface pour accéder à la configuration de manière sécurisée.
|
| 7 |
+
"""
|
| 8 |
+
import os
|
| 9 |
+
from functools import lru_cache
|
| 10 |
+
|
| 11 |
+
from dotenv import load_dotenv
|
| 12 |
+
|
| 13 |
+
# Charger .env au démarrage du module
|
| 14 |
+
load_dotenv()
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
class Settings:
|
| 18 |
+
"""
|
| 19 |
+
Configuration de l'application.
|
| 20 |
+
|
| 21 |
+
Toutes les valeurs sensibles (API keys, etc.) sont chargées depuis
|
| 22 |
+
les variables d'environnement ou le fichier .env.
|
| 23 |
+
"""
|
| 24 |
+
|
| 25 |
+
# ===== SÉCURITÉ =====
|
| 26 |
+
API_KEY: str = os.getenv("API_KEY", "dev-key-change-me-in-production")
|
| 27 |
+
|
| 28 |
+
# ===== API =====
|
| 29 |
+
API_VERSION: str = os.getenv("API_VERSION", "1.0.0")
|
| 30 |
+
API_HOST: str = os.getenv("API_HOST", "0.0.0.0")
|
| 31 |
+
API_PORT: int = int(os.getenv("API_PORT", "8000"))
|
| 32 |
+
|
| 33 |
+
# ===== MODÈLE =====
|
| 34 |
+
HF_MODEL_REPO: str = os.getenv(
|
| 35 |
+
"HF_MODEL_REPO", "ASI-Engineer/employee-turnover-model"
|
| 36 |
+
)
|
| 37 |
+
MODEL_FILENAME: str = os.getenv("MODEL_FILENAME", "model/model.pkl")
|
| 38 |
+
|
| 39 |
+
# ===== ENVIRONNEMENT =====
|
| 40 |
+
DEBUG: bool = os.getenv("DEBUG", "False").lower() == "true"
|
| 41 |
+
LOG_LEVEL: str = os.getenv("LOG_LEVEL", "INFO")
|
| 42 |
+
|
| 43 |
+
@property
|
| 44 |
+
def is_api_key_required(self) -> bool:
|
| 45 |
+
"""
|
| 46 |
+
Vérifie si l'API key est requise.
|
| 47 |
+
|
| 48 |
+
Returns:
|
| 49 |
+
False en mode DEBUG, True en production.
|
| 50 |
+
"""
|
| 51 |
+
return not self.DEBUG
|
| 52 |
+
|
| 53 |
+
|
| 54 |
+
@lru_cache()
|
| 55 |
+
def get_settings() -> Settings:
|
| 56 |
+
"""
|
| 57 |
+
Retourne l'instance singleton des settings.
|
| 58 |
+
|
| 59 |
+
Le décorateur @lru_cache() assure qu'on ne crée qu'une seule instance.
|
| 60 |
+
|
| 61 |
+
Returns:
|
| 62 |
+
Settings: Configuration de l'application.
|
| 63 |
+
"""
|
| 64 |
+
return Settings()
|
src/logger.py
ADDED
|
@@ -0,0 +1,223 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env python3
|
| 2 |
+
"""
|
| 3 |
+
Module de logging structuré pour l'API Employee Turnover.
|
| 4 |
+
|
| 5 |
+
Fournit un système de logging centralisé avec :
|
| 6 |
+
- Logs structurés en JSON
|
| 7 |
+
- Rotation automatique des fichiers
|
| 8 |
+
- Niveaux de log configurables
|
| 9 |
+
- Intégration FastAPI
|
| 10 |
+
"""
|
| 11 |
+
import logging
|
| 12 |
+
import sys
|
| 13 |
+
from pathlib import Path
|
| 14 |
+
from typing import Any, Dict
|
| 15 |
+
|
| 16 |
+
from pythonjsonlogger import jsonlogger
|
| 17 |
+
|
| 18 |
+
from src.config import get_settings
|
| 19 |
+
|
| 20 |
+
settings = get_settings()
|
| 21 |
+
|
| 22 |
+
# Créer le dossier logs s'il n'existe pas
|
| 23 |
+
LOG_DIR = Path("logs")
|
| 24 |
+
LOG_DIR.mkdir(exist_ok=True)
|
| 25 |
+
|
| 26 |
+
# Fichiers de logs
|
| 27 |
+
LOG_FILE = LOG_DIR / "api.log"
|
| 28 |
+
ERROR_LOG_FILE = LOG_DIR / "error.log"
|
| 29 |
+
|
| 30 |
+
|
| 31 |
+
class CustomJsonFormatter(jsonlogger.JsonFormatter):
|
| 32 |
+
"""
|
| 33 |
+
Formatter JSON personnalisé avec champs supplémentaires.
|
| 34 |
+
"""
|
| 35 |
+
|
| 36 |
+
def add_fields(
|
| 37 |
+
self,
|
| 38 |
+
log_record: Dict[str, Any],
|
| 39 |
+
record: logging.LogRecord,
|
| 40 |
+
message_dict: Dict[str, Any],
|
| 41 |
+
) -> None:
|
| 42 |
+
"""
|
| 43 |
+
Ajoute des champs personnalisés aux logs JSON.
|
| 44 |
+
"""
|
| 45 |
+
super().add_fields(log_record, record, message_dict)
|
| 46 |
+
|
| 47 |
+
# Ajouter des métadonnées
|
| 48 |
+
log_record["level"] = record.levelname
|
| 49 |
+
log_record["logger"] = record.name
|
| 50 |
+
log_record["module"] = record.module
|
| 51 |
+
log_record["function"] = record.funcName
|
| 52 |
+
log_record["line"] = record.lineno
|
| 53 |
+
|
| 54 |
+
# Timestamp ISO 8601
|
| 55 |
+
if not log_record.get("timestamp"):
|
| 56 |
+
log_record["timestamp"] = self.formatTime(record, self.datefmt)
|
| 57 |
+
|
| 58 |
+
|
| 59 |
+
def setup_logger(name: str = "employee_turnover_api") -> logging.Logger:
|
| 60 |
+
"""
|
| 61 |
+
Configure et retourne un logger structuré.
|
| 62 |
+
|
| 63 |
+
Args:
|
| 64 |
+
name: Nom du logger.
|
| 65 |
+
|
| 66 |
+
Returns:
|
| 67 |
+
Logger configuré avec handlers console et fichiers.
|
| 68 |
+
|
| 69 |
+
Examples:
|
| 70 |
+
>>> logger = setup_logger()
|
| 71 |
+
>>> logger.info("API démarrée", extra={"version": "2.0.0"})
|
| 72 |
+
"""
|
| 73 |
+
logger = logging.getLogger(name)
|
| 74 |
+
|
| 75 |
+
# Éviter duplication si déjà configuré
|
| 76 |
+
if logger.handlers:
|
| 77 |
+
return logger
|
| 78 |
+
|
| 79 |
+
# Niveau de log depuis configuration
|
| 80 |
+
log_level = getattr(logging, settings.LOG_LEVEL.upper(), logging.INFO)
|
| 81 |
+
logger.setLevel(log_level)
|
| 82 |
+
|
| 83 |
+
# === HANDLER CONSOLE (stdout) ===
|
| 84 |
+
console_handler = logging.StreamHandler(sys.stdout)
|
| 85 |
+
console_handler.setLevel(log_level)
|
| 86 |
+
|
| 87 |
+
# Format simple pour la console en dev, JSON en prod
|
| 88 |
+
if settings.DEBUG:
|
| 89 |
+
console_format = logging.Formatter(
|
| 90 |
+
"%(asctime)s - %(name)s - %(levelname)s - %(message)s",
|
| 91 |
+
datefmt="%Y-%m-%d %H:%M:%S",
|
| 92 |
+
)
|
| 93 |
+
else:
|
| 94 |
+
console_format = CustomJsonFormatter(
|
| 95 |
+
"%(timestamp)s %(level)s %(name)s %(message)s"
|
| 96 |
+
)
|
| 97 |
+
|
| 98 |
+
console_handler.setFormatter(console_format)
|
| 99 |
+
logger.addHandler(console_handler)
|
| 100 |
+
|
| 101 |
+
# === HANDLER FICHIER (tous les logs) ===
|
| 102 |
+
file_handler = logging.FileHandler(LOG_FILE, encoding="utf-8")
|
| 103 |
+
file_handler.setLevel(log_level)
|
| 104 |
+
file_handler.setFormatter(
|
| 105 |
+
CustomJsonFormatter("%(timestamp)s %(level)s %(name)s %(message)s")
|
| 106 |
+
)
|
| 107 |
+
logger.addHandler(file_handler)
|
| 108 |
+
|
| 109 |
+
# === HANDLER ERREURS UNIQUEMENT ===
|
| 110 |
+
error_handler = logging.FileHandler(ERROR_LOG_FILE, encoding="utf-8")
|
| 111 |
+
error_handler.setLevel(logging.ERROR)
|
| 112 |
+
error_handler.setFormatter(
|
| 113 |
+
CustomJsonFormatter("%(timestamp)s %(level)s %(name)s %(message)s")
|
| 114 |
+
)
|
| 115 |
+
logger.addHandler(error_handler)
|
| 116 |
+
|
| 117 |
+
# Éviter propagation au root logger
|
| 118 |
+
logger.propagate = False
|
| 119 |
+
|
| 120 |
+
return logger
|
| 121 |
+
|
| 122 |
+
|
| 123 |
+
def log_request(
|
| 124 |
+
method: str,
|
| 125 |
+
path: str,
|
| 126 |
+
status_code: int,
|
| 127 |
+
duration_ms: float,
|
| 128 |
+
**kwargs: Any,
|
| 129 |
+
) -> None:
|
| 130 |
+
"""
|
| 131 |
+
Log une requête HTTP avec métadonnées.
|
| 132 |
+
|
| 133 |
+
Args:
|
| 134 |
+
method: Méthode HTTP (GET, POST...).
|
| 135 |
+
path: Chemin de l'endpoint.
|
| 136 |
+
status_code: Code de statut HTTP.
|
| 137 |
+
duration_ms: Durée de la requête en millisecondes.
|
| 138 |
+
**kwargs: Métadonnées additionnelles.
|
| 139 |
+
|
| 140 |
+
Examples:
|
| 141 |
+
>>> log_request("POST", "/predict", 200, 45.3, user_id="123")
|
| 142 |
+
"""
|
| 143 |
+
logger = logging.getLogger("employee_turnover_api")
|
| 144 |
+
|
| 145 |
+
log_data = {
|
| 146 |
+
"method": method,
|
| 147 |
+
"path": path,
|
| 148 |
+
"status_code": status_code,
|
| 149 |
+
"duration_ms": round(duration_ms, 2),
|
| 150 |
+
**kwargs,
|
| 151 |
+
}
|
| 152 |
+
|
| 153 |
+
# Niveau selon status code
|
| 154 |
+
if status_code >= 500:
|
| 155 |
+
logger.error(f"Request {method} {path}", extra=log_data)
|
| 156 |
+
elif status_code >= 400:
|
| 157 |
+
logger.warning(f"Request {method} {path}", extra=log_data)
|
| 158 |
+
else:
|
| 159 |
+
logger.info(f"Request {method} {path}", extra=log_data)
|
| 160 |
+
|
| 161 |
+
|
| 162 |
+
def log_prediction(
|
| 163 |
+
employee_id: str | None,
|
| 164 |
+
prediction: int,
|
| 165 |
+
probability: float,
|
| 166 |
+
risk_level: str,
|
| 167 |
+
duration_ms: float,
|
| 168 |
+
) -> None:
|
| 169 |
+
"""
|
| 170 |
+
Log une prédiction effectuée.
|
| 171 |
+
|
| 172 |
+
Args:
|
| 173 |
+
employee_id: ID de l'employé (optionnel).
|
| 174 |
+
prediction: Prédiction (0 ou 1).
|
| 175 |
+
probability: Probabilité de turnover.
|
| 176 |
+
risk_level: Niveau de risque ("low", "medium", "high").
|
| 177 |
+
duration_ms: Durée du preprocessing + pr��diction.
|
| 178 |
+
|
| 179 |
+
Examples:
|
| 180 |
+
>>> log_prediction("EMP123", 1, 0.87, "high", 23.4)
|
| 181 |
+
"""
|
| 182 |
+
logger = logging.getLogger("employee_turnover_api")
|
| 183 |
+
|
| 184 |
+
logger.info(
|
| 185 |
+
"Prediction made",
|
| 186 |
+
extra={
|
| 187 |
+
"employee_id": employee_id,
|
| 188 |
+
"prediction": prediction,
|
| 189 |
+
"probability": round(probability, 4),
|
| 190 |
+
"risk_level": risk_level,
|
| 191 |
+
"duration_ms": round(duration_ms, 2),
|
| 192 |
+
},
|
| 193 |
+
)
|
| 194 |
+
|
| 195 |
+
|
| 196 |
+
def log_model_load(model_type: str, duration_ms: float, success: bool) -> None:
|
| 197 |
+
"""
|
| 198 |
+
Log le chargement du modèle.
|
| 199 |
+
|
| 200 |
+
Args:
|
| 201 |
+
model_type: Type de modèle chargé.
|
| 202 |
+
duration_ms: Durée du chargement.
|
| 203 |
+
success: Si le chargement a réussi.
|
| 204 |
+
|
| 205 |
+
Examples:
|
| 206 |
+
>>> log_model_load("XGBoost Pipeline", 1234.5, True)
|
| 207 |
+
"""
|
| 208 |
+
logger = logging.getLogger("employee_turnover_api")
|
| 209 |
+
|
| 210 |
+
log_data = {
|
| 211 |
+
"model_type": model_type,
|
| 212 |
+
"duration_ms": round(duration_ms, 2),
|
| 213 |
+
"success": success,
|
| 214 |
+
}
|
| 215 |
+
|
| 216 |
+
if success:
|
| 217 |
+
logger.info("Model loaded successfully", extra=log_data)
|
| 218 |
+
else:
|
| 219 |
+
logger.error("Model loading failed", extra=log_data)
|
| 220 |
+
|
| 221 |
+
|
| 222 |
+
# Créer le logger global
|
| 223 |
+
logger = setup_logger()
|
src/models.py
ADDED
|
@@ -0,0 +1,153 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env python3
|
| 2 |
+
"""
|
| 3 |
+
Module de chargement et gestion du modèle MLflow.
|
| 4 |
+
|
| 5 |
+
Ce module encapsule la logique de chargement du modèle depuis Hugging Face Hub
|
| 6 |
+
via MLflow, avec gestion des erreurs et versioning.
|
| 7 |
+
"""
|
| 8 |
+
from typing import Any, Optional
|
| 9 |
+
|
| 10 |
+
from fastapi import HTTPException
|
| 11 |
+
from huggingface_hub import hf_hub_download
|
| 12 |
+
|
| 13 |
+
# Configuration
|
| 14 |
+
HF_MODEL_REPO = "ASI-Engineer/employee-turnover-model"
|
| 15 |
+
MODEL_FILENAME = "model/model.pkl"
|
| 16 |
+
|
| 17 |
+
# Cache global du modèle
|
| 18 |
+
_model_cache: Optional[Any] = None
|
| 19 |
+
|
| 20 |
+
|
| 21 |
+
def load_model(force_reload: bool = False) -> Any:
|
| 22 |
+
"""
|
| 23 |
+
Charge le modèle depuis Hugging Face Hub via MLflow.
|
| 24 |
+
|
| 25 |
+
Cette fonction implémente un système de cache pour éviter de recharger
|
| 26 |
+
le modèle à chaque appel. Le modèle est chargé une seule fois au démarrage
|
| 27 |
+
de l'application et mis en cache.
|
| 28 |
+
|
| 29 |
+
Args:
|
| 30 |
+
force_reload: Si True, force le rechargement du modèle même s'il est en cache.
|
| 31 |
+
|
| 32 |
+
Returns:
|
| 33 |
+
Le modèle MLflow chargé et prêt pour l'inférence.
|
| 34 |
+
|
| 35 |
+
Raises:
|
| 36 |
+
HTTPException: 500 si le modèle ne peut pas être chargé.
|
| 37 |
+
|
| 38 |
+
Examples:
|
| 39 |
+
>>> model = load_model()
|
| 40 |
+
>>> # Utiliser le modèle pour prédiction
|
| 41 |
+
>>> predictions = model.predict(X)
|
| 42 |
+
"""
|
| 43 |
+
global _model_cache
|
| 44 |
+
|
| 45 |
+
# Retourner le modèle en cache si disponible
|
| 46 |
+
if _model_cache is not None and not force_reload:
|
| 47 |
+
return _model_cache
|
| 48 |
+
|
| 49 |
+
try:
|
| 50 |
+
import joblib
|
| 51 |
+
|
| 52 |
+
print(f"🔄 Chargement du modèle depuis HF Hub: {HF_MODEL_REPO}")
|
| 53 |
+
|
| 54 |
+
# Télécharger le modèle depuis Hugging Face Hub
|
| 55 |
+
model_path = hf_hub_download(
|
| 56 |
+
repo_id=HF_MODEL_REPO, filename=MODEL_FILENAME, repo_type="model"
|
| 57 |
+
)
|
| 58 |
+
|
| 59 |
+
print(f"📦 Modèle téléchargé: {model_path}")
|
| 60 |
+
|
| 61 |
+
# Charger le modèle avec joblib
|
| 62 |
+
model = joblib.load(model_path)
|
| 63 |
+
|
| 64 |
+
# Mettre en cache
|
| 65 |
+
_model_cache = model
|
| 66 |
+
|
| 67 |
+
print(f"✅ Modèle chargé avec succès: {type(model).__name__}")
|
| 68 |
+
return model
|
| 69 |
+
|
| 70 |
+
except Exception as e:
|
| 71 |
+
error_msg = f"❌ Erreur lors du chargement du modèle: {str(e)}"
|
| 72 |
+
print(error_msg)
|
| 73 |
+
raise HTTPException(
|
| 74 |
+
status_code=500,
|
| 75 |
+
detail={
|
| 76 |
+
"error": "Model loading failed",
|
| 77 |
+
"message": str(e),
|
| 78 |
+
"model_repo": HF_MODEL_REPO,
|
| 79 |
+
"solution": "Vérifiez que le modèle est disponible sur HF Hub et correctement entraîné",
|
| 80 |
+
},
|
| 81 |
+
)
|
| 82 |
+
|
| 83 |
+
|
| 84 |
+
def get_model_info() -> dict:
|
| 85 |
+
"""
|
| 86 |
+
Retourne les informations sur le modèle chargé.
|
| 87 |
+
|
| 88 |
+
Returns:
|
| 89 |
+
Dict contenant les métadonnées du modèle.
|
| 90 |
+
|
| 91 |
+
Raises:
|
| 92 |
+
HTTPException: 500 si le modèle n'est pas chargé.
|
| 93 |
+
"""
|
| 94 |
+
try:
|
| 95 |
+
model = load_model()
|
| 96 |
+
|
| 97 |
+
return {
|
| 98 |
+
"status": "✅ Modèle chargé",
|
| 99 |
+
"model_type": type(model).__name__,
|
| 100 |
+
"hf_hub_repo": HF_MODEL_REPO,
|
| 101 |
+
"model_file": MODEL_FILENAME,
|
| 102 |
+
"cached": _model_cache is not None,
|
| 103 |
+
}
|
| 104 |
+
|
| 105 |
+
except Exception as e:
|
| 106 |
+
raise HTTPException(
|
| 107 |
+
status_code=500,
|
| 108 |
+
detail={"error": "Model info unavailable", "message": str(e)},
|
| 109 |
+
)
|
| 110 |
+
|
| 111 |
+
|
| 112 |
+
def load_preprocessing_artifacts(run_id: str) -> dict:
|
| 113 |
+
"""
|
| 114 |
+
Charge les artifacts de preprocessing (scaler, encoders) depuis MLflow.
|
| 115 |
+
|
| 116 |
+
Args:
|
| 117 |
+
run_id: ID du run MLflow contenant les artifacts.
|
| 118 |
+
|
| 119 |
+
Returns:
|
| 120 |
+
Dict contenant les artifacts de preprocessing.
|
| 121 |
+
|
| 122 |
+
Raises:
|
| 123 |
+
HTTPException: 500 si les artifacts ne peuvent pas être chargés.
|
| 124 |
+
|
| 125 |
+
Note:
|
| 126 |
+
Cette fonction sera implémentée quand les preprocessing artifacts
|
| 127 |
+
seront disponibles dans le modèle HF Hub.
|
| 128 |
+
"""
|
| 129 |
+
raise NotImplementedError(
|
| 130 |
+
"Le chargement des preprocessing artifacts sera implémenté "
|
| 131 |
+
"lors de l'intégration complète avec MLflow"
|
| 132 |
+
)
|
| 133 |
+
|
| 134 |
+
|
| 135 |
+
if __name__ == "__main__":
|
| 136 |
+
# Test de chargement du modèle
|
| 137 |
+
print("=" * 80)
|
| 138 |
+
print("TEST DE CHARGEMENT DU MODÈLE")
|
| 139 |
+
print("=" * 80)
|
| 140 |
+
|
| 141 |
+
try:
|
| 142 |
+
model = load_model()
|
| 143 |
+
print("\n✅ Test réussi!")
|
| 144 |
+
print(f"Type de modèle: {type(model).__name__}")
|
| 145 |
+
|
| 146 |
+
# Afficher les infos
|
| 147 |
+
info = get_model_info()
|
| 148 |
+
print("\nInformations du modèle:")
|
| 149 |
+
for key, value in info.items():
|
| 150 |
+
print(f" {key}: {value}")
|
| 151 |
+
|
| 152 |
+
except Exception as e:
|
| 153 |
+
print(f"\n❌ Test échoué: {e}")
|
src/preprocessing.py
ADDED
|
@@ -0,0 +1,243 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env python3
|
| 2 |
+
"""
|
| 3 |
+
Module de preprocessing pour transformer les données d'entrée avant prédiction.
|
| 4 |
+
|
| 5 |
+
Ce module applique les mêmes transformations que le pipeline d'entraînement :
|
| 6 |
+
- Feature engineering (ratios, moyennes)
|
| 7 |
+
- Encoding (OneHot, Ordinal)
|
| 8 |
+
- Scaling (StandardScaler)
|
| 9 |
+
"""
|
| 10 |
+
import numpy as np
|
| 11 |
+
import pandas as pd
|
| 12 |
+
from sklearn.preprocessing import OneHotEncoder, OrdinalEncoder, StandardScaler
|
| 13 |
+
|
| 14 |
+
from src.schemas import EmployeeInput
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
def create_input_dataframe(employee: EmployeeInput) -> pd.DataFrame:
|
| 18 |
+
"""
|
| 19 |
+
Convertit un objet EmployeeInput Pydantic en DataFrame pandas.
|
| 20 |
+
|
| 21 |
+
Args:
|
| 22 |
+
employee: Données validées d'un employé.
|
| 23 |
+
|
| 24 |
+
Returns:
|
| 25 |
+
DataFrame avec une seule ligne contenant toutes les features.
|
| 26 |
+
"""
|
| 27 |
+
data = {
|
| 28 |
+
# SONDAGE
|
| 29 |
+
"nombre_participation_pee": [employee.nombre_participation_pee],
|
| 30 |
+
"nb_formations_suivies": [employee.nb_formations_suivies],
|
| 31 |
+
"nombre_employee_sous_responsabilite": [
|
| 32 |
+
employee.nombre_employee_sous_responsabilite
|
| 33 |
+
],
|
| 34 |
+
"distance_domicile_travail": [employee.distance_domicile_travail],
|
| 35 |
+
"niveau_education": [employee.niveau_education],
|
| 36 |
+
"domaine_etude": [employee.domaine_etude],
|
| 37 |
+
"ayant_enfants": [employee.ayant_enfants],
|
| 38 |
+
"frequence_deplacement": [employee.frequence_deplacement],
|
| 39 |
+
"annees_depuis_la_derniere_promotion": [
|
| 40 |
+
employee.annees_depuis_la_derniere_promotion
|
| 41 |
+
],
|
| 42 |
+
"annes_sous_responsable_actuel": [employee.annes_sous_responsable_actuel],
|
| 43 |
+
# EVALUATION
|
| 44 |
+
"satisfaction_employee_environnement": [
|
| 45 |
+
employee.satisfaction_employee_environnement
|
| 46 |
+
],
|
| 47 |
+
"note_evaluation_precedente": [employee.note_evaluation_precedente],
|
| 48 |
+
"niveau_hierarchique_poste": [employee.niveau_hierarchique_poste],
|
| 49 |
+
"satisfaction_employee_nature_travail": [
|
| 50 |
+
employee.satisfaction_employee_nature_travail
|
| 51 |
+
],
|
| 52 |
+
"satisfaction_employee_equipe": [employee.satisfaction_employee_equipe],
|
| 53 |
+
"satisfaction_employee_equilibre_pro_perso": [
|
| 54 |
+
employee.satisfaction_employee_equilibre_pro_perso
|
| 55 |
+
],
|
| 56 |
+
"note_evaluation_actuelle": [employee.note_evaluation_actuelle],
|
| 57 |
+
"heure_supplementaires": [employee.heure_supplementaires],
|
| 58 |
+
"augementation_salaire_precedente": [employee.augementation_salaire_precedente],
|
| 59 |
+
# SIRH
|
| 60 |
+
"age": [employee.age],
|
| 61 |
+
"genre": [employee.genre],
|
| 62 |
+
"revenu_mensuel": [employee.revenu_mensuel],
|
| 63 |
+
"statut_marital": [employee.statut_marital],
|
| 64 |
+
"departement": [employee.departement],
|
| 65 |
+
"poste": [employee.poste],
|
| 66 |
+
"nombre_experiences_precedentes": [employee.nombre_experiences_precedentes],
|
| 67 |
+
"nombre_heures_travailless": [employee.nombre_heures_travailless],
|
| 68 |
+
"annee_experience_totale": [employee.annee_experience_totale],
|
| 69 |
+
"annees_dans_l_entreprise": [employee.annees_dans_l_entreprise],
|
| 70 |
+
"annees_dans_le_poste_actuel": [employee.annees_dans_le_poste_actuel],
|
| 71 |
+
}
|
| 72 |
+
|
| 73 |
+
return pd.DataFrame(data)
|
| 74 |
+
|
| 75 |
+
|
| 76 |
+
def engineer_features(df: pd.DataFrame) -> pd.DataFrame:
|
| 77 |
+
"""
|
| 78 |
+
Applique le feature engineering (mêmes transformations que l'entraînement).
|
| 79 |
+
|
| 80 |
+
Args:
|
| 81 |
+
df: DataFrame avec les colonnes brutes.
|
| 82 |
+
|
| 83 |
+
Returns:
|
| 84 |
+
DataFrame avec les features engineered ajoutées.
|
| 85 |
+
"""
|
| 86 |
+
df = df.copy()
|
| 87 |
+
|
| 88 |
+
# Ratios (+ 1 pour éviter division par zéro)
|
| 89 |
+
df["revenu_par_anciennete"] = df["revenu_mensuel"] / (
|
| 90 |
+
df["annees_dans_l_entreprise"] + 1
|
| 91 |
+
)
|
| 92 |
+
df["experience_par_anciennete"] = df["annee_experience_totale"] / (
|
| 93 |
+
df["annees_dans_l_entreprise"] + 1
|
| 94 |
+
)
|
| 95 |
+
df["promo_par_anciennete"] = df["annees_depuis_la_derniere_promotion"] / (
|
| 96 |
+
df["annees_dans_l_entreprise"] + 1
|
| 97 |
+
)
|
| 98 |
+
|
| 99 |
+
# Moyenne de satisfaction
|
| 100 |
+
df["satisfaction_moyenne"] = df[
|
| 101 |
+
[
|
| 102 |
+
"satisfaction_employee_environnement",
|
| 103 |
+
"satisfaction_employee_nature_travail",
|
| 104 |
+
"satisfaction_employee_equipe",
|
| 105 |
+
"satisfaction_employee_equilibre_pro_perso",
|
| 106 |
+
]
|
| 107 |
+
].mean(axis=1)
|
| 108 |
+
|
| 109 |
+
return df
|
| 110 |
+
|
| 111 |
+
|
| 112 |
+
def encode_and_scale(df: pd.DataFrame) -> pd.DataFrame:
|
| 113 |
+
"""
|
| 114 |
+
Encode les variables catégorielles et scale les numériques.
|
| 115 |
+
IMPORTANT: Doit correspondre EXACTEMENT au pipeline d'entraînement.
|
| 116 |
+
|
| 117 |
+
Args:
|
| 118 |
+
df: DataFrame avec features engineered.
|
| 119 |
+
|
| 120 |
+
Returns:
|
| 121 |
+
DataFrame transformé avec 50 colonnes (comme training).
|
| 122 |
+
"""
|
| 123 |
+
df = df.copy()
|
| 124 |
+
|
| 125 |
+
# === ENCODING ===
|
| 126 |
+
|
| 127 |
+
# NOTE: ayant_enfants et heure_supplementaires sont SUPPRIMÉS
|
| 128 |
+
# (ne font pas partie des features du modèle d'entraînement)
|
| 129 |
+
cols_to_drop = ["ayant_enfants", "heure_supplementaires"]
|
| 130 |
+
df = df.drop(columns=[col for col in cols_to_drop if col in df.columns])
|
| 131 |
+
|
| 132 |
+
# OneHot pour variables catégorielles non-ordonnées
|
| 133 |
+
# IMPORTANT: Utiliser les mêmes catégories que lors de l'entraînement
|
| 134 |
+
cat_non_ord = ["genre", "statut_marital", "departement", "poste", "domaine_etude"]
|
| 135 |
+
|
| 136 |
+
# Définir toutes les catégories possibles (depuis training data)
|
| 137 |
+
categories_dict = {
|
| 138 |
+
"genre": ["F", "M"],
|
| 139 |
+
"statut_marital": ["Célibataire", "Divorcé(e)", "Marié(e)"],
|
| 140 |
+
"departement": ["Commercial", "Consulting", "Ressources Humaines"],
|
| 141 |
+
"poste": [
|
| 142 |
+
"Assistant de Direction",
|
| 143 |
+
"Cadre Commercial",
|
| 144 |
+
"Consultant",
|
| 145 |
+
"Directeur Technique",
|
| 146 |
+
"Manager",
|
| 147 |
+
"Représentant Commercial",
|
| 148 |
+
"Ressources Humaines",
|
| 149 |
+
"Senior Manager",
|
| 150 |
+
"Tech Lead",
|
| 151 |
+
],
|
| 152 |
+
"domaine_etude": [
|
| 153 |
+
"Autre",
|
| 154 |
+
"Entrepreunariat",
|
| 155 |
+
"Infra & Cloud",
|
| 156 |
+
"Marketing",
|
| 157 |
+
"Ressources Humaines",
|
| 158 |
+
"Transformation Digitale",
|
| 159 |
+
],
|
| 160 |
+
}
|
| 161 |
+
|
| 162 |
+
onehot = OneHotEncoder(
|
| 163 |
+
sparse_output=False,
|
| 164 |
+
handle_unknown="ignore",
|
| 165 |
+
categories=[categories_dict[col] for col in cat_non_ord],
|
| 166 |
+
)
|
| 167 |
+
|
| 168 |
+
encoded_non_ord = pd.DataFrame(
|
| 169 |
+
onehot.fit_transform(df[cat_non_ord]),
|
| 170 |
+
columns=onehot.get_feature_names_out(cat_non_ord),
|
| 171 |
+
index=df.index,
|
| 172 |
+
)
|
| 173 |
+
|
| 174 |
+
# Ordinal pour fréquence déplacement
|
| 175 |
+
ordinal = OrdinalEncoder(categories=[["Aucun", "Occasionnel", "Frequent"]])
|
| 176 |
+
df["frequence_deplacement"] = ordinal.fit_transform(
|
| 177 |
+
df[["frequence_deplacement"]]
|
| 178 |
+
).flatten()
|
| 179 |
+
|
| 180 |
+
# Supprimer les colonnes catégorielles originales
|
| 181 |
+
df = df.drop(columns=cat_non_ord)
|
| 182 |
+
|
| 183 |
+
# Concaténer les encodages OneHot
|
| 184 |
+
df = pd.concat([df, encoded_non_ord], axis=1)
|
| 185 |
+
|
| 186 |
+
# === SCALING ===
|
| 187 |
+
|
| 188 |
+
# Colonnes numériques à scaler
|
| 189 |
+
quantitative_cols = df.select_dtypes(include=[np.number]).columns.tolist()
|
| 190 |
+
|
| 191 |
+
# Retirer les colonnes OneHot du scaling (elles sont déjà 0/1)
|
| 192 |
+
cols_to_scale = [
|
| 193 |
+
col
|
| 194 |
+
for col in quantitative_cols
|
| 195 |
+
if df[col].nunique() > 2 # Exclut colonnes binaires (0/1)
|
| 196 |
+
]
|
| 197 |
+
|
| 198 |
+
# Appliquer le scaling uniquement s'il y a des colonnes
|
| 199 |
+
if cols_to_scale:
|
| 200 |
+
scaler = StandardScaler()
|
| 201 |
+
df[cols_to_scale] = scaler.fit_transform(df[cols_to_scale])
|
| 202 |
+
|
| 203 |
+
return df
|
| 204 |
+
|
| 205 |
+
|
| 206 |
+
def preprocess_for_prediction(employee: EmployeeInput) -> np.ndarray:
|
| 207 |
+
"""
|
| 208 |
+
Pipeline complet de preprocessing pour une prédiction.
|
| 209 |
+
|
| 210 |
+
Args:
|
| 211 |
+
employee: Données validées d'un employé.
|
| 212 |
+
|
| 213 |
+
Returns:
|
| 214 |
+
Array numpy transformé prêt pour model.predict().
|
| 215 |
+
|
| 216 |
+
Examples:
|
| 217 |
+
>>> from src.schemas import EmployeeInput
|
| 218 |
+
>>> employee = EmployeeInput(...)
|
| 219 |
+
>>> X = preprocess_for_prediction(employee)
|
| 220 |
+
>>> prediction = model.predict(X)
|
| 221 |
+
"""
|
| 222 |
+
# 1. Créer DataFrame
|
| 223 |
+
df = create_input_dataframe(employee)
|
| 224 |
+
|
| 225 |
+
# 2. Feature engineering
|
| 226 |
+
df = engineer_features(df)
|
| 227 |
+
|
| 228 |
+
# 3. Encoding et scaling
|
| 229 |
+
df = encode_and_scale(df)
|
| 230 |
+
|
| 231 |
+
# 4. Convertir en numpy array (le modèle attend un array)
|
| 232 |
+
return df.values
|
| 233 |
+
|
| 234 |
+
|
| 235 |
+
# TODO: Implémenter le chargement des artifacts sauvegardés
|
| 236 |
+
# def load_preprocessing_artifacts(run_id: str) -> dict:
|
| 237 |
+
# """
|
| 238 |
+
# Charge les encoders et scaler depuis MLflow.
|
| 239 |
+
#
|
| 240 |
+
# Returns:
|
| 241 |
+
# dict avec keys: 'onehot_encoder', 'ordinal_encoder', 'scaler'
|
| 242 |
+
# """
|
| 243 |
+
# pass
|
src/rate_limit.py
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env python3
|
| 2 |
+
"""
|
| 3 |
+
Module de rate limiting pour protéger l'API contre les abus.
|
| 4 |
+
|
| 5 |
+
Utilise SlowAPI pour limiter le nombre de requêtes par IP/utilisateur.
|
| 6 |
+
"""
|
| 7 |
+
from slowapi import Limiter
|
| 8 |
+
from slowapi.util import get_remote_address
|
| 9 |
+
|
| 10 |
+
from src.config import get_settings
|
| 11 |
+
|
| 12 |
+
settings = get_settings()
|
| 13 |
+
|
| 14 |
+
# Créer le limiter avec stratégie par IP
|
| 15 |
+
limiter = Limiter(
|
| 16 |
+
key_func=get_remote_address,
|
| 17 |
+
default_limits=["100/minute"] if not settings.DEBUG else [],
|
| 18 |
+
storage_uri="memory://", # En production: utiliser Redis
|
| 19 |
+
strategy="fixed-window",
|
| 20 |
+
)
|
| 21 |
+
|
| 22 |
+
|
| 23 |
+
def get_rate_limit_key(request):
|
| 24 |
+
"""
|
| 25 |
+
Fonction pour obtenir la clé de rate limiting.
|
| 26 |
+
|
| 27 |
+
En production, on pourrait utiliser l'API Key au lieu de l'IP.
|
| 28 |
+
|
| 29 |
+
Args:
|
| 30 |
+
request: Requête FastAPI.
|
| 31 |
+
|
| 32 |
+
Returns:
|
| 33 |
+
Clé unique pour identifier l'utilisateur.
|
| 34 |
+
"""
|
| 35 |
+
# Priorité: API Key > IP
|
| 36 |
+
api_key = request.headers.get("X-API-Key")
|
| 37 |
+
if api_key:
|
| 38 |
+
return f"api_key:{api_key}"
|
| 39 |
+
|
| 40 |
+
return get_remote_address(request)
|
src/schemas.py
ADDED
|
@@ -0,0 +1,232 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env python3
|
| 2 |
+
"""
|
| 3 |
+
Schémas Pydantic pour validation des données d'entrée de l'API.
|
| 4 |
+
|
| 5 |
+
Ces schémas correspondent aux colonnes brutes du dataset avant preprocessing,
|
| 6 |
+
permettant une validation stricte des inputs avec messages d'erreur clairs.
|
| 7 |
+
"""
|
| 8 |
+
from enum import Enum
|
| 9 |
+
from typing import Literal
|
| 10 |
+
|
| 11 |
+
from pydantic import BaseModel, Field, field_validator
|
| 12 |
+
|
| 13 |
+
|
| 14 |
+
# Enums pour les valeurs catégorielles
|
| 15 |
+
class GenreEnum(str, Enum):
|
| 16 |
+
"""Genre de l'employé."""
|
| 17 |
+
|
| 18 |
+
M = "M"
|
| 19 |
+
F = "F"
|
| 20 |
+
|
| 21 |
+
|
| 22 |
+
class StatutMaritalEnum(str, Enum):
|
| 23 |
+
"""Statut marital de l'employé."""
|
| 24 |
+
|
| 25 |
+
CELIBATAIRE = "Célibataire"
|
| 26 |
+
MARIE = "Marié(e)"
|
| 27 |
+
DIVORCE = "Divorcé(e)"
|
| 28 |
+
|
| 29 |
+
|
| 30 |
+
class DepartementEnum(str, Enum):
|
| 31 |
+
"""Département de l'employé."""
|
| 32 |
+
|
| 33 |
+
COMMERCIAL = "Commercial"
|
| 34 |
+
CONSULTING = "Consulting"
|
| 35 |
+
|
| 36 |
+
|
| 37 |
+
class DomaineEtudeEnum(str, Enum):
|
| 38 |
+
"""Domaine d'études de l'employé."""
|
| 39 |
+
|
| 40 |
+
INFRA_CLOUD = "Infra & Cloud"
|
| 41 |
+
TRANSFORMATION_DIGITALE = "Transformation Digitale"
|
| 42 |
+
AUTRE = "Autre"
|
| 43 |
+
|
| 44 |
+
|
| 45 |
+
class FrequenceDeplacementEnum(str, Enum):
|
| 46 |
+
"""Fréquence des déplacements professionnels."""
|
| 47 |
+
|
| 48 |
+
AUCUN = "Aucun"
|
| 49 |
+
OCCASIONNEL = "Occasionnel"
|
| 50 |
+
FREQUENT = "Frequent"
|
| 51 |
+
|
| 52 |
+
|
| 53 |
+
class EmployeeInput(BaseModel):
|
| 54 |
+
"""
|
| 55 |
+
Schéma de validation pour les données d'entrée d'un employé.
|
| 56 |
+
|
| 57 |
+
Tous les champs correspondent aux colonnes brutes des 3 fichiers CSV
|
| 58 |
+
(sondage, eval, sirh) avant preprocessing.
|
| 59 |
+
"""
|
| 60 |
+
|
| 61 |
+
# === Données SONDAGE ===
|
| 62 |
+
nombre_participation_pee: int = Field(
|
| 63 |
+
..., ge=0, description="Nombre de participations au PEE"
|
| 64 |
+
)
|
| 65 |
+
nb_formations_suivies: int = Field(
|
| 66 |
+
..., ge=0, le=10, description="Nombre de formations suivies"
|
| 67 |
+
)
|
| 68 |
+
nombre_employee_sous_responsabilite: int = Field(
|
| 69 |
+
..., ge=0, description="Nombre d'employés sous responsabilité"
|
| 70 |
+
)
|
| 71 |
+
distance_domicile_travail: int = Field(
|
| 72 |
+
..., ge=0, le=50, description="Distance domicile-travail en km"
|
| 73 |
+
)
|
| 74 |
+
niveau_education: int = Field(
|
| 75 |
+
..., ge=1, le=5, description="Niveau d'éducation (1-5)"
|
| 76 |
+
)
|
| 77 |
+
domaine_etude: DomaineEtudeEnum = Field(..., description="Domaine d'études")
|
| 78 |
+
ayant_enfants: Literal["Y", "N"] = Field(..., description="A des enfants (Y/N)")
|
| 79 |
+
frequence_deplacement: FrequenceDeplacementEnum = Field(
|
| 80 |
+
..., description="Fréquence des déplacements"
|
| 81 |
+
)
|
| 82 |
+
annees_depuis_la_derniere_promotion: int = Field(
|
| 83 |
+
..., ge=0, description="Années depuis la dernière promotion"
|
| 84 |
+
)
|
| 85 |
+
annes_sous_responsable_actuel: int = Field(
|
| 86 |
+
..., ge=0, description="Années sous le responsable actuel"
|
| 87 |
+
)
|
| 88 |
+
|
| 89 |
+
# === Données EVALUATION ===
|
| 90 |
+
satisfaction_employee_environnement: int = Field(
|
| 91 |
+
..., ge=1, le=4, description="Satisfaction environnement (1-4)"
|
| 92 |
+
)
|
| 93 |
+
note_evaluation_precedente: int = Field(
|
| 94 |
+
..., ge=1, le=5, description="Note évaluation précédente (1-5)"
|
| 95 |
+
)
|
| 96 |
+
niveau_hierarchique_poste: int = Field(
|
| 97 |
+
..., ge=1, le=5, description="Niveau hiérarchique (1-5)"
|
| 98 |
+
)
|
| 99 |
+
satisfaction_employee_nature_travail: int = Field(
|
| 100 |
+
..., ge=1, le=4, description="Satisfaction nature du travail (1-4)"
|
| 101 |
+
)
|
| 102 |
+
satisfaction_employee_equipe: int = Field(
|
| 103 |
+
..., ge=1, le=4, description="Satisfaction équipe (1-4)"
|
| 104 |
+
)
|
| 105 |
+
satisfaction_employee_equilibre_pro_perso: int = Field(
|
| 106 |
+
..., ge=1, le=4, description="Satisfaction équilibre pro/perso (1-4)"
|
| 107 |
+
)
|
| 108 |
+
note_evaluation_actuelle: int = Field(
|
| 109 |
+
..., ge=1, le=5, description="Note évaluation actuelle (1-5)"
|
| 110 |
+
)
|
| 111 |
+
heure_supplementaires: Literal["Oui", "Non"] = Field(
|
| 112 |
+
..., description="Fait des heures supplémentaires"
|
| 113 |
+
)
|
| 114 |
+
augementation_salaire_precedente: float = Field(
|
| 115 |
+
..., ge=0, le=100, description="Augmentation salaire précédente (%)"
|
| 116 |
+
)
|
| 117 |
+
|
| 118 |
+
# === Données SIRH ===
|
| 119 |
+
age: int = Field(..., ge=18, le=70, description="Âge de l'employé")
|
| 120 |
+
genre: GenreEnum = Field(..., description="Genre")
|
| 121 |
+
revenu_mensuel: float = Field(..., ge=1000, description="Revenu mensuel (€)")
|
| 122 |
+
statut_marital: StatutMaritalEnum = Field(..., description="Statut marital")
|
| 123 |
+
departement: DepartementEnum = Field(..., description="Département")
|
| 124 |
+
poste: str = Field(..., min_length=3, description="Intitulé du poste")
|
| 125 |
+
nombre_experiences_precedentes: int = Field(
|
| 126 |
+
..., ge=0, description="Nombre d'expériences précédentes"
|
| 127 |
+
)
|
| 128 |
+
nombre_heures_travailless: int = Field(
|
| 129 |
+
..., ge=35, le=80, description="Nombre d'heures travaillées par semaine"
|
| 130 |
+
)
|
| 131 |
+
annee_experience_totale: int = Field(
|
| 132 |
+
..., ge=0, description="Années d'expérience totale"
|
| 133 |
+
)
|
| 134 |
+
annees_dans_l_entreprise: int = Field(
|
| 135 |
+
..., ge=0, description="Années dans l'entreprise"
|
| 136 |
+
)
|
| 137 |
+
annees_dans_le_poste_actuel: int = Field(
|
| 138 |
+
..., ge=0, description="Années dans le poste actuel"
|
| 139 |
+
)
|
| 140 |
+
|
| 141 |
+
@field_validator("augementation_salaire_precedente")
|
| 142 |
+
@classmethod
|
| 143 |
+
def validate_augmentation(cls, v: float) -> float:
|
| 144 |
+
"""Nettoie le format de l'augmentation (enlève % si présent)."""
|
| 145 |
+
if isinstance(v, str):
|
| 146 |
+
v = float(v.replace(" %", "").replace("%", ""))
|
| 147 |
+
return v
|
| 148 |
+
|
| 149 |
+
class Config:
|
| 150 |
+
"""Configuration Pydantic."""
|
| 151 |
+
|
| 152 |
+
json_schema_extra = {
|
| 153 |
+
"example": {
|
| 154 |
+
# Exemple basé sur la première ligne des CSV
|
| 155 |
+
"nombre_participation_pee": 0,
|
| 156 |
+
"nb_formations_suivies": 0,
|
| 157 |
+
"nombre_employee_sous_responsabilite": 1,
|
| 158 |
+
"distance_domicile_travail": 1,
|
| 159 |
+
"niveau_education": 2,
|
| 160 |
+
"domaine_etude": "Infra & Cloud",
|
| 161 |
+
"ayant_enfants": "Y",
|
| 162 |
+
"frequence_deplacement": "Occasionnel",
|
| 163 |
+
"annees_depuis_la_derniere_promotion": 0,
|
| 164 |
+
"annes_sous_responsable_actuel": 5,
|
| 165 |
+
"satisfaction_employee_environnement": 2,
|
| 166 |
+
"note_evaluation_precedente": 3,
|
| 167 |
+
"niveau_hierarchique_poste": 2,
|
| 168 |
+
"satisfaction_employee_nature_travail": 4,
|
| 169 |
+
"satisfaction_employee_equipe": 1,
|
| 170 |
+
"satisfaction_employee_equilibre_pro_perso": 1,
|
| 171 |
+
"note_evaluation_actuelle": 3,
|
| 172 |
+
"heure_supplementaires": "Oui",
|
| 173 |
+
"augementation_salaire_precedente": 11.0,
|
| 174 |
+
"age": 41,
|
| 175 |
+
"genre": "F",
|
| 176 |
+
"revenu_mensuel": 5993.0,
|
| 177 |
+
"statut_marital": "Célibataire",
|
| 178 |
+
"departement": "Commercial",
|
| 179 |
+
"poste": "Cadre Commercial",
|
| 180 |
+
"nombre_experiences_precedentes": 8,
|
| 181 |
+
"nombre_heures_travailless": 80,
|
| 182 |
+
"annee_experience_totale": 8,
|
| 183 |
+
"annees_dans_l_entreprise": 6,
|
| 184 |
+
"annees_dans_le_poste_actuel": 4,
|
| 185 |
+
}
|
| 186 |
+
}
|
| 187 |
+
|
| 188 |
+
|
| 189 |
+
class PredictionOutput(BaseModel):
|
| 190 |
+
"""Schéma de sortie pour les prédictions."""
|
| 191 |
+
|
| 192 |
+
prediction: int = Field(..., description="Classe prédite (0=reste, 1=part)")
|
| 193 |
+
probability_0: float = Field(
|
| 194 |
+
..., ge=0, le=1, description="Probabilité de rester (classe 0)"
|
| 195 |
+
)
|
| 196 |
+
probability_1: float = Field(
|
| 197 |
+
..., ge=0, le=1, description="Probabilité de partir (classe 1)"
|
| 198 |
+
)
|
| 199 |
+
risk_level: str = Field(..., description="Niveau de risque (Low/Medium/High)")
|
| 200 |
+
|
| 201 |
+
class Config:
|
| 202 |
+
"""Configuration Pydantic."""
|
| 203 |
+
|
| 204 |
+
json_schema_extra = {
|
| 205 |
+
"example": {
|
| 206 |
+
"prediction": 1,
|
| 207 |
+
"probability_0": 0.35,
|
| 208 |
+
"probability_1": 0.65,
|
| 209 |
+
"risk_level": "High",
|
| 210 |
+
}
|
| 211 |
+
}
|
| 212 |
+
|
| 213 |
+
|
| 214 |
+
class HealthCheck(BaseModel):
|
| 215 |
+
"""Schéma pour le endpoint health check."""
|
| 216 |
+
|
| 217 |
+
status: str = Field(..., description="Status de l'API")
|
| 218 |
+
model_loaded: bool = Field(..., description="Modèle chargé ou non")
|
| 219 |
+
model_type: str = Field(..., description="Type du modèle")
|
| 220 |
+
version: str = Field(..., description="Version de l'API")
|
| 221 |
+
|
| 222 |
+
class Config:
|
| 223 |
+
"""Configuration Pydantic."""
|
| 224 |
+
|
| 225 |
+
json_schema_extra = {
|
| 226 |
+
"example": {
|
| 227 |
+
"status": "healthy",
|
| 228 |
+
"model_loaded": True,
|
| 229 |
+
"model_type": "Pipeline",
|
| 230 |
+
"version": "1.0.0",
|
| 231 |
+
}
|
| 232 |
+
}
|