175 lines
No EOL
6.1 KiB
Python
175 lines
No EOL
6.1 KiB
Python
import os
|
|
import sys
|
|
import pickle
|
|
from dotenv import load_dotenv
|
|
from fastapi import FastAPI, HTTPException
|
|
from pydantic import BaseModel
|
|
import mlflow
|
|
from mlflow.tracking import MlflowClient
|
|
from tensorflow.keras.models import load_model
|
|
from tensorflow.keras.preprocessing.sequence import pad_sequences
|
|
|
|
# A DECOMMENTER POUR DEVELOPPER. AUTREMENT, LAISSER AINSI
|
|
# Accéder à .env
|
|
PARENT_DIR = os.path.abspath(os.path.join(os.path.dirname(__file__), "../../"))
|
|
sys.path.append(PARENT_DIR)
|
|
# 🔐 Charger les variables d'environnement (.env à deux niveaux au-dessus)
|
|
load_dotenv()
|
|
|
|
|
|
# 🔁 Import de la fonction de prétraitement (LOCAL)
|
|
from preprocessing import preprocess_tweet
|
|
|
|
# 📧 Import du service email
|
|
from email_service import send_error_report_email
|
|
|
|
|
|
# 🔗 Configurer MLflow
|
|
mlflow_tracking_uri = os.getenv("MLFLOW_TRACKING_URI")
|
|
if mlflow_tracking_uri:
|
|
mlflow.set_tracking_uri(mlflow_tracking_uri)
|
|
else:
|
|
print("⚠️ MLFLOW_TRACKING_URI non défini dans .env")
|
|
|
|
# 📌 Nom du modèle et stage
|
|
MODEL_NAME = "SentimentAnalysisLSTM"
|
|
STAGE = "Production"
|
|
|
|
# 🧾 Structures des requêtes POST
|
|
class TweetRequest(BaseModel):
|
|
tweet: str
|
|
|
|
class ReportRequest(BaseModel):
|
|
tweet: str
|
|
prediction: str
|
|
probability: float
|
|
|
|
# 🧠 Stockage des signalements d'erreurs
|
|
error_reports = {}
|
|
|
|
# 🔍 Fonction pour charger modèle + tokenizer
|
|
def load_model_and_artifacts(model_name=MODEL_NAME, stage=STAGE):
|
|
client = MlflowClient()
|
|
versions = client.get_latest_versions(model_name, [stage])
|
|
|
|
if not versions:
|
|
raise RuntimeError(f"Aucune version du modèle '{model_name}' en stage '{stage}'")
|
|
|
|
version_info = versions[0]
|
|
run_id = version_info.run_id
|
|
|
|
if not run_id:
|
|
raise RuntimeError("Run ID introuvable.")
|
|
|
|
# Extraire le tag d'embedding
|
|
model_version_info = client.get_model_version(model_name, version_info.version)
|
|
# glove est ici la valeur par défaut si embedding_type n'est pas trouvé.
|
|
embedding_type = model_version_info.tags.get("embedding_type", "glove").lower()
|
|
|
|
# Déterminer les chemins des artefacts
|
|
local_dir = "./downloaded_artifacts"
|
|
os.makedirs(local_dir, exist_ok=True)
|
|
base_path = "local_artifacts"
|
|
tokenizer_file = "tokenizer.pickle"
|
|
|
|
if "glove" in embedding_type:
|
|
keras_file = "final_model_LSTM_GloVe-300d-Fige.keras"
|
|
elif "word2vec" in embedding_type:
|
|
keras_file = "final_model_LSTM_Word2Vec-Fige.keras"
|
|
else:
|
|
raise ValueError(f"Embedding type inconnu : {embedding_type}")
|
|
|
|
print(f"Modèle {embedding_type} trouvé. Téléchargement du modèle lié au run_id suivant : {run_id}")
|
|
|
|
def dl(path):
|
|
return client.download_artifacts(run_id=run_id, path=f"{base_path}/{path}", dst_path=local_dir)
|
|
|
|
# 🔽 Télécharger les artefacts
|
|
model_path = dl(keras_file)
|
|
tokenizer_path = dl(tokenizer_file)
|
|
print("Contenu de downloaded_artifacts :", os.listdir(local_dir))
|
|
|
|
|
|
# 📦 Chargement
|
|
model = load_model(model_path)
|
|
with open(tokenizer_path, "rb") as f:
|
|
tokenizer = pickle.load(f)
|
|
|
|
return model, tokenizer, embedding_type
|
|
|
|
# 🚀 Lancer l'API
|
|
app = FastAPI(title="API Prédiction Sentiment - Air Paradis")
|
|
# app = FastAPI(title="API Prédiction Sentiment - Air Paradis",
|
|
# root_path="/api")
|
|
|
|
try:
|
|
model, tokenizer, embedding_type = load_model_and_artifacts()
|
|
except Exception as e:
|
|
print(f"\n❌ Erreur lors du chargement des artefacts : {e}")
|
|
model, tokenizer = None, None
|
|
|
|
@app.get("/")
|
|
def root():
|
|
return {"message": "API Sentiment - Air Paradis - en ligne"}
|
|
|
|
@app.post("/predict")
|
|
def predict_sentiment(request: TweetRequest):
|
|
if model is None or tokenizer is None:
|
|
raise HTTPException(status_code=503, detail="Modèle ou artefacts indisponibles.")
|
|
|
|
try:
|
|
processed_tweet = preprocess_tweet(request.tweet)
|
|
except Exception as e:
|
|
raise HTTPException(status_code=500, detail=f"Erreur de prétraitement : {e}")
|
|
|
|
if not processed_tweet.strip():
|
|
raise HTTPException(status_code=400, detail="Tweet vide ou invalide après prétraitement.")
|
|
|
|
try:
|
|
sequence = tokenizer.texts_to_sequences([processed_tweet])
|
|
sequence_padded = pad_sequences(sequence, maxlen=model.input_shape[1])
|
|
prediction_prob = model.predict(sequence_padded)[0][0]
|
|
if prediction_prob >=0.5:
|
|
sentiment = "positif"
|
|
else:
|
|
sentiment = "negatif"
|
|
# prediction_prob = 1-prediction_prob
|
|
except Exception as e:
|
|
raise HTTPException(status_code=500, detail=f"Erreur d'inférence : {e}")
|
|
|
|
return {
|
|
"sentiment": sentiment,
|
|
"probability": float(prediction_prob)
|
|
}
|
|
|
|
@app.post("/report_error")
|
|
def report_error(request: ReportRequest):
|
|
tweet = request.tweet.strip()
|
|
prediction = request.prediction.strip().lower()
|
|
probability = request.probability
|
|
|
|
if not tweet or not prediction:
|
|
raise HTTPException(status_code=400, detail="Tweet ou prédiction manquants.")
|
|
if probability < 0 or probability > 1:
|
|
raise HTTPException(status_code=400, detail="Probabilité invalide. Elle doit être entre 0 et 1.")
|
|
|
|
# Ajouter ou mettre à jour le signalement du tweet avec la probabilité
|
|
error_reports[tweet] = prediction + f' (p = {probability:.2f} ) '
|
|
|
|
report_sent = False
|
|
|
|
# Condition : À chaque multiple de 3 signalements, un mail est envoyé
|
|
try:
|
|
if len(error_reports) % 3 == 0: # La condition reste
|
|
report_sent = True
|
|
print(f"📧 Tentative d'envoi d'email pour {len(error_reports)} signalements...")
|
|
email_success = send_error_report_email(error_reports.copy())
|
|
if email_success:
|
|
print(f"✅ Email envoyé avec succès pour {len(error_reports)} signalements")
|
|
else:
|
|
print(f"❌ Échec de l'envoi de l'email pour {len(error_reports)} signalements")
|
|
except Exception as e:
|
|
print(f"❌ Erreur lors de l'envoi de l'email : {e}")
|
|
# On continue même si l'email échoue pour ne pas bloquer l'API
|
|
|
|
return {"report_sent": report_sent} |