Primera prueba de código de microservicios

src/flask.py

@@ -0,0 +1,85 @@
+from flask import Flask, jsonify, request
+import os
+import psycopg2
+import pandas as pd
+import xgboost as xgb
+import joblib
+
+app = Flask(__name__)
+
+# Configuración de la base de datos
+DB_CONFIG = {
+    'dbname': os.getenv('DB_NAME', 'postgres'),
+    'user': os.getenv('DB_USER', 'postgres'),
+    'password': os.getenv('DB_PASSWORD', 'tfmuocdfcarvajal'),
+    'host': os.getenv('DB_HOST', '10.10.5.32'),
+    'port': os.getenv('DB_PORT', '5432')
+}
+
+# Ruta para entrenar o cargar el modelo
+MODEL_PATH = 'modelo_xgb.joblib'
+
+def get_db_connection():
+    conn = psycopg2.connect(**DB_CONFIG)
+    return conn
+
+def fetch_data():
+    conn = get_db_connection()
+    query = """
+        SELECT h3, hour, dow, total_events, total_nulos, nulo_rate
+        FROM demanda_h3_hour
+    """
+    df = pd.read_sql(query, conn)
+    conn.close()
+    return df
+
+def train_model(df):
+    # Variables predictoras
+    X = df[['hour', 'dow', 'total_events']]
+    # Variable objetivo: nulo_rate > 0 -> 1, else 0
+    y = (df['nulo_rate'] > 0).astype(int)
+    
+    model = xgb.XGBClassifier(
+        max_depth=4, n_estimators=100, learning_rate=0.1, use_label_encoder=False, eval_metric='logloss'
+    )
+    model.fit(X, y)
+    joblib.dump(model, MODEL_PATH)
+    return model
+
+def load_model():
+    try:
+        model = joblib.load(MODEL_PATH)
+        return model
+    except:
+        df = fetch_data()
+        return train_model(df)
+
+model = load_model()
+
+@app.route('/predict', methods=['GET'])
+def predict():
+    # Parámetros de la consulta
+    hour = int(request.args.get('hour'))
+    dow = int(request.args.get('dow'))
+    total_events = int(request.args.get('total_events', 1))  # valor por defecto si no se pasa
+    
+    X_pred = pd.DataFrame([[hour, dow, total_events]], columns=['hour', 'dow', 'total_events'])
+    prob = model.predict_proba(X_pred)[0][1]  # Probabilidad de nulo
+    return jsonify({
+        'hour': hour,
+        'dow': dow,
+        'total_events': total_events,
+        'predicted_nulo_prob': float(prob)
+    })
+
+@app.route('/demand', methods=['GET'])
+def demand():
+    df = fetch_data()
+    X_pred = df[['hour', 'dow', 'total_events']]
+    df['predicted_nulo_prob'] = model.predict_proba(X_pred)[:,1].astype(float)
+    # Convertimos a JSON
+    result = df.to_dict(orient='records')
+    return jsonify(result)
+
+if __name__ == '__main__':
+    app.run(host='0.0.0.0', port=5000)

src/pyproject.toml

@@ -0,0 +1,33 @@
+[build-system]
+requires = ["setuptools >= 61.0"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "microservicios"
+version = "1.0.1"
+requires-python = ">= 3.8"
+description = "Microservicios para Estimaciones"
+license={text = "EULA"}
+
+authors = [
+  {name = "Diego Fernández", email = "carvajal.diego@gmail.com"}
+]
+
+dependencies = [
+   'flask',
+   'psycopg2-binary',
+   'pandas',
+   'xgboost',
+   'scikit-learn',
+   'joblib'
+]
+
+[tool.setuptools]
+include-package-data = true
+py-modules = ["flask"]
+
+[tool.setuptools.packages]
+find = {}
+
+[project.urls]
+    Documentation='https://https://gitea.tfmuocdfcarvajal.duckdns.org/TFM/microservicios_python'