diff --git a/app/job-recommend/preprocessed_data/embeddings.npy b/app/job-recommend/preprocessed_data/embeddings.npy new file mode 100644 index 0000000..609c69d Binary files /dev/null and b/app/job-recommend/preprocessed_data/embeddings.npy differ diff --git a/app/job-recommend/preprocessed_data/jobs_preprocessed.parquet b/app/job-recommend/preprocessed_data/jobs_preprocessed.parquet new file mode 100644 index 0000000..4b44cca Binary files /dev/null and b/app/job-recommend/preprocessed_data/jobs_preprocessed.parquet differ diff --git a/app/job-recommend/requirements.txt b/app/job-recommend/requirements.txt new file mode 100644 index 0000000..35ed53f --- /dev/null +++ b/app/job-recommend/requirements.txt @@ -0,0 +1,8 @@ +Flask==2.1.1 +Flask-CORS==3.1.1 +pandas==1.3.3 +numpy==1.21.2 +geopy==2.2.0 +scikit-learn==1.0.2 +sentence-transformers==2.1.0 +waitress==2.1.0 diff --git a/app/job-recommend/server.py b/app/job-recommend/server.py new file mode 100644 index 0000000..e16db40 --- /dev/null +++ b/app/job-recommend/server.py @@ -0,0 +1,183 @@ +from flask import Flask, request, jsonify, send_file +from flask_cors import CORS +import numpy as np +import pandas as pd +from geopy.geocoders import Nominatim +from sklearn.neighbors import BallTree +from sentence_transformers import SentenceTransformer +import logging +import os +from datetime import datetime + +# Configuración inicial +app = Flask(__name__) +CORS(app, resources={r"/*": {"origins": "*"}}) # CORS más restrictivo para producción + +# Configurar logging avanzado +logging.basicConfig( + level=logging.INFO, + format='%(asctime)s - %(name)s - %(levelname)s - %(message)s', + handlers=[ + logging.FileHandler('server.log'), + logging.StreamHandler() + ] +) +logger = logging.getLogger(__name__) + +class EnhancedJobRecommender: + def __init__(self, data_path='preprocessed_data'): + try: + logger.info("Inicializando recomendador...") + + # Cargar datos + self.jobs_data = pd.read_parquet(os.path.join(data_path, 'jobs_preprocessed.parquet')) + self.embeddings = np.load(os.path.join(data_path, 'embeddings.npy')) + + # Inicializar modelos + self.geolocator = Nominatim(user_agent="enhanced_job_recommender", timeout=10) + self.geo_cache = {} + self.model = SentenceTransformer('paraphrase-multilingual-MiniLM-L12-v2') + + # Construir índices + self._build_geo_index() + + logger.info(f"Recomendador listo con {len(self.jobs_data)} empleos cargados") + + except Exception as e: + logger.error(f"Error de inicialización: {str(e)}") + raise + + def _build_geo_index(self): + """Construir índice geoespacial optimizado""" + coords = np.deg2rad(self.jobs_data[['Latitude', 'Longitude']].dropna().values) + self.geo_index = BallTree(coords, metric='haversine', leaf_size=40) + + def _get_coordinates(self, location_str): + """Geocodificación con caché y reintentos""" + if not location_str: + return None + + if location_str in self.geo_cache: + return self.geo_cache[location_str] + + try: + location = self.geolocator.geocode(location_str, exactly_one=True) + if location: + coords = (location.latitude, location.longitude) + self.geo_cache[location_str] = coords + logger.info(f"Geocodificado: {location_str} -> {coords}") + return coords + return None + except Exception as e: + logger.warning(f"Error en geocodificación: {str(e)}") + return None + + def recommend(self, user_query, user_location, top_n=5): + """Generar recomendaciones con validación mejorada""" + try: + # Validación de entrada + if not user_query or not user_location: + raise ValueError("Parámetros incompletos") + + # Geocodificación + start_time = datetime.now() + coords = self._get_coordinates(user_location) + if not coords: + return [] + + # Búsqueda geoespacial + distances, indices = self.geo_index.query( + np.deg2rad([coords]), + k=min(1000, len(self.jobs_data)) + ) + + # Procesamiento semántico + query_embedding = self.model.encode([user_query]) + candidate_embeddings = self.embeddings[indices[0]] + + # Cálculo de similitud + text_scores = np.dot(candidate_embeddings, query_embedding.T).flatten() + geo_scores = 1 / (1 + (distances * 6371).flatten()) # Conversión a km + + # Combinación de puntuaciones + combined_scores = 0.7 * text_scores + 0.3 * geo_scores + + # Selección de resultados + valid_n = min(top_n, len(indices[0])) + top_indices = np.argsort(combined_scores)[-valid_n:][::-1] + + # Formateo de resultados + results = self.jobs_data.iloc[indices[0][top_indices]].copy() + results['Match_Score'] = (combined_scores[top_indices]).round(2) + + logger.info(f"Recomendación generada en {(datetime.now()-start_time).total_seconds():.2f}s") + return results[['JobCategory', 'Latitude', 'Longitude', 'Avg_salary', 'URL']].to_dict('records') + + except Exception as e: + logger.error(f"Error en recomendación: {str(e)}", exc_info=True) + return [] + +# Inicialización del sistema +recommender = EnhancedJobRecommender() + +# Endpoints +@app.route('/') +def serve_interface(): + """Servir la interfaz web local""" + return send_file('static/index.html') + + +@app.route('/api/recommend', methods=['POST']) +def handle_recommendation(): + try: + data = request.get_json() + + # Validación de parámetros + if not data or 'description' not in data or 'location' not in data: + return jsonify({ + "error": "Parámetros requeridos: 'description' y 'location'", + "example_request": { + "description": "Job description or skills", + "location": "City, State or Country" + } + }), 400 + + # Procesar recomendación + recommendations = recommender.recommend( + user_query=data['description'], + user_location=data['location'], + top_n=5 + ) + + return jsonify({ + "count": len(recommendations), + "results": recommendations, + "status": "success" + }) + + except KeyError as ke: + return jsonify({ + "error": f"Campo faltante en los datos: {str(ke)}" + }), 400 + + except Exception as e: + app.logger.error(f"Error interno: {str(e)}") + return jsonify({ + "error": "Error procesando la solicitud", + "details": str(e) + }), 500 + +@app.route('/health') +def health_check(): + """Endpoint de verificación de salud""" + return jsonify({ + "status": "OK", + "timestamp": datetime.now().isoformat(), + "jobs_loaded": len(recommender.jobs_data) + }) + +if __name__ == '__main__': + if os.getenv("AZURE_FUNCTIONS_ENVIRONMENT") != "Production": + from waitress import serve + logger.info("Iniciando servidor Flask localmente...") + serve(app, host="0.0.0.0", port=5000, threads=4) \ No newline at end of file diff --git a/app/job-recommend/static/image/inclusiveai_logo.png b/app/job-recommend/static/image/inclusiveai_logo.png new file mode 100644 index 0000000..2a0d5dd Binary files /dev/null and b/app/job-recommend/static/image/inclusiveai_logo.png differ diff --git a/app/job-recommend/static/index.html b/app/job-recommend/static/index.html new file mode 100644 index 0000000..1c1f253 --- /dev/null +++ b/app/job-recommend/static/index.html @@ -0,0 +1,343 @@ + + + + + + + + + +
+
+ Logo + 🤖 Recommend job assistant +
+ +
+
+ Hello! I am your job search assistant +

I can help you find the best possible job. Here are some examples of how you can ask me:

+
    +
  • "I have cooking experience and I'm looking for a job in Alabama"
    • +
  • "I have knowledge of engineering and electric fundamentals in Washington"
    • +
+
+
+ +
+
Cargando...
+ + +
+
+ + + + \ No newline at end of file