#!/usr/bin/env python3
"""
Flask App with Gunicorn for Deep Modal Files
Economics Chat Application using Qwen2 model
"""
from flask import Flask, request, jsonify, render_template_string
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
import os
import logging
# Configure logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
app = Flask(__name__)
# Global variables for model and tokenizer
model = None
tokenizer = None
# HTML template
HTML_TEMPLATE = """
AEGIS Economics AI
🏛️ AEGIS Economics AI
Advanced Economic Analysis & Policy Insights
Hello! I'm AEGIS Economics AI. Ask me about economic policies, market analysis, or financial strategies.
Checking model status...
"""
def load_model():
"""Load the Qwen2 model and tokenizer from HF repository"""
global model, tokenizer
try:
logger.info("Loading model and tokenizer from Hugging Face...")
# Load from the deployed model repository
model_repo = "Gaston895/Aegisecon1"
logger.info(f"Loading tokenizer from {model_repo}...")
tokenizer = AutoTokenizer.from_pretrained(
model_repo,
trust_remote_code=True,
use_auth_token=False
)
logger.info(f"Loading model from {model_repo}...")
model = AutoModelForCausalLM.from_pretrained(
model_repo,
torch_dtype=torch.float16, # Changed from bfloat16 for better compatibility
device_map="cpu", # Force CPU for HF Spaces compatibility
trust_remote_code=True,
use_auth_token=False,
low_cpu_mem_usage=True
)
logger.info("Model loaded successfully from HF repository!")
return True
except Exception as e:
logger.error(f"Error loading model from HF: {str(e)}")
# Try alternative loading method
try:
logger.info("Trying alternative loading method...")
tokenizer = AutoTokenizer.from_pretrained(
"Qwen/Qwen2-1.5B", # Fallback to base model
trust_remote_code=True
)
model = AutoModelForCausalLM.from_pretrained(
"Qwen/Qwen2-1.5B",
torch_dtype=torch.float16,
device_map="cpu",
trust_remote_code=True,
low_cpu_mem_usage=True
)
logger.info("Fallback model loaded successfully!")
return True
except Exception as e2:
logger.error(f"Fallback loading also failed: {str(e2)}")
return False
def generate_response(prompt):
"""Generate response using the loaded model"""
try:
if model is None or tokenizer is None:
return "Model is still loading, please wait a moment and try again..."
# Economics-focused system prompt
system_prompt = """You are AEGIS Economics AI, an expert economic analyst and policy advisor.
Provide clear, accurate, and insightful responses about economics, finance, markets, and policy.
Focus on practical analysis and actionable insights."""
full_prompt = f"{system_prompt}\n\nUser: {prompt}\nAssistant:"
# Tokenize input
inputs = tokenizer(full_prompt, return_tensors="pt", truncation=True, max_length=1024)
# Generate response
with torch.no_grad():
outputs = model.generate(
inputs.input_ids,
max_new_tokens=256, # Reduced for faster generation
temperature=0.7,
do_sample=True,
pad_token_id=tokenizer.eos_token_id,
repetition_penalty=1.1,
no_repeat_ngram_size=3
)
# Decode response
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
# Extract only the assistant's response
if "Assistant:" in response:
response = response.split("Assistant:")[-1].strip()
return response
except Exception as e:
logger.error(f"Error generating response: {str(e)}")
return "I apologize, but I'm having trouble processing your request right now. Please try again in a moment."
@app.route('/')
def home():
"""Serve the main chat interface"""
return render_template_string(HTML_TEMPLATE)
@app.route('/chat', methods=['POST'])
def chat():
"""Handle chat messages"""
try:
data = request.get_json()
user_message = data.get('message', '')
if not user_message:
return jsonify({'error': 'No message provided'}), 400
# Generate AI response
ai_response = generate_response(user_message)
return jsonify({'response': ai_response})
except Exception as e:
logger.error(f"Error in chat endpoint: {str(e)}")
return jsonify({'error': 'Internal server error'}), 500
@app.route('/health')
def health():
"""Health check endpoint"""
return jsonify({
'status': 'healthy',
'model_loaded': model is not None,
'tokenizer_loaded': tokenizer is not None,
'model_info': 'Gaston895/Aegisecon1' if model is not None else 'Not loaded'
})
@app.route('/load_model', methods=['POST'])
def load_model_endpoint():
"""Endpoint to trigger model loading"""
try:
success = load_model()
return jsonify({
'success': success,
'model_loaded': model is not None,
'tokenizer_loaded': tokenizer is not None
})
except Exception as e:
return jsonify({'error': str(e)}), 500
if __name__ == '__main__':
# Load model on startup
logger.info("Starting AEGIS Economics AI...")
# Try to load model, but don't fail if it doesn't work
logger.info("Attempting to load model...")
model_loaded = load_model()
if model_loaded:
logger.info("Model loaded successfully, starting server...")
else:
logger.warning("Model failed to load, starting server anyway. Model can be loaded via /load_model endpoint.")
app.run(host='0.0.0.0', port=7860, debug=False)