app.py

import flask # Gradio nahi, Flask hi rakhte hain
from flask import request, jsonify
# from transformers import pipeline # अब इसकी ज़रूरत नहीं
# import torch # अब इसकी ज़रूरत नहीं

from ctransformers import AutoModelForCausalLM # ctransformers से मॉडल लोड करेंगे

app = flask.Flask(__name__)

# ===========================
# LOAD MODEL
# ===========================
model_id = "TheBloke/Mistral-7B-Instruct-v0.2-GGUF"
print("🔄 Loading model...")

try:
    # ctransformers का उपयोग करके GGUF मॉडल को CPU पर लोड करें
    ai = AutoModelForCausalLM.from_pretrained(
        model_id,
        model_file="mistral-7b-instruct-v0.2.Q4_K_M.gguf", # GGUF फ़ाइल का नाम
        model_type="mistral",
        gpu_layers=0 # CPU पर चलाने के लिए
    )
    print("✅ Model loaded!")
except Exception as e:
    print(f"❌ Error loading model: {e}")
    # Fallback/Exit strategy here if loading fails

# ===========================
# CHAT API
# ===========================
@app.route('/chat', methods=['POST'])
def chat():
    try:
        data = request.get_json()
        msg = data.get("message", "")
        if not msg:
            return jsonify({"error": "No message sent"}), 400

        # ctransformers से response generate करें
        output = ai(msg, max_new_tokens=200, temperature=0.7)
        return jsonify({"reply": output})
    except Exception as e:
        return jsonify({"error": str(e)}), 500

# ===========================
# RUN SERVER
# ===========================
if __name__ == "__main__":
    app.run(host='0.0.0.0', port=7860)