app.py

import os
import threading
import gradio as gr
from openai import OpenAI
from dotenv import load_dotenv

# Load API keys from .env file
load_dotenv()
API_KEY_LLAMA = os.getenv("OPENROUTER_API_KEY1")  # Llama API Key
API_KEY_GEMMA = os.getenv("OPENROUTER_API_KEY2")  # Gemma API Key
API_KEY_DEEPSEEK1 = os.getenv("OPENROUTER_API_KEY3")  # DeepSeek First Query
API_KEY_DEEPSEEK2 = os.getenv("OPENROUTER_API_KEY4")  # DeepSeek Final Refinement

# Create OpenAI Clients for each model
llama_client = OpenAI(base_url="https://openrouter.ai/api/v1", api_key=API_KEY_LLAMA)
gemma_client = OpenAI(base_url="https://openrouter.ai/api/v1", api_key=API_KEY_GEMMA)
deepseek_client1 = OpenAI(base_url="https://openrouter.ai/api/v1", api_key=API_KEY_DEEPSEEK1)
deepseek_client2 = OpenAI(base_url="https://openrouter.ai/api/v1", api_key=API_KEY_DEEPSEEK2)

# Function to query Llama model
def query_llama(user_input, results):
    try:
        completion = llama_client.chat.completions.create(
            model="meta-llama/llama-3.2-3b-instruct:free",
            messages=[{"role": "user", "content": user_input}]
        )
        results["Llama"] = completion.choices[0].message.content
    except Exception as e:
        results["Llama"] = f"Error: {str(e)}"

# Function to query Gemma model
def query_gemma(user_input, results):
    try:
        completion = gemma_client.chat.completions.create(
            model="google/gemma-2-9b-it:free",
            messages=[{"role": "user", "content": user_input}]
        )
        results["Gemma"] = completion.choices[0].message.content
    except Exception as e:
        results["Gemma"] = f"Error: {str(e)}"

# Function to query DeepSeek for additional context
def query_deepseek_1(user_input, results):
    try:
        completion = deepseek_client1.chat.completions.create(
            model="deepseek/deepseek-r1:free",
            messages=[{"role": "user", "content": user_input}]
        )
        results["DeepSeek1"] = completion.choices[0].message.content
    except Exception as e:
        results["DeepSeek1"] = f"Error: {str(e)}"

# Function to refine responses using DeepSeek-R1 (Final API)
def refine_response(user_input):
    try:
        results = {}

        # Create threads for parallel API calls
        threads = [
            threading.Thread(target=query_llama, args=(user_input, results)),
            threading.Thread(target=query_gemma, args=(user_input, results)),
            threading.Thread(target=query_deepseek_1, args=(user_input, results))
        ]

        # Start all threads
        for thread in threads:
            thread.start()

        # Wait for all threads to complete
        for thread in threads:
            thread.join()

        # Ensure all responses are received
        valid_responses = {k: v.strip() for k, v in results.items() if v and "Error" not in v}

        if len(valid_responses) < 2:
            return "\n\n".join(f"{k} Response: {v}" for k, v in valid_responses.items())

        # Prepare refinement prompt
        improvement_prompt = f"""
        Here are three AI-generated responses:

        Response 1 (Llama): {results.get("Llama", "N/A")}
        Response 2 (Gemma): {results.get("Gemma", "N/A")}
        Response 3 (DeepSeek1): {results.get("DeepSeek1", "N/A")}

        Please combine the best elements of all three, improve clarity, and provide a final refined answer.
        """

        # Query DeepSeek-R1 for refinement using API key 4
        try:
            refined_completion = deepseek_client2.chat.completions.create(
                model="deepseek/deepseek-r1:free",
                messages=[{"role": "user", "content": improvement_prompt}]
            )
            refined_content = refined_completion.choices[0].message.content
            return refined_content if refined_content.strip() else "Refinement failed, returning best response."
        except Exception as e:
            return f"Error refining response: {str(e)}"

    except Exception as e:
        return f"Unexpected error: {str(e)}"

# Create Gradio interface
iface = gr.Interface(
    fn=refine_response, 
    inputs=gr.Textbox(lines=2, placeholder="Ask me anything..."), 
    outputs="text",
    title="Multi-Model AI Enhancer (4 API Keys)",
    description="Llama (API 1) + Gemma (API 2) + DeepSeek (API 3) → Final Refinement with DeepSeek (API 4)"
)

# Launch app
iface.launch()