Spaces:
Sleeping
Sleeping
| import os | |
| import json | |
| import httpx | |
| import gradio as gr | |
| from huggingface_hub import InferenceClient | |
| from openai import OpenAI | |
| from dotenv import load_dotenv | |
| # Load API keys from .env file | |
| load_dotenv() | |
| HF_API_KEY = os.getenv("HF_API_KEY") | |
| OPENROUTER_API_KEY = os.getenv("OPENROUTER_API_KEY") | |
| # Initialize Hugging Face Gemma Client | |
| hf_client = InferenceClient( | |
| provider="hf-inference", | |
| api_key=HF_API_KEY | |
| ) | |
| # Initialize OpenRouter DeepSeek Client | |
| openrouter_client = OpenAI( | |
| base_url="https://openrouter.ai/api/v1", | |
| api_key=OPENROUTER_API_KEY | |
| ) | |
| # Function to query Gemma-2-27B (Hugging Face) | |
| def query_gemma(user_input): | |
| try: | |
| messages = [{"role": "user", "content": user_input}] | |
| completion = hf_client.chat.completions.create( | |
| model="google/gemma-2-27b-it", | |
| messages=messages, | |
| max_tokens=500 | |
| ) | |
| return completion.choices[0].message["content"] | |
| except Exception as e: | |
| return f"Error querying Gemma: {str(e)}" | |
| # Function to query DeepSeek-R1 (OpenRouter) | |
| def query_deepseek(user_input): | |
| try: | |
| completion = openrouter_client.chat.completions.create( | |
| model="deepseek/deepseek-r1:free", | |
| messages=[{"role": "user", "content": user_input}] | |
| ) | |
| return completion.choices[0].message.content | |
| except Exception as e: | |
| return f"Error querying DeepSeek: {str(e)}" | |
| # Function to refine responses using DeepSeek | |
| def refine_response(user_input): | |
| try: | |
| # Get responses from both models | |
| gemma_response = query_gemma(user_input) | |
| deepseek_response = query_deepseek(user_input) | |
| # If either response failed, return the available one | |
| if "Error" in gemma_response: | |
| return f"Only DeepSeek Response:\n{deepseek_response}" | |
| if "Error" in deepseek_response: | |
| return f"Only Gemma Response:\n{gemma_response}" | |
| # Prepare refinement prompt | |
| improvement_prompt = f""" | |
| Here are two AI-generated responses: | |
| Response 1 (Gemma): {gemma_response} | |
| Response 2 (DeepSeek): {deepseek_response} | |
| Please combine the best elements of both, improve clarity, and provide a final refined answer. | |
| """ | |
| # Send request to OpenRouter | |
| response = httpx.post( | |
| "https://openrouter.ai/api/v1/chat/completions", | |
| headers={ | |
| "Authorization": f"Bearer {OPENROUTER_API_KEY}", | |
| "Content-Type": "application/json" | |
| }, | |
| json={ | |
| "model": "deepseek/deepseek-r1:free", | |
| "messages": [{"role": "user", "content": improvement_prompt}] | |
| } | |
| ) | |
| # Print raw response for debugging | |
| print("OpenRouter Response:", response.text) | |
| # Check if response is valid JSON | |
| response_json = response.json() | |
| return response_json["choices"][0]["message"]["content"] | |
| except Exception as e: | |
| return f"Error refining response: {str(e)}" | |
| # Create Gradio interface | |
| iface = gr.Interface( | |
| fn=refine_response, | |
| inputs=gr.Textbox(lines=2, placeholder="Ask me anything..."), | |
| outputs="text", | |
| title="AI Response Enhancer", | |
| description="Get responses from both Gemma and DeepSeek, then receive an improved final answer." | |
| ) | |
| # Launch app | |
| iface.launch() | |