# !pip install diffusers --upgrade # !pip install invisible_watermark transformers accelerate safetensors import torch from diffusers import DiffusionPipeline import os import random import gradio as gr import time import openai import matplotlib.pyplot as plt from dotenv import load_dotenv, find_dotenv load_dotenv(find_dotenv()) openai.api_key = os.getenv('OPENAI_API_KEY') def chat_completion(messages: list) -> str: completion = openai.ChatCompletion.create( model='gpt-3.5-turbo', messages=messages ) return completion['choices'][0]['message']['content'] def generate_messages(messages: list, query: str) -> list: formatted_messages = [ {'role': 'system', 'content': 'You are a helpful Network Design assistant in brief.'} ] for m in messages: formatted_messages.append({'role': 'user', 'content': m[0]}) formatted_messages.append({'role': 'assistant', 'content': m[1]}) formatted_messages.append({'role': 'user', 'content': query}) return formatted_messages def generate_response(query: str, chat_history: list) -> tuple: messages = generate_messages(chat_history, query) bot_message = chat_completion(messages) chat_history.append((query, bot_message)) time.sleep(random.randint(0, 5)) return '', chat_history # Example Usage: initial_message = "Hello, how can you assist me with network design?" chat_history = [(initial_message, '')] # Image generator pipe = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16, use_safetensors=True, variant="fp16") pipe.to("cuda") def generate_and_display_images(query): # Generate a response based on user input response, chat_history = generate_response(query, chat_history) # Generate images based on the assistant's response generated_images = pipe(prompt=response).images[0] # Display the generated images plt.imshow(generated_images) plt.axis('off') plt.show() # Example Usage: initial_chat_message = "Hello, how can you assist me with network design?" chat_history = [(initial_chat_message, '')] # Create a Gradio interface iface = gr.Interface(fn=generate_and_display_images, inputs="text", outputs="image") iface.launch()