from transformers import MarianMTModel, AutoModelForSeq2SeqLM, AutoTokenizer, GPTNeoForCausalLM, GPT2Tokenizer
import gradio as gr
import requests
import io
from PIL import Image
import os  # Import os to access environment variables

# Load MarianMT model and tokenizer for Tamil to English translation
model_name = "Helsinki-NLP/opus-mt-mul-en"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)

# Load GPT-Neo model and tokenizer
gpt_neo_model_name = "EleutherAI/gpt-neo-125M"
gpt_neo_model = GPTNeoForCausalLM.from_pretrained(gpt_neo_model_name)
gpt_neo_tokenizer = GPT2Tokenizer.from_pretrained(gpt_neo_model_name)

# Retrieve the API URL and headers for Flux.1 from environment variables
API_URL = "https://api-inference.huggingface.co/models/black-forest-labs/FLUX.1-dev"
headers = {"Authorization": f"Bearer {os.environ.get('HUGGINGFACE_API_KEY')}"}  # Use the environment variable

def generate_image_from_text(english_text):
    payload = {"inputs": english_text}
    response = requests.post(API_URL, headers=headers, json=payload)
    
    if response.status_code == 200:
        image_bytes = response.content
        image = Image.open(io.BytesIO(image_bytes))
        return image
    else:
        return None  # Handle error appropriately

def translate_tamil_to_english(tamil_text):
    # Tokenize input and generate translation
    inputs = tokenizer(tamil_text, return_tensors="pt", padding=True)
    translated_tokens = model.generate(**inputs)
    translated_text = tokenizer.decode(translated_tokens[0], skip_special_tokens=True)
    return translated_text

def generate_creative_text(english_text):
    input_ids = gpt_neo_tokenizer.encode(english_text, return_tensors='pt')
    output = gpt_neo_model.generate(input_ids, max_length=150, num_return_sequences=1)
    return gpt_neo_tokenizer.decode(output[0], skip_special_tokens=True)

def process_input(tamil_text):
    # Step 1: Translate Tamil to English
    translated_text = translate_tamil_to_english(tamil_text)
    
    # Step 2: Generate Image from Translated English Text
    image = generate_image_from_text(translated_text)
    
    # Step 3: Generate Creative Text
    creative_text = generate_creative_text(translated_text)
    
    # Return results (translated text, image, and creative text)
    return translated_text, image, creative_text

# Create a Gradio interface with input and output components
interface = gr.Interface(
    fn=process_input, 
    inputs=gr.Textbox(lines=2, placeholder="Enter Tamil text..."), 
    outputs=[gr.Textbox(label="Translated Text (English)"), 
             gr.Image(label="Generated Image"), 
             gr.Textbox(label="Creative Text")],
    title="Tamil to Creative Text & Image Generator",
    description="Enter Tamil text to translate, generate an image, and produce creative content."
)

# Launch the Gradio app
interface.launch(debug=True)