app.py

import gradio as gr
import requests
from transformers import MarianMTModel, MarianTokenizer, AutoModelForCausalLM, AutoTokenizer
from PIL import Image
import torch
import io
import os
from typing import Tuple

# Load HF token
HF_API_KEY = os.getenv("HF_API_KEY") or "your_hf_token_here"  # Replace this with your token if local
if not HF_API_KEY:
    raise ValueError("HF_API_KEY is not set.")

# Hugging Face image model
IMAGE_GEN_URL = "https://api-inference.huggingface.co/models/black-forest-labs/FLUX.1-schnell"
HEADERS = {"Authorization": f"Bearer {HF_API_KEY}"}

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

# Translation model (Tamil to English)
translator_model = "Helsinki-NLP/opus-mt-mul-en"
translator = MarianMTModel.from_pretrained(translator_model).to(device)
translator_tokenizer = MarianTokenizer.from_pretrained(translator_model)

# Text generation model
generator_model = "EleutherAI/gpt-neo-1.3B"
generator = AutoModelForCausalLM.from_pretrained(generator_model).to(device)
generator_tokenizer = AutoTokenizer.from_pretrained(generator_model)
generator_tokenizer.pad_token = generator_tokenizer.eos_token

def translate_tamil_to_english(text: str) -> str:
    inputs = translator_tokenizer(text, return_tensors="pt", padding=True, truncation=True).to(device)
    output = translator.generate(**inputs)
    return translator_tokenizer.decode(output[0], skip_special_tokens=True)

def generate_text(prompt: str) -> str:
    inputs = generator_tokenizer(prompt, return_tensors="pt", padding=True, truncation=True).to(device)
    output = generator.generate(**inputs, max_length=100, num_return_sequences=1)
    return generator_tokenizer.decode(output[0], skip_special_tokens=True)

def generate_image(prompt: str) -> Image.Image:
    response = requests.post(IMAGE_GEN_URL, headers=HEADERS, json={"inputs": prompt})
    try:
        if response.status_code == 200 and response.headers["content-type"].startswith("image"):
            return Image.open(io.BytesIO(response.content))
    except Exception as e:
        print("Image generation failed:", e)
    return Image.new("RGB", (300, 300), color="gray")

def process_input(tamil_text: str) -> Tuple[str, str, Image.Image]:
    english_text = translate_tamil_to_english(tamil_text)
    creative_text = generate_text(english_text)
    image = generate_image(english_text)
    return english_text, creative_text, image

# Gradio app
with gr.Blocks() as demo:
    gr.Markdown("## Tamil to English Translator with Text and Image Generator")

    tamil_input = gr.Textbox(label="Enter Tamil Text")
    translate_btn = gr.Button("Translate & Generate")

    english_output = gr.Textbox(label="Translated English")
    creative_output = gr.Textbox(label="Creative Text")
    image_output = gr.Image(label="Generated Image")

    translate_btn.click(fn=process_input, inputs=tamil_input, outputs=[english_output, creative_output, image_output])

demo.launch()