app.py

import gradio as gr
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

# Model
MODEL_NAME = "ai4bharat/indictrans2-indic-indic-1B"

# Load tokenizer and model
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, trust_remote_code=True)
model = AutoModelForSeq2SeqLM.from_pretrained(MODEL_NAME, trust_remote_code=True)

# Supported languages: full name -> code
LANGUAGES = {
  "Assamese": "asm",
    "Bengali": "ben",
    "Gujarati": "guj",
    "Hindi": "hin",
    "Kannada": "kan",
    "Malayalam": "mal",
    "Marathi": "mar",
    "Odia": "ori",
    "Punjabi": "pan",
    "Tamil": "tam",
    "Telugu": "tel",
    "Urdu": "urd",
    "English": "eng"
}

def translate(text: str, src_lang_name: str, tgt_lang_name: str) -> str:
    if not text.strip():
        return "⚠️ Please enter some text to translate."

    try:
        # Convert full name to code
        src_lang = LANGUAGES[src_lang_name]
        tgt_lang = LANGUAGES[tgt_lang_name]

        formatted_text = f"{src_lang}>>{tgt_lang} {text}"
        inputs = tokenizer(formatted_text, return_tensors="pt")

        output_tokens = model.generate(**inputs, max_length=512)
        translation = tokenizer.decode(output_tokens[0], skip_special_tokens=True)

        return translation
    except Exception as e:
        return f"❌ Error: {str(e)}"

# Gradio interface
demo = gr.Interface(
    fn=translate,
    inputs=[
        gr.Textbox(label="Text", placeholder="Enter your text here..."),
        gr.Dropdown(label="Source Language", choices=list(LANGUAGES.keys()), value="Tamil"),
        gr.Dropdown(label="Target Language", choices=list(LANGUAGES.keys()), value="English")
    ],
    outputs=gr.Textbox(label="Translated Text"),
    title="IndicTrans2 Language Translator",
    description=(
        "🌐 Translate text between Indian languages (and English) using "
        "[ai4bharat/indictrans2-indic-indic-1B](https://huggingface.co/ai4bharat/indictrans2-indic-indic-1B)."
    )
)

if __name__ == "__main__":
    demo.launch()