app.py

# app.py

import os
import gradio as gr
import spaces
from datasets import load_dataset
from transformers import MarianTokenizer, MarianMTModel
from huggingface_hub import login

# Set cache paths for models
os.environ['TRANSFORMERS_CACHE'] = "/tmp/cache/models"
os.makedirs(os.environ['TRANSFORMERS_CACHE'], exist_ok=True)

# Log in to Hugging Face Hub if HF_TOKEN is available
hf_token = os.getenv("HF_TOKEN")
if hf_token:
    try:
        login(token=hf_token)
        print("✅ Authenticated with Hugging Face Hub")
    except Exception as e:
        print(f"[ERROR] Hugging Face login failed: {e}")
else:
    print("[WARNING] HF_TOKEN not found — dataset may be inaccessible")

# Load Dataset from Hub (Private or Public)
DATASET_NAME = "ajsbsd/legalese-sentences_estonian-english"
try:
    ds = load_dataset(DATASET_NAME)
    print("✅ Dataset loaded successfully")
except Exception as e:
    print(f"[ERROR] Failed to load dataset '{DATASET_NAME}': {e}")
    ds = None

# Load Translation Model
MODEL_NAME = "Helsinki-NLP/opus-mt-et-en"
try:
    tokenizer = MarianTokenizer.from_pretrained(MODEL_NAME, use_fast=True)
    model = MarianMTModel.from_pretrained(MODEL_NAME, use_safetensors=True)
    print("✅ Translation model loaded successfully")
except Exception as e:
    print(f"[ERROR] Failed to load model '{MODEL_NAME}': {e}")
    tokenizer = None
    model = None


def get_sample(split="train", index=0):
    """Fetch a sample from the dataset."""
    print(f"[DEBUG] Loading sample from '{split}' split, index {index}")
    if not ds:
        return {"error": "Dataset not loaded or unavailable."}
    try:
        sample = ds[split][index]
        return {
            "Estonian": sample["input"],
            "English": sample["output"]
        }
    except IndexError:
        return {"error": f"Index {index} out of range for '{split}' split"}
    except Exception as e:
        return {"error": str(e)}


@spaces.GPU
def translate(text):
    """Translate Estonian legal text to English."""
    if not tokenizer or not model:
        return "[Error] Translation model failed to load."
    try:
        inputs = tokenizer(text, return_tensors="pt", padding=True, truncation=True)
        outputs = model.generate(**inputs)
        translation = tokenizer.decode(outputs[0], skip_special_tokens=True)
        return translation
    except Exception as e:
        return f"[Error] Translation failed: {str(e)}"


# --- Gradio Interfaces ---

# Dataset Explorer Tab
dataset_interface = gr.Interface(
    fn=get_sample,
    inputs=[
        gr.Dropdown(choices=["train", "test"], label="Split", value="train"),
        gr.Number(label="Sample Index", value=0, precision=0)
    ],
    outputs=gr.JSON(label="Sample"),
    title="📚 Dataset Explorer",
    description="Browse Estonian–English legal sentence pairs.",
    flagging_mode="never"
)

# Translation Demo Tab
translate_interface = gr.Interface(
    fn=translate,
    inputs=gr.Textbox(lines=5, placeholder="Enter Estonian legal text...", label="Estonian Input"),
    outputs=gr.Textbox(label="English Translation"),
    title="🔤 Translate Legal Text",
    description="Translate Estonian legal sentences to English using Helsinki-NLP/opus-mt-et-en",
    flagging_mode="never"
)

# Combined App with Tabs
demo = gr.TabbedInterface(
    interface_list=[dataset_interface, translate_interface],
    tab_names=["📚 Dataset Explorer", "🔤 Translation Demo"],
)

# For Hugging Face Spaces compatibility
app = demo.app

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