app.py

# app.py

import gradio as gr
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline

model_id = "google/flan-t5-base"  # ~250MB, loads fast

# Load tokenizer and model correctly for FLAN-T5
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForSeq2SeqLM.from_pretrained(model_id)

# Create pipeline for text2text generation
polisher = pipeline("text2text-generation",
                    model=model,
                    tokenizer=tokenizer)

def oxford_polish_strict(sentence: str) -> str:
    prompt = (
        "Rewrite this sentence in formal written English, following the Oxford University Style Guide. "
        "Ensure tense matches time expressions (e.g. 'tomorrow' → future, 'yesterday' → past), "
        "use British spelling, apply the Oxford comma, and correct uncountable nouns naturally. "
        "Sentence: " + sentence
    )
    out = polisher(prompt, max_new_tokens=80, do_sample=False)
    return out[0]["generated_text"].strip()

# Gradio interface
demo = gr.Interface(
    fn=oxford_polish_strict,
    inputs=gr.Textbox(lines=2, placeholder="Enter a sentence to correct..."),
    outputs=gr.Textbox(label="Oxford-style Correction"),
    title="Oxford Grammar Polisher",
    description="Rewrite sentences in formal written English using Oxford grammar rules. Powered by FLAN-T5."
)

demo.launch()