app.py

import streamlit as st
import pdfplumber
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

# Load the model
@st.cache_resource
def load_model():
    tokenizer = AutoTokenizer.from_pretrained("OFA-Sys/mineru-base")
    model = AutoModelForSeq2SeqLM.from_pretrained("OFA-Sys/mineru-base")
    return tokenizer, model

tokenizer, model = load_model()

# UI
st.title("📄 MinerU: Ask Questions from PDF")
uploaded_file = st.file_uploader("Upload a PDF", type="pdf")
question = st.text_input("Enter your question:")

if uploaded_file and question:
    with pdfplumber.open(uploaded_file) as pdf:
        text = ''
        for page in pdf.pages:
            text += page.extract_text()

    # Prepare input for MinerU (usually expects a prompt)
    input_text = f"question: {question} context: {text[:3000]}"  # MinerU has token limit

    inputs = tokenizer(input_text, return_tensors="pt", truncation=True)
    outputs = model.generate(**inputs, max_new_tokens=128)
    answer = tokenizer.decode(outputs[0], skip_special_tokens=True)

    st.markdown(f"### Answer:\n{answer}")