app.py

import gradio as gr
from transformers import AutoTokenizer, AutoModelForQuestionAnswering
import torch
from pypdf import PdfReader

MODEL_NAME = "harishforaiandml/my-pretrained-qa-model"

tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
model = AutoModelForQuestionAnswering.from_pretrained(MODEL_NAME)
model.eval()


# ----------------------------
# PDF TEXT EXTRACTION
# ----------------------------
def extract_text_from_pdf(pdf_file):
    reader = PdfReader(pdf_file)
    text = ""

    for page in reader.pages:
        text += page.extract_text() + "\n"

    return text


# ----------------------------
# SIMPLE CHUNKING (mini RAG)
# ----------------------------
def chunk_text(text, chunk_size=500):
    words = text.split()
    chunks = []

    for i in range(0, len(words), chunk_size):
        chunk = " ".join(words[i:i + chunk_size])
        chunks.append(chunk)

    return chunks


# ----------------------------
# FIND BEST CHUNK (retrieval step)
# ----------------------------
def get_best_chunk(chunks, question):
    question_words = set(question.lower().split())

    best_chunk = ""
    best_score = 0

    for chunk in chunks:
        chunk_words = set(chunk.lower().split())
        score = len(question_words.intersection(chunk_words))

        if score > best_score:
            best_score = score
            best_chunk = chunk

    return best_chunk


# ----------------------------
# QA FUNCTION
# ----------------------------
def ask_pdf(pdf_file, question):

    if pdf_file is None:
        return "Please upload a PDF"

    text = extract_text_from_pdf(pdf_file)
    chunks = chunk_text(text)

    context = get_best_chunk(chunks, question)

    inputs = tokenizer(
        question,
        context,
        return_tensors="pt",
        truncation=True
    )

    with torch.no_grad():
        outputs = model(**inputs)

    start = torch.argmax(outputs.start_logits)
    end = torch.argmax(outputs.end_logits) + 1

    answer_tokens = inputs["input_ids"][0][start:end]
    answer = tokenizer.decode(answer_tokens, skip_special_tokens=True)

    if answer.strip() == "":
        return "Not found in document"

    return answer


# ----------------------------
# GRADIO UI
# ----------------------------
demo = gr.Interface(
    fn=ask_pdf,

    inputs=[
        gr.File(label="Upload PDF"),
        gr.Textbox(label="Question")
    ],

    outputs="text",

    title="📄 PDF QA (Mini RAG System)",
    description="Upload a PDF and ask questions. System finds best chunk and extracts answer using DistilBERT."
)

demo.launch()