app.py

import os

# ✅ Always use /tmp for Hugging Face cache in Spaces
os.environ["HF_HOME"] = "/tmp/huggingface"
os.environ["TRANSFORMERS_CACHE"] = "/tmp/transformers"
os.environ["HF_HUB_CACHE"] = "/tmp/hfhub"

# Create dirs (now in /tmp, which is writable)
os.makedirs(os.environ["HF_HOME"], exist_ok=True)
os.makedirs(os.environ["TRANSFORMERS_CACHE"], exist_ok=True)
os.makedirs(os.environ["HF_HUB_CACHE"], exist_ok=True)

import streamlit as st
import json
import re
import fitz  # PyMuPDF
from langdetect import detect, DetectorFactory
from googletrans import Translator
from transformers import pipeline

# for model serialization
import joblib

# for creating a folder
import os

# for hugging face space authentication to upload files
from huggingface_hub import login, HfApi

DetectorFactory.seed = 0

# Initialize Translator & Summarizer
# -------------------------
# Note: Initializing models here will load them when the app starts.
# Consider caching or lazy loading for performance in production.
translator = Translator()
summarizer = pipeline("summarization", model="sshleifer/distilbart-cnn-12-6")

# -------------------------
# Extract text from PDF
# ------------------------

def extract_text_from_pdf(pdf_file):
    text = ""
    # ✅ Save the uploaded file into /tmp (writable in Spaces/Docker)
    temp_file_path = os.path.join("/tmp", "temp.pdf")
    with open(temp_file_path, "wb") as f:
        f.write(pdf_file.getvalue())

    try:
        with fitz.open(temp_file_path) as doc:
            for page in doc:
                text += page.get_text("text")
    finally:
        # Ensure the temporary file is removed
        if os.path.exists(temp_file_path):
            os.remove(temp_file_path)

    return text.strip()


# Translate text to English using Google Translate
# -------------------------
def translate_text_google(text):
    if not text:
        return ""

    max_chunk = 5000  # Google Translate handles large text but splitting is safer
    chunks = [text[i:i+max_chunk] for i in range(0, len(text), max_chunk)]
    translations = []
    for chunk in chunks:
        translated = translator.translate(chunk, dest='en')
        translations.append(translated.text)
    return " ".join(translations)

# Summarize text safely
# -------------------------
def safe_summarize(text, max_length=150, min_length=30):
    if not text or len(text.split()) < 10:
        return text  # too short to summarize
    try:
        summary = summarizer(text, max_length=max_length, min_length=min_length, do_sample=False)
        return summary[0]['summary_text']
    except Exception as e:
        st.warning(f"⚠️ Summarization failed: {e}")
        return text

# Extract entities
# -------------------------
def extract_entities(text):
    entities = {}

    # PAN format: AAAAA9999A
    # pan_match = re.search(r"\b[A-Z]{5}\d{4}[A-Z]\b", text, re.IGNORECASE)
    pan_match = re.search(r"PAN (\w{5}\d{4}\w)", text)

    # Account Number
    # acc_match = re.search(r"account\s*number\s*[:\-]?\s*([A-Za-z0-9]+)", text, re.IGNORECASE)
    acc_match = re.search(r"Account Number (\w+)", text, re.IGNORECASE)

    # Penalty (accepts 'penalty' or 'penalties')
    # penalty_match = re.search(r"\bpenalt(?:y|ies)\s*[:\-]?\s*([\d,]+)", text, re.IGNORECASE)
    penalty_match = re.search(r"INR ([\d,]+)", text, re.IGNORECASE)

    # Deactivation keywords
    deactivate_match = re.search(r"\bdeactivat(?:e|ed|ion)\b", text, re.IGNORECASE)

    if pan_match:
        #entities["PAN"] = pan_match.group(0).upper()
        entities["PAN"] = pan_match.group(1) if pan_match else None

    if acc_match:
        entities["Account_Number"] = acc_match.group(1) if acc_match else None

    if penalty_match:
        #entities["Penalty"] = penalty_match.group(1).replace(",", "")
        entities["Penalty"] = penalty_match.group(1) if penalty_match else None

    if deactivate_match:
        entities["Deactivate"] = deactivate_match.group(0).lower()

    return entities

 # Trigger actions
# -------------------------
def trigger_action(entities):
    if "Penalty" in entities:
        return f"Penalty of {entities['Penalty']} recorded for account {entities.get('Account_Number', 'N/A')} (PAN: {entities.get('PAN', 'N/A')})"
    elif "Deactivate" in entities:
        return f"Kindy Deactivate {entities.get('Account_Number', 'N/A')} as per request having (PAN: {entities.get('PAN', 'N/A')})"
    elif "Account_Number" in entities:
        return f"Account {entities['Account_Number']} flagged for review."
    else:
        return "No action required"

# Process single PDF - adapted for Streamlit FileUploader
# -------------------------
def process_uploaded_pdf(pdf_file):
    raw_text = extract_text_from_pdf(pdf_file)
    lang = detect(raw_text)
    translated_text = translate_text_google(raw_text) if lang != "en" else raw_text
    summary = safe_summarize(translated_text)
    entities = extract_entities(translated_text)
    action_result = trigger_action(entities)

    result = {
        "file_name": pdf_file.name,
        "detected_language": lang,
        "raw_text_snippet": raw_text[:500] + ("..." if len(raw_text) > 500 else ""),
        "translated_text_snippet": translated_text[:500] + ("..." if len(translated_text) > 500 else ""),
        "summary": summary,
        "entities": entities,
        "action_triggered": action_result
    }

    return result


st.title("PDF Document Processor")
st.write("Upload a PDF file to extract text, translate (if needed), summarize, identify key entities, and suggest actions.")

uploaded_file = st.file_uploader("Choose a PDF file", type="pdf")

if uploaded_file is not None:
    st.write("Processing PDF...")
    try:
        processed_data = process_uploaded_pdf(uploaded_file)

        st.subheader("Processing Results:")
        st.write(f"**File Name:** {processed_data['file_name']}")
        st.write(f"**Detected Language:** {processed_data['detected_language']}")
        st.write(f"**Raw Text Snippet:** {processed_data['raw_text_snippet']}")
        st.write(f"**Translated Text Snippet:** {processed_data['translated_text_snippet']}")
        st.write(f"**Summary:** {processed_data['summary']}")
        st.write(f"**Extracted Entities:**")
        for key, value in processed_data['entities'].items():
            st.write(f"- {key}: {value}")
        st.write(f"**Action Triggered:** {processed_data['action_triggered']}")

    except Exception as e:
        st.error(f"An error occurred during processing: {e}")