app.py

import streamlit as st
from transformers import pipeline
import pandas as pd
import re
from PyPDF2 import PdfReader

# Cache the NER model for performance
@st.cache_resource
def load_ner_model():
    return pipeline("ner", model="dslim/bert-base-NER", grouped_entities=True)

ner_model = load_ner_model()

# Function to extract text from PDF
def extract_text_from_pdf(file):
    pdf_reader = PdfReader(file)
    text = ""
    for page in pdf_reader.pages:
        text += page.extract_text() or ""
    return text

# Main app
def main():
    st.title("Invoice Data Extraction (NER)")
    st.write("Extract entities like company names, dates, amounts, and other details from invoice text or uploaded files using Named Entity Recognition.")

    # Input options: Text or File Upload
    input_method = st.radio("Choose input method:", ("Text Input", "Upload File"))

    invoice_text = ""
    if input_method == "Text Input":
        invoice_text = st.text_area("Invoice Text", placeholder="Paste your invoice text here...", height=200)
    elif input_method == "Upload File":
        uploaded_file = st.file_uploader("Upload Invoice (PDF or TXT)", type=["pdf", "txt"])
        if uploaded_file is not None:
            if uploaded_file.type == "application/pdf":
                invoice_text = extract_text_from_pdf(uploaded_file)
            else:  # txt
                invoice_text = uploaded_file.read().decode("utf-8")
            st.text_area("Extracted Text", value=invoice_text, height=200, disabled=True)

    # Button and output
    if st.button("Extract"):
        if not invoice_text:
            st.warning("Please enter text or upload a file to extract!")
            return

        with st.spinner("Extracting entities..."):
            # Perform NER
            entities = ner_model(invoice_text)
            
            # Initialize entity dictionary with broader fields
            entity_dict = {
                "Organization": [],
                "Date": [],
                "Amount": [],
                "Supplier": [],
                "Item": [],
                "Due By": [],
                "Invoice Number": []
            }
            
            # Process NER entities
            for entity in entities:
                if entity["entity_group"] == "ORG":
                    entity_dict["Organization"].append(entity["word"])
                elif entity["entity_group"] in ["DATE", "TIME"]:
                    entity_dict["Date"].append(entity["word"])

            # Enhanced heuristic rules for invoice-specific fields
            # Supplier (look for "Supplier:", "From:", or "From" prefix)
            supplier_match = re.search(r"(?:Supplier|From):?\s*([^\n]+)", invoice_text, re.IGNORECASE)
            if supplier_match and "Global Trading Ltd" not in entity_dict["Organization"]:  # Avoid duplication
                entity_dict["Supplier"].append(supplier_match.group(1).strip())
            
            # Date (enhanced regex for multiple formats)
            date_patterns = [
                r"\b\d{1,2}[/-]\d{1,2}[/-]\d{4}\b",  # e.g., 03/22/2025
                r"\b\d{1,2}\s+(?:January|February|March|April|May|June|July|August|September|October|November|December)\s+\d{4}\b",  # e.g., 12 December 2024
                r"\b\d{1,2}-[A-Za-z]{3}-\d{4}\b"  # e.g., 15-Jan-2025
            ]
            for pattern in date_patterns:
                dates = re.findall(pattern, invoice_text)
                if dates:
                    entity_dict["Date"].extend(dates)
            
            # Item (look for "Item:" prefix)
            item_matches = re.findall(r"Item:?\s*([^\n$€]+)(?:\s*[$€]\d+\.\d{2})?", invoice_text, re.IGNORECASE)
            if item_matches:
                entity_dict["Item"].extend([item.strip() for item in item_matches if item.strip()])
            
            # Due By (look for "Due by:", "Payment due by:", or similar)
            due_by_match = re.search(r"(?:Due by|Payment due by):?\s*([^\n]+)", invoice_text, re.IGNORECASE)
            if due_by_match:
                entity_dict["Due By"].append(due_by_match.group(1).strip())
            
            # Invoice Number (look for "Invoice #")
            invoice_match = re.search(r"Invoice #?(\w+)", invoice_text, re.IGNORECASE)
            if invoice_match:
                entity_dict["Invoice Number"].append(invoice_match.group(1).strip())
            
            # Amount (improved regex for currency and numbers)
            amount_pattern = r'(?:\$\d+\.\d{2}|\€\d+\.\d{2})'
            amounts = re.findall(amount_pattern, invoice_text)
            if amounts:
                entity_dict["Amount"].extend(amounts)
            
            # Clean up empty categories
            entity_dict = {k: list(set(v)) if v else ["Not found"] for k, v in entity_dict.items()}  # Remove duplicates

            # Display results as a table
            df = pd.DataFrame({
                "Entity Type": list(entity_dict.keys()),
                "Extracted Value": [", ".join(v) for v in entity_dict.values()]
            })
            st.success("Extracted Entities:")
            st.table(df)

            # Footer (shown only after extraction)
            st.markdown("""
                <p style="font-size: small; color: grey; text-align: center; margin-top: 20px; border-top: 1px solid #eee; padding-top: 10px;">
                    Developed By: Krishna Prakash
                    <a href="https://www.linkedin.com/in/krishnaprakash-profile/" target="_blank">
                        <img src="https://img.icons8.com/ios-filled/30/0077b5/linkedin.png" alt="LinkedIn" style="vertical-align: middle; margin: 0 5px;"/>
                    </a>
                </p>
            """, unsafe_allow_html=True)

if __name__ == "__main__":
    main()