app.py

import streamlit as st
from PIL import Image
import os
import base64
import fitz  # PyMuPDF
from helper import (
    custom_file_uploader, resize_image, convert_image_to_base64, post_request_and_parse_response, 
    draw_bounding_boxes_for_textract, extract_text_from_textract_blocks, ChatGPTClient
)
import io

# Load OpenAI API Key from environment variable
OPENAI_API_KEY = os.environ["OPENAI_API_KEY"]
TEXTRACT_API_URL = os.environ["TEXTRACT_API_URL"]

st.set_page_config(page_title="💬 Chat with OCR 📝", layout="wide")

# Initialize chat history if not in session state
if "messages" not in st.session_state:
    st.session_state.messages = []

# Sidebar for image upload
with st.sidebar:
    st.title("🖼️ Upload and Display Images")
    
    # Display a placeholder for uploaded image
    st.warning("Please upload an image or a single-page PDF file!")
    uploaded_file = st.file_uploader("Upload an Image or PDF", type=['PDF'], label_visibility="collapsed")
    
    pil_image = None
    if uploaded_file:
        # Handle PDF file
        if uploaded_file.type == "application/pdf":
            try:
                # Read PDF as bytes
                pdf_bytes = uploaded_file.read()
                pdf_document = fitz.open(stream=pdf_bytes, filetype="pdf")
                
                # Check if the PDF has only one page
                if pdf_document.page_count != 1:
                    st.warning("Please upload a PDF with only one page!")
                else:
                    # Convert the first page to an image
                    page = pdf_document.load_page(0)
                    pix = page.get_pixmap()
                    image_bytes = pix.tobytes()
                    pil_image = Image.open(io.BytesIO(image_bytes))
            except Exception as e:
                st.error(f"Failed to convert PDF to image: {e}")
        else:
            # Handle image file
            pil_image = Image.open(uploaded_file)

        if pil_image:
            resized_image = resize_image(pil_image)

            with st.expander("Original Image", expanded=False):
                st.image(pil_image, caption="Uploaded Image", use_column_width=True)
            
            # Convert image to base64 and send to Textract API
            image_base64 = convert_image_to_base64(resized_image)
            payload = {"image": image_base64}
            result_dict = post_request_and_parse_response(TEXTRACT_API_URL, payload)
            
            # Draw bounding boxes
            image_with_boxes = draw_bounding_boxes_for_textract(resized_image.copy(), result_dict)
            
            with st.expander("Image with Bounding Boxes", expanded=True):
                st.image(image_with_boxes, caption="Image with Bounding Boxes", use_column_width=True)
            
            # Extract text from Textract
            cleaned_up_body = extract_text_from_textract_blocks(result_dict['body'])

            # Display JSON body in the sidebar inside an expander (default not expanded)
            with st.expander("View JSON Body", expanded=False):
                st.json(result_dict)

            # Display cleaned-up body (text extracted from JSON) in the sidebar inside an expander (default not expanded)
            with st.expander("View Cleaned-up Text", expanded=True):
                st.text_area("Cleaned-up Text", cleaned_up_body, height=200, key="cleaned_text_area", help="Drag the bottom right corner to resize")
    
    # Add some space at the bottom of the sidebar before the "Clear Session" button
    st.sidebar.markdown("<br><br><br><br>", unsafe_allow_html=True)
    
    # Clear session button at the bottom of the sidebar
    if st.sidebar.button("Clear Session"):
        st.session_state.messages = []

# Main chat interface
st.title("Chat with OCR Output")

# Display previous messages from session state
for message in st.session_state.messages:
    with st.chat_message(message["role"]):
        st.markdown(message["content"])

# Initialize ChatGPTClient with session state history
if uploaded_file and pil_image:
    history_copy = st.session_state.messages.copy()

    if cleaned_up_body:
        history_copy.append({"role": "system", "content": cleaned_up_body})

    bot = ChatGPTClient(
        api_key=OPENAI_API_KEY,
        protocol="You are fed with the text portion of json file that come out of OCR after scanning an image. User will ask you questions about this json file.",
        body=cleaned_up_body
    )
    bot.history = history_copy  # Set ChatGPT history to session state messages

# React to user input
if prompt := st.chat_input("Ask me about the image"):
    # Display user message in chat container
    st.chat_message("user").markdown(prompt)
    st.session_state.messages.append({"role": "user", "content": prompt})

    # Generate a response using ChatGPTClient
    if uploaded_file and pil_image:
        response = bot.generate_response(prompt)
    else:
        response = "Please upload an image before asking questions."

    # Display assistant message in chat container
    st.chat_message("assistant").markdown(response)
    st.session_state.messages.append({"role": "assistant", "content": response})