app.py

import streamlit as st
import os
import time
import re
import json
import requests
from PIL import Image
from openai import OpenAI

# ------------------ App Configuration ------------------
st.set_page_config(page_title="Document AI Assistant", layout="wide")
st.title("📄 Document AI Assistant")
st.caption("Chat with an AI Assistant on your medical/pathology documents")

# ------------------ Load API Key and Assistant ID ------------------
OPENAI_API_KEY = os.environ.get("OPENAI_API_KEY")
ASSISTANT_ID = os.environ.get("ASSISTANT_ID")

if not OPENAI_API_KEY or not ASSISTANT_ID:
    st.error("Missing secrets. Please ensure both OPENAI_API_KEY and ASSISTANT_ID are set in your Hugging Face Space secrets.")
    st.stop()

client = OpenAI(api_key=OPENAI_API_KEY)

# ------------------ Session State Initialization ------------------
if "messages" not in st.session_state:
    st.session_state.messages = []
if "thread_id" not in st.session_state:
    st.session_state.thread_id = None
if "image_url" not in st.session_state:
    st.session_state.image_url = None
if "image_updated" not in st.session_state:
    st.session_state.image_updated = False

# ------------------ Sidebar Controls ------------------
st.sidebar.header("🔧 Settings")
if st.sidebar.button("🔄 Clear Chat"):
    st.session_state.messages = []
    st.session_state.thread_id = None
    st.session_state.image_url = None
    st.session_state.image_updated = False
    st.rerun()

show_image = st.sidebar.checkbox("📖 Show Document Image", value=True)

# ------------------ Load Structured Summary/FAQ ------------------
with open("51940670-Manual-of-Surgical-Pathology-Third-Edition_1_structured_output.json", "r") as f:
    structured_data = json.load(f)  # This is a list of dicts, not a dict

# ------------------ Three-Column Layout ------------------
left, center, right = st.columns([1, 2, 1])

# ------------------ Left Column: Document Image ------------------
with left:
    st.subheader("📄 Document Image")
    if show_image and st.session_state.image_url:
        try:
            image = Image.open(requests.get(st.session_state.image_url, stream=True).raw)
            st.image(image, caption="📑 Extracted Page", use_container_width=True)
            st.session_state.image_updated = False
        except Exception as e:
            st.warning("⚠️ Could not load image.")

# ------------------ Center Column: Chat UI ------------------
with center:
    st.subheader("💬 Document AI Assistant")
    for message in st.session_state.messages:
        role, content = message["role"], message["content"]
        st.chat_message(role).write(content)

    if prompt := st.chat_input("Type your question about the document..."):
        st.session_state.messages.append({"role": "user", "content": prompt})
        st.chat_message("user").write(prompt)

        try:
            if st.session_state.thread_id is None:
                thread = client.beta.threads.create()
                st.session_state.thread_id = thread.id

            thread_id = st.session_state.thread_id

            client.beta.threads.messages.create(
                thread_id=thread_id,
                role="user",
                content=prompt
            )

            run = client.beta.threads.runs.create(
                thread_id=thread_id,
                assistant_id=ASSISTANT_ID
            )

            with st.spinner("Assistant is thinking..."):
                while True:
                    run_status = client.beta.threads.runs.retrieve(
                        thread_id=thread_id,
                        run_id=run.id
                    )
                    if run_status.status == "completed":
                        break
                    time.sleep(1)

            messages = client.beta.threads.messages.list(thread_id=thread_id)
            assistant_message = None
            for message in reversed(messages.data):
                if message.role == "assistant":
                    assistant_message = message.content[0].text.value
                    break

            st.chat_message("assistant").write(assistant_message)
            st.session_state.messages.append({"role": "assistant", "content": assistant_message})

            # Extract GitHub image URL
            image_match = re.search(
                r'https://raw\.githubusercontent\.com/AndrewLORTech/surgical-pathology-manual/main/[\w\-/]*\.png',
                assistant_message
            )
            if image_match:
                st.session_state.image_url = image_match.group(0)
                st.session_state.image_updated = True
                st.rerun()

        except Exception as e:
            st.error(f"❌ Error: {str(e)}")

# ------------------ Right Column: Summary and FAQ ------------------
with right:
    st.subheader("📌 Summary")

    # Parse page number from image URL if available
    if st.session_state.image_url:
        match = re.search(r'page_(\d+)', st.session_state.image_url)
        page_number = int(match.group(1)) if match else 151
    else:
        page_number = 151  # default

    # Get entry from structured data
    page_entry = next((entry for entry in structured_data if entry.get("page") == page_number), None)
    if page_entry:
        summary_text = page_entry.get("summary", "No summary available.")
        faq_list = page_entry.get("faqs", [])
    else:
        summary_text = "No summary available."
        faq_list = []

    st.markdown(summary_text)

    st.subheader("❓ Auto-Generated FAQ")
    if faq_list:
        for faq in faq_list:
            st.markdown(f"**Q:** {faq.get('question', '')}\n\n**A:** {faq.get('answer', '')}")
    else:
        st.info("No FAQs available for this page.")