app.py

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

# ------------------ Authentication ------------------
VALID_USERS = {
    "[email protected]": "Pass.123",
    "[email protected]": "Pass.123",
    "[email protected]": "Pass.123",
    "[email protected]": "Pass.123",
}

def login():
    st.title("🔐 Login Required")
    email = st.text_input("Email")
    password = st.text_input("Password", type="password")
    if st.button("Login"):
        if VALID_USERS.get(email) == password:
            st.session_state.authenticated = True
            st.rerun()
        else:
            st.error("❌ Incorrect email or password.")

if "authenticated" not in st.session_state:
    st.session_state.authenticated = False

if not st.session_state.authenticated:
    login()
    st.stop()

# ------------------ App Config ------------------
st.set_page_config(page_title="AI Pathology Assistant", layout="wide", initial_sidebar_state="collapsed")
st.title("🧬 AI Pathology Assistant")

# ------------------ Load OpenAI ------------------
OPENAI_API_KEY = os.environ.get("OPENAI_API_KEY")
if not OPENAI_API_KEY:
    st.error("❌ Missing OPENAI_API_KEY environment variable.")
    st.stop()

client = OpenAI(api_key=OPENAI_API_KEY)

# ------------------ Assistant Setup ------------------
ASSISTANT_ID = "asst_jXDSjCG8LI4HEaFEcjFVq8KB"

# ------------------ State ------------------
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_urls" not in st.session_state:
    st.session_state.image_urls = []
if "pending_prompt" not in st.session_state:
    st.session_state.pending_prompt = None

# ------------------ Tabs ------------------
tab1, tab2 = st.tabs(["💬 Chat Assistant", "🖼️ Visual Reference Search"])

# ------------------ Tab 1: Chat Assistant ------------------
with tab1:
    # Sidebar
    with st.sidebar:
        st.header("🧪 Pathology Tools")

        if st.button("🧹 Clear Chat"):
            for k in ["messages", "thread_id", "image_urls", "pending_prompt"]:
                st.session_state[k] = [] if k.endswith("s") else None
            st.rerun()

        show_image = st.toggle("📸 Show Images", value=True)
        keyword = st.text_input("Keyword Search", placeholder="e.g. mitosis, carcinoma")
        if st.button("🔎 Search") and keyword:
            st.session_state.pending_prompt = f"Find clauses or references related to: {keyword}"

        section = st.text_input("Section Lookup", placeholder="e.g. Connective Tissue")
        if section:
            st.session_state.pending_prompt = f"Summarize or list key points from section: {section}"

        action = st.selectbox("Common Pathology Queries", [
            "Select an action...",
            "List histological features of inflammation",
            "Summarize features of carcinoma",
            "List muscle types and features",
            "Extract diagnostic markers",
            "Summarize embryology stages"
        ])
        if action != "Select an action...":
            st.session_state.pending_prompt = action

    chat_col, image_col = st.columns([2, 1])

    with chat_col:
        st.markdown("### 💬 Ask a Pathology-Specific Question")
        user_input = st.chat_input("Example: What are features of squamous cell carcinoma?")

        if user_input:
            st.session_state.messages.append({"role": "user", "content": user_input})
        elif st.session_state.pending_prompt:
            st.session_state.messages.append({"role": "user", "content": st.session_state.pending_prompt})
            st.session_state.pending_prompt = None

        # Assistant logic
        if st.session_state.messages and st.session_state.messages[-1]["role"] == "user":
            try:
                if not st.session_state.thread_id:
                    st.session_state.thread_id = client.beta.threads.create().id

                client.beta.threads.messages.create(
                    thread_id=st.session_state.thread_id,
                    role="user",
                    content=st.session_state.messages[-1]["content"]
                )

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

                with st.spinner("🔬 Analyzing..."):
                    while True:
                        status = client.beta.threads.runs.retrieve(thread_id=st.session_state.thread_id, run_id=run.id)
                        if status.status in ("completed", "failed", "cancelled"):
                            break
                        time.sleep(1)

                if status.status == "completed":
                    responses = client.beta.threads.messages.list(thread_id=st.session_state.thread_id).data
                    for m in reversed(responses):
                        if m.role == "assistant":
                            full_reply = m.content[0].text.value.strip()

                            # ✂️ Extract and remove "Some Possible Questions"
                            question_block_match = re.search(
                                r"Some Possible Questions:\s*(.*?)(\n{2,}|Other References:|$)",
                                full_reply,
                                re.DOTALL
                            )
                            question_block = None
                            if question_block_match:
                                question_block = question_block_match.group(1)
                                full_reply = full_reply.replace(f"Some Possible Questions:\n{question_block}", "").strip()

                            reply = full_reply

                            if not any(reply in msg["content"] or msg["content"] in reply
                                       for msg in st.session_state.messages if msg["role"] == "assistant"):
                                st.session_state.messages.append({"role": "assistant", "content": reply})

                                # Extract image URLs
                                images = re.findall(
                                    r'https://raw\.githubusercontent\.com/AndrewLORTech/witspathologai/main/[^\s\n"]+\.png',
                                    reply
                                )
                                st.session_state.image_urls = images
                            break
                else:
                    st.error("❌ Assistant failed to complete.")
                st.rerun()
            except Exception as e:
                st.error(f"❌ Error: {e}")

        # Display messages
        for msg in st.session_state.messages:
            with st.chat_message(msg["role"]):
                st.markdown(msg["content"], unsafe_allow_html=True)

        # ✅ Follow-Up Suggestions (from stripped question block)
        if 'question_block' in locals() and question_block:
            questions = [
                line.strip(" -•") for line in question_block.splitlines()
                if line.strip().startswith(("-", "•")) and line.strip().endswith("?")
            ]
            if questions:
                st.markdown("#### 💡 Follow-Up Suggestions")
                for q in questions:
                    if st.button(f"📌 {q}"):
                        st.session_state.pending_prompt = q
                        st.rerun()

    with image_col:
        if show_image and st.session_state.image_urls:
            st.markdown("### 🖼️ Images")
            for url in st.session_state.image_urls:
                try:
                    img = Image.open(BytesIO(requests.get(url, timeout=5).content))
                    st.image(img, caption=url.split("/")[-1], use_container_width=True)
                except Exception:
                    st.warning(f"⚠️ Failed to load image: {url}")

# ------------------ Tab 2: Visual Reference Search ------------------
with tab2:
    ASSISTANT_ID = "asst_9v09zgizdcuuhNdcFQpRo9RO"

    if "image_thread_id" not in st.session_state:
        st.session_state.image_thread_id = None
    if "image_response" not in st.session_state:
        st.session_state.image_response = None
    if "image_results" not in st.session_state:
        st.session_state.image_results = []
    if "image_lightbox" not in st.session_state:
        st.session_state.image_lightbox = None

    image_input = st.chat_input("Ask for histology visual references (e.g. ovary histology, mitosis)")
    if image_input:
        st.session_state.image_response = None
        st.session_state.image_results = []
        st.session_state.image_lightbox = None

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

            client.beta.threads.messages.create(
                thread_id=st.session_state.image_thread_id,
                role="user",
                content=image_input
            )

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

            with st.spinner("🔬 Searching for histology references..."):
                while True:
                    run_status = client.beta.threads.runs.retrieve(
                        thread_id=st.session_state.image_thread_id,
                        run_id=run.id
                    )
                    if run_status.status in ("completed", "failed", "cancelled"):
                        break
                    time.sleep(1)

            if run_status.status == "completed":
                messages = client.beta.threads.messages.list(thread_id=st.session_state.image_thread_id)
                for msg in reversed(messages.data):
                    if msg.role == "assistant":
                        response_text = msg.content[0].text.value
                        st.session_state.image_response = response_text

                        # ✅ FINAL FIXED PARSER: Handles Image URL on next line
                        lines = response_text.splitlines()
                        image_urls = []
                        expecting_url = False

                        for line in lines:
                            line_clean = line.strip().replace("**", "")
                            if "Image URL:" in line_clean:
                                parts = line_clean.split("Image URL:")
                                if len(parts) > 1 and parts[1].strip().startswith("http"):
                                    image_urls.append(parts[1].strip())
                                else:
                                    expecting_url = True  # Flag next line
                            elif expecting_url:
                                if line_clean.startswith("http"):
                                    image_urls.append(line_clean)
                                expecting_url = False

                        st.session_state.image_results = [{"image": url} for url in image_urls]

                        if image_urls and not st.session_state.image_lightbox:
                            st.session_state.image_lightbox = image_urls[0]
                        break
        except Exception as e:
            st.error(f"❌ Visual Assistant Error: {e}")

    text_col, image_col = st.columns([2, 1])

    with text_col:
        if st.session_state.image_response:
            st.markdown("### 🧠 Assistant Response")
            st.markdown(st.session_state.image_response, unsafe_allow_html=True)
            # Optional: Debug
            # st.code(st.session_state.image_response, language="markdown")

    with image_col:
        if st.session_state.image_results:
            st.markdown("### 🖼️ Image Preview(s)")
            for i, item in enumerate(st.session_state.image_results):
                image_url = item.get("image")
                if image_url:
                    try:
                        encoded_url = requests.utils.requote_uri(image_url)
                        r = requests.get(encoded_url, timeout=10)
                        r.raise_for_status()
                        img = Image.open(BytesIO(r.content))
                        st.image(img, caption=encoded_url.split("/")[-1], use_container_width=True)
                        if st.button("🔍 View Full Image", key=f"full_img_{i}"):
                            st.session_state.image_lightbox = encoded_url
                    except Exception as e:
                        st.warning(f"⚠️ Could not load: {image_url}")
                        st.error(f"{e}")
        else:
            st.info("ℹ️ No image references found yet.")

    if st.session_state.image_lightbox:
        st.markdown("### 🔬 Full Image View")
        try:
            img_url = st.session_state.image_lightbox
            r = requests.get(img_url, timeout=10)
            r.raise_for_status()
            full_img = Image.open(BytesIO(r.content))
            st.image(full_img, caption=img_url.split("/")[-1], use_container_width=True)
        except Exception as e:
            st.warning("⚠️ Could not load full image.")
            st.error(str(e))
        if st.button("❌ Close Preview"):
            st.session_state.image_lightbox = None
            st.rerun()