app.py

import streamlit as st
import os
import time
import re
import requests
import tempfile
import wave
import numpy as np
from openai import OpenAI
from streamlit_audio_recorder import audio_recorder

# ------------------ Page Config ------------------
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 Secrets ------------------
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 set both OPENAI_API_KEY and ASSISTANT_ID in Hugging Face Space settings.")
    st.stop()

client = OpenAI(api_key=OPENAI_API_KEY)

# ------------------ Session State Init ------------------
for key in ["messages", "thread_id", "image_url", "transcript"]:
    if key not in st.session_state:
        st.session_state[key] = [] if key == "messages" else None

# ------------------ Whisper Transcription ------------------
def transcribe_audio(file_path, api_key):
    with open(file_path, "rb") as f:
        response = requests.post(
            "https://api.openai.com/v1/audio/transcriptions",
            headers={"Authorization": f"Bearer {api_key}"},
            files={"file": f},
            data={"model": "whisper-1"}
        )
    return response.json().get("text", None)

# ------------------ Sidebar & Layout ------------------
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.transcript = None
    st.rerun()

show_image = st.sidebar.checkbox("📖 Show Document Image", value=True)
col1, col2 = st.columns([1, 2])

# ------------------ Image Panel ------------------
with col1:
    if show_image and st.session_state.image_url:
        st.image(st.session_state.image_url, caption="📑 Extracted Page", use_container_width=True)

# ------------------ Chat + Mic Panel ------------------
with col2:
    for message in st.session_state.messages:
        st.chat_message(message["role"]).write(message["content"])

    st.subheader("🎙️ Ask with Your Voice")

    audio_bytes = audio_recorder(pause_threshold=3.0, energy_threshold=-1.0, sample_rate=44100)
    
    if audio_bytes:
        # Save temporary WAV file
        with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmpfile:
            tmpfile.write(audio_bytes)
            tmp_path = tmpfile.name

        st.audio(tmp_path, format="audio/wav")

        with st.spinner("🧠 Transcribing..."):
            transcript = transcribe_audio(tmp_path, OPENAI_API_KEY)

        if transcript:
            st.success("📝 Transcript: " + transcript)
            st.session_state.transcript = transcript

    # Submit Transcript to Assistant
    if st.session_state.transcript:
        if st.button("✅ Send Transcript to Assistant"):
            user_input = st.session_state.transcript
            st.session_state.transcript = None  # reset

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

            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=user_input)
                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 = next(
                    (m.content[0].text.value for m in reversed(messages.data) if m.role == "assistant"), None
                )

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

                # Extract GitHub image if available
                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)

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

    # Fallback text input
    if prompt := st.chat_input("💬 Or type your question..."):
        st.session_state.messages.append({"role": "user", "content": prompt})
        st.chat_message("user").write(prompt)
        st.session_state.transcript = prompt  # Treat like voice input for now