app.py

import streamlit as st
import os
import time
import re
import requests
import tempfile
from openai import OpenAI
from streamlit_webrtc import webrtc_streamer, WebRtcMode, ClientSettings
import av
import numpy as np
import wave

# ------------------ 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")

# ------------------ 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 your Hugging Face Space settings.")
    st.stop()

client = OpenAI(api_key=OPENAI_API_KEY)

# ------------------ Session 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_url" not in st.session_state:
    st.session_state.image_url = None
if "audio_buffer" not in st.session_state:
    st.session_state.audio_buffer = []

# ------------------ 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)

# ------------------ Audio Recorder ------------------
class AudioProcessor:
    def __init__(self):
        self.frames = []

    def recv(self, frame):
        audio = frame.to_ndarray()
        self.frames.append(audio)
        return av.AudioFrame.from_ndarray(audio, layout="mono")

def save_wav(frames, path, rate=48000):
    audio_data = np.concatenate(frames)
    with wave.open(path, 'wb') as wf:
        wf.setnchannels(1)
        wf.setsampwidth(2)
        wf.setframerate(rate)
        wf.writeframes(audio_data.tobytes())

# ------------------ Sidebar & Image Panel ------------------
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.rerun()

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

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 & Voice Panel ------------------
with col2:
    # Display previous messages
    for message in st.session_state.messages:
        st.chat_message(message["role"]).write(message["content"])

    # 🎤 Real-time voice recorder
    st.subheader("🎙️ Ask with your voice")
    audio_ctx = webrtc_streamer(
        key="speech",
        mode=WebRtcMode.SENDONLY,
        in_audio_enabled=True,
        audio_receiver_size=256,
        client_settings=ClientSettings(
            media_stream_constraints={"audio": True, "video": False},
            rtc_configuration={"iceServers": [{"urls": ["stun:stun.l.google.com:19302"]}]},
        ),
    )

    if audio_ctx.audio_receiver:
        audio_processor = AudioProcessor()
        result = audio_ctx.audio_receiver.recv()
        audio_data = result.to_ndarray()
        st.session_state.audio_buffer.append(audio_data)

        # ⏱️ Auto stop after short time
        if len(st.session_state.audio_buffer) > 30:  # about 3s
            tmp_path = tempfile.NamedTemporaryFile(delete=False, suffix=".wav").name
            save_wav(st.session_state.audio_buffer, tmp_path)
            st.session_state.audio_buffer = []

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

            if transcript:
                st.success("📝 " + transcript)
                st.session_state.messages.append({"role": "user", "content": transcript})
                st.chat_message("user").write(transcript)
                prompt = transcript

                # ---- Assistant interaction ----
                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})

                    # Image link extract
                    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)
        # Send prompt to assistant logic follows same flow above (you can wrap in a function)