app.py

import streamlit as st
import base64
from openai import OpenAI
from PIL import Image
import io
import cv2
import numpy as np

# Configure app
st.set_page_config(
    page_title="AI Vision Assistant",
    page_icon="🔍",
    layout="wide",
    initial_sidebar_state="expanded"
)

# Custom CSS (keep your existing CSS here)
st.markdown("""
<style>
    /* Your existing CSS styles */
</style>
""", unsafe_allow_html=True)

# App title and description
st.title("🔍 Optimus Alpha | Live Vision Assistant")

# Initialize OpenAI client (keep your existing cached function)
@st.cache_resource
def get_client():
    return OpenAI(
        base_url="https://openrouter.ai/api/v1",
        api_key='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'
    )

# ===== New Live Camera Section =====
st.subheader("Live Camera Feed")
run_camera = st.checkbox("Enable Camera", value=False)

FRAME_WINDOW = st.empty()
captured_image = None

if run_camera:
    cap = cv2.VideoCapture(0)
    
    capture_button = st.button("Capture Image")
    stop_button = st.button("Stop Camera")
    
    if stop_button:
        run_camera = False
        cap.release()
        st.experimental_rerun()
    
    while run_camera:
        ret, frame = cap.read()
        if not ret:
            st.error("Failed to access camera")
            break
        
        # Display the live feed
        frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
        FRAME_WINDOW.image(frame)
        
        if capture_button:
            captured_image = frame
            run_camera = False
            cap.release()
            break
else:
    FRAME_WINDOW.info("Camera is currently off")

# ===== Image Processing Section =====
col1, col2 = st.columns([1, 2])

with col1:
    st.subheader("Image Source")
    
    # Option to use captured image or upload
    if captured_image is not None:
        st.image(captured_image, caption="Captured Image", width=300)
        use_captured = True
    else:
        use_captured = False
    
    uploaded_file = st.file_uploader(
        "Or upload an image",
        type=["jpg", "jpeg", "png"],
        disabled=use_captured
    )
    
    # Determine which image to use
    if use_captured:
        image = Image.fromarray(captured_image)
    elif uploaded_file:
        image = Image.open(uploaded_file)
    else:
        image = None

with col2:
    st.subheader("AI Analysis")
    
    user_prompt = st.text_input(
        "Your question about the image:",
        placeholder="e.g. 'What objects do you see?' or 'Explain this diagram'",
        key="user_prompt"
    )
    
    if st.button("Analyze", type="primary") and image:
        try:
            # Convert image to base64
            buffered = io.BytesIO()
            image.save(buffered, format="JPEG")
            image_base64 = base64.b64encode(buffered.getvalue()).decode("utf-8")
            
            # Prepare messages
            messages = [
                {
                    "role": "system",
                    "content": """You are a real-time vision assistant. Analyze the current camera feed or uploaded image and:
1. Identify objects, people, text clearly
2. Answer follow-up questions precisely
3. Format responses with bullet points
4. Highlight urgent/important findings"""
                },
                {
                    "role": "user",
                    "content": [
                        {
                            "type": "text",
                            "text": user_prompt if user_prompt else "Describe what you see in detail"
                        },
                        {
                            "type": "image_url",
                            "image_url": {
                                "url": f"data:image/jpeg;base64,{image_base64}"
                            }
                        }
                    ]
                }
            ]
            
            # Stream the response
            response_container = st.empty()
            full_response = ""
            
            client = get_client()
            stream = client.chat.completions.create(
                model="openrouter/optimus-alpha",
                messages=messages,
                stream=True
            )
            
            for chunk in stream:
                if chunk.choices[0].delta.content is not None:
                    full_response += chunk.choices[0].delta.content
                    response_container.markdown(f"""
                        <div class="markdown-text">
                        {full_response}
                        </div>
                    """, unsafe_allow_html=True)
        
        except Exception as e:
            st.error(f"Error: {str(e)}")

# Sidebar (keep your existing sidebar)
with st.sidebar:
    st.image("blob.png", width=200)
    st.markdown("""
    *Powered by OpenRouter*
    """)
    st.markdown("---")
    st.markdown("Made with ❤️ by Koshur AI")