src/app.py

import streamlit as st
import requests
import pandas as pd
from together import Together
import os

# =============================================================================
# CONFIGURATION - Using Secrets Management
# =============================================================================
NOCODB_URL = "https://app.nocodb.com"  # Base URL

# Get sensitive data from Streamlit secrets or environment variables
def get_api_credentials():
    """Get API credentials from secrets or environment"""
    try:
        # Try Streamlit secrets first (for Hugging Face Spaces)
        api_token = st.secrets.get("NOCODB_API_TOKEN", os.environ.get("NOCODB_API_TOKEN", ""))
        together_key = st.secrets.get("TOGETHER_API_KEY", os.environ.get("TOGETHER_API_KEY", ""))
        endpoint_path = st.secrets.get("NOCODB_ENDPOINT_PATH", os.environ.get("NOCODB_ENDPOINT_PATH", ""))
        
        return api_token, together_key, endpoint_path
    except:
        # Fallback to environment variables
        api_token = os.environ.get("NOCODB_API_TOKEN", "")
        together_key = os.environ.get("TOGETHER_API_KEY", "")
        endpoint_path = os.environ.get("NOCODB_ENDPOINT_PATH", "")
        
        return api_token, together_key, endpoint_path

# Initialize Together AI client
@st.cache_resource
def get_ai_client():
    """Initialize Together AI client"""
    _, together_key, _ = get_api_credentials()
    if not together_key:
        st.error("Together AI API key not found. Please configure it in the secrets.")
        return None
    return Together(api_key=together_key)

# =============================================================================
# HELPER FUNCTIONS
# =============================================================================
def safe_int(value, default=0):
    """Safely convert value to integer"""
    try:
        return int(float(value)) if value else default
    except (ValueError, TypeError):
        return default

def safe_float(value, default=0.0):
    """Safely convert value to float"""
    try:
        return float(value) if value else default
    except (ValueError, TypeError):
        return default

@st.cache_data(ttl=300)  # Cache for 5 minutes
def get_properties():
    """Fetch properties from NocoDB"""
    api_token, _, endpoint_path = get_api_credentials()
    
    if not api_token or not endpoint_path:
        st.error("NocoDB credentials not configured. Please set up your secrets.")
        return []
    
    headers = {"xc-token": api_token}
    
    try:
        response = requests.get(
            f"{NOCODB_URL}{endpoint_path}?limit=1000",  # Get more records
            headers=headers
        )
        
        if response.status_code == 200:
            data = response.json()
            return data.get('list', [])
        else:
            st.error(f"Failed to fetch data: {response.status_code}")
            return []
            
    except Exception as e:
        st.error(f"Error connecting to database: {e}")
        return []

def filter_properties(properties, filters):
    """Apply filters to properties list"""
    filtered = []
    
    for prop in properties:
        # Price filter
        price = safe_int(prop.get('cash_price'))
        if price > filters['max_price']:
            continue
            
        # Rooms filter
        rooms = safe_int(prop.get('rooms'))
        if rooms < filters['min_rooms']:
            continue
            
        # Energy rating filter
        if filters['energy_ratings'] and prop.get('energy_rating') not in filters['energy_ratings']:
            continue
            
        # City filter
        if filters['cities'] and prop.get('city') not in filters['cities']:
            continue
            
        filtered.append(prop)
    
    return filtered

def create_property_context(properties):
    """Create context string about current properties for AI"""
    if not properties:
        return "No properties match the current filters."
    
    total = len(properties)
    prices = [safe_int(p.get('cash_price')) for p in properties if safe_int(p.get('cash_price')) > 0]
    
    if prices:
        avg_price = sum(prices) / len(prices)
        min_price = min(prices)
        max_price = max(prices)
        
        context = f"""Currently showing {total} Danish villas. 
        Price range: {min_price:,} - {max_price:,} DKK. 
        Average price: {avg_price:,.0f} DKK. """
    else:
        context = f"Currently showing {total} Danish villas. "
    
    # Add some location info
    cities = list(set([p.get('city', 'Unknown') for p in properties[:10]]))
    if cities:
        context += f"Cities include: {', '.join(cities[:5])}. "
    
    return context

def get_ai_response(client, question, context, model_name):
    """Get response from Together AI"""
    try:
        # Create a comprehensive prompt
        prompt = f"""You are a helpful Danish real estate assistant. Based on the current property data, please answer the user's question accurately and helpfully.

Current Property Data Context:
{context}

User Question: {question}

Please provide a helpful, accurate response based on the data provided. Keep your answer concise but informative."""

        response = client.chat.completions.create(
            model=model_name,
            messages=[
                {"role": "system", "content": "You are a helpful Danish real estate assistant with expertise in property analysis and market insights."},
                {"role": "user", "content": prompt}
            ],
            max_tokens=300,
            temperature=0.7,
        )
        
        return response.choices[0].message.content
        
    except Exception as e:
        raise Exception(f"Together AI Error: {str(e)}")

def test_together_models():
    """Test different Together AI models"""
    # Include both Gemma and other reliable serverless models
    models_to_test = [
        # Gemma models (Google's lightweight models)
        "google/gemma-2b-it", 
        # Other reliable models
        "mistralai/Mistral-7B-Instruct-v0.1",
        "NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO",
        "mistralai/Mixtral-8x7B-Instruct-v0.1"
    ]
    
    results = {}
    client = get_ai_client()
    
    if not client:
        return {"error": "Could not initialize AI client"}
    
    for model_name in models_to_test:
        try:
            test_response = client.chat.completions.create(
                model=model_name,
                messages=[
                    {"role": "system", "content": "You are a helpful assistant."},
                    {"role": "user", "content": "Hello, can you help me analyze real estate data?"}
                ],
                max_tokens=50,
                temperature=0.7,
            )
            
            results[model_name] = {
                "status": "✅ Success", 
                "response": test_response.choices[0].message.content[:100]
            }
            
        except Exception as e:
            results[model_name] = {"status": "❌ Error", "response": str(e)[:100]}
    
    return results

# =============================================================================
# MAIN APP
# =============================================================================
def main():
    # Page config
    st.set_page_config(
        page_title="Danish Villa Assistant",
        page_icon="🏡",
        layout="wide"
    )
    
    # Header
    st.title("🏡 Danish Villa Assistant")
    st.write("Explore Danish villas with AI-powered insights using Together AI!")
    
    # Check API credentials
    api_token, together_key, endpoint_path = get_api_credentials()
    
    if not together_key:
        st.error("⚠️ Together AI API key not configured!")
        st.info("Please set your TOGETHER_API_KEY in the Hugging Face Spaces secrets.")
        st.stop()
    
    if not api_token or not endpoint_path:
        st.error("⚠️ NocoDB credentials not configured!")
        st.info("Please set NOCODB_API_TOKEN and NOCODB_ENDPOINT_PATH in the Hugging Face Spaces secrets.")
        st.stop()
    
    # Add model testing section
    with st.expander("🧪 Test Together AI Models (for debugging)"):
        if st.button("Test Different Models"):
            with st.spinner("Testing models..."):
                test_results = test_together_models()
                for model, result in test_results.items():
                    st.write(f"**{model}:** {result['status']}")
                    if result['status'] == "✅ Success":
                        st.success(f"Response preview: {result['response']}")
                    else:
                        st.error(f"Error: {result['response']}")
    
    # Initialize AI client
    try:
        client = get_ai_client()
        if not client:
            st.stop()
    except Exception as e:
        st.error(f"Failed to initialize Together AI client: {e}")
        st.stop()
    
    # Sidebar filters
    st.sidebar.header("🔍 Filter Properties")
    
    # Get all properties first to populate filter options
    with st.spinner("Loading properties..."):
        all_properties = get_properties()
    
    if not all_properties:
        st.error("Could not load properties. Please check your NocoDB connection.")
        st.stop()
    
    # Extract unique values for filters
    all_cities = sorted(list(set([p.get('city', 'Unknown') for p in all_properties if p.get('city')])))
    all_energy_ratings = sorted(list(set([p.get('energy_rating') for p in all_properties if p.get('energy_rating')])))
    
    # Sidebar filter controls
    max_price = st.sidebar.slider(
        "Maximum Price (DKK)", 
        min_value=0, 
        max_value=20000000, 
        value=10000000, 
        step=500000,
        format="%d"
    )
    
    min_rooms = st.sidebar.slider(
        "Minimum Rooms", 
        min_value=1, 
        max_value=15, 
        value=3
    )
    
    selected_cities = st.sidebar.multiselect(
        "Cities", 
        options=all_cities,
        default=[]
    )
    
    selected_energy_ratings = st.sidebar.multiselect(
        "Energy Ratings", 
        options=all_energy_ratings,
        default=[]
    )
    
    # Create filter dictionary
    filters = {
        'max_price': max_price,
        'min_rooms': min_rooms,
        'cities': selected_cities,
        'energy_ratings': selected_energy_ratings
    }
    
    # Apply filters
    filtered_properties = filter_properties(all_properties, filters)
    
    # Main content area
    col1, col2 = st.columns([2, 1])
    
    with col1:
        # Property listings
        st.subheader(f"📋 Found {len(filtered_properties)} Properties")
        
        if filtered_properties:
            # Show first 10 properties
            for i, prop in enumerate(filtered_properties[:10]):
                with st.expander(
                    f"{prop.get('address', 'N/A')} - {safe_int(prop.get('cash_price')):,} DKK"
                ):
                    # Property details in columns
                    detail_col1, detail_col2, detail_col3 = st.columns(3)
                    
                    with detail_col1:
                        st.write(f"**🏙️ City:** {prop.get('city', 'N/A')}")
                        st.write(f"**🚪 Rooms:** {prop.get('rooms', 'N/A')}")
                        st.write(f"**📐 Living Area:** {prop.get('living_area', 'N/A')} m²")
                    
                    with detail_col2:
                        st.write(f"**⚡ Energy Rating:** {prop.get('energy_rating', 'N/A')}")
                        st.write(f"**📅 Year Built:** {prop.get('year_built', 'N/A')}")
                        st.write(f"**🏛️ Municipality:** {prop.get('municipal', 'N/A')}")
                    
                    with detail_col3:
                        price_per_sqm = safe_int(prop.get('square_meter_price'))
                        st.write(f"**💰 Price/m²:** {price_per_sqm:,} DKK" if price_per_sqm else "**💰 Price/m²:** N/A")
                        
                        plot_area = safe_int(prop.get('area'))
                        st.write(f"**🌿 Plot Area:** {plot_area:,} m²" if plot_area else "**🌿 Plot Area:** N/A")
                        
                        st.write(f"**🏠 Type:** {prop.get('legal_type', 'N/A')}")
            
            if len(filtered_properties) > 10:
                st.info(f"Showing first 10 of {len(filtered_properties)} properties. Adjust filters to narrow results.")
        else:
            st.info("No properties match your current filters. Try adjusting the criteria.")
    
    with col2:
        # AI Chat Section
        st.subheader("🤖 Ask AI Assistant")
        st.write("Ask questions about the Danish villa market!")
        
        # Model selection for Together AI
        model_choice = st.selectbox(
            "Select AI Model:",
            [
                # Gemma models (Google's efficient models)
                "google/gemma-2b-it",
                # Other reliable models
                "mistralai/Mistral-7B-Instruct-v0.1",
                "NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO",
                "mistralai/Mixtral-8x7B-Instruct-v0.1"
            ],
            help="Gemma models are Google's efficient, lightweight models."
        )
        
        # Example questions
        with st.expander("💡 Example Questions"):
            st.write("• What's the average price range?")
            st.write("• Tell me about energy ratings in the data")
            st.write("• Which areas have the most expensive properties?")
            st.write("• How many properties are available in each city?")
            st.write("• What's the price per square meter trend?")
        
        user_question = st.text_area(
            "Your Question:",
            placeholder="Ask about prices, locations, energy ratings, market trends...",
            height=100
        )
        
        if st.button("🔍 Ask AI", type="primary"):
            if user_question:
                with st.spinner("AI is analyzing the data..."):
                    # Create context from current filtered data
                    context = create_property_context(filtered_properties)
                    
                    try:
                        # Get AI response
                        ai_response = get_ai_response(client, user_question, context, model_choice)
                        
                        st.success("**AI Assistant Response:**")
                        st.write(ai_response)
                        
                        # Show debug info
                        with st.expander("Debug Info"):
                            st.write(f"Model used: {model_choice}")
                            st.write(f"Properties analyzed: {len(filtered_properties)}")
                            st.write(f"Context: {context[:150]}...")
                        
                    except Exception as e:
                        st.error(f"AI Error: {str(e)}")
                        
                        # Fallback response with data analysis
                        st.info("**Fallback Analysis:**")
                        if filtered_properties:
                            avg_price = sum(safe_int(p.get('cash_price')) for p in filtered_properties) / len(filtered_properties)
                            st.write(f"• Found {len(filtered_properties)} properties")
                            st.write(f"• Average price: {avg_price:,.0f} DKK")
                            
                            cities = list(set(p.get('city') for p in filtered_properties if p.get('city')))
                            if cities:
                                st.write(f"• Cities: {', '.join(cities[:3])}")
                                
                            energy_ratings = list(set(p.get('energy_rating') for p in filtered_properties if p.get('energy_rating')))
                            if energy_ratings:
                                st.write(f"• Energy ratings: {', '.join(energy_ratings[:3])}")
            else:
                st.warning("Please enter a question first!")
    
    # Footer stats
    st.markdown("---")
    if all_properties:
        total_props = len(all_properties)
        filtered_props = len(filtered_properties)
        
        stat_col1, stat_col2, stat_col3, stat_col4 = st.columns(4)
        
        with stat_col1:
            st.metric("Total Properties", total_props)
        
        with stat_col2:
            st.metric("Filtered Results", filtered_props)
        
        with stat_col3:
            if filtered_properties:
                avg_price = sum(safe_int(p.get('cash_price')) for p in filtered_properties) / len(filtered_properties)
                st.metric("Avg Price", f"{avg_price:,.0f} DKK")
        
        with stat_col4:
            unique_cities = len(set(p.get('city') for p in filtered_properties if p.get('city')))
            st.metric("Cities", unique_cities)

if __name__ == "__main__":
    main()