app.py

import streamlit as st
import pandas as pd
import numpy as np
import plotly.express as px
import plotly.graph_objects as go
import os
import json
import logging
import time
from datetime import datetime, timedelta
from typing import Dict, List, Optional, Union, Any
from dotenv import load_dotenv

# Configure logging - Reduce verbosity and improve performance
logging.basicConfig(
    level=logging.WARNING,  # Only show warnings and errors by default
    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
)

# Create a custom filter to suppress repetitive Gemini API errors
class SuppressRepetitiveErrors(logging.Filter):
    def __init__(self):
        super().__init__()
        self.error_counts = {}
        self.max_errors = 3  # Show at most 3 instances of each error
        
    def filter(self, record):
        if record.levelno < logging.WARNING:
            return True
            
        # If it's a Gemini API error for non-existent tokens, suppress it after a few occurrences
        if 'Error fetching historical prices from Gemini API' in record.getMessage():
            key = 'gemini_api_error'
            self.error_counts[key] = self.error_counts.get(key, 0) + 1
            
            # Only allow the first few errors through
            return self.error_counts[key] <= self.max_errors
            
        return True

# Apply the filter
logging.getLogger().addFilter(SuppressRepetitiveErrors())

from modules.api_client import ArbiscanClient, GeminiClient
from modules.data_processor import DataProcessor
from modules.visualizer import Visualizer
from modules.detection import ManipulationDetector

# Load environment variables
load_dotenv()

# Set page configuration
st.set_page_config(
    page_title="Whale Wallet AI - Market Manipulation Detection",
    page_icon="🐳",
    layout="wide",
    initial_sidebar_state="expanded"
)

# Add custom CSS
st.markdown("""
<style>
    .main-header {
        font-size: 2.5rem;
        color: #1E88E5;
        text-align: center;
        margin-bottom: 1rem;
    }
    .sub-header {
        font-size: 1.5rem;
        color: #424242;
        margin-bottom: 1rem;
    }
    .info-text {
        background-color: #E3F2FD;
        padding: 1rem;
        border-radius: 0.5rem;
        margin-bottom: 1rem;
    }
    .stButton>button {
        width: 100%;
    }
</style>
""", unsafe_allow_html=True)

# Initialize Streamlit session state for persisting data between tab navigation
if 'transactions_data' not in st.session_state:
    st.session_state.transactions_data = pd.DataFrame()
    
if 'patterns_data' not in st.session_state:
    st.session_state.patterns_data = None
    
if 'price_impact_data' not in st.session_state:
    st.session_state.price_impact_data = None
    
# Performance metrics tracking
if 'performance_metrics' not in st.session_state:
    st.session_state.performance_metrics = {
        'api_calls': 0,
        'data_processing_time': 0,
        'visualization_time': 0,
        'last_refresh': None
    }
    
# Function to track performance
def track_timing(category: str):
    def timing_decorator(func):
        def wrapper(*args, **kwargs):
            start_time = time.time()
            result = func(*args, **kwargs)
            elapsed = time.time() - start_time
            
            if category in st.session_state.performance_metrics:
                st.session_state.performance_metrics[category] += elapsed
            else:
                st.session_state.performance_metrics[category] = elapsed
                
            return result
        return wrapper
    return timing_decorator
    
if 'alerts_data' not in st.session_state:
    st.session_state.alerts_data = None

# Initialize API clients
arbiscan_client = ArbiscanClient(os.getenv("ARBISCAN_API_KEY"))
# Set debug mode to False to reduce log output
arbiscan_client.verbose_debug = False
gemini_client = GeminiClient(os.getenv("GEMINI_API_KEY"))

# Initialize data processor and visualizer
data_processor = DataProcessor()
visualizer = Visualizer()

# Apply performance tracking to key instance methods after initialization
original_fetch_whale = arbiscan_client.fetch_whale_transactions
arbiscan_client.fetch_whale_transactions = track_timing('api_calls')(original_fetch_whale)

original_identify_patterns = data_processor.identify_patterns
data_processor.identify_patterns = track_timing('data_processing_time')(original_identify_patterns)

original_analyze_price_impact = data_processor.analyze_price_impact
data_processor.analyze_price_impact = track_timing('data_processing_time')(original_analyze_price_impact)
detection = ManipulationDetector()

# Initialize crew system (for AI-assisted analysis)
try:
    from modules.crew_system import WhaleAnalysisCrewSystem
    crew_system = WhaleAnalysisCrewSystem(arbiscan_client, gemini_client, data_processor)
    CREW_ENABLED = True
    logging.info("CrewAI system loaded successfully")
except Exception as e:
    CREW_ENABLED = False
    logging.error(f"Failed to load CrewAI system: {str(e)}")
    st.sidebar.error("CrewAI features are disabled due to an error.")

# Sidebar for inputs
st.sidebar.header("Configuration")

# Wallet tracking section
st.sidebar.subheader("Track Wallets")
wallet_addresses = st.sidebar.text_area(
    "Enter wallet addresses (one per line)",
    placeholder="0x1234abcd...\n0xabcd1234..."
)

threshold_type = st.sidebar.radio(
    "Threshold Type",
    ["Token Amount", "USD Value"]
)

if threshold_type == "Token Amount":
    threshold_value = st.sidebar.number_input("Minimum Token Amount", min_value=0.0, value=1000.0)
    token_symbol = st.sidebar.text_input("Token Symbol", placeholder="ETH")
else:
    threshold_value = st.sidebar.number_input("Minimum USD Value", min_value=0.0, value=100000.0)

# Time period selection
st.sidebar.subheader("Time Period")
time_period = st.sidebar.selectbox(
    "Select Time Period",
    ["Last 24 hours", "Last 7 days", "Last 30 days", "Custom"]
)

if time_period == "Custom":
    start_date = st.sidebar.date_input("Start Date", datetime.now() - timedelta(days=7))
    end_date = st.sidebar.date_input("End Date", datetime.now())
else:
    # Calculate dates based on selection
    end_date = datetime.now()
    if time_period == "Last 24 hours":
        start_date = end_date - timedelta(days=1)
    elif time_period == "Last 7 days":
        start_date = end_date - timedelta(days=7)
    else:  # Last 30 days
        start_date = end_date - timedelta(days=30)

# Manipulation detection settings
st.sidebar.subheader("Manipulation Detection")
enable_manipulation_detection = st.sidebar.toggle("Enable Manipulation Detection", value=True)
if enable_manipulation_detection:
    sensitivity = st.sidebar.select_slider(
        "Detection Sensitivity",
        options=["Low", "Medium", "High"],
        value="Medium"
    )

# Price impact analysis settings
st.sidebar.subheader("Price Impact Analysis")
enable_price_impact = st.sidebar.toggle("Enable Price Impact Analysis", value=True)
if enable_price_impact:
    lookback_minutes = st.sidebar.slider("Lookback (minutes)", 1, 60, 5)
    lookahead_minutes = st.sidebar.slider("Lookahead (minutes)", 1, 60, 5)

# Action buttons
track_button = st.sidebar.button("Track Transactions", type="primary")
pattern_button = st.sidebar.button("Analyze Patterns")
if enable_manipulation_detection:
    detect_button = st.sidebar.button("Detect Manipulation")

# Main content area
tab1, tab2, tab3, tab4, tab5 = st.tabs([
    "Transactions", "Patterns", "Price Impact", "Alerts", "Reports"
])

with tab1:
    st.header("Whale Transactions")
    if track_button and wallet_addresses:
        with st.spinner("Fetching whale transactions..."):
            # Function to track whale transactions
            def track_whale_transactions(wallets, start_date, end_date, threshold_value, threshold_type, token_symbol=None):
                # Direct API call since CrewAI is temporarily disabled
                try:
                    min_token_amount = None
                    min_usd_value = None
                    if threshold_type == "Token Amount":
                        min_token_amount = threshold_value
                    else:
                        min_usd_value = threshold_value
                    
                    # Add pagination control to prevent infinite API requests
                    max_pages = 5  # Limit the number of pages to prevent excessive API calls
                    transactions = arbiscan_client.fetch_whale_transactions(
                        addresses=wallets,
                        min_token_amount=min_token_amount,
                        max_pages=5,
                        min_usd_value=min_usd_value
                    )
                    
                    if transactions.empty:
                        st.warning("No transactions found for the specified addresses")
                    
                    return transactions
                except Exception as e:
                    st.error(f"Error fetching transactions: {str(e)}")
                    return pd.DataFrame()

            wallet_list = [addr.strip() for addr in wallet_addresses.split("\n") if addr.strip()]
            
            # Use cached data or fetch new if not available
            if st.session_state.transactions_data is None or track_button:
                with st.spinner("Fetching transactions..."):
                    transactions = track_whale_transactions(
                        wallets=wallet_list,
                        start_date=start_date,
                        end_date=end_date,
                        threshold_value=threshold_value,
                        threshold_type=threshold_type,
                        token_symbol=token_symbol
                    )
                    # Store in session state
                    st.session_state.transactions_data = transactions
            else:
                transactions = st.session_state.transactions_data
            
            if not transactions.empty:
                st.success(f"Found {len(transactions)} transactions matching your criteria")
                
                # Display transactions
                if len(transactions) > 0:
                    st.dataframe(transactions, use_container_width=True)
                    
                    # Add download button
                    csv = transactions.to_csv(index=False).encode('utf-8')
                    st.download_button(
                        "Download Transactions CSV",
                        csv,
                        "whale_transactions.csv",
                        "text/csv",
                        key='download-csv'
                    )
                    
                    # Volume by day chart
                    st.subheader("Transaction Volume by Day")
                    try:
                        st.plotly_chart(visualizer.plot_volume_by_day(transactions), use_container_width=True)
                    except Exception as e:
                        st.error(f"Error generating volume chart: {str(e)}")
                        
                    # Transaction flow visualization
                    st.subheader("Transaction Flow")
                    try:
                        flow_chart = visualizer.plot_transaction_flow(transactions)
                        st.plotly_chart(flow_chart, use_container_width=True)
                    except Exception as e:
                        st.error(f"Error generating flow chart: {str(e)}")
            else:
                st.warning("No transactions found matching your criteria. Try adjusting the parameters.")
    else:
        st.info("Enter wallet addresses and click 'Track Transactions' to view whale activity")

with tab2:
    st.header("Trading Patterns")
    if track_button and wallet_addresses:
        with st.spinner("Analyzing trading patterns..."):
            # Function to analyze trading patterns
            def analyze_trading_patterns(wallets, start_date, end_date):
                # Direct analysis
                try:
                    transactions_df = arbiscan_client.fetch_whale_transactions(addresses=wallets, max_pages=5)
                    if transactions_df.empty:
                        st.warning("No transactions found for the specified addresses")
                        return []
                        
                    return data_processor.identify_patterns(transactions_df)
                except Exception as e:
                    st.error(f"Error analyzing trading patterns: {str(e)}")
                    return []

            wallet_list = [addr.strip() for addr in wallet_addresses.split("\n") if addr.strip()]
            
            # Use cached data or fetch new if not available
            if st.session_state.patterns_data is None or track_button:
                with st.spinner("Analyzing trading patterns..."):
                    patterns = analyze_trading_patterns(
                        wallets=wallet_list,
                        start_date=start_date,
                        end_date=end_date
                    )
                    # Store in session state
                    st.session_state.patterns_data = patterns
            else:
                patterns = st.session_state.patterns_data
            
            if patterns:
                for i, pattern in enumerate(patterns):
                    pattern_card = st.container()
                    with pattern_card:
                        # Pattern header with name and risk profile
                        header_cols = st.columns([3, 1])
                        with header_cols[0]:
                            st.subheader(f"Pattern {i+1}: {pattern['name']}")
                        with header_cols[1]:
                            risk_color = "green"
                            if pattern.get('risk_profile') == "Medium":
                                risk_color = "orange"
                            elif pattern.get('risk_profile') in ["High", "Very High"]:
                                risk_color = "red"
                            st.markdown(f"<h5 style='color:{risk_color};'>Risk: {pattern.get('risk_profile', 'Unknown')}</h5>", unsafe_allow_html=True)
                        
                        # Pattern description and details
                        st.markdown(f"**Description:** {pattern['description']}")
                        
                        # Additional strategy information
                        if 'strategy' in pattern:
                            st.markdown(f"**Strategy:** {pattern['strategy']}")
                            
                        # Time insight
                        if 'time_insight' in pattern:
                            st.info(pattern['time_insight'])
                        
                        # Metrics
                        metric_cols = st.columns(3)
                        with metric_cols[0]:
                            st.markdown(f"**Occurrences:** {pattern['occurrence_count']} instances")
                        with metric_cols[1]:
                            st.markdown(f"**Confidence:** {pattern.get('confidence', 0):.2f}")
                        with metric_cols[2]:
                            st.markdown(f"**Volume:** {pattern.get('volume_metric', 'N/A')}")
                        
                        # Display main chart first
                        if 'charts' in pattern and 'main' in pattern['charts']:
                            st.plotly_chart(pattern['charts']['main'], use_container_width=True)
                        elif 'chart_data' in pattern and pattern['chart_data'] is not None:  # Fallback for old format
                            st.plotly_chart(pattern['chart_data'], use_container_width=True)
                        
                        # Create two columns for additional charts
                        if 'charts' in pattern and len(pattern['charts']) > 1:
                            charts_col1, charts_col2 = st.columns(2)
                            
                            # Hourly distribution chart
                            if 'hourly_distribution' in pattern['charts']:
                                with charts_col1:
                                    st.plotly_chart(pattern['charts']['hourly_distribution'], use_container_width=True)
                            
                            # Value distribution chart
                            if 'value_distribution' in pattern['charts']:
                                with charts_col2:
                                    st.plotly_chart(pattern['charts']['value_distribution'], use_container_width=True)
                        
                        # Advanced metrics in expander
                        if 'metrics' in pattern and pattern['metrics']:
                            with st.expander("Detailed Metrics"):
                                metrics_table = []
                                for k, v in pattern['metrics'].items():
                                    if v is not None:
                                        if isinstance(v, float):
                                            metrics_table.append([k.replace('_', ' ').title(), f"{v:.4f}"])
                                        else:
                                            metrics_table.append([k.replace('_', ' ').title(), v])
                                
                                if metrics_table:
                                    st.table(pd.DataFrame(metrics_table, columns=["Metric", "Value"]))
                        
                        # Display example transactions
                        if 'examples' in pattern and not pattern['examples'].empty:
                            with st.expander("Example Transactions"):
                                # Format the dataframe for better display
                                display_df = pattern['examples'].copy()
                                # Convert timestamp to readable format if needed
                                if 'timeStamp' in display_df.columns and not pd.api.types.is_datetime64_any_dtype(display_df['timeStamp']):
                                    display_df['timeStamp'] = pd.to_datetime(display_df['timeStamp'], unit='s')
                                
                                st.dataframe(display_df, use_container_width=True)
                        
                        st.markdown("---")
            else:
                st.info("No significant trading patterns detected. Try expanding the date range or adding more addresses.")
    else:
        st.info("Track transactions to analyze trading patterns")

with tab3:
    st.header("Price Impact Analysis")
    if enable_price_impact and track_button and wallet_addresses:
        with st.spinner("Analyzing price impact..."):
            # Function to analyze price impact
            def analyze_price_impact(wallets, start_date, end_date, lookback_minutes, lookahead_minutes):
                # Direct analysis
                transactions_df = arbiscan_client.fetch_whale_transactions(addresses=wallets, max_pages=5)
                # Get token from first transaction
                if not transactions_df.empty:
                    token_symbol = transactions_df.iloc[0].get('tokenSymbol', 'ETH')
                    # For each transaction, get price impact
                    price_impacts = {}
                    progress_bar = st.progress(0)
                    for idx, row in transactions_df.iterrows():
                        progress = int((idx + 1) / len(transactions_df) * 100)
                        progress_bar.progress(progress, text=f"Analyzing transaction {idx+1} of {len(transactions_df)}")
                        if 'timeStamp' in row:
                            try:
                                tx_time = datetime.fromtimestamp(int(row['timeStamp']))
                                impact_data = gemini_client.get_price_impact(
                                    symbol=f"{token_symbol}USD",
                                    transaction_time=tx_time,
                                    lookback_minutes=lookback_minutes,
                                    lookahead_minutes=lookahead_minutes
                                )
                                price_impacts[row['hash']] = impact_data
                            except Exception as e:
                                st.warning(f"Could not get price data for transaction: {str(e)}")
                    
                    progress_bar.empty()
                    if price_impacts:
                        return data_processor.analyze_price_impact(transactions_df, price_impacts)
                
                # Create an empty chart for the default case
                empty_fig = go.Figure()
                empty_fig.update_layout(
                    title="No Price Impact Data Available",
                    xaxis_title="Time",
                    yaxis_title="Price Impact (%)",
                    height=400,
                    template="plotly_white"
                )
                empty_fig.add_annotation(
                    text="No transactions found with price impact data",
                    showarrow=False,
                    font=dict(size=14)
                )
                
                return {
                    "avg_impact_pct": 0, 
                    "max_impact_pct": 0, 
                    "min_impact_pct": 0,
                    "significant_moves_count": 0,
                    "total_transactions": 0,
                    "transactions_with_impact": pd.DataFrame(),
                    "charts": {
                        "main_chart": empty_fig,
                        "impact_distribution": empty_fig,
                        "cumulative_impact": empty_fig,
                        "hourly_impact": empty_fig
                    },
                    "insights": [],
                    "impact_summary": "No price impact data available"
                }

            wallet_list = [addr.strip() for addr in wallet_addresses.split("\n") if addr.strip()]
            
            # Use cached data or fetch new if not available
            if st.session_state.price_impact_data is None or track_button:
                with st.spinner("Analyzing price impact..."):
                    impact_analysis = analyze_price_impact(
                        wallets=wallet_list,
                        start_date=start_date,
                        end_date=end_date,
                        lookback_minutes=lookback_minutes,
                        lookahead_minutes=lookahead_minutes
                    )
                    # Store in session state
                    st.session_state.price_impact_data = impact_analysis
            else:
                impact_analysis = st.session_state.price_impact_data
            
            if impact_analysis:
                # Display impact summary
                if 'impact_summary' in impact_analysis:
                    st.info(impact_analysis['impact_summary'])
                
                # Summary metrics in two rows
                metrics_row1 = st.columns(4)
                with metrics_row1[0]:
                    st.metric("Avg. Price Impact (%)", f"{impact_analysis.get('avg_impact_pct', 0):.2f}%")
                with metrics_row1[1]:
                    st.metric("Max Impact (%)", f"{impact_analysis.get('max_impact_pct', 0):.2f}%")
                with metrics_row1[2]:
                    st.metric("Min Impact (%)", f"{impact_analysis.get('min_impact_pct', 0):.2f}%")
                with metrics_row1[3]:
                    st.metric("Std Dev (%)", f"{impact_analysis.get('std_impact_pct', 0):.2f}%")
                
                metrics_row2 = st.columns(4)
                with metrics_row2[0]:
                    st.metric("Significant Moves", impact_analysis.get('significant_moves_count', 0))
                with metrics_row2[1]:
                    st.metric("High Impact Moves", impact_analysis.get('high_impact_moves_count', 0))
                with metrics_row2[2]:
                    st.metric("Positive/Negative", f"{impact_analysis.get('positive_impacts_count', 0)}/{impact_analysis.get('negative_impacts_count', 0)}")
                with metrics_row2[3]:
                    st.metric("Total Transactions", impact_analysis.get('total_transactions', 0))
                
                # Display insights if available
                if 'insights' in impact_analysis and impact_analysis['insights']:
                    st.subheader("Key Insights")
                    for insight in impact_analysis['insights']:
                        st.markdown(f"**{insight['title']}**: {insight['description']}")
                
                # Display the main chart
                if 'charts' in impact_analysis and 'main_chart' in impact_analysis['charts']:
                    st.subheader("Price Impact Over Time")
                    st.plotly_chart(impact_analysis['charts']['main_chart'], use_container_width=True)
                
                # Create two columns for secondary charts
                col1, col2 = st.columns(2)
                
                # Distribution chart
                if 'charts' in impact_analysis and 'impact_distribution' in impact_analysis['charts']:
                    with col1:
                        st.plotly_chart(impact_analysis['charts']['impact_distribution'], use_container_width=True)
                
                # Cumulative impact chart
                if 'charts' in impact_analysis and 'cumulative_impact' in impact_analysis['charts']:
                    with col2:
                        st.plotly_chart(impact_analysis['charts']['cumulative_impact'], use_container_width=True)
                
                # Hourly impact chart
                if 'charts' in impact_analysis and 'hourly_impact' in impact_analysis['charts']:
                    st.plotly_chart(impact_analysis['charts']['hourly_impact'], use_container_width=True)
                
                # Detailed transactions with impact
                if not impact_analysis['transactions_with_impact'].empty:
                    st.subheader("Transactions with Price Impact")
                    # Convert numeric columns to have 2 decimal places for better display
                    display_df = impact_analysis['transactions_with_impact'].copy()
                    for col in ['impact_pct', 'pre_price', 'post_price', 'cumulative_impact']:
                        if col in display_df.columns:
                            display_df[col] = display_df[col].apply(lambda x: f"{float(x):.2f}%" if pd.notnull(x) else "N/A")
                    
                    st.dataframe(display_df, use_container_width=True)
                else:
                    st.info("No transaction-specific price impact data available")
            else:
                st.info("No price impact data available for the given parameters")
    else:
        st.info("Enable Price Impact Analysis and track transactions to see price effects")

with tab4:
    st.header("Manipulation Alerts")
    if enable_manipulation_detection and detect_button and wallet_addresses:
        with st.spinner("Detecting potential manipulation..."):
            wallet_list = [addr.strip() for addr in wallet_addresses.split("\n") if addr.strip()]
            
            # Function to detect manipulation
            def detect_manipulation(wallets, start_date, end_date, sensitivity):
                try:
                    transactions_df = arbiscan_client.fetch_whale_transactions(addresses=wallets, max_pages=5)
                    if transactions_df.empty:
                        st.warning("No transactions found for the specified addresses")
                        return []
                        
                    pump_dump = detection.detect_pump_and_dump(transactions_df, sensitivity)
                    wash_trades = detection.detect_wash_trading(transactions_df, wallets, sensitivity)
                    return pump_dump + wash_trades
                except Exception as e:
                    st.error(f"Error detecting manipulation: {str(e)}")
                    return []

            alerts = detect_manipulation(
                wallets=wallet_list,
                start_date=start_date,
                end_date=end_date,
                sensitivity=sensitivity
            )
            
            if alerts:
                for i, alert in enumerate(alerts):
                    alert_color = "red" if alert['risk_level'] == "High" else "orange" if alert['risk_level'] == "Medium" else "blue"
                    
                    with st.expander(f" {alert['type']} - Risk: {alert['risk_level']}", expanded=i==0):
                        st.markdown(f"<h4 style='color:{alert_color}'>{alert['title']}</h4>", unsafe_allow_html=True)
                        st.write(f"**Description:** {alert['description']}")
                        st.write(f"**Detection Time:** {alert['detection_time']}")
                        st.write(f"**Involved Addresses:** {', '.join(alert['addresses'])}")
                        
                        # Display evidence
                        if 'evidence' in alert and alert['evidence'] is not None and not (isinstance(alert['evidence'], pd.DataFrame) and alert['evidence'].empty):
                            st.subheader("Evidence")
                            try:
                                evidence_df = alert['evidence']
                                if isinstance(evidence_df, str):
                                    # Try to convert from JSON string if needed
                                    evidence_df = pd.read_json(evidence_df)
                                st.dataframe(evidence_df, use_container_width=True)
                            except Exception as e:
                                st.error(f"Error displaying evidence: {str(e)}")
                        
                        # Display chart if available
                        if 'chart' in alert and alert['chart'] is not None:
                            try:
                                st.plotly_chart(alert['chart'], use_container_width=True)
                            except Exception as e:
                                st.error(f"Error displaying chart: {str(e)}")
            else:
                st.success("No manipulation tactics detected for the given parameters")
    else:
        st.info("Enable Manipulation Detection and click 'Detect Manipulation' to scan for suspicious activity")

with tab5:
    st.header("Reports & Visualizations")
    
    # Report type selection
    report_type = st.selectbox(
        "Select Report Type",
        ["Transaction Summary", "Pattern Analysis", "Price Impact", "Manipulation Detection", "Complete Analysis"]
    )
    
    # Export format
    export_format = st.radio(
        "Export Format",
        ["CSV", "PDF", "PNG"],
        horizontal=True
    )
    
    # Generate report button
    if st.button("Generate Report"):
        if wallet_addresses:
            with st.spinner("Generating report..."):
                wallet_list = [addr.strip() for addr in wallet_addresses.split("\n") if addr.strip()]
                
                if CREW_ENABLED and crew_system is not None:
                    try:
                        with st.spinner("Generating AI analysis report..."):
                            # Check if crew_system has llm attribute defined
                            if not hasattr(crew_system, 'llm') or crew_system.llm is None:
                                raise ValueError("LLM not initialized in crew system")
                                
                            report = crew_system.generate_market_manipulation_report(wallet_addresses=wallet_list)
                            st.markdown(f"## AI Analysis Report")
                            st.markdown(report['content'])
                            
                            if 'charts' in report and report['charts']:
                                for i, chart in enumerate(report['charts']):
                                    st.plotly_chart(chart, use_container_width=True)
                    except Exception as e:
                        st.error(f"CrewAI report generation failed: {str(e)}")
                        st.warning("Using direct analysis instead")
                        
                        # Fallback to direct analysis
                        with st.spinner("Generating basic analysis..."):
                            insights = detection.generate_manipulation_insights(transactions=st.session_state.transactions_data)
                            st.markdown(f"## Potential Manipulation Insights")
                            
                            for insight in insights:
                                st.markdown(f"**{insight['title']}**\n{insight['description']}")
                else:
                    st.error("Failed to generate report: CrewAI is not enabled")
        else:
            st.error("Please enter wallet addresses to generate a report")

# Footer with instructions
st.markdown("---")
with st.expander("How to Use"):
    st.markdown("""
    ### Typical Workflow

    1. **Input wallet addresses** in the sidebar - these are the whale wallets you want to track
    2. **Set the minimum threshold** for transaction size (token amount or USD value)
    3. **Select time period** for analysis
    4. **Click 'Track Transactions'** to see large transfers for these wallets
    5. **Enable additional analysis** like pattern recognition or manipulation detection
    6. **Export reports** for further analysis or record-keeping
    
    ### API Keys
    
    This app requires two API keys to function properly:
    - **ARBISCAN_API_KEY** - For accessing Arbitrum blockchain data
    - **GEMINI_API_KEY** - For real-time token price data
    
    These should be stored in a `.env` file in the project root.
    """)