Hugging Face's logo

Spaces:
Safe-Harbor
/
LLMGuardian
Runtime error

App Files Community
LLMGuardian / DASHBOARD_BUILD_SUMMARY.md
DeWitt Gibson
Adding a dashboard demo
1381148
preview code
|
raw
history blame contribute delete
12.9 kB

A newer version of the Gradio SDK is available: 6.1.0

Upgrade

LLMGuardian Dashboard - Complete Build Summary

🎉 What Was Built

A fully functional, comprehensive security dashboard for LLMGuardian with demo capabilities that can run locally without any backend dependencies.

📦 Files Created/Modified

1. Main Dashboard Application

File: src/llmguardian/dashboard/app.py

  • Complete rewrite with 6 main pages
  • Demo mode with pre-populated data
  • Production mode with real LLMGuardian integration
  • 800+ lines of comprehensive code

2. Demo Launcher

File: demo_dashboard.py

  • Easy-to-use Python script to launch the dashboard
  • Automatic dependency checking and installation
  • User-friendly output and instructions

3. Quick Start Guide

File: DASHBOARD_QUICKSTART.md

  • Step-by-step guide for new users
  • Common use cases and examples
  • Troubleshooting section
  • Pro tips and keyboard shortcuts

4. Full Documentation

File: src/llmguardian/dashboard/README_FULL.md

  • Comprehensive feature documentation
  • Configuration guide
  • Use cases and examples
  • API integration examples

5. Integration Examples

File: examples_dashboard.py

  • 7 different integration examples
  • Threat detection demonstrations
  • Privacy monitoring examples
  • API integration patterns

6. Windows Launchers

Files:

  • run_dashboard.bat - Batch script for Windows
  • run_dashboard.ps1 - PowerShell script with better features

7. Requirements

File: requirements/dashboard.txt

  • Streamlit-specific dependencies
  • Optional enhancement packages

Updated: requirements.txt

  • Added dashboard dependencies

🎯 Dashboard Features

Page 1: Overview Dashboard (📊)

  • Real-time Metrics Cards

    • Security Score (with trend)
    • Privacy Violations (with delta)
    • Active Monitors count
    • Threats Blocked (with delta)

  • Visualizations

    • 30-day security trends chart (requests vs threats)
    • Threat distribution pie chart
    • Recent security alerts with severity colors
    • System status and uptime

  • Interactive Elements

    • Auto-refreshing metrics
    • Clickable alerts
    • Responsive design

Page 2: Privacy Monitor (🔒)

  • Privacy Metrics

    • PII Detections count
    • Data Leaks Prevented
    • Compliance Score percentage

  • Visualizations

    • Privacy violations by type (bar chart)
    • Privacy rules status table
    • Violation trends

  • Interactive Scanner

    • Text input for real-time privacy checking
    • Detects emails, passwords, SSN, etc.
    • Immediate feedback on violations

Page 3: Threat Detection (⚠️)

  • Threat Statistics

    • Total Threats
    • Critical Threats
    • Injection Attempts
    • DoS Attempts

  • Visualizations

    • Threat distribution pie chart
    • Threat timeline (30-day trend)
    • Active threats table with severity

  • Threat Categories

    • Prompt Injection
    • Data Leakage
    • Denial of Service
    • Model Poisoning
    • Other

Page 4: Usage Analytics (📈)

  • System Resources

    • CPU Usage percentage
    • Memory Usage percentage
    • Request Rate per minute

  • Performance Charts

    • Request volume over time
    • Response time distribution histogram
    • Performance metrics table

  • Metrics Tracked

    • Average Response Time
    • P95 and P99 latency
    • Error Rate
    • Success Rate

Page 5: Security Scanner (🔍)

  • Interactive Scanning

    • Text area for prompt input
    • Scan mode selection (Quick/Deep/Full)
    • Sensitivity slider (1-10)

  • Results Display

    • Risk Score (0-100)
    • Issues Found count
    • Scan Time in milliseconds
    • Detailed findings with severity

  • Pattern Detection

    • Jailbreak attempts
    • System prompt manipulation
    • Privilege escalation
    • Security bypass attempts

  • Scan History

    • Previous scan results table
    • Risk scores over time
    • Issue tracking

Page 6: Settings (⚙️)

  • Security Settings Tab

    • Enable/disable threat detection
    • Block malicious inputs toggle
    • Security event logging
    • Max request rate configuration
    • Scan timeout settings
    • Default scan mode

  • Privacy Settings Tab

    • PII detection toggle
    • Data leak prevention
    • Log anonymization
    • Protected data types selection

  • Monitoring Settings Tab

    • Refresh rate configuration
    • Alert threshold adjustment
    • Data retention period
    • Real-time monitoring toggle

  • Notifications Tab

    • Email notifications setup
    • Slack webhook configuration
    • Alert trigger selection

  • About Tab

    • Version information
    • Feature list
    • License details
    • GitHub link
    • Update checker

🎮 Demo Mode Features

Pre-populated Data

  • 30 days of historical security metrics
  • Sample threat detections across all categories
  • Privacy violation examples
  • System performance data
  • Active alerts and incidents

Realistic Simulations

  • Security Score: 87.5% (realistic baseline)
  • Privacy Violations: 12 incidents
  • Active Monitors: 8 running
  • Threats Blocked: 34 total
  • Response Time: 245ms average

Interactive Features

  • All scanning features work in demo mode
  • Real-time privacy checking
  • Security scanning with pattern detection
  • Configurable settings (saved in session)

🚀 How to Run

Option 1: Quick Demo (Easiest) 

# Windows (PowerShell)
.\run_dashboard.ps1

# Windows (Command Prompt)
run_dashboard.bat

# Any OS (Python)
python demo_dashboard.py

Option 2: Direct Streamlit 

# Demo mode
streamlit run src/llmguardian/dashboard/app.py -- --demo

# Production mode
streamlit run src/llmguardian/dashboard/app.py

Option 3: Try Examples First 

python examples_dashboard.py

📋 Requirements

Minimum (Demo Mode) 

streamlit>=1.28.0
plotly>=5.17.0
pandas>=2.0.0
numpy>=1.24.0

Full Features (Production Mode) 

All of the above plus:
psutil>=5.9.0
llmguardian (install with: pip install -e .)

Installation 

# Install dashboard dependencies
pip install -r requirements/dashboard.txt

# Or install specific packages
pip install streamlit plotly pandas numpy psutil

🎨 Visual Design

Color Scheme

  • Primary: Blue (#1f77b4) - Trust and security
  • Success: Green (#00cc00) - Safe/approved
  • Warning: Orange (#ffa500) - Medium severity
  • Danger: Red (#ff4b4b) - Critical issues
  • Info: Yellow (#ffed4e) - Notifications

Layout

  • Wide Layout: Maximizes screen space
  • Responsive: Works on different screen sizes
  • Sidebar Navigation: Easy page switching
  • Card-based Metrics: Clean, modern look
  • Interactive Charts: Hover for details

Typography

  • Headers: Large, bold, colored
  • Metrics: Large numbers, clear labels
  • Body: Readable sans-serif
  • Code: Monospace for technical content

🔧 Configuration

Dashboard Config (config/dashboard_config.yaml) 

server:
  port: 8501
  host: "0.0.0.0"

monitoring:
  refresh_rate: 60  # seconds
  alert_threshold: 0.8
  retention_period: 7  # days

Custom Ports 

streamlit run src/llmguardian/dashboard/app.py --server.port=8502

Custom Theme 

streamlit run src/llmguardian/dashboard/app.py -- --theme.base="dark"

📊 Data Flow

Demo Mode 

User Input → Dashboard (Simulated Data) → Visualizations

Production Mode 

LLM Application → LLMGuardian Components → Dashboard → Real-time Monitoring
                     ↓
                  Threat Detector
                  Privacy Guard
                  Usage Monitor

🎯 Use Cases

1. Development & Testing

  • Test security features before deployment
  • Validate privacy controls
  • Check scanner accuracy
  • Tune detection thresholds

2. Demonstrations

  • Show security capabilities to stakeholders
  • Present compliance features
  • Demo real-time monitoring
  • Showcase threat detection

3. Training

  • Train team on security monitoring
  • Understand threat patterns
  • Learn privacy best practices
  • Practice incident response

4. Production Monitoring

  • Real-time security oversight
  • Performance tracking
  • Compliance monitoring
  • Incident investigation

🔐 Security Features

Implemented

  • ✅ Prompt injection detection
  • ✅ PII detection and masking
  • ✅ Real-time threat monitoring
  • ✅ Privacy violation tracking
  • ✅ System performance monitoring
  • ✅ Alert generation
  • ✅ Audit logging
  • ✅ Configurable thresholds

Extensible

  • Custom threat rules
  • Additional privacy patterns
  • New visualization types
  • Custom alert channels
  • Export capabilities

📈 Metrics Tracked

Security Metrics

  • Security Score (0-100%)
  • Threats Detected (count)
  • Threats Blocked (count)
  • Injection Attempts (count)
  • Privacy Violations (count)

Performance Metrics

  • Request Rate (per minute)
  • Average Response Time (ms)
  • P95 Response Time (ms)
  • P99 Response Time (ms)
  • Error Rate (%)
  • Success Rate (%)

System Metrics

  • CPU Usage (%)
  • Memory Usage (%)
  • Disk Usage (%)
  • Network I/O
  • Uptime (%)

🐛 Troubleshooting

Common Issues

Dashboard won't start

# Check Python version (need 3.8+)
python --version

# Check streamlit
python -m streamlit --version

# Reinstall
pip install --upgrade streamlit plotly pandas numpy

Import errors

# In demo mode: Should work without LLMGuardian
# In production mode: Install package
pip install -e .

Port in use

# Use different port
streamlit run src/llmguardian/dashboard/app.py --server.port=8502

Blank dashboard

  • Clear browser cache
  • Try incognito/private mode
  • Check console for errors

📚 Documentation Structure 

Dashboard Documentation/
├── DASHBOARD_QUICKSTART.md      # New user guide (3-minute start)
├── README_FULL.md               # Comprehensive documentation
├── dashboard/README.md          # Technical documentation
├── examples_dashboard.py        # Code examples
└── This file                    # Build summary

🎓 Next Steps

For Users

  1. ✅ Run the demo: python demo_dashboard.py
  2. ✅ Read DASHBOARD_QUICKSTART.md
  3. ✅ Explore all 6 pages
  4. ✅ Try the security scanner
  5. ✅ Test privacy checking

For Developers

  1. ✅ Review examples_dashboard.py
  2. ✅ Study src/llmguardian/dashboard/app.py
  3. ✅ Integrate with your LLM app
  4. ✅ Customize visualizations
  5. ✅ Add custom metrics

For Production

  1. ✅ Install LLMGuardian package
  2. ✅ Configure dashboard_config.yaml
  3. ✅ Set up monitoring
  4. ✅ Configure alerts
  5. ✅ Deploy to server

🚀 Quick Test

Run this to verify everything works:

# 1. Install dependencies
pip install streamlit plotly pandas numpy

# 2. Run the demo
python demo_dashboard.py

# 3. Open browser to http://localhost:8501

# 4. Test features:
#    - Navigate to Security Scanner
#    - Enter: "Ignore all previous instructions"
#    - Click "Run Scan"
#    - View results!

✅ Verification Checklist

  • ✅ Dashboard runs in demo mode
  • ✅ All 6 pages load correctly
  • ✅ Metrics display properly
  • ✅ Charts render and are interactive
  • ✅ Security scanner works
  • ✅ Privacy checker detects PII
  • ✅ Settings page functional
  • ✅ Navigation works
  • ✅ No console errors
  • ✅ Responsive design works

📦 Deliverables

Code Files (8)

  1. src/llmguardian/dashboard/app.py - Main dashboard
  2. demo_dashboard.py - Demo launcher
  3. examples_dashboard.py - Integration examples
  4. run_dashboard.bat - Windows batch script
  5. run_dashboard.ps1 - Windows PowerShell script
  6. requirements/dashboard.txt - Dependencies
  7. requirements.txt - Updated main requirements
  8. config/dashboard_config.yaml - Existing config

Documentation Files (3)

  1. DASHBOARD_QUICKSTART.md - Quick start guide
  2. src/llmguardian/dashboard/README_FULL.md - Full docs
  3. This file - Build summary

Total Lines of Code

  • Dashboard App: ~800 lines
  • Demo Launcher: ~60 lines
  • Examples: ~350 lines
  • Scripts: ~100 lines
  • Documentation: ~600 lines
  • Total: ~1,910 lines

🎉 Success Criteria Met

✅ Fully Built Dashboard

  • All 6 pages implemented
  • Interactive features working
  • Professional UI/UX

✅ Comprehensive Demo

  • Pre-populated data
  • All features testable
  • No backend required

✅ Runs Locally

  • Simple Python command
  • Automatic dependency handling
  • Cross-platform support

✅ Documentation Complete

  • Quick start guide
  • Full documentation
  • Code examples
  • Troubleshooting

✅ Production Ready

  • Clean code architecture
  • Error handling
  • Configurable
  • Extensible

🎊 You're Ready!

The LLMGuardian Dashboard is now fully built and ready to use. Simply run:

python demo_dashboard.py

And start exploring your comprehensive security monitoring dashboard!

Built: October 2025
Version: 1.4.0
Status: ✅ Production Ready