Voice Conversation Statistics Dashboard

Voice Conversation Statistics Dashboard

The voice-mode system now includes a comprehensive live statistics dashboard that tracks and displays conversation performance metrics in real-time.

Features

Real-time Performance Tracking

• Time to First Audio (TTFA): How quickly voice synthesis begins
• TTS Generation Time: Time to generate speech audio
• TTS Playback Time: Time to play the generated audio
• STT Processing Time: Time to process speech-to-text
• Total Turnaround Time: Complete conversation round-trip time

Session Statistics

• Total conversation count
• Success/failure rate tracking
• Session duration monitoring
• Provider usage analytics
• Voice and model usage patterns

Live Dashboard

• Real-time updates as conversations occur
• Min/max/average performance metrics
• Recent interaction history
• Provider performance comparison

MCP Tools

voice_statistics

Display the complete live statistics dashboard with all performance metrics, provider usage, and recent interaction history.

# Example usage in Claude
> Use the voice_statistics tool to show current performance

voice_statistics_summary

Get a concise summary of key performance indicators including session duration, success rate, and average turnaround times.

voice_statistics_recent

Show detailed timing information for recent voice interactions.

# Show last 5 interactions
voice_statistics_recent(limit=5)

voice_statistics_reset

Clear all tracked statistics and start a new session. Useful for testing performance changes.

voice_statistics_export

Export all statistics and raw metrics as JSON for external analysis.

MCP Resources

The statistics are also available as MCP resources for programmatic access:

voice://statistics/current

Real-time JSON data with complete session statistics, performance metrics, and recent interactions.

voice://statistics/summary/json

Lightweight JSON summary with key performance indicators.

voice://statistics/export/latest

Complete dataset export including all raw metrics for analysis.

Integration

The statistics system automatically tracks all voice interactions from the converse tool:

• Speak-only interactions: Tracks TTS performance
• Full conversations: Tracks complete round-trip timing
• LiveKit interactions: Tracks transport-level metrics
• Error cases: Records failures and error messages

Example Dashboard Output

🎙️ VOICE CONVERSATION STATISTICS
==================================================

📊 SESSION OVERVIEW
Duration: 0:05:23
Total Interactions: 12
Successful: 11
Failed: 1
Success Rate: 91.7%

⚡ PERFORMANCE METRICS (seconds)
------------------------------
Time to First Audio:   0.65s  (min:  0.31s, max:  1.20s)
TTS Generation:        1.24s  (min:  0.80s, max:  2.10s)
TTS Playback:          2.15s  (min:  1.20s, max:  3.80s)
STT Processing:        0.78s  (min:  0.50s, max:  1.10s)
Total Turnaround:     18.50s  (min: 12.50s, max: 27.00s)

🔧 PROVIDER USAGE
------------------------------
Voice Providers:
  openai: 7 uses
  kokoro: 4 uses
Transports:
  local: 10 uses
  livekit: 1 uses
Voices:
  nova: 4 uses
  af_sky: 3 uses
  shimmer: 3 uses

📝 RECENT INTERACTIONS (5 of 12)
------------------------------
1. 15:42:18 ✅  12.5s [kokoro] What time is it?...
2. 15:41:45 ✅  19.0s [openai] Hello, how are you?...
3. 15:41:20 ❌   N/A  [openai] Can you hear me?...
4. 15:40:55 ✅  27.0s [openai] Tell me a joke...
5. 15:40:30 ✅  15.2s [livekit] Test LiveKit...

Performance Analysis

The dashboard helps identify:

• Fastest providers: Compare TTS generation times across OpenAI and Kokoro
• Optimal configurations: Track which voice/model combinations perform best
• Session patterns: Monitor success rates and identify failure causes
• Timing bottlenecks: Identify whether delays are in TTS, STT, or network

Usage Recommendations

1. Before performance testing: Use voice_statistics_reset to start fresh tracking
2. During conversations: Monitor real-time metrics with voice_statistics_summary
3. After testing: Use voice_statistics for complete analysis
4. For debugging: Check voice_statistics_recent for detailed timing of failed interactions
5. For reporting: Export data with voice_statistics_export for external analysis

Technical Details

• Thread-safe: Statistics tracking works across concurrent voice operations
• Memory efficient: Maintains last 1000 interactions to prevent memory bloat
• Automatic integration: No manual calls needed - statistics track automatically
• Error resilient: Statistics tracking failures don't interrupt voice operations

The statistics system provides valuable insights into voice conversation performance and helps optimize the voice-mode system for the best user experience.