Debugging Voice Agents
Use this page when something is clearly broken in a way a caller would notice. Start by reproducing the issue in Test Agent, then use Conversations to see exactly what happened and where the interaction went off track. Use Chat or Web call in the Test Agent popover to reproduce quickly, then confirm behavior in Phone call mode before launch.Where To Look First
Conversation History
Access the full conversation history including transcripts, actions, and metadata
Real-Time Transcript
Monitor live transcription and agent responses during test calls
Tool Activity
Review details about transfers, bookings, custom actions, and other connected steps
Error Details
See messages that explain why a step failed or did not complete
Systematic Debugging Approach
When something goes wrong, follow this systematic process:Reproduce the issue
Test again to confirm the problem is consistentNote exact conditions when it occurs
Identify the component
Determine which part appears to be wrong:
- Speech recognition
- Response quality
- Voice output
- Tool or integration execution
- Knowledge retrieval
Review logs
Open Conversations and find the problematic callExamine the transcript, tool activity, and any visible error messages
Test components individually
Isolate the failing component:
- Try different transcriber
- Try a simpler prompt or questions
- Try different voice
- Trigger one tool at a time
- If needed, ask the owner of the connected system to confirm it is available
Component-Level Debugging
Use the conversation log to identify the failing component, then switch to the specialist page that owns the fix.| If the log suggests… | Check in the log | Specialist guide |
|---|---|---|
| Transcription is wrong | Transcript text vs. what the caller said, missing words, language mismatch | Transcriber and Call Quality |
| AI response is wrong | Full conversation context, retrieved knowledge, prompt behavior | Prompt Engineering Guide and Knowledge Issues |
| Voice output sounds wrong | Recording, spoken text, pronunciation, voice settings | Voice Settings and Call Quality |
| Tool execution failed | Tool call event, request summary, response summary, error message, duration | Tools & Integrations and Custom Actions |
| Knowledge retrieval failed | Retrieved chunks, assigned knowledge base, item status | Knowledge Issues |
| Conversation timing feels wrong | Turn timing, interruptions, silence, response latency | Call Quality |
Using Conversation Logs for Debugging
Every test call creates a detailed log accessible in Conversations.What’s in the logs:
Basic information:- Call date, time, duration
- Agent used
- Phone number (if phone test)
- Call status (completed, failed, etc.)
- Full transcript (user + agent)
- Timestamps for each message
- Audio recording (if available)
- Tools triggered with details
- Phone keypad inputs captured
- Goal analysis results
- Gather Insights responses
- Additional metadata
- Error messages
- Filter by agent name to find test calls
- Open specific call to see full details
- Read transcript to identify where it went wrong
- Check tool details if tools failed
- Listen to audio if transcript looks correct but audio was wrong
- Review timestamps to identify response time issues
Getting Help
When you need additional support:Review Documentation
Check specific feature docs for configuration details
Check Provider Status
Check the service status of any providers your agent depends on
Contact Support
Email support@itellico.ai with call logs and error details
- Agent ID or name
- Conversation ID from logs
- Specific error messages
- Steps to reproduce
- Screenshots if applicable
Next Steps
Launch Checklist
Review the launch checklist to prepare your agent for production