Diagnosis

Toggle diagnostic layers to inspect tool calls, flows, latency, and topic citations on the Transcription tab.

Use Conversation Diagnosis when you need to understand why the agent behaved a certain way on a specific turn – which tools ran, which topics were cited, and where latency occurred.

Diagnosis is a toggle group on the Transcription tab of the Conversation review side panel. Switch any layer on or off to overlay extra information on each turn. Layers persist between conversations, so the views you care about stay enabled as you move through the table.

Available diagnosis layers

Conversation variables

Displays live variable values captured during the call (for example, booking IDs, customer names, flags), with diff markers when a variable changes turn-to-turn.

Flows and steps

Tracks the agent’s navigation through flows and steps, showing the execution path and the decisions made at each branch.

Tool calls

Shows the tools the agent triggered during the call, including call parameters and outcomes.

Topic citations

Highlights the Knowledge topics the agent matched for each response.

Sources

Shows which Connected Knowledge source files the agent retrieved for each turn. Click a source name to open an inline preview panel showing the retrieved content. Use Open in Knowledge in the panel to navigate directly to the source.

Transcript corrections

Displays where the automatic transcript was edited for clarity or accuracy. Use it to distinguish ASR errors from agent logic problems.

Turn latency

Measures how long the agent took to respond at each turn.

Latency visualization includes component-level timing breakdowns.

Latency breakdown

When viewing turn latency, you can inspect timing for:

Component	Description
LLM requests	Time spent waiting for the language model to generate a response
Function calls	Time spent executing functions, including API calls and data processing
Total response time	Combined time from user speech end to agent response start

Use these breakdowns to:

• Identify slow function calls that need optimization
• Understand LLM response times for different query types
• Find bottlenecks causing user-perceived delays
• Compare latency across different conversation types

High LLM latency may indicate complex prompts that could be simplified. High function call latency often points to slow external API dependencies.

Interruptions

Shows when the caller interrupted the agent, or when barge-in was detected. Use this to understand if caller behavior affected the conversation flow.

Variants

Identifies which variant handled each part of the call.

Logs

Displays structured runtime logs emitted from your functions during the call. The Logs layer surfaces:

• Entries from conv.log.info(), conv.log.warning(), and conv.log.error().
• API response logs from conv.log_api_response().

Use the Logs layer to debug function behavior, track external API responses, and correlate runtime events with what the caller experienced. See Conversation log for the full conv.log API.

Entities

Lists extracted entities captured from the user, like booking numbers, account IDs, or city names. Especially useful in transactional scenarios where the agent needs to capture structured data from free text.

Using diagnosis for optimization

Combine multiple layers to understand agent behavior:

1. Enable Turn latency to identify slow responses.
2. Check Tool calls for those turns to see if external calls are causing delays.
3. Use Sources to verify the content the agent retrieved from Connected Knowledge.
4. Use Flows and steps to confirm the agent followed the expected path.
5. Use Logs to read structured conv.log output and API response payloads emitted by your functions.

Related pages

Conversation review

Full conversation analysis across voice and webchat.

Conversation log

The conv.log API that powers the Logs diagnosis layer.

Performance monitoring

Ongoing performance management workflows.