Theme
Conversation History
The Conversation History view lets you review past agent-customer interactions along with the suggestions that Agent Assist provided. Use it to audit assist quality, investigate escalations, and understand how the RAG pipeline performed for specific conversations.
Browsing Conversations
Conversation history list showing date, agent, duration, messages, suggestions, and feedback columns
The conversation list shows all past conversations with key metadata:
| Column | Description |
|---|---|
| Date/Time | When the conversation started |
| Agent | The agent who handled the conversation |
| Duration | Total conversation length |
| Messages | Number of messages exchanged |
| Suggestions | Number of assist suggestions delivered |
| Feedback | Summary of agent feedback (accepted/rejected counts) |
Filtering
Narrow the conversation list using these filters:
- Date range: Select a start and end date
- Agent: Filter by a specific agent
- Queue: Filter by CCaaS queue
- Has feedback: Show only conversations where the agent provided feedback
Transcript Viewer
Transcript viewer showing chat-style message layout with inline suggestion cards and feedback indicators
Clicking a conversation opens the transcript viewer, which displays the full conversation in a chat-style layout. Messages are rendered with markdown support, preserving formatting such as lists, code blocks, and links.
Agent Assist suggestions appear inline alongside the conversation, showing:
- The suggested answer text
- Whether the agent accepted or rejected the suggestion
- The timestamp of the suggestion relative to the customer message
Search Log Viewer
The search log viewer shows every RAG query that was executed during the conversation. For each query:
| Field | Description |
|---|---|
| Query text | The text sent to the vector search (may differ from the raw customer message after preprocessing) |
| Knowledge base | Which KB was searched |
| Results returned | Number of chunks that met the similarity threshold |
| Processing time | End-to-end latency for the search and answer generation |
Source Citations
Each suggestion includes source citations linking back to the document chunks that were used to generate the answer. Citations display:
| Detail | Description |
|---|---|
| Document name | The source document filename |
| Chunk excerpt | A preview of the matched chunk text |
| Relevance score | The cosine similarity score between the query and the chunk |
| Page/section | Location within the source document, when available |
TIP
Low relevance scores on accepted suggestions may indicate that your similarity threshold is set appropriately but the knowledge base content could be improved. Use these insights to refine your documents.
Exporting Data
Conversation transcripts and search logs can be exported as CSV or JSON for offline analysis. Use the export button at the top of the conversation detail view.