Chat history
The Chat history surface (previously called Inbox) is your read-only audit log of every AI ↔ visitor conversation, across email, the Web SDK / chat widget, and the hosted portal. It is not the primary operator queue — that's Tickets. (Voice is currently disabled behind a feature flag; voice transcripts will reappear here when voice is re-enabled.)
Navigate to Support → Chat history (sits below Analytics in the sidebar).
What it's for
Use Chat history to:
- Audit Lira's accuracy. Spot-check a sample of recent chats. Did the AI give the right answer?
- Investigate complaints. A customer says "your AI told me wrong info" — pull the thread and see exactly what was said.
- Find drift. If CSAT drops, browsing recent chats often reveals what Lira got confused about.
- Discover KB gaps. Patterns of vague or hedged replies signal missing content in your Knowledge Base.
Use Tickets instead when you need to:
- Reply to a customer
- Track open work
- Mark something resolved
Why both exist
| Chat history | Tickets | |
|---|---|---|
| Volume | Every chat (huge — most resolved by AI silently) | Only when Lira can't answer (small) |
| Purpose | QA / audit | Operator work queue |
| State | open / pending / resolved | open / in_progress / resolved / closed |
| Replies | Read-only by default | Reply triggers visitor email |
A single conversation in Chat history may have spawned one or more tickets (the agent calls lira_create_support_ticket). Where this happens, the ticket number is shown on the conversation detail page so you can jump between the two views.
Conversation statuses
| Status | What it means |
|---|---|
| Open | Active — Lira is handling it or waiting for a customer reply |
| Pending | Waiting — Lira sent a response and is waiting to hear back |
| Escalated | (Legacy) Flagged for human review. The new ticketing flow uses Tickets, not this status. |
| Resolved | Closed — the conversation has finished |
Filtering and search
Use the status filter buttons to narrow the view. The search bar matches on subject, customer name, email, or detected intent.
Reading a conversation row
Each row shows:
- Customer name (or "Chat visitor" for anonymous widget conversations)
- Sentiment emoji (😊 positive, 😐 neutral, 😠 frustrated)
- Status badge
- Subject / detected intent
- Tags (up to four)
- Short preview of the last message
- Time-ago
- CSAT stars (if rated)
Viewing a conversation
Click any row to open the full thread. You can:
- Read the full message history
- See the channel (email / chat / portal; voice when re-enabled)
- View Lira's confidence score for each AI reply
- See the customer profile (name, email, history)
- (If applicable) Jump to the spawned ticket
Replies from this surface are de-emphasised by design — if a thread needs human follow-up, it should be a ticket, not a manual reply on an in-flight AI chat. Use Tickets for active operator work.
Tips
Sample, don't read everything. Pick 20 chats from the last week and audit them. Look for hedges, "I don't know" replies, and tonal misses.
Filter by CSAT < 3. Customers who rated below 3 stars often signal a gap in your Knowledge Base or a wrong answer.
Cross-reference with Tickets. If a ticket exists for a conversation, the chat history is the "before" of how Lira tried before opening the ticket. That's gold for prompt tuning and KB authoring.