Chatbots

Chat history

Find, read, tag, rate and export the conversations your chatbots and agents have had with visitors.

History in the sidebar lists every conversation a visitor or tester had with a chatbot or agent. Use it to see what people ask, where the agent fails and what the answers cost.

The page has three tabs: Chatbots history (the sessions), Chat logs (individual events) and Chatbot tags (the tags you can put on sessions).

The sessions list#

By default you see the last 30 days, and only sessions where the visitor sent at least one message. Both defaults show as removable filter chips.

Each row shows the agent or workflow and chatbot, the session, the first message, the last message time and the visitor. The detail cells show tags, message count, duration, credits, error count, positive and negative feedback, and an Escalated mark when the conversation was handed to a human. Click a row to open it.

Filter the list#

  • Chatbot filter and Agent or workflow filter
  • Messages feedback: positive or negative
  • Error messages: only sessions with errors
  • Escalation: Escalated or Not escalated
  • Min customer messages: counts only what the visitor sent, not bot or system messages
  • tags, and a session date range

Read one conversation#

A session page shows the transcript and a summary: agent, number of messages, errors, whether it was escalated, tags, credits split between model usage and other usage, the visitor’s IP address, the page the chat started on, and when it started, ended and how long it lasted. Newer session and Older session move through your filtered list.

Visitors can rate an answer with thumbs up or down, and the rating shows on that message and in the session totals.

To label a session, add tags from the Tags cell in the list, if you have permission to edit. Create and remove tags on the Chatbot tags tab with Add Tag.

Improve the prompt from a session#

If your role can edit, a session page has Improve prompt. Describe what the agent should have done, choose the agent if the session used more than one, and click Suggest a change. FlowHunt proposes a revised system prompt. Applying it overwrites the stored prompt in the agent’s draft. It takes effect only when you publish, so you can back out by not publishing.

Metrics#

Above the list, Show metrics opens charts for the filtered sessions: total sessions, positive and negative feedback, errors, escalations, average messages and average cost per session. Hide them with Hide metrics. They are not available on a cancelled subscription.

Export to CSV#

Owners and admins see Export CSV. It exports the sessions that match your current filters, so set the date range and filters first. Choose the columns to include, then Download CSV. If the export takes too long, narrow the date range. Columns cover the session and chatbot IDs and names, times, duration, counts, credits, tags, page URL, IP address, first message, variables and a link to the session.

Delete a session#

Use the delete button on a row and confirm Remove chatbot history. This removes the session from the list.

How far back you can look#

Your plan sets the window. A request that reaches past it is refused by the backend, and the page keeps its default range inside the window.

PlanHistory window
Cancelled7 days
Trial90 days
Starter180 days
Pro365 days
Premium and EnterpriseNo limit

Personal data

Transcripts, IP addresses and page URLs are visitor data. Limit who in your workspace can open History, and tell visitors that conversations are stored.

Next steps#

Spot a bad answer? Fix the source in Knowledge sources or use Improve prompt. For handoffs to people, see Human handoff.