Conversations
The Conversations module is where every exchange between your users and Rational AI's agents comes together. Whenever someone interacts with a touchpoint, the messages are captured here as a single conversation and automatically enriched with a title, topic and sentiment. Use this module to monitor activity, review what your agents are saying, step into a live chat when a user needs a human, and audit individual messages down to the raw tool calls.
Navigate your conversations
Open the Conversations section from the top navigation bar of the Control Room. The page presents a paginated table in which each row is one conversation. From here you can:
- Search across conversations using the search box in the top‑right corner.
- Sort and filter the table by column to focus on what matters.
- Open a conversation to read the full thread and take action.
- Select rows with their checkboxes to run bulk Actions.
The conversations table
Each row summarises a single conversation across the following columns:
| Column | Description |
|---|---|
| User | The end user who took part in the conversation. |
| Conversation | A preview of the most recent message in the thread. |
| Topic | The topic detected for the conversation (for example, Market Research/SEO, Grant Application or Data/Security Concern). Conversations with no clear match show No Topic. |
| Sentiment | The estimated emotional tone of the conversation, shown as a coloured dot with a label (for example, Neutral). |
| Last message | The date and time of the most recent message. The table is sorted by this column, newest first, by default. |
| Status | Whether the conversation is Open or Closed. |
| Favourite | A star you can click to flag a conversation so it's easy to return to later. |
The title, topic and sentiment shown here are produced by the automatic analyses you configure in Conversations settings. All labels can be re‑run. Useful after you change your topic list or models in settings.
Open a conversation
Click any row (or choose Open from the row's options menu) to open the conversation in a panel on the right. The panel shows the full thread, with the user's messages and the agent's replies in order. Hover over a message to reveal quick actions such as regenerate, edit and copy.
The panel header gives you several tools:
- Search within the conversation.
- Star to add the conversation to your favourites.
- Debug mode (
{}) toggles a developer view that exposes the underlying message objects: tool calls, JSON payloads,tool_call_ids and any errors the agent encountered. Use it to understand exactly what happened behind a reply. - Info opens a profile panel for the participant, with their Basic info (given and family name), the date of their last message and the total number of conversations they've had.
At the bottom of the panel you can:
- Ask a question: type into the message box to send a reply into the conversation yourself, optionally attaching a file, code or a Canvas.
- Take over conversation: hand the conversation from the agent to a human, so you can respond directly when a user needs personal attention.
- Summary: generate an AI summary of the conversation in a pop‑up for a quick recap of a long thread.
Best practices
- Triage with filters: combine the Sentiment and Status filters to surface conversations that may need attention, such as open threads with a Frustrated or Angry sentiment.
- Take over when it counts: when an agent is struggling or a user asks for a person, use Take over conversation rather than letting the thread stall.
- Keep labels current: after editing your topic list or switching models, select the affected conversations and Refresh labels so reporting stays accurate.
- Debug with the raw view: when a reply looks wrong, open the
{}view to inspect the tool calls and errors behind it before changing a prompt or model.
Related resources
- Conversations settings: choose the default conversational model and configure title generation, topic detection and sentiment analysis.
- Governance: review aggregated metrics and audit logs for conversational activity.
- Knowledge: manage the information your agents draw on during conversations.