Chat Sessions¶

Overview¶

Chat sessions in Open Chat Studio define the scope of conversations between a participant and a chatbot within a specific channel. Sessions are isolated, ensuring data privacy and contextual continuity for the duration of an interaction.

Session Scope¶

A session is uniquely defined by:

• Participant: The individual engaging with the chatbot.
• Channel: The platform through which the chat occurs (e.g., WhatsApp, Telegram, Web, Slack). See channels.
• Chatbot: The specific chatbot handling the conversation.

Each session is independent, meaning:

• The session's data is bound to that session only and is not shared with other sessions.
• When a participant interacts with a chatbot, the chatbot receives the session's history to maintain context.
• Multi-Session Channels: Channels such as Web, API, and Slack allow multiple active sessions per participant, enabling parallel conversations.
• Single-Session Channels: Platforms like WhatsApp, Telegram, and SureAdhere support only one active session per participant at a time.

Every session has a status that tracks the progress of the conversation. See Session Status.

History Management¶

• As conversations progress, all previous messages within a session are stored as history, regardless of how the pipeline is configured.
• What subset of that stored history reaches an LLM node, and whether it is summarized or trimmed as it grows, is controlled per node. See Conversation History for how this works.

Participant Data¶

Aside from session-specific data, Open Chat Studio maintains participant data, which:

• Persists across sessions.
• Is tied to the same Participant, Channel, Chatbot scope.
• Helps retain long-term participant preferences and contextual information beyond a single session.

Anonymous Sessions¶

On the Web channel, participants can have anonymous sessions, where:

• Participant data is only available for the duration of the session.
• Since participant identity cannot be verified, data cannot persist beyond the session.

Resetting Sessions¶

For Single-Session Channels like WhatsApp and Telegram, the current session continues until it is explicitly ended.

Sessions can be reset either manually by the participant or automatically.

For task-focused details, see How to reset sessions.

Annotation Queues on the Sessions List¶

The sessions list shows which annotation queues each session is in, so you don't need to open the session. The Annotation Queues column shows one badge per queue, in the format Queue · Status, where the status is the session's item status in that queue. Queues of every status appear, and only session-level items count.

To filter by membership, use the Annotation Queue filter:

• Select one or more queues to show sessions in any of them.
• Select Not in any queue, the first value in the list, to show sessions that are in no queue.
• Select Not in any queue together with queue names to combine them with OR. For example, Not in any queue and Safety QA shows unreviewed sessions plus the sessions already in Safety QA.
• Use the excludes operator to get the complement of the selection.

The filter stores queue IDs, so saved filters keep working after you rename a queue. The same filter is available in the dataset session pickers, the participants page and the queue's Add Sessions picker.

The sessions CSV export has a matching Annotation Queues column after Session Tags, for example Safety QA (Completed), Tone (Pending). Columns after it move one position to the right. The header is always present so the layout is stable, but cells are empty for teams without the feature flag.

See also¶

• Session Status
• Participant Data
• Channels
• Conversation History
• Annotation Queues