Data Hub: automated session analysis
Data Hub only shows up on AI Agent projects. As each session completes, it's run through an automated analysis pass — Data Hub is where those results get aggregated across every session, with charts you can click straight into as filters.
What gets analysed on every session
Quality, Relevance & Engagement
Three independent 0–10 scores per session, each charted as its own histogram.
Sentiment
A positive / neutral / negative split, averaged across every analysed session into one bar.
Emotion
A single dominant emotion per session (e.g. 'excitement'), ranked by how often each one occurs.
Tags
Free-form labels like 'high-effort' or 'short_session' — the most frequent ones are ranked and clickable, same as Emotion.
Each analysed session also gets a short written rationale explaining why it scored the way it did — that lives on the Sessions tab alongside the individual transcript, rather than in the Data Hub charts, which stay aggregate.
Reading the dashboard
A row of stat tiles up top (total sessions, completed, completion rate, average quality, average duration) reacts to whatever's currently filtered. Below that: a state filter (Completed / Open / Terminated / Quota full), then histograms for Quality, Relevance and Engagement, a Sentiment bar, and ranked bars for Emotion and Tags.
If your current filters match sessions that haven't been analysed yet, Data Hub shows an empty state pointing you to the Sessions tab instead of a blank chart — analysis has to have actually run on a session before it can contribute to any of these numbers.
Next up
- Writing a discussion guideComing soon
- Coding open-ended responsesComing soon