Workspace Insights
Workspace Insights is a project Teramot keeps in your workspace with a record of how the workspace itself is used: three results tables that each gain a day of activity every day, and a dashboard that sums them up. It is there for the people who run the workspace to see who is using Teramot and what it costs, with the same tools they use on their own data.
It is being turned on workspace by workspace, so yours may not have it yet.
Where: the Projects page of the workspace. The project is called
workspace-insightsand its card carries a System badge.
Who sees it
Only the workspace’s Owner and Admins see the project. Analysts, Read only members and people invited to a single project do not.
Nobody can change it, Owners included: Teramot writes it. Inside the project every action that would change it is greyed out, and its card has no ⋯ menu, so it cannot be deleted, left or shared with more members.
What it holds
Three results tables, each with a snapshot_date column for the day the row describes (UTC).
Their names start as below and end in a code of the project.
insights_workspace has one row per day for the workspace as a whole:
| Column | What it holds |
|---|---|
tier | The workspace’s plan. |
seats_limit | The most members the plan allows; empty when it sets no limit. |
members | Members of the workspace. |
active_users | Members who did anything that day. |
credits | Credits spent that day, by people and by what no person caused, such as scheduled refreshes. |
projects_total, sources_total | Projects and sources in the workspace. |
golds_total | Results tables in the workspace. |
dashboards | Dashboards in the workspace. |
alerts_active | Active alerts in the workspace. |
storage_table_bytes, storage_upload_bytes | Bytes stored in tables and in uploaded files that day. |
refreshes_ok, refreshes_failed | Refreshes that succeeded and that failed that day. |
insights_users has one row per person per day:
| Column | What it holds |
|---|---|
user_id, user_email, role | Who, and their role in the workspace. |
active | Whether the person did anything that day, AI use included. A new version of one of their results tables does not count on its own, since someone else may have made it. |
credits | Credits the person spent that day. |
golds_created | Results tables the person created that day. |
golds_edited | Results tables of the person’s that got a new version that day, whoever made it. |
dashboards_created, alerts_created | Dashboards and alerts the person created that day. |
insights_ai has one row per person per day, plus one a day with no person for what no
one caused, such as building tables or calls made with an API key:
| Column | What it holds |
|---|---|
user_id, user_email | Who; empty on the row no person caused. |
conversations, messages | Conversations started with the agent, and messages sent to it. |
mcp_calls | Calls made to Teramot from an AI assistant. |
llm_calls, llm_input_tokens, llm_output_tokens | AI model calls, and their tokens. |
Activity in this project is not counted: its own results tables do not add to
golds_total, nor its dashboard to dashboards.
The workspace counts are what existed at the end of each day, so a day added later still
reads as it stood then. Two are the exception: members is the count when the day was
added, and alerts_active counts the alerts that existed that day by whether they are
active now.
When it updates
Once a day, early in the morning UTC, Teramot adds the day before. The history grows a day at a time from the day the project was created; when a day is sent again it replaces the earlier copy rather than adding a second one.
The dashboard
The project’s dashboard, Workspace Insights, condenses the three tables: the last 30 days (active people, credits, messages to the agent, calls from AI assistants, results tables created), each with its change against the 30 days before; where the workspace stands (members, results tables, dashboards, storage, AI model tokens, failed refreshes); active people, calls from AI assistants and credits per day; and the 20 people with the most active days in the last 30. It updates with the tables and does not count toward your plan’s dashboards.
Using it
Open the project and read the tables as you would any results table, or ask the agent or an AI assistant connected to Teramot — for example “credits spent per person last month” or “who created results tables this week”.