Skip to main content

Data tabs in Context

What the Data tabs show from Databricks, how to read the Genie caps, and how to set them up.

Context's Data group shows what your Databricks costs, who generates that cost and how much each person delivers. It answers three questions: how much do you spend on compute and on Genie, and on what; who is close to their Genie cap this month; and whether each person's cap is right for what they deliver. It needs the Databricks integration.


​

The tabs


​

  • Genie — Genie caps this month: what each person spent this month against their cap, how much of their 150 free DBUs they used, the tasks they delivered and the cost per task. Genie: who used Genie (Genie Code, Genie One and Genie Agents), what was billed, the SQL of its queries and the free usage. Genie's cost by day: its three parts day by day — the billed LLM, the SQL warehouse of its queries and the free usage at what it would have cost. Questions to Genie by space: the questions to each Genie space and to Genie One, from the interface or by API; it needs the Genie events view.

  • Cost — Cost: what Databricks billed by product, at list price and without your contract's discounts. SQL queries: queries by source (dashboards, Genie Agents, SQL editor, notebooks, jobs and alerts), their estimated cost, the failed ones and the ones served from cache. SQL warehouses: each warehouse's queries, estimated cost, average duration, queue wait and share failed. Most expensive queries: up to 50 of the period by estimated cost, with their ID to find them in Query History.

  • Jobs — Jobs and pipelines: runs by outcome, the most common failure causes and how long recovery took.

  • People and value — Genie spending and delivery: what Genie costs each person beside the tasks they delivered, the biggest spender first. By person and By team: queries, Genie cost, free usage, compute, runs, tasks delivered and cost per task.


​

Tasks delivered are the ones the Workflow's Throughput counts in the view's boards.


​

Add the tabs


​

In New view, choose the Data idea, or click I’d rather build it myself and pick the Data topic. You can also add any of the tabs from the + of a view's tabs. Whoever sees Performance Delivery can open them.


​

The figures cover the view's dates, in UTC days. The view's teams and people narrow everything to what runs under their Databricks accounts, so jobs that run with a service account are left out.


​

Genie caps


​

Databricks does not let Leanmote read your Genie caps: they are only in your Databricks account console. So that Leanmote can compare what each person spends with their cap, whoever connected Databricks types them in Settings → Organization Settings → Tabs → Data, as they are in Databricks:


​

  • Cap for everybody: each person's monthly cap, in US dollars. Empty means no cap.

  • Exceptions: up to 20 people with a cap of their own. Click Add an exception, pick the person and type their cap.


​

If Databricks has an exception for a group, add it here for each person in the group. Without caps, Leanmote still shows what each person spent in the month.


​

Reading the caps


​

Spending counts as Databricks' caps do: only Genie's billed LLM, at list price, in the calendar month the view ends in. It does not count free usage or the SQL warehouse. Under Delivered is the table's median, and each row can carry one of these pills:


​

  • N % over the cap: spent more than their cap. Click the pill to see the amounts.

  • Cap reached: spent all of their cap. Databricks cuts their Genie until someone raises the cap or the next month starts.

  • Delivers more than the median: review the cap: reached their cap and delivered more tasks than the median. Their cap may be low for what they deliver.


​

Who is «To review»


​

In Genie spending and delivery, the To review pill marks whoever spends more than the median on Genie and delivers fewer tasks than the median. Click the pill to see their figures beside both medians. The card needs at least two Leanmote people who used Genie.


​

People and teams


​

Databricks accounts are matched to Leanmote people by email. An account with no match shows its email and «not a Leanmote person» in By person: click Link and pick which Leanmote person it is. Whoever can link people in Users sees it.


​

Until it is linked, that account counts under «No team» and has no tasks delivered. A Leanmote person with no team also says «No team», and whoever manages teams can click Add to a team right there. Someone in several teams adds their figures to each.


​

Show details, filters and copies


​

  • Each Show … link opens the rows behind its figure, with a filter on every column and Export to CSV.

  • A filter applied from the details measures the card with that filter for everyone who opens the view. Remove filter goes back.

  • Whoever edits the view can compare workspaces or warehouses side by side: open ⋮ › Duplicate by workspace or Duplicate by SQL warehouse, tick the ones you want and click Make N copies. The copies go to a tab called «<card> by workspace» (or «by SQL warehouse»), each measured on its own.


​

Frequently asked questions


​

How often is the data updated? Leanmote reads Databricks once a day, and the first read starts a few minutes after connecting. Databricks publishes its billing about 12 hours later. Each card says up to which day it has data.


​

Why doesn't the spending match Databricks' console exactly? Leanmote uses Databricks' billing, and Databricks warns that its console's estimate can differ a little from it. Leanmote also shows list prices, without your contract's discounts.


​

Why is a query's cost estimated? Databricks puts no price on each query. Leanmote splits each SQL warehouse's daily cost by its queries' compute time.


​

What is Genie's free usage? Each person gets 150 free DBUs a month. Leanmote shows what that usage would have cost, but it is not money spent and does not count toward the cap. Service accounts get no free usage.


​

Why is there no LLM cost by Genie space? Databricks' billing does not say which space an LLM charge comes from. By space you only see the SQL of its queries.


​

Can I see the SQL of the most expensive queries? No: Leanmote never stores the text of your queries. Use each one's ID to find it in Databricks' Query History.


​

Are Leanmote's own reads counted? No. They show apart in SQL queries and are not added to the totals.


​

Related articles


​


​

Did this answer your question?