Databricks October 2026 update: window measures, Genie Code CLI and the features that matter

Databricks ships changes almost every day, which makes it hard to see what actually matters. October 2026 has been a busy month: nine features reached general availability in the first two weeks alone, alongside new AI tools and models. This round-up picks out what changes day-to-day work for analysts and analytics engineers, with examples you can copy.

Window measures in metric views (GA)

Metric views are Databricks' semantic layer: you define a measure such as revenue once, and dashboards, Genie and SQL all use the same definition. Until now, rolling and cumulative calculations had to be rewritten in every query. Window measures move them into the definition itself.

version: 1.1
source: samples.tpch.orders
dimensions:
  - name: date
    expr: o_orderdate
measures:
  - name: t7d_customers            # rolling 7-day distinct customers
    expr: COUNT(DISTINCT o_custkey)
    window:
      - order: date
        range: trailing 7 day
        semiadditive: last
  - name: running_total_sales      # cumulative revenue
    expr: SUM(o_totalprice)
    window:
      - order: date
        range: cumulative
        semiadditive: last

SELECT
  DATE_TRUNC('month', date)        AS month,
  MEASURE(t7d_customers)           AS customers_7d,
  MEASURE(running_total_sales)     AS revenue_to_date
FROM my_metric_view
WHERE date >= DATE'2026-01-01'
GROUP BY ALL;

The range can be a trailing window such as trailing 7 day, cumulative, or a comparison period such as trailing 1 day for day-over-day growth. Because the logic lives in the metric view, every dashboard and every AI question that uses it gets the same answer.

AI and agents

Here is how chart extraction looks in SQL. It is ideal for turning reports and slide decks that only exist as PDFs back into data:

-- Parse every PDF in a volume and keep the extracted chart data
SELECT
  path,
  ai_parse_document(content) AS parsed
FROM READ_FILES('/Volumes/finance/reports/quarterly/', format => 'binaryFile');

Governance and data engineering

Models: what's new and what's retiring

What to do this month

New to semantic layers? Read Star schema data modelling first, and see how Databricks compares with Snowflake in The modern data stack in 2026.

Source: Databricks platform release notes, October 2026, and the metric view window measures documentation. Features roll out gradually by region and workspace.

Alessandro Ecclesie AgazziWritten by Alessandro Ecclesie Agazzi, analytics engineer in London. Updated 14 October 2026. About me