Reference · Chapter 40
What Everything Used to Be Called
Databricks renames things. Not occasionally, and not only at the margins: the pipeline framework, the orchestrator and the vector database have all changed names, some of them more than once, and the documentation for the old name is often the first result you get.
The short version
If you learned this platform more than a year ago, several things you know are now called something else. Delta Live Tables is Lakeflow pipelines. Workflows is Lakeflow Jobs. Vector Search has been through Mosaic AI Vector Search and Databricks Vector Search and now answers to Databricks AI Search. Your code mostly still runs, because the renames have been careful about backward compatibility, but your searches do not, because a query for the old name returns documentation written for a product that has moved. This chapter is the translation table.
The renames that matter #
| What you probably call it | What the documentation calls it now |
|---|---|
| Delta Live Tables, DLT | Lakeflow pipelines, built on Spark Declarative Pipelines |
| Databricks Workflows | Lakeflow Jobs |
| Mosaic AI Vector Search, Databricks Vector Search | Databricks AI Search |
The Mosaic name is the one worth pausing on, because it has quietly thinned out. It arrived with the MosaicML acquisition and spread across the AI stack, so a great deal of writing from 2024 and 2025 refers to Mosaic AI Vector Search, Mosaic AI Model Serving and Mosaic AI Agent Framework. Read the current documentation for those products and the prefix is largely gone. The vector search page now opens by calling itself "Databricks AI Search (formerly Databricks Vector Search)", and the word Mosaic does not appear on it at all.
Nothing was taken away. The capabilities are all still there and mostly improved. What changed is the label, which means the label is no longer a reliable way to find them.
What still carries the old name #
This is where a rename stops being cosmetic and starts costing somebody an afternoon.
Your Python still imports the old module unless you change it. The migration is real but small: import dlt becomes from pyspark import pipelines as dp, @view becomes @temporary_view, and there's a new @materialized_view decorator for the thing the old API expressed differently. Existing code keeps running, so this is a tidy-up rather than an emergency.
The bill did not get the memo. Databricks notes that the classic SKUs for Lakeflow pipelines still begin with DLT, so a finance team reading a line item is reading a product name that the documentation no longer uses. When somebody asks why they are paying for Delta Live Tables on a platform that has no such product, this is the answer.
And the internet is worse than either. Conference talks, blog posts, Stack Overflow answers and a good deal of training material all still use the old names, and none of it is wrong about how the thing works. It's only wrong about what to type into a search box.
How to keep up without watching for it #
Databricks publishes the answer for the biggest of these itself. There's a documentation page called "What happened to Delta Live Tables (DLT)?", which exists because enough people asked. When a name you know stops returning results, search for the old name plus the word "renamed", and the vendor's own migration note is usually the first thing back.
Two habits help more than vigilance. Read the release notes for your cloud rather than the general blog, because that's where a rename lands with a date attached. And when you write anything internal that will outlive the quarter, name the capability as well as the product: "the declarative pipeline framework" survives a rename, and "DLT" does not.