Settings → Search indexing controls what ⌘K search can find: each data plane’s catalogs, schemas, tables and columns. It’s a separate card per data plane, because each plane can hold a different set of connected sources.
This page needs metadata search on your plan. If your organization doesn’t have it, the page explains that and points you to your account team instead of showing settings.
Anyone in your organization can see this page. Changing a setting needs permission to manage query engines; everyone else sees the same cards with the controls disabled and a note explaining why.

Turning indexing on for a plane

Indexing is per plane, and it’s turned on for you when you get started so search works from day one:
  • When you create a query engine on a plane you’ve never set up for indexing, indexing is switched on automatically for that plane (as long as your plan includes metadata search). This happens both in the console and on dip engine create, so you don’t have to visit this page first.
  • dip engine create accepts --no-index to skip it.
This only ever applies to a plane you haven’t configured yet. Once you’ve adjusted a plane here — turned it on, turned it off, or chosen which engine crawls it — creating more engines never changes that choice. This page is where you manage indexing by hand from then on. Turning a plane on costs engine time on every crawl, and embedding costs too if your deployment has an embedding provider configured, so if you’d rather not index a plane, turn it off here (or create its engine with --no-index). A plane that’s never been set up and whose engine opted out starts unindexed, not “switched off”: the page says so rather than showing a toggle that looks like somebody already turned it off. For each plane you can set:
  • Whether it’s indexed at all, with a switch.
  • Which engine to crawl through. Pick a specific engine, or leave it on Automatic and the platform picks one for you. If the engine you’d previously chosen is no longer on this plane, the page tells you and falls back to automatic until you choose again.
Turning a plane on starts its first crawl right away. You don’t need to also trigger a crawl by hand: the index begins building the moment you switch indexing on, and search fills in as soon as that first crawl finishes.

How fresh the index stays

An enabled plane is re-indexed automatically as soon as its catalogs change or an integration is added or removed. Adding an integration normally re-crawls only that integration’s catalogs. If the plane has never been indexed (for example, you add sample data to a new workspace), that crawl covers every catalog instead, so search fills in without a crawl by hand. When you delete an integration, its catalogs, tables and columns leave search right away, without waiting for a catalog apply on the engine. If a crawl is running at that moment, they leave when that crawl finishes. Separately, every enabled plane is refreshed in the background about once a week regardless of whether anything visible changed. That weekly pass is what picks up columns and comments a connected system changed on its own, underneath a catalog set the platform never saw move. Turning a plane off stops it from being refreshed. Search keeps answering from whatever index already exists; it stops getting newer, nothing more.
A single crawl only re-inventories up to 500 catalogs. If a plane mounts more than that, the rest are carried forward from the previous crawl rather than dropped. They stay searchable, only not as fresh as the ones the crawl reached this time. The next crawl picks up wherever the last one left off.

Triggering a crawl yourself

Turning a plane on from this page already starts its first crawl. There’s no separate “crawl now” button here, because switching indexing on is the crawl. To start another crawl later, or to start one without opening this page:
  • In the catalog browser, when nothing has been indexed yet and indexing is still off, a Turn on indexing button in the empty rail switches it on and starts the first crawl in one step (if you can manage query engines).
  • In the catalog browser, once indexing is on, Refresh from engine starts a fresh crawl of every enabled plane’s engines. This is the one control that wakes a parked engine, and it says so before you use it. While indexing is off it is disabled and links back here.
  • In search, when nothing has been indexed yet, an Index now button appears in the palette’s empty state (if you can manage query engines). It starts a crawl, which runs once indexing is on for a plane with an engine.
  • From the CLI, dip crawl start dispatches a crawl for one engine, one integration’s catalogs, or every eligible engine, and dip crawl status reports whether it’s currently crawling, when it last completed, and which catalogs it couldn’t reach.
A crawl reports that it started, not that it finished. Crawls run in the background on the data plane itself, so check back (or re-run dip crawl status) to see the result land.

Why does search look stuck or out of date?

If search seems to be missing recent changes, or a crawl seems to run forever, dip crawl status is the most direct way to check. It shows, per engine, whether it has ever been crawled, when the last complete crawl finished, how many documents and catalogs it’s currently serving, which catalogs the last crawl couldn’t read, whether that engine’s connected catalogs have changed since the index was built, and whether indexing is on for that engine’s plane. A plane with indexing off keeps serving whatever it indexed before it was switched off, so its engines can still read indexed. The indexing line is what tells you that index is frozen and that no crawl will run there until you turn it back on. An empty search result and an index that’s never been built look identical from the console alone; dip crawl status is what tells them apart.
Check indexing status
Start a crawl for one engine

Search your data

What indexing actually makes findable, and what an unindexed tenant sees.

Browse the catalog

Where the one control that wakes a parked engine to re-crawl lives.