Skip to main content
Enterprise only
The Console is a web UI for LanceDB Enterprise — an alternative to the SDK for people who want to explore and operate a deployment from a browser instead of writing code. It ships with every Managed and BYOC deployment, reachable at the URL your deployment provisions.

What you can do in the Console

  • Run SQL. A workspace for writing and running SQL against your tables, with results rendered as a browsable grid — the same engine the SDK and lancedb CLI use, reachable without either one. Multimodal columns render inline, so an image column shows as a thumbnail in the grid instead of raw bytes.
  • Browse tables. Inspect a table’s schema, row count, and recent versions, and preview rows, without writing a query first.
  • Monitor jobs. Follow the running, queued, completed, and failed jobs behind refreshes, materialized views, and maintenance, including per-job progress and errors.
  • Manage access. Sign in with your organization’s identity provider, and manage who else on your team can reach the deployment.
Reach out to us if you’d like to see the Console as part of a LanceDB Enterprise deployment. For camelCase columns such as movieId, see SQL identifiers and strings for the quoting rules in SQL statements, SDK filters, and update expressions.

Table history

Open a table’s History tab to browse its retained versions on the selected branch. Tag badges identify named versions; Create tag pins a new name to the selected branch and exact version without moving an existing tag. Switching branches changes the history and matching tags shown. Select a version’s details to inspect its schema and recorded operation. Function refreshes also show their job, Function version, and output columns when that provenance is recorded. Older versions may have less metadata. View data reads the first 20 rows from that historical snapshot, including up to 12 scalar columns. Blob, binary, vector, list, and nested payloads are omitted. Compare versions shows schema additions, removals, and type changes within the selected branch. It also shows the net row-count change when both versions’ counts are available; this does not identify individual inserted, deleted, or modified rows. Load row changes to compare inserted, deleted and updated row counts between an ordered pair of versions with stable row IDs. The server refuses ranges containing restore operations; schema and net-count comparisons remain available. These counts exclude temporary inserts deleted within the range and do not return row values. Restore creates a new latest version from the selected historical snapshot after confirmation. It leaves existing historical versions available. The server refuses restoration to or from a version with real-time streaming enabled and validates Function bindings before committing. Restore requires table ownership.

Functions

Open Functions, then select a Function version to inspect its definition and Used by panel. Each binding shows the table, output column, pinned Function version, declared input columns, and computed columns that directly read its output, including SQL columns. Enable All versions to include bindings to other versions of the same Function. Usage discovery checks accessible tables’ current main branches. Other branches and historical table versions are not included. The panel warns when discovery is incomplete and labels bindings whose Function identity cannot be verified. Use Refresh usage to check again after changing a table. Deletion confirmations summarize known dependencies and any incomplete discovery. Choose Compare versions to compare another registered version with the selected version. The read-only diff shows Python source or Full definition, including signatures, images, and metadata. Source comparison requires source to be available for both versions; full definitions can still be compared when source is unavailable.

Function errors

Open a table’s Errors tab to inspect retained Function failures on the selected branch. Groups preserve the Function version, source snapshot and recorded row or fragment coordinates. These are historical records, not counts of rows currently failing. Select a Function column to see its current freshness status separately. An incomplete-listing warning means the displayed counts cover only returned records. Preview recorded input reads bounded primitive fields from the recorded source version. Long text is shortened and Blob payloads are not loaded. Missing or inaccessible snapshots produce an unavailable state; the preview does not fall back to the latest table. Failed-only retries, Function upgrades and quarantine require server APIs that the current Errors workflow does not expose. The Function column’s freshness panel also offers Retrieve failing inputs when that control is available. It copies a Python SDK lookup using the recorded source version and row address. Use an SDK that supports versioned open_table and row-address filters. The source snapshot must still be retained; missing row IDs or tracebacks do not prevent a lookup when physical coordinates are recorded.

Related guides

LanceDB Enterprise

What Enterprise is, and why teams use it.

SQL reference

The SQL the Console’s workspace runs.

Job management

The job states and history the Console’s job view surfaces.

Deployment and governance

Managed and BYOC deployment models the Console ships with.