Browse¶

The Browse surface is how you navigate the database hierarchy in Studio — profile → database (or catalog) → schema → table → column. Every level supports inline-edit on the description text, a per-asset Generate action that drafts a comment through the same human-in-the-loop queue the CLI uses, and a scope-aware Generate description dialog for bulk runs.

Routes¶

Studio supports two URL shapes, one per backend hierarchy:

Shape	When	Routes
2-level (`/db/…`)	PostgreSQL, MySQL, Snowflake, Oracle, SQL Server, Redshift, ClickHouse, DuckDB	`/db/:profile/:database`, `/db/:profile/:database/:schema`, `/db/:profile/:database/:schema/:table`
3-level (`/cat/…`)	Databricks (Unity Catalog), BigQuery	`/cat/:profile/:catalog/:schema`, `/cat/:profile/:catalog/:schema/:table`

Encoding the profile and the catalog/database into the URL means two browser tabs on different profiles never share scope. Bookmarks survive profile renames as long as the URL is regenerated.

The left sidebar lists every saved DB profile as its own expandable row.

A profile collapses to show just its name; expanding it lazy-loads the database/catalog tier from the live connection.
Each level (profile, database/catalog, schema) gets a Refresh (🔄) button that re-syncs from the backend. A small spinner replaces the icon while the refresh is in flight.
A search box at the top filters by substring across profile names, databases, schemas, and tables — useful in workspaces with many profiles.
A pinned-catalog or pinned-database badge marks a profile whose config restricts the visible scope (set via /use-db or the profile editor).

The DB cache entry near the bottom links to the DB cache inspection page.

Database page (`/db/:profile/:database`, `/cat/:profile/:catalog`)¶

Header: database / catalog name, breadcrumbs, an inline-edit description box, and a Generate description button.

The Schemas card below lists every schema reachable in the database with a folder icon, schema name, current comment (line-clamped to one line), and a per-row Gen button that drafts the schema comment in the background.

Schema page¶

Same layout, one level deeper. The card lists every table, view, and materialised view — each prefixed by an asset-kind icon coloured by kind (accent for TABLE, positive for VIEW, warning for MATERIALIZED_VIEW). The row is clickable to the table page.

Table page¶

The header is the same edit + generate pair, applied to the table-level comment. The body is a Columns card rendered as a responsive DataTable:

Column	Notes
Column name	Mono font, left-pinned on desktop
Data type
Current comment	Click to inline-edit
Status	Icon: commented vs uncommented
Actions	Per-row Gen button

The summary line above the table reads N columns total, M commented. The DataTable supports sort, filter, and search; non-critical columns hide on mobile via the hideOnMobile flag (see Responsive Studio).

Generate description dialog¶

Triggered from the header Generate description button on any Browse page. Two options:

Just this database / schema / table. Single LLM call, inline. The result lands directly in the description box.
Everything (bulk run). Spawns a full background /run over the asset's children — equivalent to running /run <scope> from the CLI. Progress lands in Runs; results route to Pending.

Optional doc / code profile chips below the two options let you pick the RAG context the agent should consult. Empty = use the linked profiles configured for the active DB scope.

Assets panel¶

Profiles that ship executable assets (Databricks notebooks / jobs / pipelines, Snowflake notebooks / Streamlit apps / streams, query history on either platform) surface a second tab next to the schema tree: Assets. See Remote code-asset ingestion for the per-backend support matrix and the underlying remote_* storage tables.

Asset list¶

The asset list is a paginated DataTable per kind. Filters along the top: kind picker (notebooks / jobs / pipelines / streamlit apps / streams / queries), substring search, and a result-count chip. Sorting on every column. The hideOnMobile flag suppresses the noisiest columns on small viewports so the list stays scannable on a phone — see Responsive Studio.

Each row links to the asset detail page (source preview, lineage links, last modification timestamp). The detail page itself uses the Lineage canvas's logo node for the host platform so the asset's provenance is obvious at a glance.

Per-asset chunking¶

Every row carries a Chunk button. Click it to override the profile-wide chunking strategy for this single asset. Useful when one notebook is a literate-programming write-up that benefits from cell-level chunking while the rest of the profile keeps the whole-notebook default.

The dialog shows the active per-kind default, the available strategies for that asset kind, and a save button that writes the override to cfg.assets_chunking.overrides keyed by asset id. Re-ingesting the asset picks up the override automatically; for already-ingested assets, click Reindex this asset in the same dialog to re-embed under the new chunking.

The profile-wide default lives in cfg.assets_chunking. Per-kind strategies and their defaults are documented under Remote code-asset ingestion → Per-kind chunking strategy.

Selective ingestion¶

The + Ingest button on the Assets header opens a picker rather than launching a blanket "ingest everything" run. The picker fans out into per-kind sections; for each kind you can:

Pick all (mirror everything on the remote platform), or
Pick specific assets by checkbox from the live list.

The picked ids land as --include-id KIND:EXTERNAL_ID on the underlying /db ingest-assets call, so a partial mirror lands in seconds instead of pulling the entire workspace tree.

The same Assets header has the Deep sync button covered on the DB cache page when the profile's catalog tier is what needs the deeper refresh.

CLI equivalents¶

Studio	CLI
Inline-edit description	`/edit <db>.<schema>.<table>.<column>`
Per-asset Gen	`/run <db>.<schema>.<table>.<column>`
Generate description (bulk)	`/run <scope>`
Refresh sidebar tier	`/sync --db-profile <name>`