Table details

The Details tab is a table’s home page. It opens by default when you select a table in the Catalog, on a project’s or workspace’s Catalog page, or in the panel beside a canvas, and gathers everything about the table in one place: the editable detail form and the table’s keys, with Suggest match in the panel’s More actions and Ask AI in the top navigation.

The Details tab: form fields, the Name from source table card, the Definition and the fields that tell an AI what the table is for, the AI suggestions section with Suggest with AI, and the Keys section.

The panel’s toolbar is the one every detail panel shares: Save and Discard changes come first, then More actions with the table’s other commands. On the Catalog page those are Customize, Reset to system-generated and Open in designer where the view supports them, Suggest match, and last Delete table (or Remove view for a managed view). In a workspace’s or project’s Catalog the menu ends with Remove from workspace or Remove from project.

What lives on the Details tab

Top to bottom:

1. The detail form: Schema, Name, Type, Business entity, Business name, the Name from source table card, Definition, and then the fields that tell an AI what the table is for: Grain sentence, AI context, AI instructions, Usage, Limitations, and Lifecycle. Your edits wait for Save, which applies them to the current branch.
2. AI suggestions: Suggest with AI proposes the table’s grain sentence, AI context, AI instructions, usage, limitations, and lifecycle. Accept or reject each suggestion, or choose Accept all. Nothing changes until you accept.
3. Keys: a read-only list of the table’s key columns.

The previous Overview tab has been merged into Details. Old links that pointed at Overview open Details automatically.

Schema, Name, and Type

The top row of the detail form controls where and how the table is identified in the Catalog.

Schema

The Schema field is a picker grouped by connection and catalog. It shows every schema your organization has discovered or registered across all connections.

Choosing a schema under a different connection moves the table to that connection. When the move is committed, the table and all its columns re-home under the new connection, and their files relocate in the commit tree to reflect the new path. To move the table back, switch to the original schema and commit again.

Name

The Name field is the table’s identifier within its schema. Renaming a table updates the Catalog entry on save; the file path in the commit tree is updated when the branch is committed.

Type

The Type list is scoped to what is valid for the table’s layer and connection type. For example:

• Landing and staging layers offer Table, View, and Query.
• A file or cloud-storage connection offers only file types (such as Parquet File or Avro File).

Changing the type to one that is incompatible with the current connection is not permitted: the list filters the options automatically so only valid choices are shown.

Keys are derived from columns

Tables don’t store their own primary or business keys. Keys belong to columns: flag a column as Primary key or Business key on the Columns tab, and the Details tab’s Keys section reflects it immediately. Templates and the Data Vault accelerator read the same flags, so what you see here is exactly what generation uses.

If the Keys section says no key columns are flagged, head to the Columns tab and set the flags there.

Business entity and business name

Linking a table to a business entity records what the table represents in your business model. Pick the entity in the Details form: the picker searches your entities by name and technical name. When an entity is linked, its technical name is what downstream consumers (such as the Data Vault accelerator) will use as the table’s effective business name.

When no entity fits, use the free-text Business name field instead. If both are set, the linked entity’s technical name wins; the free text is the fallback.

When Draft table and column descriptions links the entity, the link is recorded as the AI’s too, but the Details form shows no badge for it: the small sparkle icon appears beside the entity in the Business entity column of a tables grid, which you can add from the column chooser.

Name from source table

Below those two fields, the Name from source table card suggests a business name. It walks back through the tables DeltaVault built this one from (its landing and staging chain) to the source table, shows the way it took under Walked back to: (this table last), and offers the source table’s name formatted with your entity naming convention, in a field you can edit. If you change the suggestion, a control beside the field puts the formatted name back. On a table DeltaVault did not build from another one (a source table, or a table imported straight into a later layer, like the landing table in the image at the top of this page), the card says it is the source table and suggests the table’s own name; when the walk back fails, it says no source table was found and uses the table’s own name. Under the default Proper Case convention a suggestion keeps a run of capitals together, so a source named CustomerID suggests Customer ID rather than Customer Id. The card never renames the table itself.

• Use as business name puts the suggestion into the Business name field.
• Create entity and select creates a business entity with that name at once and selects it in the Business entity picker; the card tells you the technical name the new entity will get. When an entity with that name already exists, the button reads Select existing entity and selects it instead of creating a duplicate.

The table’s link to the entity waits for Save, like every other field. If you choose Discard changes instead, an entity the card created stays in your business model, unlinked, and a notice says so. Once an entity is selected, the card folds to one line; its chevron opens it again.

Saving a newly chosen entity here also links the rest of the table’s ingest chain (its source, landing, staging, and persistent staging tables) to the same entity, and a notice says how many tables were linked. A table in the chain that is already linked to a different entity is skipped and named in the notice. When the source table is linked to a different entity, or the chain cannot be worked out, only the table you edited is linked, and the notice says why.

Definition

The Definition field says what the table is, in one or two sentences that stand on their own, as the info icon beside the label reminds you. It is stored as part of the table’s metadata, included in YAML exports, and read by the app’s AI whenever a skill works on the table. You can write the definition by hand or let Draft table and column descriptions (under Ask AI) propose one. A definition you applied from that proposal, or an AI suggestion you saved unchanged from the field’s Edit with AI editor, shows the AI badge beside Definition, with no percentage because neither gives a confidence score. Edit the definition and save, and it is yours: the badge goes.

Tell an AI what the table is for

Below the Definition, the Details tab of an existing table carries the fields an AI needs and a definition should not hold. Each one has an info icon with its guidance, and each waits for Save like the rest of the form.

Grain sentence says what one row of the table stands for, such as “one row is one customer as currently known from record source crm”. It’s an AI-assisted field, so Edit with AI can draft or improve it.

AI context is the background the app’s own AI reads before it proposes or drafts anything for this table. A skill that creates or enriches the table fills it only while it’s empty, never over your text. Suggest with AI, in the AI suggestions section below the form, can also propose a new AI context, which replaces yours only when you accept it. Either way the box then shows the AI badge until you edit it and save. It’s a plain box with no AI editor of its own, because the app’s AI is what writes it. Skills read it together with the project’s and the upstream tables’ context, as described in How AI context compounds.

AI instructions are rules for an AI that answers questions over the table’s data: the default to assume, which measure wins when a word is ambiguous, what to ask before answering. You write them, with the field’s Edit with AI help if you want it, in at most 1,000 characters, or accept what Suggest with AI proposes in the AI suggestions section below the form.

Usage says who uses the table and for which questions, and Limitations says what it doesn’t cover or must not be used for. Both are AI-assisted fields of up to 2,000 characters, and they keep that guidance out of the Definition.

Lifecycle says whether the table is still in use. It reads Not set until you choose Active or Deprecated. Choose Deprecated and two more fields appear: Replaced by, a picker for the table that takes this one’s place (any other table on the branch, emptied again with Clear), and Deprecation note, for why the table is deprecated and what to use instead. Choose Active and both are cleared when you save. A deprecated table keeps building: the flag tells people and the AI, and switches nothing off.

Ask AI and Suggest match

Suggest match sits in the table header row, visible from every tab on the table page. Ask AI is not a table-level button: it lives in the top navigation and picks the table up as context from your selection.

• Ask AI opens the AI pane with the selected table as context. Skills available there include:
  • Draft table and column descriptions: proposes a definition for the table and its undocumented columns, a business name, and a matching business entity, shown as a preview you apply or discard.
  • Review schema quality of selected tables: reviews the table for naming, normalization, and missing-constraint issues.
  • Analyze impact of a column change: assesses the downstream impact of changing a column.
• Suggest match searches for a business entity whose name or technical name fits this table and proposes a link. You review the suggestion before anything is saved.

Applying a proposal requires the editor role and a feature branch. Nothing touches the table until you apply.