Browse & edit

Browse

A paged, sortable grid whose columns are inferred by sampling documents — nested paths like shipTo.city become first-class columns.

Numeric filters and sorts compare numerically, so 9 < 10 rather than "9" > "10".

Filtering

Search matches across the string fields found by sampling — the quick way in when you know a value but not which field holds it.

+ Filter builds conditions on any envelope column or JSON path: =, ≠, contains, starts with, ends with, comparisons, and null checks. Several conditions combine with AND.

Columns picks which of the sampled paths to show, for a type whose documents are wider than a screen.

For anything the filter row can’t express — an OR, a function, a projection — use the query console, which speaks DocumentDb’s own string grammar over the same connection.

The JSON view

Any row expands into a read-only, syntax-highlighted view of its whole body, with objects and arrays collapsible (native <details> elements, so it works with scripting off) and a toggle to the raw pretty-printed text.

Every value is HTML-escaped on the way out: document content belongs to whoever wrote it, and this markup is rendered raw.

Long numeric arrays — embeddings — render as float[1536] [0.021, -0.114, …] in both the grid and the formatted view rather than component by component. Nothing is hidden: the raw toggle still shows the whole body. It is the difference between a summary and tens of kilobytes of markup per row.

Row links into the sidecars

Every sidecar is keyed by document id, so each row carries a History, Blobs and Geometry link beside its JSON and Edit buttons. They open the same tab the header would, already narrowed to that one document — and each of those tabs keeps its own control to widen back out to the whole type.

A link only appears when the thing behind it exists: History when the table has a {table}_history sidecar, Blobs when it has {table}_blobs, Geometry when the type stores GeoJSON. A plain type gets a plain row rather than three dead ends.

The trip back works too: the document id in the History, Blobs and Geometry listings is a link, and clicking it opens the live document body in a read-only overlay. A sidecar row is about a document it does not show, and an id on its own rarely says which one. If nothing comes back — a removed document still carrying history, a payload that outlived its owner — the overlay says so instead of failing.

Edit

A JSON editor with format and validation. Insert, edit, duplicate-by-id, single and bulk delete.

Writes carry the sidecars with them, because nothing in the database does it for you:

• deleting a document clears its blob rows,
• a temporal type records the version, stamped with the actor shiny-docdb-myadmin,
• a vector-mapped type has its embedding re-indexed — or, where the tool cannot write the sidecar, says so instead of leaving a search index quietly disagreeing with the documents,
• full-text indexes need nothing, because the database maintains them.

Clear is the one exception, matching the library: a bulk delete is not tracked in history.

Create

The + Document button writes the first document of a type — which is how a type comes into existence in a schema-free store. Creating the documents table itself is on the database overview, using the provider’s own DDL, so a database whose store has never run isn’t a dead end.