Install

UI guide ​

One page per screen of the console. Each screen works on the connection at the top of the side menu; click it to pick another one. Click a screenshot to see it at full size.

Connections ​

Connections: one card per daemon with status, collections and version, and the Discover card

Each card shows if the daemon answers, its collections, version and uptime. Open opens it, Edit changes it. The Discover card finds a daemon on this machine or on a host.

New connection and Discover ​

+ New connection opens a drawer with three fields:

  • Name, optional. Empty takes local for a socket, or the host for TCP.
  • URL or host. Type the full URL, or a host and press Discover. Click a result to fill the URL.
  • Token, only for TCP with a token. If a result needs one, the cursor moves there.

Test checks the connection, Save and open keeps it.

What Discover searches:

  • Empty, localhost, 127.0.0.1, or ::1: unix sockets: listen.unix of ndb.yml (NDB_CONFIG, ./ndb.yml, /etc/ndb/ndb.yml), /run/ndb/ndb.sock, and each *.sock in /run/ndb, /var/run, /run, /tmp, $XDG_RUNTIME_DIR, and the subfolders of /tmp with ndb in the name (at most 50). It also tries TCP on 127.0.0.1.
  • Another host: TCP on the port in the field (10.10.0.1:7071), or on 7070.
  • Each candidate gets GET /health with an 800 ms timeout. ok shows the version and the collection count, needs a token is a 401, starting is a 503 while ndb recovers. Anything else is dropped. The probe never sends a saved token.

Dashboard ​

Dashboard: documents, disk, memory and version cards, traffic, the disk bar and the collections table

The state of the daemon: health, quarantined lines, and the notes that need attention (for example, the encryption key in the data dir). The cards show documents, disk, memory and version. Traffic comes from /_metrics. The disk bar splits segments, WAL, indexes and meta, and blobs and cursors. The table lists each collection and stream with its kind, documents, indexes, TTL and status.

Collections ​

Collections: the list and the orders collection with its general information, definition and storage

Find a collection by name or by a pattern (logs-2026-*), and mark favorites with the star. + makes a collection. A collection has the tabs Overview, Indexes, Storage, Locks and Backups. Actions has the changes to the collection, and Explore documents opens the browser.

Browser ​

Browser: the document list and the selected order as a JSON tree editor

  • Open a collection and select Explore documents, filter with a query (JSON, the same as a search), and page with Next. The line under the filter shows the plan and the rows examined.
  • The selected document opens as a tree: click a value to edit it, + Field adds one, × removes one. JSON shows the raw text, which you can edit too.
  • Save sends only the changed fields as $set and $unset, with the _v that the page was loaded with as a guard. If somebody else wrote first, nothing is written and the page says so.
  • + Insert adds a document; Delete asks first. A direct collection is read-only.
  • The collection, the filter, the page and the selected _id are in the URL, so a link shows the same view.

Aggregate ​

Aggregate: a $group stage and its result as a chart

Build a pipeline stage by stage on one collection, a list, or a pattern. Each stage shows how many rows leave it. The result shows as Chart, Table or JSON. Saved pipelines keeps a pipeline, Copy as curl and Open in Dev Tools give the request.

Dev Tools ​

Dev Tools: the request editor and the answer as a tree

  • Write requests as METHOD /path with an optional JSON body under it. ▶ marks each request; Cmd/Ctrl + Enter sends the one under the cursor.
  • The editor indents for you: Enter keeps the indent and opens { } blocks, } dedents, Tab and Shift+Tab move lines. The { } button on hover formats the JSON body.
  • The answer shows the status, time, size, the X-Ndb-* headers, and the body as Tree, Raw or Headers. An NDJSON answer shows one node per line.
  • History keeps the last 50 requests, and the editor text is kept per connection in the browser. Copy as curl writes $NDB_TOKEN as a placeholder, never the token.
  • Every route of the HTTP API can be sent, except /_backup and _watch, which have their own tabs.

Watch ​

Watch: live put events of the orders collection

Pick a collection, optionally a query filter, and Start: each change arrives as a put or del row; click one to see the document. Pause, Clear, and Start at (now or the beginning) control the stream. The list keeps the last 500 events.

The tail only reads: it never acks. Each stream opens a watch with a consumer named ndb-ui-xxxxxxxx, and ndb stores a position row for it in cursors.bolt. With wal_retention, the WAL files from the oldest such position are kept. ndb has no call to remove a consumer yet.

Streams ​

Streams: entries, groups, pending and lag of the jobs stream

A stream shows its entries, groups, pending entries, lag and dead letter. Append entry, New group and Trim change it. The tabs show the entries and the dead letter; a group shows its pending entries.

Backups ​

Backups: point-in-time recovery per collection and a full backup

Point-in-time recovery shows the window that each collection can restore, with Restore and Rename a collection.

Download backup streams POST /_backup to a file ndb-<name>-<date>.ndb, or .ndb.gz with gzip. It works on a unix socket connection only (ndb answers 403 over TCP; run ndb backup on the host). Restore is not in the console, because it replaces the data directory while ndb is stopped: copy the command from the page. See Backup and restore.

Operations ​

Operations: logs, encryption key, settings, security and tokens

The log level and the log limits (body_max, slow_ms): Apply stores them in the data dir, so they hold across a restart. The page also shows the encryption key (masked), the settings of /_cluster/settings, the security checks and the token of the connection.