Skip to main content
Open Streams in the sidebar and click Create Stream. The builder is a three-step wizard; Next stays disabled while the current step has validation errors, so you always know where you stand.

​
Step 1 — Stream Properties

The stream’s identity and configuration:
  • application.id (required) — the Kafka Streams application id. It names the consumer group, prefixes the internal topics, and identifies the stream everywhere in Alginte (including its detail-page URL). Pick it deliberately: redeploying under the same id resumes from the same committed offsets and state.
  • bootstrap.servers (required) — usually your cluster’s default.
  • Serde defaults — the default key/value types records are read and written with; individual nodes can override them later.
  • Optional groups for resiliency (replication factor, standby replicas, state directory) and error handling (the dead-letter-queue topic), plus any other Kafka Streams property.
The Create Stream wizard on the Stream Properties step, with grouped config sections and Upload in the header

Step 1 — grouped configuration; the badge on step 2 shows the canvas still has unresolved errors.

​
Step 2 — Stream Flow: the canvas

The node palette sits beside the canvas, grouped by category — Source, Stateless, Group, Aggregate, Window, Join, Branch / Merge, Convert, Suppress, Terminal. Drag an item onto the canvas to add it, then draw edges between node handles to define the record flow.
  • Start with a Source (Stream Source, Table Source, or Global Table Source) reading from a topic.
  • Add processors — Map, Filter, Group By, Aggregate, Join, windowing and the rest of the operator set (see the overview).
  • End each branch in a Terminal node — typically a Sink writing to a topic.
Every node carries a small toolbar: Show Properties Panel, Duplicate, Delete. The properties panel opens as a drawer. Above the editor sit two collapsed sections — Help for the operator, and Editor type hints; the node’s SpEL expression editors are always visible in the middle; below them come the node’s own settings (serde overrides, join windowing), then Sample preview (in → out) — a real record from the source topic run through your expressions, showing what arrives here and what leaves — and Declared output, where you state this node’s output type for everything downstream. Both are described on the SpEL page.
The Stream Flow canvas with the word-count topology, the node palette, and the FlatMap Values drawer open

The canvas at work: the palette on the left, the word-count topology, and a node's drawer open with its expression editor.

​
Validation as you build

Nodes, wizard steps, and palette groups carry error badges while something is incomplete — a missing topic, an unconnected handle, an invalid expression. Hover a badge to read the messages. When the canvas is clean, Next unlocks. Three badges look past the node itself, at what the graph would do deployed: a join whose inputs are not co-partitioned, a sink whose value serde cannot serialize what reaches it, and a sink that a windowed key reaches with nothing re-keying in between. Each names the fix, and the deploy refuses the same graph with the same message.

​
Editing shortcuts

  • Copy / paste: Ctrl+C / Ctrl+V — works for multi-node selections, including their connections.
  • Undo / redo: Ctrl+Z / Ctrl+Y.
  • Multi-select: hold Shift and drag a box around nodes (box selection — not Shift+click).
  • Alignment toolbar for tidying the selected nodes.

​
Save your work: Download / Upload

The wizard header has Download (export the whole design — configuration and canvas layout — as JSON) and Upload (import one back, ready to edit or submit). Export is your durable copy: designs live in the file, not in the server. Download early, download often. A design carries its logic, not its cluster. If an uploaded file has bootstrap.servers, a Schema Registry URL or security settings in its configuration, the upload leaves them out and a notice names which, so the connection that applies is this console’s own. Set them on Stream Properties only if this stream needs values of its own. That is what lets a design made in the playground deploy on the console that runs against your cluster (From the playground to your cluster). Available from Alginte 0.13.0.

​
Step 3 — Submit Stream

Review and hit Submit. The topology is built server-side, deployed into your Alginte instance, and started. What happens next — lifecycle, state, and edition behaviour — is on Deploy & lifecycle.