Alginte 0.8.0 Beta

Released 2026-08-13.

Alginte is in beta (0.x): configuration keys, the topology export format, and APIs may change between releases; breaking changes are called out in release notes with migration guidance.

Watch it run, per node

Because your streams run inside Alginte, the console reads their runtime metrics directly — no agent, no scraping. The Stream Flow tab now joins those metrics back to the nodes you drew: live badges show throughput per node and state-store size on aggregates, the canvas frame is colored by the stream's health, and a one-switch DEBUG metrics toggle lights up real msg/s for one stream without restarting anything. Details: Monitor per node.

The editor knows your data

Pick a source topic with a registered schema and the whole builder becomes schema-aware: expression editors offer the schema's fields as complete, typed calls (get('customerId') — customerId : string), the Sample preview runs a real record from your topic through your expressions in the same sandbox deployed streams use, and downstream of a map the completions include the fields your expressions computed, marked (inferred). Filters show their verdict on the sample before you deploy anything.

Behind it sits a new differential test harness that runs every expression through the editor's validator, the preview, and a real deployed topology, and asserts the three agree — six fidelity bugs were found and fixed by the harness before this release, none by a user.

Produce against any schema version

When producing a test message with a schema serde, picking a subject reveals a Schema version selector listing that subject's registered versions, defaulting to the latest. Choose an older one and the record is encoded against that version and carries its schema id — exactly what a client that has not been upgraded would send — so you can prove your consumers still handle it before anything is upgraded. (Shipped in 0.8.0; added to these notes on 2026-09-08.)

Breaking: stable operator names (one-time migration)

Topology operators, state stores, changelog and repartition topics now get deterministic, readable names derived from your design (enrichOrder-map1 instead of KSTREAM-MAP-0000000007) — the join that makes per-node telemetry possible, and better identities in logs, JMX, and on your cluster.

The one-time consequence: a stream deployed before 0.8.0 gets new store and internal-topic names on its first redeploy. Committed offsets survive (processing continues from where it left off), but stateful nodes start with empty state — forward-only, never an automatic recompute. If an aggregate must be rebuilt from history: stop the stream, reset the consumer group's offsets to earliest (the Consumer Groups page has the seek tools), and start it again. The superseded <appId>-*-changelog / -repartition topics — and any registry subjects auto-registered under their names — are orphaned on the cluster: harmless, but worth cleaning up, and the subjects are the half people forget.

Housekeeping

• Deploying a topology whose application.id is already deployed or registered is now rejected with a clear message.
• Schema-registry-backed topologies deploy using the app's configured registry — no more re-entering schema.registry.url per stream (per-stream overrides still win).
• Canvas polish: drag-and-drop lands nodes exactly where the ghost shows, source topics preselect their registered schema, a rejected submit no longer locks the wizard, snap to grid, zoom to selection.
• Broker compatibility re-verified for this release: Apache Kafka 3.7–4.2, Confluent Platform 8.2, and Redpanda — including deployed topologies.

Get it

docker pull alginte/alginte:latest

Full install guide in the documentation — or try the playground: one command, no Kafka needed.