Changelog

Added ARIMA-based forecasting for disk usage, consumer lag, and throughput metrics with automatic model selection and confidence intervals.

Introduced Kafka Watchdog: stateful anomaly rules with open/acknowledged/resolved event lifecycle and MTTR tracking.

Forecast-based predictive breach detection — watchdog rules can now fire before a metric actually exceeds a threshold.

Seasonal pattern decomposition view added to the forecasting dashboard — visualize trend, seasonal, and residual components separately.

Changepoint detection now surfaces structural breaks on the metric timeline alongside deployment markers.

Fixed an issue where MSK cluster metrics stopped refreshing after a CloudWatch API rate limit error.

Resolved a UI freeze when the consumer group table loaded more than 500 groups simultaneously.

Amazon MSK monitoring via CloudWatch API — agentless integration using IAM role-based authentication.

Unified cluster selector now treats MSK Provisioned, MSK Serverless, and self-managed Kafka identically in the UI.

MCP (Model Context Protocol) server published — connect Claude, Cursor, and other AI tools to live Kafka cluster state.

API key management redesigned: per-key permission scoping (read-only, read-write, admin), configurable expiry, and zero-downtime rotation grace periods.

OpenAPI 3.0 spec published at /api/openapi.json with full endpoint documentation and example responses.

Alert rule form now shows a live preview of matched topics/brokers/groups when a regex pattern is entered.

Fixed incorrect partition count displayed for topics with offline partitions in the topic list view.

Server-side pagination introduced for all major list views (topics, consumer groups, connectors, schemas) — supports clusters with 10k+ topics without performance degradation.

Column resizing added to all data tables with per-table persistence in localStorage.

Protobuf message deserialization support added to the topic browser alongside existing Avro and JSON support.

Topic browser now supports header inspection — key, value, and header columns are independently resizable.

Consumer group lag chart now supports zoom and pan with keyboard shortcuts for time range selection.

Kafka Connect task restart and pause/resume actions now available directly from the connector detail page.

Fixed a race condition causing duplicate alert notifications when two notification channels were configured for the same rule.

Corrected the ISR count display for topics where the replication factor is higher than the number of online brokers.

Kafka topology visualization now shows live message flow rates on edges between producers, topics, and consumer groups.

Stream processing monitoring added: Kafka Streams topology graph, task distribution, and per-task processing rate and latency.

ksqlDB integration: view running queries, persistent queries, and push query status from the KLogic dashboard.

Cluster health score algorithm updated to weight under-replication and offline partition events more heavily.

Dark mode contrast improvements across all chart components for better readability at low brightness.

Slack alert notifications now include a direct link to the affected resource in KLogic for faster incident navigation.

Fixed log compaction monitoring showing incorrect dirty ratio for topics with segment.bytes set below the default.

Schema Registry monitoring added — browse subjects, view schema versions, inspect compatibility settings, and validate schema evolution.

Audit log viewer added to the enterprise dashboard — full record of user actions, configuration changes, and API access.

RBAC roles introduced: Admin, Operator, and Viewer — workspace owners can now assign roles per member.

Producer performance dashboard now shows end-to-end latency percentiles (p50, p95, p99) in addition to averages.

Multi-cluster filter bar extended to support regex matching on topic names, broker IDs, and consumer group IDs simultaneously.

JWT token generation script added to the CLI for programmatic authentication in CI/CD environments.

Resolved an issue where the alert history page failed to load when more than 10,000 historical alerts were present.

Fixed timezone handling in scheduled alert suppression windows — suppressions now correctly respect the configured workspace timezone.