Version 0.5.3
stable
A saved-data and custom-view release that adds optional Bunny Cloud Sync across computers, adds Confluent Cloud retained-topic metrics, centralizes durable user content in the Java backend, unifies calculated columns with saved layouts, adds guided array and AI-assisted column building, expands SL with find(), flatten(), and split(), hardens packaged startup with collision-free local API ports and clearer service progress, and polishes topic operations, connection recovery, and data-table consistency.
What changed
- Moved calculated columns into the custom view being edited so normal, lookup, and calculated columns share one selected-column list, ordering, preview, and saved preset.
- Added an Array Data workflow that discovers direct and nested arrays, distinguishes object and primitive arrays, and creates count, sum, average, minimum, maximum, and find-value columns.
- Added SL find(predicate) for returning the first matching array item, flatten() for navigating arrays nested beneath parent arrays, and split(delimiter) for turning literal-delimited text into arrays for element-level work and counting.
- Expanded and aligned SL validation, evaluation, autocomplete, syntax highlighting, help content, and AI method inventories for find(), flatten(), split(), and the complete supported function set, while favoring the simplest valid expression.
- Added view-level AI Assist that reviews loaded message samples, proposes normal and calculated columns with rationale, and requires user selection before applying changes.
- Added provider-aware structured view-column generation for ChatGPT/OpenAI, Claude, Gemini, and OpenRouter, including backend validation of generated SL expressions.
- Added batched calculated-column evaluation for visible result rows and integrated calculated values into both result lists and message-detail views.
- Hardened JSON mapping inference so numeric and boolean fields are never selected as column identifiers and low-confidence array mappings are rejected instead of guessed.
- Manual topic reloads now refresh topic disk-size data immediately while leaving message-count refresh on its configured schedule.
- Upgraded the Edit View Add topic picker with twice-wider controls and topic-filter syntax for boolean operators, negation, grouping, wildcards, and inline syntax errors.
- Rebuilt all Connection Manager dropdowns as themed, keyboard-accessible menus with neutral hover states, selected checks, responsive portal positioning, consistent input heights, and no native Windows blue selection.
- Refined Connection Manager recovery by moving visibility and help actions into the title bar, keeping the recovery controls together on the left, and widening the Backup and Connection pickers.
- Standardized muted uppercase column headings across topic, partition, consumer, broker-config, connector, ACL, task, criteria-preview, connection, help, and Create Topic configuration tables.
- Added icons to the Topic List filter command menu and clarified record-count labels with message-oriented wording.
- Centralized Live Views, scripts, playbooks, sample messages, and saved message-list view presets in typed Java REST repositories under the default ~/.kafka-streamyard directory, replacing their former Electron or browser-storage ownership.
- Added synchronized atomic JSON writes with validation, normalization, canonical responses, per-connection Live View isolation, cross-window refresh notifications, and direct repository unit coverage.
- Made playbook, Sample message builder, and Live View writes await backend persistence, surface save and delete failures, and restore persisted Views after a failed request.
- Made saved message-list views use immutable GUID identities so duplicate display names remain distinct, active selections restore the exact server record, and same-named views keep independent column widths.
- Added optional Bunny Saved Data Cloud Sync for criteria and topic filters, message-list views, Live Views, playbooks and Data samples, with encrypted AccessKey storage, checksum-based startup, focus, and periodic synchronization, local-first queued uploads, backups before cloud replacement, status notifications, and manual Sync now controls while preserving normal local-only behavior whenever sync is disabled.
- Made packaged macOS, Windows, and Linux applications request an available loopback port for the local Java backend at each startup, avoiding conflicts with other software while keeping development on port 8082 by default and supporting fixed-port overrides through KAFKA_STREAMYARD_API_PORT.
- Improved startup feedback with separate Starting services and Connecting phases, propagated the selected backend port to HTTP and WebSocket clients, reported the active port in the Debug Log, and added Technical help for finding and overriding it.
- Reorganized Settings UI into clearly described Interface, Integrations, Performance & storage, and Diagnostics sections with consistent nested panels, a fixed action footer, and reliable scrolling for long configurations.
- Expanded Debug Log diagnostics for topic-list refreshes to show whether each fetch requests disk-size and record-count metrics.
- Made Live View topic-flow palettes manually collapsible into a top-left chart control, while locked diagrams hide the palette, resize handle, hover previews, and reopen control so read-only charts use the full available width.
- Added optional Confluent Cloud Metrics API integration using dedicated resource-management credentials with MetricsViewer access, surfacing delayed retained topic sizes in the Topic List, Cluster Dashboard, Topic Information, and Partition Information while preserving broker-reported disk usage for other clusters.
- Made Empty Topic reliable for compact-only topics by temporarily enabling delete cleanup, retrying while the policy propagates, clearing only the selected partitions, and restoring the original cleanup policy afterward.
- Kept Create Topic drafts stable while Live View messages arrive so live counters, result refreshes, and broker metadata updates no longer wipe in-progress names or configuration edits.
- Added a nearly full-screen in-place Data Sample editor launched from the Message Details sample area in both right and bottom dock layouts, replacing the one-click Update action with an explicit Edit flow while keeping sample creation separate.
- Reused the full sample editing workspace in update-only mode with name, parent, tags, headers, session variables, key, message, and AI Assist editing; Test, Close, and Save footer actions; Cmd/Ctrl+S support; and guarded handling of unsaved changes.