2026 W39

Bruin Cloud

Dashboards

• New Image tables and widgets → In dashboard YAML, set a column or widget to type: image, map src, title, caption, and alt fields, and choose fit: contain or cover to render URL-backed images. Also available in DAC.
• Improved Query result previews → Select a SQL chart or table widget → Results to inspect the raw rows behind a visualization without leaving the dashboard.
• New Tabbed widgets → In dashboard YAML, set type: tabs to combine complete widgets in one container, with a separate query, encoding, notes, and download view for each tab. Also available in DAC.
• New Tab-scoped filters → In dashboard YAML, add tab to a filter to show it in that tab's filter bar; applying a filter now refreshes only widgets that use it. Also available in DAC.
• New Dashboard notes → In dashboard YAML, define dimension-scoped notes and reuse them across widgets; publish the dashboard before agents use new or changed note definitions. Also available in DAC.
• New Drag-and-drop widget ordering → Open a dashboard in edit mode and drag widgets directly on the canvas, including widgets inside tabs.
• Improved Full-width widget defaults → Widgets without an explicit width now use all 12 columns in Cloud and repo-defined dashboards.
• Improved Dashboard filter validation → Dashboard filters validate declared values and reuse a shared options cache, so dependent dropdowns update more reliably and invalid selections fail clearly.
• Fix Dashboard editing fixes → Deleting a widget now reflows its row, row menus stay within the viewport, and move-to-folder remains available from the menu.
• Improved YAML editor indentation → Pressing Enter or pasting a multiline block keeps nested mappings and list mappings aligned with the current indentation.

Pipelines and runs

• New Pipeline tags and filters → In Pipelines, use tag badges, search by owner or tags, include unowned pipelines, and combine the filters with text search.
• Improved Next-run and status details → Pipeline and run pages now distinguish scheduled, continuous, unscheduled, loading, and unavailable states, keep statuses visible beside long run IDs, and show blocking or non-blocking check failures.
• Fix Run task status updates → Cloud run actions now mark task instances by default when a run is marked successful or failed, including bulk, API, backfill, and MCP actions.
• Fix Backfill run durations → Marked backfill runs without a real execution start no longer display inflated multi-day durations.

Data apps

• New Access controls → Use private or team visibility and grant individual viewer or editor access from the Share menu in the data-app list or builder.

Scheduled agents

• New Pipeline success triggers → In a scheduled agent's Schedule section, select one or more pipelines to start the agent when any scheduled run succeeds. Pipeline triggers work independently of cron schedules and Cloud notifications.
• Improved Advanced trigger conditions → Open Advanced to customize the event types and conditions used by a scheduled-agent trigger, while simple pipeline selections stay focused on scheduled runs.
• Improved Run history and trigger details → Browse scheduled-agent runs with a total count and Previous/Next pagination, and see whether each run started from a schedule, a manual action, or a pipeline.

Connections and GitHub

• Improved GCP service account uploads → In GCP connection forms and connection sets, choose Upload JSON file to populate the editable service-account field, with inline validation for invalid files.
• New Multiple GitHub accounts → Connect personal and organization GitHub installations together, filter repositories by account, and use the selected project's installation for GitOps writes.
• Improved GitHub connection controls → Team Settings now shows one clear GitHub connection status and one Refresh action that re-discovers installations without a browser round-trip.
• Fix Connection validation messages → Snowflake forms support password-only and key-pair authentication again, and rejected connection details are distinguished from an unreachable validation service.

Notifications and agents

• Improved Run-tag filters → Notification rules can filter on run tags such as manual, scheduled, and backfill, plus pipeline and asset owners and tags, with suggestions scoped to the selected projects, pipelines, and assets.
• Improved Pipeline notification emails → Notification emails now use clearer success, warning, and failure cards with event details and direct links to the relevant run, asset, or quality check.
• Improved Permission guidance → Capability-gated dashboard, scheduled-run, and agent-chat pages now explain which Cloud CLI permission is missing and link directly to the settings needed to enable it.
• Improved Custom MCP secret handling → The custom MCP server form now directs secrets to Environment variables with ${VAR} references and keeps argument values out of saved labels.
• Improved Warehouse-first data answers → When MCP servers are attached, agents now check the warehouse first for data questions and use MCP directly for actions or offered live pulls.

Bruin CLI

• New Cloud browser login → Run bruin cloud login to approve access in your browser, choose teams and a permission level, set an expiration, and store the credential locally; use bruin auth status or bruin logout to manage it.
• New Scheduled-agent pipeline triggers → Use bruin cloud scheduled-agents pipeline-trigger get|set|delete to configure a scheduled agent's pipeline-success trigger with standard Cloud authentication and JSON output.
• New Asset-level run status → Use bruin cloud runs mark-status --asset to mark one asset instance successful or failed without changing the rest of the run.
• Improved ClickHouse SCD2 and table options → Track row history with scd2_by_time or scd2_by_column materializations, and set table engines, sort keys, TTL, and partitions in ClickHouse asset definitions. Materializations can also run across a cluster, and connections accept query settings.
• New HubSpot reverse ETL options → HubSpot reverse ETL assets accept incremental_strategy: update to change existing records only and delete to archive them. Set reject_mode and write_nulls to control rejected rows and empty fields.
• Improved Filtered pipeline runs → bruin run now validates dependencies against the selected assets when using --tag or --exclude-tag, so excluded assets no longer block discovery runs unnecessarily.
• Fix Faster low-volume logs → Secret masking no longer holds ordinary low-volume CLI output for minutes before flushing.
• New Semantic note definitions → Define reusable, dimension-aware notes in semantic models with validated dimension options.

Bruin VS Code

• Fix Readable table-diff errors → The Table Diff panel now shows the actual CLI error message instead of the raw JSON error object when a diff fails.

DAC

• Fix Dark export backgrounds → PNG and PDF dashboard exports now preserve the active dark theme background.

ingestr

• New HubSpot and CleverTap reverse ETL → Write warehouse rows to HubSpot CRM objects and associations or CleverTap profiles and events through reverse ETL destinations, with matching and reject handling for each API.
• Improved Safer column renaming → --columns renames now preserve values for schema-less sources, and automatic naming no longer mutates source schemas or primary keys during single-table and CDC runs.
• Fix Fail-fast reverse ETL writes → A reverse ETL write error with fail_fast now stops the source read immediately instead of consuming the rest of the stream.
• Improved BigQuery staging location → When a BigQuery destination omits location, _bruin_staging now follows the target dataset's region. A mismatched existing staging dataset fails fast with an actionable error.
• Fix Quieter Snowflake first runs → Handled destination-table probes no longer print misleading driver errors during normal ingestion; use --debug when you need the driver's details.