V2 Node Types: SQL-First Authoring

You can now build transformations by writing SQL directly in Coalesce. Node Type V2 lets you paste existing queries, author CTE-heavy logic, or drop in AI-generated SQL and get everything Coalesce offers in return: automatic column extraction, full DAG and column-level lineage, template-driven deployment, and governance. Configuration moves into the file itself via inline annotations like @insertStrategy("MERGE") or @isBusinessKey, keeping your transformation and its metadata in one place. V2 nodes work alongside your existing V1 nodes in the same pipeline, so you can migrate at your own pace, one node type at a time.

Learn more about V2 Node Types.

COA CLI: Local-First Pipeline Development

You can now build and run Coalesce pipelines entirely from your terminal. The COA CLI brings local-first development to Coalesce, letting you author .sql transformation nodes in your editor of choice, run them directly against your warehouse, and deploy to your environment without touching the UI. Install it via npm, point it at your warehouse credentials, and you have the full pipeline workflow at your fingertips: coa create generates your DDL, coa run populates your tables, and coa plan and coa deploy push changes to your Coalesce environment. COA works with your existing Git workflow and supports the same V2 SQL nodes, node types, and ref() dependency graph you already know. Whether you prefer coding locally, integrating with CI/CD, or just want faster iteration cycles, the COA CLI is how you bring that workflow to Coalesce.

Get coa version 7.33.
Read the documentation.

Updates

• Copilot Downstream Column Descriptions: Copilot now carries column descriptions into downstream columns it creates.

Packages

BigQuery

BigQuery Base Node Types 1.1.0: Added a Last Modified Comparison option to MERGE logic for Persistent Stage, Dimension, and Fact Nodes; UUID visibility moved into the Node definition for column visibility; package tags for marketplace visibility; documentation updates including a config matrix.

Databricks

Databricks Base Node Types 1.1.0: Zero-key Dimension support, a last-modified toggle on Dimension, Persistent Stage, and Fact, soft delete and hard delete on Dimension, documentation updates, and an updated get_clause() macro.

Snowflake

• Streams and Tasks 2.6.1: Replaces ALLOW_OVERLAPPING_EXECUTION with OVERLAP_POLICY, small UI fixes, and Git documentation updates with impact analysis for task-enabled Node types.

• Iceberg Tables 3.2.1: Error and success notification integrations, execute as user, suspend after failures, task auto-retry, overlapping task runs, expanded task configuration and schedule fields, serverless warehouse size limits for tasks, and related documentation for Iceberg external tables.

• Functional Node Types 4.3.2: Node Type switch support and documentation updates.

• Deferred Merge 2.0.3: Task notification integrations, execute as user, suspend after failures, auto-retry, overlapping runs, richer schedule and configuration fields, serverless task size bounds, and schedule unit expansion.

• Create or Alter package 1.6.1: Same task-focused improvements as Deferred Merge 2.0.3 for supported Node types.

• Cortex 4.0.0: Replaces deprecated !PREDICT with AI_EXTRACT, renames Document AI to AI EXTRACT, and updates configuration to match the new function.

Bug Fixes

• Databricks Token Authentication: Databricks token authentication now shows clearer errors to help with troubleshooting.

• setNode With Duplicate Table Names: setNode now resolves the correct table when two tables share a name but live in different storage locations.

• Presync Errors: Presync deploy errors show clearer messaging to help with troubleshooting.

• Workspace Wizard and Git Permissions: When Git credentials lack permission, the Workspace creation wizard keeps your entries and shows a specific error instead of a generic failure.

• Re-sync Columns and Descriptions: Re-syncing columns now picks up description changes from your data platform so your Node metadata stays current.

• Run Details Sorting: Sorting ascending or descending in run details works when some values are missing instead of mis-ordering rows.

Updates

• Pause and Resume for Scheduled Jobs: You can pause and resume scheduled jobs from the action menu instead of deleting them. When a Job is paused the next run shows Paused.

• Create Job Schedules Without an Immediate Run: When you create or edit a schedule, you can turn off the automatic run on save so the first execution waits for the next scheduled run.

• Faster Clones for Large Git Repositories: We’ve enhanced the Git cloning process during Workspace creation, significantly improving speed for large repositories.

• Add Sources Modal and Data Preview: The Add Sources flow now updates the Data Preview layout to match the Build tab, with tighter header spacing, rounded borders, improved background color, and consistent padding across the Builder.

• Copilot Data Quality Tests on Columns: When asked, Copilot can create data quality tests on single columns (for example, not null or unique checks) using the same test types you configure on Nodes.

• Copilot Context for Large Graphs: Copilot can now handle large graphs improving Copilots context window.

• SQL Node Support on Databricks: You can create, edit, validate, and deploy SQL Nodes in Databricks projects.

Packages

BigQuery

• BigQuery Base Node Types - Advanced Deploy 1.1.0: UUID visibility is driven from Node definition. Deploy metadata handles SET NOT NULL on BigQuery. Cluster keys compare by internal ID on redeploy so matches stay stable. Join metadata resolves correctly when macros use {% %} Jinja blocks.

Snowflake

• Base Node Types 1.6.0 : Adds the SQL Stage Node type.

• External Data Package 2.1.3: Copy Into, Snowpipe, and Inferschema parse input correctly when ESCAPE_UNENCLOSED_FIELD = NONE.

• Streams and Tasks 2.6.0: Supports task notification integrations, execute as user, suspend after failures, auto-retry, overlapping task runs, richer schedule and config fields, task comments, and serverless warehouse size limits for tasks.

Bug Fixes

• Consistent Graph Layout After Reload: Reloading the graph from the menu keeps the same layout as the initial load, with source Nodes aligned on the left as expected.

• System Columns After Node Type Switch: When you switch Node types through re-initialization, the mapping grid drops the old type’s system columns and only shows columns for the new type.

• SQL Node Columns Grid : Selected expressions, including aggregates, appear in the Columns grid even when you omit an AS alias. The grid fills in default column names where the database would.

• SQL Editor When Switching Nodes: Fixed intermittent Model is disposed errors from Monaco when switching between Nodes in the SQL editor.

• Presync Plan Scrolling: The presync deploy plan panel scrolls when the diff list is long, so rows at the bottom stay visible.

• Copilot and Object Descriptions: Copilot can read table, view, and column descriptions from the environment, including descriptions that come from attachments or that propagate between Nodes.

Updates

• SkipParsing Query Parameter for Nodes API: The Nodes GET endpoints now accept optional skipParsing=true, which skips parsing Node transforms for a faster response. You can use skipParsing=true on its own or together with detail=true. Both combinations are supported. When you pass detail=true, responses include full Node detail but omit parsed transform data, which helps integrations and dashboards that do not need that payload.

  • List Environment Nodes

  • List Workspace Nodes

  • Get Environment Node

  • Get Workspace Node

Updates

• SkipParsing Query Parameter for Nodes API: The Nodes GET endpoints now accept optional skipParsing=true, which skips parsing Node transforms for a faster response. You can use skipParsing=true on its own or together with detail=true. Both combinations are supported. When you pass detail=true, responses include full Node detail but omit parsed transform data, which helps integrations and dashboards that do not need that payload.

  • List Environment Nodes

  • List Workspace Nodes

  • Get Environment Node

  • Get Workspace Node

Bug Fixes

• CLI Refresh Performance Regression: Fixed an issue where running refresh jobs with coa CLI would run for too long. CLI refresh jobs now complete in expected time.

This release includes BigQuery OAuth, Copilot enhancements, Job Scheduler and Deploy updates, and bug fixes across the platform.

Updates

• Auto-Preview Control: You can now control whether Coalesce automatically fetches and displays data in the Data Viewer after a successful Node run. Set the default in Project Settings under the Preview Control tab, or override it for individual Nodes during your session. When disabled, Coalesce skips the preview query to reduce compute costs. Use Load Preview when you want to manually load data. Single-node runs respect your setting; multi-node runs never auto-fetch regardless of settings.

  • 

• Custom Instructions for Copilot: Organization administrators and project administrators can set custom instructions in Project Settings so Copilot generates code matching your team's naming and structural conventions.

• Copilot Node Config and Validation: Copilot now injects top Node configurations into AI context and validates generated configurations for better alignment with your Node definitions.

• OAuth Display of Authenticated User: BigQuery OAuth now displays who is authenticated after connection setup. The authenticated user's email appears next to the Authenticated badge in the onboarding wizard and Workspace and Environment Settings.

• Job Scheduler Test Failures on Deploy: The Deploy screen now shows whether refreshes contain any failed table tests.

• Update Credentials for Saved Jobs: You can now update credentials for the environment and the change will apply to every job.

Packages

Snowflake

• Semantic View Node 1.0.0: Initial release with Semantic View and Semantic Query node types for a business-friendly semantic layer on Snowflake

• Functional Node Types 4.3.1: Simplified Pivot mode addition,`get_clause()` macro changes to support keyword columns, and package tags updated.

• Incremental Loading 2.1.1: Node Type switching support.

• Create or Alter 1.6.0: Node Type Switch support and `get_clause()` macro support for keyword columns.

• Streams and Tasks 2.4.3: Node Type Switch support and package tags.

• Dynamic Tables 2.1.7: Node Type Switch support and package tags.

• Iceberg Tables 3.1.0: Node Type switch support and package tags.

• Base Node Types Advanced Deploy 2.5.0: `get_clause()` macro supports column names with keywords in the join tab for Dim, Pstage, and Fact, and handles single and block comments.

• Base Node Types 1.5.0: `get_clause()` macro supports column names with keywords in the join tab for Dim, Pstage, and Fact, and handles single and block comments.

Bug Fixes

• Nodes with Forward Slash in Name: Fixed an issue where nodes named with a`/` character were not visible in the browser, subgraphs, or jobs.

• Self-Service Password Reset Window: Fixed an issue where people provisioned with "set user password" unchecked were blocked from self-service password reset if they missed the 24-hour window.

• Stage Node Multisource Toggle: Fixed an issue where turning Multisource OFF on a Stage Node did not reset the Insert Strategy from UNION back to Insert. It now resets correctly.

• Refresh Triggered Twice: Fixed an issue where pressing the refresh button in the Workspace once could cause the query to run twice in Snowflake.

• Databricks Authentication: Fixed an issue where Databricks authentication could fail after a while when using Token and U2M.

• Git Merge Cancellation: Fixed an issue where merging specific branches in the Git modal failed with a "Merge from remotes/origin cancelled" error.

Updates

• Copilot File Upload: Copilot now accepts file uploads and drag and drop. This makes it easier to convert existing SQL.

• Copilot Organization Instructions: Organization administrators can now set instructions for Copilot such as naming conventions and architectural patterns. Configure Copilot in the Organization Settings**.**

• BigQuery Run Error Visibility: BigQuery run errors are now clearly surfaced in the UI. Rows with errors are highlighted with an icon, and a badge at the top displays detailed error information on hover. The modal distinguishes between empty but accessible schemas and inaccessible schemas that return errors, and tables with no rows show a clear placeholder. Specific handling is included for permission denied, invalid or missing storage mappings or schemas, and general errors, making troubleshooting significantly easier and more self-serve.

• Run Page Render Failure Visibility: Failures that occur during the rendering stage of a refresh are now surfaced directly on the Run page, similar to other run failures. Errors collected by the renderer are no longer limited to logs or the Deploy Wizard.

• Workspace Settings Dropdown Icons: Added icons for Copy Objects, Duplicate Settings, and Delete Workspace to make actions easier to recognize at a glance.

• Readable Workspace Creation Dates: Workspace creation dates now use a new format, for example, “August 9, 2024” instead of “8/9/24,” reducing regional ambiguity.

• Run Details Full Step Visibility: Run Details now display queued, running, successful, failed, and skipped stages. For both deploys and refreshes, all Node steps are visible, even before they start, making it easier to track overall progress.

• Redesigned Copilot Composer: The Copilot composer features a refreshed layout with a dedicated input area, separate actions bar, context usage indicator, relocated Allow edits toggle, square buttons with rounded corners, and an auto-resizing textarea for a more natural typing experience.

• Builder Deep Links with Node ID: Builder URLs now include the Node ID at /projects/<project-id>/workspaces/<workspace-id>/build/nodes/<node-id>, enabling reliable deep linking directly to specific Nodes from documentation, tickets, or shared links.

Packages

BigQuery

• Base Node Types - Advanced Deploy 1.0.0: Initial release.

Databricks

• External Data Package 1.0.1: Added support for loading Excel file format.

• Lakeflow Declarative Pipelines (LDP) Node Type 1.1.0: Added support for AVRO, ORC, TEXT, and XML file formats. Advanced file format options are now available across all supported file formats.

Snowflake

• External Data Package 2.1.2: Added support for API key authentication, including a toggle to forcibly add credentials in the request header.

• Incremental Loading 2.1.0: Added data quality support and Test Passed/Failed records node type.

• Base Node Types Advanced Deploy 2.4.0: Added node type switch support, package tags, and node config matrix with business summary in readme. Fixed truncate description in readme.

• Materialized View 2.0.1: Added node type switch support and package tags. Added node type switch details in readme.

• Create or Alter 1.5.0: Added Last Modified Comparison toggle.

• Data Quality 2.2.0: Added node type switch support in Advanced Deploy Templates and package tags. Fixed truncate description in readme.

Bug Fixes

• Authentication Type Dropdown Behavior: Fixed an issue where the Authentication Type dropdown caret in Workspace or Environment Settings under User Credentials did not reliably open. Alignment issues in the tab were also corrected.

• Refresh Results Missing Nodes: SQL Nodes that do not execute SQL, such as views now appear in refresh results, ensuring the results list accurately reflects all participating nodes.

• Storage Mapping on Partial Deployment Failures: When deployments fail partway through, successfully deployed nodes now correctly write their location and mapping to environment metadata. This prevents environments from entering a blocked state and reduces manual intervention.

• Encrypted SSH Key Reliability: Improved the SSH private key experience with clearer input states, real-time validation, encrypted key detection, and required passphrase enforcement. Encrypted keys now work reliably across sessions.

• listRuns API Failure Reasons: The listRuns API now returns runFailureReason, including when detail: true is set, enabling external dashboards and monitoring tools to surface failure reasons without additional per-run API calls.

• Logging Redaction Performance: Optimized the redactSensitiveData logging layer to prevent long delays when processing large error payloads, resulting in more responsive query runs and error handling.

BigQuery in Private Preview

Big Query support in Coalesce Transform is now available in private preview. Reach out to your account team to enable it.

UI Changes

• Project and Workspace ID in URL — URLs now include the Project and Workspace ID, making it easy to share direct links to specific Workspaces. For example /projects/<project-id>/workspaces/<workspace-id>/build

• Redesigned Copilot Chat History — The Copilot chat history now uses a card-style design with improved hover states. History items show "X messages · Xh ago" instead of full timestamps, and the delete button only appears on hover for a cleaner look.

Copilot

• Message Feedback — You can now provide feedback on specific Copilot messages using like and dislike buttons, helping improve AI response quality.

• Improved Abort Functionality — The Copilot abort button now properly stops the UI spinner when you cancel a request. Previously, the spinner continued even after aborting.

Packages

BigQuery

• Base Node Types 1.0.0 is available — Sign up for our private preview by contacting your account manager.

Bug Fixes

• Preserved Thread Titles — Fixed an issue where Copilot thread titles were overwritten when starting a new chat. Existing thread titles are now preserved correctly.

• Runtime Parameter Warning Consistency — Fixed an issue in the Node Type Editor where after correcting invalid runtime parameters and saving, the error popup wouldn't automatically clear until you left and re-entered the editor. The warning now updates correctly when parameters are fixed.

• Copilot Text Overflow — Fixed an issue where text content in Copilot responses overflowed horizontally beyond the shaded content area. Content now displays within proper boundaries.

• Subgraph Tab Handling — Fixed an issue where deleting a Subgraph while its tab was active resulted in a blank white tab. The interface now handles Subgraph deletion gracefully.

• Code Editor Line Numbers — Fixed a CSS conflict where duplicate line numbers would appear.

Navigating complex data pipelines just got easier. We've introduced several enhancements to the Node graph view that help you visualize, filter, and explore your DAGs more efficiently. Whether you're tracking down dependencies or understanding your pipeline's structure, these tools work together to give you better control over how you view and interact with your data transformations.

Mini Map

The Mini Map gives you a bird's-eye view of your entire graph at once. Instead of panning around to understand your DAG's size and structure, you can now see the complete layout in a compact overlay. Click any area on the Mini Map to jump directly to that section of your graph.

The colored Nodes in the Mini Map match the Node types in your main view, making it easy to identify sources and targets as you navigate.

Filter Your DAG

You can now refine your view by dragging Nodes directly from the Node list into the filter. Drag a Node into the Include area to focus on it, or into Exclude to hide it from view. This makes filtering faster and keeps your graph focused on what matters most.

Explore Lineage From Any Direction

You now have more control over how you expand lineage directly from the context menu:

• Upstream: Shows the data sources that flow into your Node.

• Downstream: Reveals where your data goes next.

• Related: Displays both directions at once for a complete picture.

This gives you more control over how much lineage you explore at a time.

Node Expansion

Click the plus icon on any Node to expand its lineage one step at a time. If a plus icon appears, there's another Node connected. If the icon isn't there, you've reached the end of that path. You can expand both upstream and downstream connections this way.

The filter automatically updates to show Custom when you add Nodes individually.

Spotlight Navigation

Spotlight helps you navigate large DAGs 4 Nodes at a time. If you have 12 Nodes, Spotlight expands them in manageable chunks. Click Spotlight again to move forward in the direction you want, 4 Nodes at a time.

Collapsed Nodes by Default

All Nodes are now collapsed by default to use screen space more efficiently. If you want to see all the columns in a Node, expand it.

Storage Location in Graph View

Storage Locations now appear directly in the graph view, making it easier to identify where each Node originates.

Alphabetically Sorted Nodes

When you right-click on a Node and select Add Node, all available Nodes now appear in alphabetical order. This matches the sorting order in the left-hand Node list, making it consistent across the interface.

View Column Lineage From the Column Grid

You can now view the lineage of any column directly from the Column Grid. Right-click on a column and select View Lineage to trace where it originates, how it transforms, and where it's used downstream.

Columns are collapsed by default but can be expanded to show full mappings. You can edit columns directly and use Propagation to apply changes consistently across connected Nodes.

Get a quick tour by watching this video or visiting our documentation.

These features are currently in Private Preview. Please reach out to your Coalesce account manager to sign up.

We've made significant improvements to Node type switching in Coalesce. When changing a Node's type (for example, from View to Dimension), Coalesce now performs a complete metadata reinitialization based on the target type definition. This ensures your Node fully adopts the target type's structure while intelligently preserving your work.

You can also now use the bulk edit to switch Nodes Types across multiple Nodes at once. Each Node displays its own preview, showing the specific metadata changes that will occur for that Node.