Subscribe to updatesSnowflake OAuth via PrivateLink: Snowflake OAuth connections configured with USE_PRIVATELINK_FOR_AUTHORIZATION_ENDPOINT now authenticate successfully without issuer mismatch errors.
Snowflake OAuth: Fixed an issue where Snowflake OAuth reconnection failed with an issuer mismatch for accounts with uppercase identifiers.
SQL (V2) Nodes on BigQuery: BigQuery workspaces now support SQL (V2) Nodes and version selection.
SQL (V2) Annotations: Built-in annotations (@id, @nodeType, @materializationType) are validated in the editor.
Desktop Parameters: Local workspaces support runtime parameters in workspace.yml.
Desktop Drag & Drop: Column creation and drag-and-drop column support in local dev mode.
Branch Validation: Workspace creation surfaces branch validation errors in an inline drawer.
Deploy Polling: Schedule polling pauses when tabs are hidden to reduce background requests.
Plan Recreate Logging: Execution plans log changed properties when dropping and recreating nodes.
Release Pipeline: Streamlined staging build and deployment workflows.
Node Duplication: Duplicated nodes remap column dropdown and tabular config references correctly.
SQL (V2) Node Fallback: Missing node types on SQL (V2) nodes show an inline banner instead of crashing.
Snowflake OAuth: Resolved iss mismatch errors during OAuth authorization callbacks.
Subdomain Feature Flags: Backend flag evaluations include customer subdomains so targeting rules match browser evaluations.
Auth0 Login Routing: Auth0 requests explicitly route to username/password auth to avoid unexpected IdP redirects.
Cross-Org User Creation: Clarified errors when inviting users across orgs with differing Auth0 configurations.
SQL (V2) Node Annotations: SQL (V2) Node Types now declare supported annotations directly in their spec, with an Annotations tab in the editor for easy discovery and copyable snippets.
Desktop App: A preview version of the Coalesce Desktop application is now available, which supports automatic background updates and customizable update preferences.
Local Project Association & Project ID: Workspaces now track Project ID in data.yml, and coa init associates projects with cloud workspaces by default. A one-time Git diff adding this value is expected and safe to commit.
Column Metadata: Column descriptions, default values, and nullability constraints now hydrate directly into column definitions.
Job Schedules: Paused schedule indicators are now displayed directly within job details.
New Login Experience: New login experience is being rolled out in phases starting with this release.
Deploy Dashboard Polling: Schedule polling on the Deploy dashboard pauses when the browser tab is hidden and refreshes at a lower frequency to prevent unnecessary server load.
V2 SQL Nodes SELECT DISTINCT Mapping: Correctly records SELECT DISTINCT clauses on node source mappings during SQL parsing.
Subgraph Selection Freezes: Selecting nodes on large subgraph canvases executes in a single dispatch pass, removing multi-second UI freezes.
Monaco Editor Command Palette: Restored Command Palette (F1 / context menu) and line navigation across all Monaco SQL editors.
Subgraph Performance: Resolved performance bottlenecks and rendering lag across large subgraphs.
Editor Command Menu: Restored the context command menu in the SQL editor.
Graph Error Handling: Fixed an issue where unhandled error objects caused crashes in graph and grid views.
Login Redirects: Resolved navigation issues when accessing the login page directly.
Plan Summary: Fixed mislabeled summary metrics during plan generation.
Auth0 Login Redirection: Navigating directly to the /login route in Auth0-enabled organizations now cleanly redirects to Auth0 hosted sign-in while preserving invitation and deep-link parameters.
Problem Scanner False Positives: Resolved false-positive circular dependency warnings in the Problem Scanner and Builder UI for valid query patterns (such as delta loads referencing target tables).
Copilot Long Responses: Copilot on Opus now uses its full 128K output limit, so large Node creation no longer gets cut off mid-response.
Copilot Tool Cards: Tool cards resolve when a response ends instead of spinning forever, and truncated responses are called out in chat.
Copilot Conversation Recovery: A failed tool call no longer breaks the rest of the conversation, and a failed Node creation cleans up after itself instead of leaving an empty Node behind.
Databricks OAuth on Azure: OAuth M2M connections to commercial Azure Databricks workspaces authenticate correctly. (Shipped in 7.39.3.)
Source Node Mapping Column Sort and Filter: On Source Nodes, you can sort Mapping tab columns alphabetically or by data type, and filter by column name. This makes wide Source tables easier to navigate.
Column-Level Node Type Config Items: V2 Node Types can add toggles, dropdowns, text boxes, and tables as per-column options, not only node-wide settings. Each column keeps its own value in SQL annotations and Node metadata.
Circular Dependency Problem Scanner: The problem scanner warns when your graph has a circular dependency, so you can catch cycles before plan or run surfaces a cycle-detected error.
BigQuery Local COA CLI: The local `coa` CLI supports BigQuery Service Account profiles for doctor, init, validate, create, and run paths that were previously Snowflake-only.
Node Type Name on Node API: Node API responses include `nodeTypeName` so callers get a friendly Node Type display name for custom and package types without changing the existing `nodeType` field.
View Plan for Deploy Runs: Anyone authorized to view a deploy run can open View Plan for that run's executed plan. Plan access is no longer limited to the user who created the plan.
Create or Alter package 1.6.3: Removes the Treat NULL as Current Timestamp toggle from Create or Alter Dimension, Fact, and Persistent Nodes. When Last Modified is enabled, a validation stage stops execution if NULL values exist in the selected timestamp column.
Cortex 4.2.0: Renames AI_EXTRACT to Cortex File Processor, adds `AI_PARSE_DOCUMENT` and `AI_TRANSCRIBE` support on Cortex File Processor, and migrates legacy `SNOWFLAKE.CORTEX` functions to AISQL `AI_*` functions on the LLM Node.
Column-Valued Annotation Serialization: Column selectors in V2 annotations such as `@orderbycolumn` and `@lastModifiedColumn` write the column name instead of `[object Object]`, so committed SQL and redeploy comparisons stay correct.
Hidden Config Options When Parent Toggle Is Off: When a parent config toggle is off, dependent config options are hidden and their annotations are stripped from the SQL Editor so you only see settings that apply.
Add a Column With Sorted Grid: After you sort the column grid, Add a Column focuses the newly created column instead of the last sorted row.
Subgraph Double-Click in Scrollable List: Double-clicking a subgraph in a scrollable Subgraphs list opens the subgraph you selected, including when the Node column panel is open.
Corrupt Metadata Errors in Column Editor: When corrupt column metadata blocks a transform save, the Column Editor shows an error instead of failing silently with no feedback.
Run Preview Hotkey During In-Flight Runs: Pressing Ctrl+Enter or Cmd+Enter while a Node run is already in progress no longer throws an unhandled "Nodes are already running" error.
Bulk Source Column Remap With Missing Columns: Bulk remap source columns returns a clear error when the source step has no columns, instead of crashing.
Snowflake View GROUP BY ALL Toggle: We removed the Group By All toggle from Snowflake View Nodes. You can still write GROUP BY ALL directly in your SQL.
V2 SQL Syntax Highlighting: The V2 Node editor highlights SQL keywords, types, functions, comments, Coalesce annotations, and Jinja template delimiters so SQL is easier to read while you edit.
Copilot Graph Search and Explore: Copilot can search your Workspace graph by Node name, fully qualified name, or description, and can run a read-only Explore subagent that investigates with read-only tools and returns a short digest in chat.
Copilot Opus Model Option: Copilot now has Fast and Smart effort modes, adaptive thinking, and read-only-by-default steering for mutations.
Snowflake View GROUP BY ALL: Snowflake View Nodes include a GROUP BY ALL configuration toggle. The toggle is mutually exclusive with Select Distinct.
COA Dry-Run Preview Status: `coa run --dry-run` distinguishes Nodes with no DML to preview (streams, dynamic tables, and views) from real failures, so managed objects no longer appear as red errors when the pipeline is healthy.
Run Details Step Duration: Run details show step durations in seconds, minutes, or hours instead of raw milliseconds, and sorting by duration still works.
Snowflake Comments During Plan: During plan, Coalesce can load existing Snowflake table and column comments from the warehouse. This supports externally managed descriptions without changing how you edit descriptions in the UI.
Create or Alter package 1.6.2: Adds a delete option to the Create or Alter Persistent Stage Node Type.
Base Node Types Advanced Deploy 3.0.0: Adds the External Sync Node Type.
Business Key and Change Tracking Git Sync: Editing Business Keys or Change Tracking after the initial commit now updates the generated SQL so Git tracks the change and redeploy carries the new configuration to other Environments.
Copilot Storage Location Instructions: Copilot honors explicit storage location requests instead of substituting the dominant Workspace pattern, and rejects unknown`ref()` locations with a clear suggestion when possible.
Node Right-Click Menu Visibility: The graph Node context menu keeps a consistent width and shows all options, including Delete Node, without scrolling when the Node is selected.
SQL Parse Errors at Plan Time: When active SQL fails to parse, plan and deploy surface the parse error instead of compiling stale Node metadata.
Account URL Paste Handling: Pasting a full Snowflake account URL into Workspace or Environment Account URL fields strips the`https://` prefix and`.snowflakecomputing.com` suffix as you paste or type, so the saved value is only the account identifier.
Create Branch with Duplicate Column IDs: Creating a new branch from a commit that contains duplicate column IDs no longer partially checks out the branch.