Changelog

Changed

• AI: Added execution plan analysis.
• AI: Added BOYK/local LLM support. Configure any OpenAI-compatible endpoint in settings under dbcode.ai.customModel. #816
• Results: Added setting to control zebra striping in the grid.
• MCP: Added configurable OAuth token expiry (dbcode.ai.mcp.tokenExpiry), default 1 day, max 90 days. #960

Fixed

• MySQL: Fixed MariaDB derivative detection failing when @@GLOBAL.version_comment doesn’t contain “mariadb” (e.g., Debian packages). #968
• SQLite: Fixed truncate using unsupported TRUNCATE TABLE syntax, now uses DELETE FROM. #965
• ClickHouse: Fixed query cancellation not killing the running query on the server. #962
• PostgreSQL: Fixed backup/restore to use local tools first if on the path. #961

Changed

• Missing WHERE Detection: Warns before executing DELETE or UPDATE statements missing a WHERE clause. Configurable per role (allow/ask/deny). Available once per session for Core users, continuously for Pro.

Fixed

• Results: Fixed right-click Filter By/Exclude not working for dates, booleans, and null values in the results grid. Also fixed context menu filters not triggering server-side queries in paging mode. #957
• Results: Fixed Exclude filter to include filtering for null values. #957
• Snowflake: Fixed introspection failing due to unquoted INCREMENT reserved keyword in sequences query. #956
• Snowflake: Fixed regular tables not appearing in the tree when hybrid, transient, or dynamic tables exist.
• Execution: Fixed standalone BEGIN/END blocks in SQL Server and Oracle grouping subsequent statements into one block. #958

Changed

• Join: Join tables from the DB Explorer or results with configurable join type and column mapping. #849
• Union: Union tables from the DB Explorer or results across tabs and connections.

Fixed

• Completions: ‘Select a connection’ prompt now only appears once per session instead of on every keystroke.
• BigQuery: Fixed Script CREATE (DDL) for functions. #955
• Notebook: Fixed cell editor losing focus after executing a cell with Ctrl+Enter.
• MongoDB: Fixed delete and update failing on documents with Binary/UUID _id fields (e.g. CosmosDB). #954

Fixed

• Results: Fixed column header Ctrl/Cmd+click and Shift+click selection not working correctly for charting and copy. #921
• Alter Table: Fixed empty columns generating invalid SQL (ALTER TABLE ... ADD ;). #916
• BigQuery: Fixed RANGE data type displaying as [object Object] instead of [start, end). #953
• Oracle: Fixed introspection failing when user tables shadow dictionary views (e.g. ALL_TABLES). #947
• MCP/Copilot: Fixed extension crash when AI tools return very large responses (e.g. Redis with large values). Redis bulk key listings now return metadata only, with full data available for single key lookups.

Changed

• Sequences: Added introspection and DDL generation (PostgreSQL, Oracle, MSSQL, DuckDB, DB2, Snowflake, MariaDB).
• Fabric: Added as stand alone connection type.
• Results: Added display density toggle (compact/normal/comfortable) in top right corner.
• DuckLake: Made data path optional for existing catalogs. Added Override Data Path option.
• Releases: Added tagged releases to public repo. #950

Fixed

• Execution: Fixed error line numbers in results panel not matching actual file position when executing from code lens or selection. #950
• DuckLake: Improved Windows path normalization (lowercase drive letter, trailing slash).

Changed

• Access: Removed from preview, added support for views (Access Queries), indexes, foreign keys, and read-only mode. #546
• H2: Removed preview badge - H2 Database driver is now stable. #371
• Results: Added chart toolbar button with dropdown for quick chart creation from selection, with persistent chart type preference (Ctrl/Cmd+G). #921

Fixed

• Autocomplete: Fixed keyword sort priority so SQL keywords (SELECT, etc.) appear before procedures/functions in completion results. #948
• Redis: Fixed raw commands (KEYS, SCAN) still blocking the extension, changed abort signal handling. #946
• DuckLake: Fixed Windows connection failure by normalizing backslashes to forward slashes in local catalog and data paths.

Changed

• DuckLake: In preview for connecting to DuckLake catalogs with local or database-backed metadata and local/S3/GCS data storage.
• Iceberg: In preview for connecting to Apache Iceberg catalogs via AWS Glue or S3 Tables.
• Formatting: Added preserveShortStatements setting to skip formatting for short SQL statements, leaving them exactly as written.
• Connections: Added “Editor Connection Idle Timeout” setting to auto-release editor connections after a period of inactivity. Useful for serverless databases to prevent unexpected billing from keepalive queries. If there are uncommitted changes, a warning notification is shown with a grace period before auto-rolling back.

Fixed

• Alter Table: Fixed alter table and hover actions not working. #928
• Redis: Fixed MCP queries hanging and blocking the extension on large scans by yielding and configurable max scan limit. #946

Changed

• SQLite: Updated better-sqlite3-multiple-ciphers to 12.6.2. #936
• DuckDB: Updated to version 1.4.4. #943
• SQL Server: Added Read Only support for AlwaysOn Availability Group routing via ApplicationIntent=ReadOnly. #940

Fixed

• BigQuery: Fixed stale credentials not being refreshed when re-authenticating via gcloud auth application-default login.
• BigQuery: Added clear error message for GCP RAPT session expiry instead of showing raw JSON error.
• DB2: Fixed SSL certificate paths not being passed correctly. #945
• DuckDB: Fixed “File Search Path” setting not applying to output operations (COPY TO, EXPORT). #944
• Grid: Fixed boolean column filtering generating invalid SQL.
• Editor: Execute shortcuts (Cmd/Win+Enter, Shift+Cmd/Win+Enter, Alt+Cmd/Win+Enter) now run the selected SQL when editing database objects (stored procedures, functions, triggers, etc), instead of applying the whole object.

Fixed

• Procedures/functions: Fixed comments above stored procedures being lost when saving in SQL Server.

Changed

• Data Compare: Right click a tab in the panel, or table in the tree, to compare data. “Compare data with…” menu option, direction toggle, and ability to generate sync scripts or apply changes directly.
• BigQuery: Real-time progress reporting for stored procedures - step completion status now displays in the output log while script jobs execute. #924
• Results: Tab reuse now reuses the currently active tab instead of the last tab. #921
• Results: Hold Alt while executing (Alt+Cmd+Enter / Alt+Windows+Enter) to combine multiple result sets into a single tab. Tabs can also be combined by dragging one onto another or Ctrl/Cmd+clicking to multi-select and using “Stack Selected” from context menu.
• Results: Added “Stack with…” right-click menu option for tab combining with visual selection mode.

Fixed

• Results: Tab reuse now reuses the currently active tab instead of the last tab.
• Results: Fixed output log not showing all recordset rows on first execution.
• Favorites: Fixed favorites disappearing when adding items while a filter was active in the connections tree.
• ERD: Fixed export to PNG/PDF failing with “Failed to fetch” error. #931
• ERD: Fixed export to web page not rendering tables for schema-less databases.

Changed

• BigQuery: Output log now displays individual SQL statements for each step when executing stored procedures. #924
• Results: Output log now has clickable row counts to navigate directly to result tabs. #924

Fixed

• Redshift: Fixed Explain button failing with FORMAT parameter error.
• Results: Fixed Shift+Execute triggering license prompt instead of refreshing the last result tab. #921
• Connection Import: Added support for JDBC connection strings (e.g., jdbc:db2://host/db). #927

Changed

• Salesforce: Preview support added for Salesforce.
• DB Explorer: Tree filter now searches cached schemas, not just active connections.
• BigQuery: Progressive introspection enabled. #925

Fixed

• BigQuery: Fixed “Premature close” error when queries return more than 40K rows by enabling driver pagination. #925
• SSH Tunnels: Fixed SSH Agent authentication with Pageant on Windows failing with “Failed to retrieve identities from agent”.
• Elasticsearch: Fixed connection errors with Elasticsearch 7.x and 8.x servers caused by incompatible version headers in client v9.
• Elasticsearch: Made Catalog field optional. Leave empty to use the local cluster.

Changed

• Inferred Relationships: Ability to define relationships for databases without native foreign key support (MongoDB, DynamoDB, etc.). Configure patterns in connection settings or visually in the entity relationship diagram to detect relationships based on column naming conventions (e.g., user_id → users.id).
• Results: Hold Shift while executing (Shift+Cmd+Enter / Shift+Window+Enter) to refresh the last result tab instead of opening a new one, preserving chart configuration. In single-tab mode (maxTabs=1), this behavior is inverted. #921
• Results: Added column selection via Shift/Cmd/Ctrl+click on column headers, or “Select Column” in context menu. Shift selects a range, Cmd/Ctrl adds to selection. #921
• Athena: Added multi-catalog support. Database list now shows all data catalogs (data sources) including federated connectors. #920
• Settings: Added dbcode.discovery.enabled setting to disable automatic zero-config connection discovery from workspace files (.env, config files, etc.).

Fixed

• Formatting: SQL formatter now properly respects .editorconfig insert_final_newline setting at file level.

Fixed

• AWS: Fixed SSO credentials not refreshing after user re-authenticates via CLI, previously requiring VS Code restart.
• DB2: Fixed connection failure on Windows due to CLI driver path not being resolved. #918
• Results: Fixed date filter on datetime/timestamp columns not matching rows when using equals with date-only values. Now expands to match any time on that date.

Changed

• H2: Added experimental support. #371 #788
• Table View: Added loading status text (“Connecting…”, “Loading data…”) to provide feedback during slow initial database connections.

Fixed

• Hive: Fixed column comments not appearing on Hive 2.x (legacy) servers. #669
• MySQL: Fixed progressive schema loading still timing out on very large table counts. #862
• PostgreSQL: Fixed introspection error on openGauss and PostgreSQL < 9.5 due to missing array_position function. #917