Changes with no impact on existing DDL/API functions

AI Copilot

Copilot Subagents: a new subagents menu with Planner and Reviewer. Planner breaks a task into steps before execution (with access to Copilot's full tool catalogue and any files you attach); Reviewer checks the work against your project standards
Copilot Skills can now be invoked from Forward Engineer full-script output, so you can apply reusable AI workflows (fix naming, add descriptions, and so on) directly against generated DDL
Copilot can now propose a new or updated skill — it opens the Skills manager with the fields pre-filled for you to review and save (handy for turning existing conventions or attached files into a review skill)
(Fix) Copilot's column and relationship tools now report results that match what actually happened — renames no longer change a column's display name unexpectedly or fail on casing, and creating a relationship no longer wrongly reports missing foreign key columns
(Fix) Merging a branch no longer resets a project's Copilot settings (global context and pre-prompts) — they're now preserved on both sides of the merge
(Fix) A table-level metric created by Copilot can now have its access modifier (public/private) changed afterward
(Fix) You can now clear a global, global-technical, or final instruction instead of hitting a validation error
(Fix) A follow-up message sent while Copilot is busy now gets a clear "busy" response instead of being silently blocked

Object Groups: group tables into named, colored containers on the diagram — drag objects in and out, collapse and expand groups, filter and search by group, and use grouping actions and clipboard support. Groups work with iframe/macros, auto-layout, compare worksheets, and check-in/check-out

Direct Connect for PostgreSQL on AWS RDS: reverse-engineer schemas directly from a PostgreSQL RDS instance instead of importing DDL files
Push to Git for PostgreSQL now includes views, functions, procedures, and sequences, in addition to tables

Compare-with-Unity now builds the target model directly from raw Unity Catalog metadata, producing more accurate diffs against the live catalog

Erwin 9 import support: import Erwin models from version 9 and above, with fixes for diagram membership, cached views, and property mapping

Collibra integration (Pull): pull data governance values from Collibra into SqlDBM (the reverse of the existing push). Governance-role users can preview incoming values, review changes per object on a compare screen, and accept them into a draft to save as a normal revision

Alter scripts now handle transformational column changes on Snowflake — column reorder, incompatible type changes, and length/precision reduction — that previously had no representation and could silently leave the model and database out of sync. When a change needs a rebuild rather than a simple ALTER, the editor asks you to confirm first, and SqlDBM generates the script to complete the change so your deployment pipeline keeps running

(Fix) Comparisons that include nulls (PostgreSQL option-bag collections and Data Governance fields) no longer produce false diffs
(Fix) Semantic models that can't be merged are now automatically unchecked in Compare & Merge, so you can't start an unmergeable merge

(Fix) Changes restored from a draft after a page reload are now saved correctly — previously a save didn't persist them and the project could revert on the next reload, in some cases losing schemas and tables

Modeler-role users can now merge external mappings (previously blocked)
(Fix) User management operations (inviting users, updating members) no longer crash with a server error
(Fix) Transferring project ownership no longer fails when the target user already has a project with the same name

Faster dashboard project list, project dropdown, and Project Team page, especially for users and companies with many projects and members
Concurrent Working operations (branch, merge, switching branches) are faster and no longer flood the backend with parallel refresh requests
(Fix) Applying naming conventions right after a large Reverse Engineering import no longer causes browser out-of-memory errors