August had strong momentum in both SQLMesh and SQLGlot repositories! This month wasn't just about incremental progress; it also included new features that are already making a significant difference.

Monthly Merge Report - August 2025

August had strong momentum in both SQLMesh and SQLGlot repositories! This month wasn't just about incremental progress; it also included new features that are already making a significant difference.

Fabric Warehouse Support

On August 19, 2025, after 128 commits and two months of intense collaboration, PR #4751 brought comprehensive Microsoft Fabric Warehouse support to SQLMesh.

Shoutout to the community contributors, @fresioAS and @mattiasthalen, who navigated Fabric's Warehouse nuances to get the PR merged. For teams invested in the Microsoft ecosystem, this brings SQLMesh's transformation capabilities, its virtual data environments, its comprehensive testing framework, and its intelligent deployment strategies directly into their Fabric warehouses.

dev_only VDE MODE

SQLMesh now supports dev_only VDE mode, where the Virtual Data Environment (VDE) is applied only in development. In production, SQLMesh bypasses the virtual layer and materializes models directly as physical tables using the original, unversioned model names. The option is virtual_environment_mode=dev_only. The default remains full, so nothing changes unless you opt in.

The SQLMesh default remains full VDE mode. With full VDE, you have versioned snapshots, instant view swaps, and rollbacks without recompute. You also get automatic change categorization and reuse of built data across environments. These capabilities reduce compute and keep prod and dev aligned without relying on engine‑specific features.

BigQuery Embedding Support

In SQLGlot, @rloredo had two successful PRs. PR #5660 and PR #5652 gained support for vector_search and ml.generate_embedding functions.

Reliability & Guardrails

SQLMesh introduced on_additive_change support (PR #5193). Previously, SQLMesh provided controls only for destructive changes (removing a column or changing to an incompatible data type). Now, model owners can define how additive changes should be handled. This control helps teams mitigate unintended consequences.

For Trino users, PR #5279 brought atomic REPLACE TABLE operations for Iceberg and Delta tables, preserving partitioning schemes and table properties through the replacement.

SQLGlot expanded its dialect coverage. Dremio received support through multiple PRs from @jasonthomassql. SingleStore gained JSON operations, time functions, and statistical helpers. Even established dialects like Teradata got attention, with new support for locking mechanisms and query banding. Each addition means another team can adopt SQLGlot in their existing workflows.

The Road Ahead

The message is clear - Tobiko supports open source. The Fivetran acquisition announcement made some wonder, “What will happen with SQLMesh and SQLGlot?” You can review the activity in the open source project repositories and see where the team’s priorities are. Both projects remain open source, and we are dedicated to continuing community development and support.

🤝 Join the Movement

These improvements came from our vibrant community. Special thanks to all contributors who made August 2025 remarkable. Your bug reports, feature requests, and pull requests drive innovation.

🔮 Looking Ahead

September is already shaping up with exciting developments across both projects. Stay tuned for more innovations that will continue to transform how we work with data!

Questions? Feedback? Reach out and join our community Slack.

‍