Skip to main content

​
🔔 Notable Changes

  • Unified Plan lifecycle view — the Plan detail page shows the full lifecycle of a database change, giving developers one place to follow a change from draft to deployment.
    • The three phases of a database change — change, review, deploy — are shown in sections, allowing developers to edit change, view approval result, and execute deployment in this single page.
    • Issue page is now dedicated primarily for review and approval. Developers will need to go to the plan page to edit changes and create a rollout when needed.
  • API breaking changes — Worksheet API moves to project scope (/v1/projects/{id}/worksheets), ListDatabases now requires workspace ID (/v1/workspaces/{id}/databases), branding_logo moves from workspace profile setting to workspace message, GRANT_REQUEST issue type renamed to ROLE_GRANT, UpdateSubscription API deprecated in favor of UploadLicense, and classification description and level description fields removed with level type changed from string to number. See Change Details.
  • CEL filter syntax change — Substring-search filters now use contains() instead of matches() across plan, database, instance, project, group, user, service account, workload identity, query history, access grant, and database metadata table filters. See Change Details.
  • Read-only data source consolidation — Each instance now allows at most one read-only data source. If more than one is configured, only the first is kept after migration; extra read-only data sources are removed.
  • Remove Pro plan 20-user seat cap — Pro plan no longer has a user limit.
  • Terraform provider — Updated for 3.17.0 API breaking changes. Supports get, create, update, and list identity providers. See Migration Guide.

​
🚀 Features

  • High Availability (HA) — Support licensed multi-replica deployments backed by a shared external PostgreSQL metadata database. Active replicas are tracked via heartbeats, and background runners coordinate safely across replicas.
  • Page Agent — Add an in-app AI assistant with tool-calling, threaded resumable conversations, DOM-aware autocomplete, and token usage tracking.
  • Issue Approved webhook event — Support a new webhook event type for issue approval, with redesigned Slack message format.
  • MCP query_database tool — Add a new MCP tool for executing SQL queries against managed databases with automatic database resolution.
  • CosmosDB — Support local emulator connection and dynamic data masking for all 13 query feature areas.

​
🎄 Enhancements

  • Support executing stored procedures (CALL / EXEC) in SQL Editor by classifying stored procedure execution as DML.
  • Support issue type filter in advanced search (Database Change, Role Grant, Database Export, Access Grant).
  • Replace data classification file upload with an inline JSON editor and enhance the data classification editor.
  • Support masking exemption by classification level, and redesign masking exemption list page.
  • Redesign data export creation UI.
  • Migrate the frontend from Vue to React across all major pages including settings, projects, databases, instances, plans, issues, and audit logs.
  • Upgrade SQL parsing engine for PostgreSQL, MySQL, MongoDB, and CosmosDB from ANTLR to omni parser, improving SQL syntax coverage, consistency across SQL review / schema diff / auto-completion, and error message quality.
  • Add configurable --timeout flag for bytebase-action for large SQL checks.
  • Expose --enable-json-logging flag in the Helm chart for structured log output.

​
🐞 Bug Fixes

  • Fix false schema diff noise caused by PostgreSQL trigger ordering and CRLF / whitespace normalization.
  • Fix SQL editor metadata sync retry flood.
  • Fix OAuth discovery to return usable URLs in self-hosted mode.
  • Fix stale debounced query reverts in advanced search.
  • Fix gh-ost directives in migration-based GitOps workflow.
  • PostgreSQL — Fix CRLF line endings causing query truncation in SQL splitter, schema load failure by excluding aggregates from function sync, array subscript handling in column reference normalization, whitespace between LIMIT and FOR UPDATE clause, and nested SELECTs in wrapper statements and CTEs for SQL review.
  • MySQL — Fix unqualified SET columns in prior backup for UPDATE JOIN, missing multiStatements param in IAM auth DSN, and role sync for anonymous users.
  • TiDB — Remove non-transaction statement handling.
  • Oracle — Strip trailing null bytes from schema definitions.

​
⚙️ Install and Upgrade

Before upgrading: 1) Back up the metadata — in-place downgrade is not supported. 2) Do not run multiple containers on the same data directory. 3) Terraform users: upgrade Bytebase server first, then apply the new Terraform config.

​
📃 Change Details

​
API Breaking Changes

1. Worksheet API moves to project scope:
MethodBeforeAfter
List/v1/worksheets/v1/{parent=projects/*}/worksheets
Get/v1/{name=worksheets/*}/v1/{name=projects/*/worksheets/*}
Create/v1/worksheets/v1/{parent=projects/*}/worksheets
Update/v1/{worksheet.name=worksheets/*}/v1/{worksheet.name=projects/*/worksheets/*}
Delete/v1/{name=worksheets/*}/v1/{name=projects/*/worksheets/*}
2. APIs that no longer accept workspaces/- (must use workspaces/{id}):
ServiceAffected Operations
DatabaseServiceListDatabases
3. Other breaking changes:
ChangeAffected RoutesDetails
Branding logoGET/PATCH /v1/workspaces/{id}branding_logo removed from workspace profile setting; use logo field on workspace message
Issue type enumPOST /v1/{parent=projects/*}/issues, GET /v1/{parent=projects/*}/issues, POST /v1/{parent=projects/*}/issues:searchGRANT_REQUEST renamed to ROLE_GRANT
Subscription APIPATCH /v1/subscriptionDeprecated; use PATCH /v1/subscription/license (UploadLicense)
Data source IDPOST /v1/{name=instances/*/databases/*}:query, POST /v1/{name=instances/*/databases/*}:exportdata_source_id auto-resolved server-side; at most one read-only data source per instance (extra removed during migration)
Classification levelGET/PATCH /v1/settings/{name}, POST /v1/queryHistories:searchLevel type changed from string to int; description field removed

​
CEL Filter Syntax Change

Substring-search CEL filters now use contains() instead of matches(). Existing clients that send filters such as title.matches("..."), name.matches("..."), resource_id.matches("..."), email.matches("..."), host.matches("..."), port.matches("..."), statement.matches("..."), query.matches("..."), or table.matches("...") must switch to the corresponding contains() form. This applies to plan, database, instance, project, group, user, service account, workload identity, query history, access grant, and database metadata table filters.