Changelog

What shipped, what changed, and what to know before upgrading.

Short highlights from the published release notes. Dates use UTC publication dates; development work is listed separately below.

UnreleasedDevelopment work · not in a published package

SQL Search, Database Documenter, and Test Data Generator

Prepared for v4.6.4, based on changes since v4.6.3. This candidate is not yet published; stable downloads remain on v4.6.3.

  • Find definitions and dependencies: search saved database definitions, follow Uses and Used by relationships, and assess column changes from SQL Search and Data Modeler.
  • Document your database: build a searchable dictionary, save descriptions with conflict checks, and export complete HTML or Markdown documentation.
  • Create repeatable test data: save generation profiles, preview every requested row, preserve supported relationships, and append in one transaction with cancellation and rollback.
  • Reliability: preserve comments and literals when binding SQL parameters, stabilize Data Modeler connectors, and correct nullable unique-key generation and database-switch state.

No database file-format revision is introduced. Description editing requires the new server capability; Studio generation supports one directly connected database and rejects unsupported schemas before inserting. An unconfirmed commit is reported without automatic retry.

SQL Search guide → · Test Data Generator guide →

Published releases

Latest release

v4.6.3 — Data Modeler and Admin improvements

  • Readable diagrams: searchable Sources, relationship-aware arrangement, key-focused cards, composite foreign-key mappings, and crow’s-foot cardinality.
  • Control your layout: named table groups, editable connectors, and middle segments that follow moved tables without unwanted splits. Positions, groups, routes, and viewport are saved with the diagram.
  • Review schema changes: richer inspection and editing, SQL review and validation, stale-plan protection, and transactional application where the client supports it.
  • Everyday modeling: diagram history, search, exports, clearer New / Save As / Delete Diagram actions, and relationship labels that avoid tables and connectors.
  • Catalog visibility and reliability: a read-only Internal Storage view, logical-catalog mappings, and fixes for diagram deletion and persistence.

Removing a table from the canvas or deleting a diagram does not drop database tables. Schema changes require a separate review and apply action.

Follow the Data Modeler tutorial →

Full v4.6.3 release notes & downloads →

Release

v4.6.2 — Observability and diagnostics

  • Inspect active and recent queries, sessions, storage, WAL, maintenance, and per-shard status in the Admin Observability workspace.
  • Connect structured events, OpenTelemetry traces and metrics, and optional Prometheus export without adding exporter dependencies to the engine.
  • Use API and daemon liveness/readiness endpoints, including standard gRPC health checks.

Before upgrading: Most telemetry and exporters are opt-in; supported host health endpoints are configured separately. SQL text capture is off by default. The failed v4.6.0 and v4.6.1 attempts are not separate published releases.

Full v4.6.2 release notes & downloads →

Release

v4.5.1 — Logical SQL types and XML queries

  • Preserve declared types, lengths, precision, and scale through storage, providers, transports, migration plans, and archives.
  • Use exact DECIMAL values, logical Boolean and integer types, temporal values, UUIDs, bit strings, and generated ROWVERSION tokens.
  • Query XML values with bounded XML_EXISTS / XMLEXISTS and XML_VALUE functions.

Before upgrading: INTEGER is now 32-bit; use BIGINT for 64-bit values. Bare BIT is Boolean, and bare TIMESTAMP means ROWVERSION. Back up before upgrading: legacy rowversion tokens are regenerated. v4.5.0 was an unpublished release attempt.

Full v4.5.1 release notes & downloads →

Release

v4.4.0 — SQL and schema coverage

  • Use named windows and explicit ROWS frames within the documented in-memory window-function limits.
  • Apply composite keys and foreign keys with the full supported immediate update/delete action matrix.
  • Use supported transactional column rewrites and inspect physical plans with EXPLAIN and EXPLAIN ANALYZE.
  • Expanded provider, transport, historical-database, and failure-recovery qualification covers the supported SQL surface.

Before upgrading: Support remains bounded: consult the SQL reference for supported rewrite shapes, key types, and resource limits. Unsupported forms fail with explicit diagnostics.

Full v4.4.0 release notes & downloads →

Earlier releases · v3.8.0–v4.3.0
Release

v4.3.0 — Migration tooling

  • Capture retained source packages, inspect mappings, and review a plan before creating a staged target database.
  • Resume interrupted applies and validate schema, row counts, and checksums before activation.
  • Export retained snapshots as CSV, JSON, or NDJSON and use read-only dual-run query comparisons.

Before upgrading: Live Access, SQL Server, and MySQL provider qualification remains bounded; candidate capture is not a claim that every source is approved for migration.

Full v4.3.0 release notes & downloads →

Release

v4.2.0 — Practical EF Core workflows

  • Expanded idempotent migrations, schema metadata, literal defaults, checks, and supported decimal mappings.
  • Added generated rowversion concurrency tokens and broader bounded LINQ translations.
  • Added warm embedded connection pooling, faster direct-client transaction sessions, and a bounded ASP.NET Core Identity readiness profile.

Before upgrading: Computed columns, SQL-expression defaults, savepoints, and unsupported migration shapes remain outside this release's contract.

Full v4.2.0 release notes & downloads →

Release

v4.1.0 — Constraints and SQL compatibility

  • Added quoted identifiers, SQL three-valued logic, literal defaults, and row-local CHECK constraints.
  • Persisted ordered composite primary/unique keys and table-level foreign keys across providers and catalogs.
  • Added first-tier window functions, lazy UNION ALL, transactional schema changes, and a public compatibility matrix.

Before upgrading: Schema edits reject known dependent views, triggers, and validation rules before mutation. A failed write makes an explicit transaction rollback-only.

Full v4.1.0 release notes & downloads →

Release

v4.0.4 — Join and read-path performance

  • Eligible broad key-only primary-key/foreign-key joins can merge-scan existing indexes without a separate join index.
  • Reduced checkpoint reader tracking and query-session allocation overhead.
  • Removed the inline-value read regression introduced with overflow support and updated vulnerable dependency pins.

Before upgrading: The specialized join path is conservative; predicates, payload projections, limits, and other unsupported shapes retain existing plans.

Full v4.0.4 release notes & downloads →

Release

v4.0.3 — Large values and desktop file dialogs

  • Large B-tree values use overflow pages across reads, writes, rollback, checkpoints, and reopen.
  • Storage diagnostics validate overflow chains, and replaced/deleted chains are reclaimed.
  • The Windows desktop Admin shell provides native Open, Save, and folder dialogs; browser sessions retain manual paths.

Before upgrading: Writing an older format-v1 database upgrades it to format v2. Older binaries cannot open the upgraded format; back up first if rollback may be needed.

Full v4.0.3 release notes & downloads →

Release

v4.0.2 — Scalable full-text postings

  • Hot-token postings use bounded chunks instead of rewriting one growing value.
  • Full-text updates and deletes rewrite only affected chunks while preserving rollback and search ordering.
  • Indexed predicates can identify DELETE and UPDATE targets without a full table scan.

Before upgrading: Legacy postings remain readable and migrate to the chunked representation on the next write for that term.

Full v4.0.2 release notes & downloads →

Release

v4.0.1 — Full-text large-posting fix

  • Full-text index stores spill oversized values into overflow chains, preventing page-capacity failures for common terms.
  • Search, deletion, and reopening are covered for large shared-token postings.

Before upgrading: Existing inline full-text values remain readable; oversized values spill on their next write.

Full v4.0.1 release notes & downloads →

Release

v4.0.0 — Route-aware sharding

  • Use explicit route contexts and shard-bound clients across direct, HTTP, gRPC, daemon, and ADO.NET access.
  • Manage master-catalog shard maps, inspect routes and shard status, and perform read-only fan-out with per-shard results.
  • Move exact route keys through manifest-driven migration workflows.

Before upgrading: Sharding is explicit and API-level: no route inference from SQL, distributed joins, distributed writes, or distributed transactions.

Full v4.0.0 release notes & downloads →

Release

v3.9.1 — Data Hygiene release follow-up

  • Published notes retain the 3.9 Data Hygiene and temporary-table feature set.
  • See the linked release record for the exact published package and notes.

Release note: The published notes reuse the 3.9.0 heading and do not identify a separate patch feature delta.

Full v3.9.1 release notes & downloads →

Release

v3.9.0 — Data Hygiene and temporary tables

  • Find duplicates, deduplicate rows, merge data, and manage validation rules through SQL.
  • Use the Admin Data Hygiene workspace and inspect validation metadata in system catalogs.
  • Create session-scoped in-memory temporary tables and explicitly persist supported temp data into a durable table.

Before upgrading: Temporary state is excluded from durable catalogs, backups, and checkpoints; supported features are intentionally limited.

Full v3.9.0 release notes & downloads →

Release

v3.8.0 — Table archives and the original Data Model workspace

  • Export native .csdbtable archives with schema, values, and optional primary-key lookup metadata.
  • Register archives as read-only external tables and query them alongside normal tables.
  • Use Admin Import / Export and the original saved Data Model canvas with staged schema changes.

Before upgrading: External archives are read-only. The later relationship-aware modeling, groups, and connector editing improvements are listed under Unreleased.

Full v3.8.0 release notes & downloads →

Older release archive · v2.8.0–v3.7.0

Original notes and downloads remain on GitHub. This archive uses published versions and dates rather than the former undated project milestones.

For older history, see all published releases. For planned work, see the roadmap; for how-to guides, visit tutorials.