ServiceStack v10.4!

Announcements News
1

ServiceStack v10.4 brings OrmLite’s biggest release ever ! We’ve packed it with the features on our wishlist, expanding what you can do with its typed, SQL-friendly approach to .NET data access.

image
image1920×1072 563 KB

Highlights include:

  • Connection-wide tenant isolation: Define filters and write rules once so typed queries, AutoQuery, updates, deletes and inserts stay within a tenant. Connections can fail closed until the tenant is known, and rules can enforce audit fields and soft deletes.

  • Safer SQL from user input: Sql.Fmt() parameterizes interpolated values across raw SQL APIs; OrderBySafe() permits customer-selected sorting using approved fields. Fragment validation and identifier escaping are also stronger.

  • Schema Diff and migration generation: Compare models with the database, inspect differences in the Admin UI, generate migrations for review, and apply changes in development. SQLite gains RebuildTable<T>() for changes it cannot make in place.

  • Faster large writes: BulkUpsert() imports and synchronizes thousands of rows; supported drivers also batch InsertAll, UpdateAll, UpsertAll and SaveAll into fewer round trips.

  • Read replicas: Route read-only queries and AutoQuery APIs to replicas, with primary fallback and a read-your-writes window after updates.

  • Automatic retries for temporary database failures: Retry eligible deadlocks, throttling, serialization failures and lost connections, including whole transactions.

  • Vector search: Store embeddings and find similar rows in typed queries, including joins and filters, with vector index and half-precision options where supported.

  • Full-text search: Create database-native text indexes from attributes, search as users type, and rank results by relevance within typed queries.

  • Query performance tools: db.Explain() exposes query plans; the Admin UI adds a Slowest Queries view with Explain, Analyze and Run Query actions.

  • Compiled queries: Generate SQL once for frequently run typed queries, reducing query-building time and allocations, including for tenant-filtered queries.

  • Scalable paging: SeekAfter() provides stable keyset pagination for large or changing result sets.

  • Large-list queries: SelectByIds, DeleteByIds and typed Contains() queries handle thousands of values automatically within database limits.

  • Asynchronous result streaming: SelectLazyAsync() and ColumnLazyAsync() process rows one at a time with await foreach.

  • Typed JSON and array queries: Query nested object properties and list contents directly, including Any, All, Count and PostgreSQL arrays.

  • System-versioned tables: Mark tables to retain row history and query their contents at an earlier time.

  • Typed SQL composition: Combine queries with Union, UnionAll, Intersect and Except; name subqueries with With; traverse hierarchies with WithRecursive; and calculate rankings, running totals and grouped results with window functions.

  • Atomic row-returning writes: UpdateOnlyReturning() and DeleteReturning() return affected rows in the same statement. Upserts now populate database-generated values and row versions on their objects.

  • Concurrency and joined updates: ForUpdate() locks selected rows, including skip-locked work queues; UpdateFrom() updates rows using values and filters from joined tables.

  • Richer model-defined schemas: Attributes now cover filtered and covering indexes, generated columns, enum checks and database comments.

  • Runtime table reads: Untyped APIs can select, fetch by ID and count rows when a table is known only by its Type, with async equivalents.

AI.Chat

image
image1920×1072 375 KB

AI.Chat also gets a major update, Highlights:

  • Decision Studio in AI.Chat: Build reusable typed decision recipes, compare saved examples, manage a recipe library, and import or share recipes.

  • Decisions API: Ask AI yes/no, choice and scale questions from C# or HTTP and receive validated answers with probabilities, usage and cost.

  • Authenticated user API keys: User API keys can call eligible authenticated APIs; built-in API key management also uses request-scoped connections for tenant filtering.

  • AI.Chat workspace and Git panels: Browse files, preview content, review diffs, stage changes, commit, inspect history, stash and sync from the interface.

  • AI.Chat repository projects: Clone repositories into new projects, reorder projects and archive or restore them.

  • ChatGPT subscription connection: Connect a personal ChatGPT account for eligible AI.Chat text requests.

  • Recoverable Gemini imports: Save portable folder and website import settings, synchronize changed sources, and resume interrupted uploads.

  • Static project publishing: Export an AI.Chat project to a folder served by your own static server, with configurable destinations.

  • Faster AI.Chat loading and smaller transfers: Features load when opened, shared modules preload, responses use compression and assets support cache revalidation.

  • AI.Chat usability updates: Searchable model pickers, improved skill navigation, refreshed provider metadata and smaller interface refinements.

:backhand_index_pointing_right: Read the full v10.4 release notes for examples, demos, benchmarks, and upgrade details.