🎉 BLite v3.6.5 — Production-Ready

[mrdevrobot]

🎉 BLite v3.6.5 — Production-Ready

After months of intensive development, testing, and hardening, BLite v3.6.5 is the first release we consider production-ready.

BLite is a fully managed, embedded, ACID-compliant document database built from scratch for .NET — no native dependencies, no P/Invoke, no C wrappers. Just pure, modern C#.

---

What is BLite?

BLite is designed for applications that need a fast, reliable, embedded document store without the complexity of a standalone server. Think LiteDB — but rewritten on modern .NET primitives with zero-allocation I/O, compile-time source generators, and a proper LINQ query engine.

Targets: net10.0 and netstandard2.1 — works with .NET 5+, Unity, Xamarin, and MAUI.

---

Performance

Benchmarks run on Windows 11, Intel Core i7-13800H, .NET 10.0.4 against LiteDB 5.0.21 and SQLite + System.Text.Json using a realistic ~1–2 KB e-commerce document:

Operation	vs LiteDB	vs SQLite+JSON
Single insert	4.7× faster	31.5× faster
Batch insert (1 000 docs)	1.77× faster	1.78× faster
FindById	5.8× faster	9.5× faster
Full scan	3.3× faster	2.9× faster
Scan allocation	3.4× less	—

---

Feature highlights

Core engine

• Full ACID transactions — Write-Ahead Logging (WAL) + Snapshot Isolation
• Zero-allocation I/O pipeline: Span<T> from disk to user objects, memory-mapped files
• Compile-time source generators — no reflection, no runtime overhead
• Hot backup (BackupAsync) without locking the engine
• Page compaction on delete

Query engine

• Full LINQ provider (IQueryable) with automatic B-Tree index usage for =, >, <, Between, StartsWith
• Support for GroupBy, Join, anonymous types, aggregates (Count, Sum, Min, Max, Average)
• BLQL — an MQL-inspired query language for schema-less collections, with 252 dedicated security tests (ReDoS protection, unknown operator rejection, zero-divisor guard)

Indexing

• B-Tree (single and composite)
• Nested property indexes via lambda (x => x.Address.City) or dot-notation
• HNSW vector index (v3.6.2: full correctness pass — neighbor linking, level distribution, persistence)
• R-Tree geospatial index — .Near() and .Within() LINQ operators

Advanced features

• Change Data Capture (CDC) — zero-overhead transactional event streaming, Channel-per-subscriber
• TimeSeries — native page type with retention pruning, available in DocumentDbContext
• Vector Source Config for RAG pipelines — weighted field normalization for embeddings
• Watch on dynamic collections (v3.6.0)
• IDocumentCollection<TId,T> abstraction (v3.5.0) for easy mocking and DI
• Custom ID ValueObjects via ValueConverter<TModel, TProvider>
• Auto ID fallback for string and Guid
• netstandard2.1 multi-targeting (v2.0.0)

Tooling

• BLite Studio — cross-platform desktop GUI (Windows MSI, macOS .dmg) for browsing, querying, and managing databases

---

Recent stability fixes (3.6.x series)

• v3.6.5 — nullable properties incorrectly stored when entity has init setter
• v3.6.3 — coerced BSON numeric reads + JSON type-preservation
• v3.6.2 — HNSW vector search correctness (full pass)
• v3.6.1 — duplicate index inserts

---

Get started

dotnet add package BLite

using var engine = new BLiteEngine("myapp.db");
var orders = engine.GetOrCreateCollection<Order>("orders");

await orders.InsertAsync(new Order { Id = "ORD-1", Total = 99.99m });
var result = orders.FindById("ORD-1");

• NuGet: https://www.nuget.org/packages/BLite
• GitHub: https://github.com/EntglDb/BLite
• Docs: https://blite.entgl.dev

---

What's next

Work is already underway on:

• A query result cache layer
• An embedding worker for automatic vector indexing

---

Thank you to everyone who filed issues, tested pre-release builds, and contributed feedback. If BLite is useful to you, consider leaving a ⭐ on the repo or sponsoring the project.

[LeoYang06]

Thanks to the author for providing such an excellent database with equally great documentation. The actual documentation URL is https://www.blitedb.com/docs, so an update is recommended.

Here’s a small suggestion: LiteDB is an embedded database with a history of over 10 years and a large user base. BLiteDB shares some similarities with LiteDB in certain aspects. To lower the learning curve for new BLite users, it would be beneficial to create a comparative documentation against LiteDB, covering differences in usage and design principles. This will help quickly build up the user base for BLite.

[mrdevrobot]

Hi @LeoYang06 thanks for the support :)
Your idea is great, we could provide a "moving from LiteDb" section that could help developers in moving codebase, along with a migration tool from LiteDb.