MongoDbRepositoryCore 5.0.4

Here are the updated documentation files for v5.0. I have incorporated all the enterprise hardening features we just implemented, including the strict Outbox transaction enforcement, BSON-native distributed caching, OOM-safe preloading, and the single-trip aggregation optimizations.

1. README.md Markdown

📖 MongoDbRepository Developer's Guide (v5.0.0)

Welcome to the daily cookbook for the MongoDbRepository framework. This guide covers how to leverage the advanced v5.0 features, including Zero-Boilerplate Source Generators, AI Semantic Search, Microservices Caching, the Outbox Pattern, and Code-First Migrations.

🌟 What's New in V5: Enterprise Hardening

Version 5 introduces massive stability and performance improvements for enterprise production environments:

• Strict ACID Outbox: The Outbox pattern now strictly enforces ambient transactions. If you attempt to save an entity with pending domain events outside of a transaction, the repository fails fast to prevent data corruption.
• BSON-Native Distributed Caching: Redis caching now utilizes native BSON wrappers instead of JSON, perfectly preserving ObjectId references, custom encryption conventions, and [BsonIgnore] attributes across your distributed cluster.
• OOM-Safe Preloading: Cache preloading now safely caps at 10,000 items to prevent startup crashes on massive collections.
• Single-Trip Aggregations: SumAsync and AverageAsync now calculate in a single network round-trip, halving database latency.
• Clean Architecture DI: BaseService now enforces explicit constructor injection, dropping the legacy Service Locator anti-pattern.

1. Zero-Boilerplate DI

Forget manually registering dozens of repositories and services. V5 uses a blazing-fast Roslyn Source Generator (C# 7.3+ compliant) that wires up your entire database layer at compile-time.

Step 1: Add the [GenerateRepository] attribute to your entity.

[GenerateRepository]
[Cacheable(IsDistributed = true)] // Optional: Stacks the BSON-native distributed caching decorator!
public class User : Entity 
{
    public string Name { get; set; }
}
Step 2: Call the auto-generated extension method in Program.cs. The generator automatically maps the interfaces, applies your decorators, and registers MediatR for background tasks.

C#
builder.Services.AddAutoGeneratedMongoRepositories();
builder.Services.AddAutoGeneratedMongoServices();
2. Generative AI & Vector Search
MongoDB Atlas supports native AI Vector Searches. This library reduces complex aggregation pipelines into a single attribute and method call.

Step 1: Add the [VectorEmbedding] attribute to a float[] property on your entity. The repository will automatically provision the Atlas Search Indexes on startup.

C#
public class SupportArticle : Entity
{
    public string Title { get; set; }
    public string Content { get; set; }

    [VectorEmbedding(Dimensions = 1536, Similarity = "cosine")]
    public float[] ContentEmbedding { get; set; } 
}
Step 2: Execute a Semantic Search natively.

C#
// Generate embedding via OpenAI, HuggingFace, etc.
float[] userQuestionVector = await _aiService.GetEmbeddingAsync("How do I reset my password?");

// Search, with optional pre-filtering!
var topArticles = await _articleRepo.SemanticSearchAsync(
    queryVector: userQuestionVector, 
    limit: 5, 
    filter: article => article.IsActive // Pre-filters using standard LINQ
);
3. The Transactional Outbox Pattern (Guaranteed Events)
In a distributed system, saving a database record and firing an event is dangerous. If the server crashes between the two, the event is lost. V5 handles this flawlessly with a Zero-Config Outbox.

When you attach an event using IHasDomainEvents, the repository serializes the MediatR event into a System_Outbox collection inside an ACID transaction. Note: In V5, attempting to save an entity with events outside of IMongoTransactionManager will throw a safeguard exception.

C#
using var transaction = await _transactionManager.BeginTransactionAsync();

user.AddDomainEvent(new UserRegisteredEvent(user.Id));
await _userRepo.AddAsync(user); // 100% Guaranteed Delivery

await transaction.CommitAsync();
4. Distributed Cluster Caching (Redis)
If you run multiple instances of your API (e.g., in Kubernetes), standard RAM caching leads to stale data. Use [Cacheable(IsDistributed = true)] to utilize the DistributedCachingRepositoryDecorator.

If Pod A updates a user, it deletes the key from Redis, meaning Pods B, C, and D instantly fetch fresh data on their next request.

C#
builder.Services.AddMongoDbRepositories(db => { ... }, cache => 
{
    cache.ApplicationName = "BillingApi"; 
});
5. Code-First Migrations
Need to seed an Admin user or rename a property across 1 million documents? Write a migration script. The framework will automatically run it on application startup and track its execution.

C#
public class Migration_001_SeedAdmin : IMongoMigration
{
    public int Version => 1;

    public async Task UpAsync(IMongoDatabase database)
    {
        var users = database.GetCollection<User>("users");
        await users.InsertOneAsync(new User { Email = "admin@system.com", Role = "SuperAdmin" });
    }
}
6. Mockless Unit Testing
Stop mocking IMongoCollection! We ship a blazing-fast, thread-safe fake that implements IRepository<T, TId> and executes your queries in memory.

C#
[Fact]
public async Task GetActiveUsers_ReturnsCorrectData()
{
    // Arrange: Use the fake repository (0 milliseconds execution!)
    var fakeRepo = new InMemoryRepository<User, Guid>();
    fakeRepo.Seed(new[] { new User { Id = Guid.NewGuid(), IsActive = true, Name = "Alice" } });

    var service = new UserService(fakeRepo);

    // Act
    var activeUsers = await service.GetActiveUsersAsync();

    // Assert
    Assert.Single(activeUsers); 
}