SlimMessageBus is a client faรงade for message brokers for .NET. It comes with implementations for specific brokers (Apache Kafka, Azure EventHub, MQTT/Mosquitto, Redis Pub/Sub) and also for in memory message passing (in-process communication). SlimMessageBus additionally provides request-response implementation over message queues.

Gitter GitHub license

Branch Build Status
master Build status
develop Build status
other Build status

Key elements of SlimMessageBus

  • Consumers:
    • IConsumer<in TMessage> - subscriber in pub/sub (or queue consumer)
    • IRequestHandler<in TRequest, TResponse> - request handler in request-response
  • Producers:
    • IPublishBus - publisher in pub/sub (or queue producer)
    • IRequestResponseBus - sender in req/resp
    • IMessageBus - extends IPublishBus and IRequestResponseBus
  • Misc:
    • IRequestMessage<TResponse> - marker for request messages
    • MessageBus - static accessor for current context IMessageBus



Name Descripton NuGet
SlimMessageBus The core API for SlimMessageBus NuGet
Transport providers
SlimMessageBus.Host.Kafka Transport provider for Apache Kafka NuGet
SlimMessageBus.Host.AzureServiceBus Transport provider for Azure Service Bus NuGet
SlimMessageBus.Host.AzureEventHub Transport provider for Azure Event Hubs NuGet
SlimMessageBus.Host.Redis Transport provider for Redis NuGet
SlimMessageBus.Host.Memory Transport provider implementation for in-process (in memory) message passing (no messaging infrastructure required) NuGet
SlimMessageBus.Host.Hybrid Bus implementation that composes the bus out of other transport providers and performs message routing NuGet
SlimMessageBus.Host.Serialization.Json Serialization plugin for JSON (Json.NET library) NuGet
SlimMessageBus.Host.Serialization.Avro Serialization plugin for Avro (Apache.Avro library) NuGet
SlimMessageBus.Host.Serialization.Hybrid Plugin that delegates serialization to other serializers based on message type NuGet
SlimMessageBus.Host.AspNetCore Integration for ASP.NET Core (DI adapter, config helpers) NuGet
SlimMessageBus.Host.Autofac DI adapter for Autofac container NuGet
SlimMessageBus.Host.Unity DI adapter for Unity container NuGet
SlimMessageBus.Host.ServiceLocator DI adapter for CommonServiceLocator NuGet

Typically your application components (business logic, domain) only need to depend on SlimMessageBus which is the facade, and ultimately your application hosting layer (ASP.NET, Windows Service, Console App) will reference and configure the other packages (SlimMessageBus.Host.*) which are the providers and plugins.


Check out the Samples folder.

Quick example

Some service (or domain layer) sends a message:

IMessageBus bus; // injected

await bus.Publish(new SomeMessage())

Another service (or application layer) handles the message:

public class SomeMessageConsumer : IConsumer<SomeMessage>
   public Task OnHandle(SomeMessage message, string name) // name = topic or queue name
       // handle the message

Note: It is also possible to avoid having to implement the interface IConsumer<T> (see here).

The bus also supports request-response implemented via queues (or topics - depending what the chosen transport provider supports). The sender side sends a request message:

var messageResponse = await bus.Send(new MessageRequest());

Note: It is possible to configure the bus to timeout a request when the response does not arrive within alloted time (see here).

The receiving side handles the request and replies back:

public class MessageRequestHandler : IRequestHandler<MessageRequest, MessageResponse>
   public async Task<MessageResponse> OnHandle(MessageRequest request, string name)
      // handle the request message and return response

The bus will ask the chosen DI to provide the consumer instances (SomeMessageConsumer, MessageRequestHandler).

The configuration somewhere in your service:

var builder = MessageBusBuilder.Create()

   // Pub/Sub example:
   .Produce<SomeMessage>(x => x.DefaultTopic("some-topic"))
   .Consume<SomeMessage>(x => x
      //.Group("some-kafka-consumer-group") //  Kafka provider specific
      //.SubscriptionName("some-azure-sb-topic-subscription") // Azure ServiceBus provider specific

   // Use JSON for message serialization                
   .WithSerializer(new JsonMessageSerializer())
   // Use DI from ASP.NET Core (or Autofac, Unity, ServiceLocator)
   .WithDependencyResolver(new AspNetCoreMessageBusDependencyResolver(serviceProvider))

   // Use Apache Kafka transport provider
   .WithProviderKafka(new KafkaMessageBusSettings("localhost:9092"));
   // Use Azure Service Bus transport provider
   // Use Azure Azure Event Hub transport provider
   // Use Redis transport provider
   // Use in-memory transport provider

// Build the bus from the builder. Message consumers will start consuming messages from the configured topics/queues of the chosen provider.
IMessageBus bus = builder.Build();

// Register bus in your DI

Basic in-process pub/sub messaging (for domain events)

This example shows how SlimMessageBus and SlimMessageBus.Host.Memory can be used to implement Domain Events pattern. The provider passes messages in the same app domain process (no external message broker is required).

The domain event is a simple POCO:

// domain event
public class OrderSubmittedEvent
   public Order Order { get; }
   public DateTime Timestamp { get; }

   public OrderSubmittedEvent(Order order) { ... }

The event handler implements the IConsumer<T> interface:

// domain event handler
public class OrderSubmittedHandler : IConsumer<OrderSubmittedEvent>
   public Task OnHandle(OrderSubmittedEvent e, string name)
      Console.WriteLine("Customer {0} {1} just placed an order for:", e.Order.Customer.Firstname, e.Order.Customer.Lastname);
      foreach (var orderLine in e.Order.Lines)
         Console.WriteLine("- {0}x {1}", orderLine.Quantity, orderLine.ProductId);

      Console.WriteLine("Generating a shipping order...");
      return Task.Delay(1000);

The domain handler (well, really the consumer) is obtained from dependency resolver at the time of event publication. It can be scoped (per web request, per unit of work) as configured in your favorite DI container.

Somewhere in your domain layer the domain event gets raised:

// aggregate root
public class Order
   public Customer Customer { get; }
   private IList<OrderLine> _lines = new List<OrderLine>();
   public OrderState State { get; private set; }

   public IEnumerable<OrderLine> Lines => _lines.AsEnumerable();

   public Order(Customer customer)
      State = OrderState.New;
      Customer = customer;

   public OrderLine Add(string productId, int quantity) { }

   public void Submit()
      State = OrderState.Submitted;

      var e = new OrderSubmittedEvent(this);
      MessageBus.Current.Publish(e).Wait(); // raise domain event

Some sample logic executed in your domain:

var john = new Customer("John", "Whick");

var order = new Order(john);
order.Add("id_machine_gun", 2);
order.Add("id_grenade", 4);

order.Submit(); // events fired here

Notice the static MessageBus.Current property might actually be configured to resolve a scoped IMessageBus instance (web request scoped).

The SlimMessageBus configuration for in-memory provider looks like this:

// Define the recipie how to create our IMessageBus
var mbb = MessageBusBuilder.Create()
   .Produce<OrderSubmittedEvent>(x => x.DefaultTopic(x.MessageType.Name))
   .Consume<OrderSubmittedEvent>(x => x.Topic(x.MessageType.Name).WithConsumer<OrderSubmittedHandler>())
   .WithDependencyResolver(new AutofacMessageBusDependencyResolver())
   .WithProviderMemory(new MemoryMessageBusSettings
      // supress serialization and pass the same event instance to subscribers (events contain domain objects we do not want serialized, also we gain abit on speed)
      EnableMessageSerialization = false

// Create the IMessageBus instance from the builder
IMessageBus bus = mbb.Build();

// Set the provider to resolve our bus - this setup will work as a singleton.
MessageBus.SetProvider(() => bus);

See the complete sample for ASP.NET Core where the handler and bus is web-request scoped.

Request-response over Kafka topics

Use case:

  • Some front-end web app needs to display downsized image (thumbnails) of large images to speed up page load.
  • The thumbnails are requested in the WebApi and are generated on demand (and cached to disk) by the Worker (unless they exist already).
  • WebApi and Worker exchange messages via Apache Kafka
  • Worker can be scaled out (more instances, more kafka partitions)

Front-end web app makes a call to resize an image DSC0862.jpg to 120x80 resolution, by using this URL:


This gets handled by the WebApi method of the ImageController

private readonly IRequestResponseBus _bus;
// ...
public async Task<HttpResponseMessage> GetImageThumbnail(string fileId, ThumbnailMode mode, int w, int h)
   var thumbFileContent = // ... try to load content for the desired thumbnail w/h/mode/fileId
   if (thumbFileContent == null)
      // Task will await until response comes back (or timeout happens). The HTTP request will be queued and IIS processing thread released.
      var thumbGenResponse = await _bus.Send(new GenerateThumbnailRequest(fileId, mode, w, h));
      thumbFileContent = await _fileStore.GetFile(thumbGenResponse.FileId);
   return ServeStream(thumbFileContent);

The GenerateThumbnailRequest request is handled by a handler in one of the pool of Worker console apps.

public class GenerateThumbnailRequestHandler : IRequestHandler<GenerateThumbnailRequest, GenerateThumbnailResponse>
   public Task<GenerateThumbnailResponse> OnHandle(GenerateThumbnailRequest request, string name)
      // some processing
      return new GenerateThumbnailResponse { FileId = thumbnailFileId };

The response gets replied onto the originating WebApi instance and the Task resolves causing the queued HTTP request to serve the resized image thumbnail.

var thumbGenResponse = await _bus.Send(new GenerateThumbnailRequest(fileId, mode, w, h));

The message bus configuration for the WebApi:

private IMessageBus BuildMessageBus()
   // unique id across instances of this application (e.g. 1, 2, 3)
   var instanceId = Configuration["InstanceId"];
   var kafkaBrokers = Configuration["Kafka:Brokers"];

   var instanceGroup = $"webapi-{instanceId}";
   var instanceReplyTo = $"webapi-{instanceId}-response";

   var mbb = MessageBusBuilder.Create()
      .Produce<GenerateThumbnailRequest>(x =>
         //x.DefaultTimeout(TimeSpan.FromSeconds(10)); // Default response timeout for this request type
         x.DefaultTopic("thumbnail-generation"); // Use this topic as default when topic is not specified in IMessageBus.Publish() for that message type
      .ExpectRequestResponses(x =>
         x.ReplyToTopic(instanceReplyTo); // Expect all responses to my reqests replied to this topic
         x.Group(instanceGroup); // Kafka consumer group        
         x.DefaultTimeout(TimeSpan.FromSeconds(30)); // Default global response timeout
      .WithDependencyResolver(new AutofacMessageBusDependencyResolver())
      .WithSerializer(new JsonMessageSerializer())
      .WithProviderKafka(new KafkaMessageBusSettings(kafkaBrokers));

   var messageBus = mbb.Build();
   return messageBus;

The message bus configuration for the Worker:

private static IMessageBus BuildMessageBus()
   // unique id across instances of this application (e.g. 1, 2, 3)
   var instanceId = Configuration["InstanceId"];
   var kafkaBrokers = Configuration["Kafka:Brokers"];

   var instanceGroup = $"worker-{instanceId}";
   var sharedGroup = "workers";

   var mbb = MessageBusBuilder.Create()
      .Handle<GenerateThumbnailRequest, GenerateThumbnailResponse>(s =>
         s.Topic("thumbnail-generation", t =>
               .Group(sharedGroup) // kafka consumer group
      .WithDependencyResolver(new AutofacMessageBusDependencyResolver())
      .WithSerializer(new JsonMessageSerializer())
      .WithProviderKafka(new KafkaMessageBusSettings(kafkaBrokers));

   var messageBus = mbb.Build();
   return messageBus;

Because topics are partitioned in Kafka, requests originating from WebApi instances will be distributed across all Worker instances. However, to fine tune this, message key providers should configured (see Kafka provider wiki and samples).

Check out the complete sample for image resizing.


  • Types of messaging patterns supported:
    • Publish-subscribe
    • Request-response
    • Queues
    • Hybrid of the above (e.g. Kafka with multiple topic consumers in one group)
  • Modern async/await syntax and TPL
  • Fluent configuration
  • Because SlimMessageBus is a facade, you have the ability to swap broker implementations
    • Using nuget pull another broker provider
    • Reconfigure SlimMessageBus and retest your app
    • Try out the messaging middleware that works best for your app (Kafka vs. Redis) without having to rewrite your app.


  • The core of SlimMessageBus is "slim"
    • Simple, common and friendly API to work with messaging systems
    • No external dependencies. Logging is done via Common.Logging, so that you can connect your favorite logger provider.
    • The core interface can be used in domain model (e.g. DomainEvents)
  • Plugin architecture:
    • DI integration (Autofac, CommonServiceLocator, Unity)
    • Message serialization (JSON, XML)
    • Use your favorite messaging broker as provider by simply pulling a nuget package
  • No threads created (pure TPL)
  • Async/Await support
  • Fluent configuration


Apache License 2.0


cd src
dotnet build
dotnet pack --output ../dist

NuGet packaged end up in dist folder


To run tests you need to update the respective appsettings.json to match your own cloud infrstructure or local infrastructure. SMB has some message brokers setup on Azure for integration tests (secrets not shared).

Run all tests:

dotnet test

Run all tests except integration tests which require local/cloud infrastructure:

dotnet test --filter Category!=Integration


Lightweight message bus interface for .NET (pub/sub and request-response) with transport plugins for popular message brokers.

Slimmessagebus Info

โญ Stars 81
๐Ÿ”— Source Code
๐Ÿ•’ Last Update 8 months ago
๐Ÿ•’ Created 5 years ago
๐Ÿž Open Issues 11
โž— Star-Issue Ratio 7
๐Ÿ˜Ž Author zarusz