A supercharged .NET SDK for Azure CosmosDB with ORM support
Clone or download
Latest commit 29f3033 Nov 8, 2018


Build Status NuGet Package Licensed under the MIT License


The word was derived from "kosmos" (Ancient Greek: κόσμος) which means world/universe and "nautes" (Ancient Greek: ναῦς) which means sailor/navigator

Cosmonaut is a supercharged SDK with object mapping capabilities that enables .NET developers to work with CosmosDB. It eliminates the need for most of the data-access code that developers usually need to write.

Getting started



The idea is pretty simple. You can have one CosmosStore per entity (POCO/dtos etc). This entity will be used to create a collection or use part of a one in CosmosDB and it will offer all the data access for this object.

Registering the CosmosStores in ServiceCollection for DI support

 var cosmosSettings = new CosmosStoreSettings("<<databaseName>>", 

//or just by using the Action extension

serviceCollection.AddCosmosStore<Book>("<<databaseName>>", "<<cosmosUri>>"), "<<authkey>>", settings =>
    settings.ConnectionPolicy = connectionPolicy;
    settings.DefaultCollectionThroughput = 5000;
    settings.IndexingPolicy = new IndexingPolicy(new RangeIndex(DataType.Number, -1),
        new RangeIndex(DataType.String, -1));

//or just initialise the object

ICosmosStore<Book> bookStore = new CosmosStore<Book>(cosmosSettings)

To use the AddCosmosStore extension methods you need to install the Cosmonaut.Extensions.Microsoft.DependencyInjection package.

Install-Package Cosmonaut.Extensions.Microsoft.DependencyInjection
dotnet add package Cosmonaut.Extensions.Microsoft.DependencyInjection
Retrieving an entity by id (and partition key)
var user = await cosmosStore.FindAsync("userId");
var user = await cosmosStore.FindAsync("userId", "partitionKey");
var user = await cosmosStore.FindAsync("userId", new RequestOptions());
Quering for entities using LINQ

In order to query for entities all you have to do is call the .Query() method and then use LINQ to create the query you want. It is HIGHLY recommended that you use one of the Async methods to get the results back, such as ToListAsync or FirstOrDefaultAsync , when available.

var user = await cosmoStore.Query().FirstOrDefaultAsync(x => x.Username == "elfocrash");
var users = await cosmoStore.Query().Where(x => x.HairColor == HairColor.Black).ToListAsync(cancellationToken);
Quering for entities using SQL
// plain sql query
var user = await cosmoStore.QueryMultipleAsync("select * from c w.Firstname = 'Smith'");

// or parameterised sql query
var user = await cosmoStore.QueryMultipleAsync("select * from c w.Firstname = @name", new { name = "Smith" });

Cosmonaut supports two types of pagination.

  • Page number + Page size
  • ContinuationToken + Page size

Both of there methods work by adding the .WithPagination() method after you used any of the Query methods.

var firstPage = await booksStore.Query().WithPagination(1, 10).OrderBy(x=>x.Name).ToListAsync();
var secondPage = await booksStore.Query().WithPagination(2, 10).OrderBy(x => x.Name).ToPagedListAsync();
var thirdPage = await booksStore.Query().WithPagination(secondPage.NextPageToken, 10).OrderBy(x => x.Name).ToPagedListAsync();
var fourthPage = await thirdPage.GetNextPageAsync();
var fifthPage = await booksStore.Query().WithPagination(5, 10).OrderBy(x => x.Name).ToListAsync();

ToListAsync() on a paged query will just return the results. ToPagedListAsync() on the other hand will return a CosmosPagedResults object. This object contains the results but also a boolean indicating whether there are more pages after the one you just got but also the continuation token you need to use to get the next page.

Pagination recommendations

Because page number + page size pagination goes though all the documents until it gets to the requested page, it's potentially slow and expensive. The recommended approach would be to use the page number + page size approach once for the first page and get the results using the .ToPagedListAsync() method. This method will return the next continuation token and it will also tell you if there are more pages for this query. Then use the continuation token alternative of WithPagination to continue from your last query.

Keep in mind that this approach means that you have to keep state on the client for the next query, but that's what you'd do if you where using previous/next buttons anyway.

Adding an entity in the entity store
var newUser = new User
    Name = "Nick"
var added = await cosmoStore.AddAsync(newUser);

var multiple = await cosmoStore.AddRangeAsync(manyManyUsers);
Updating entities

When it comes to updating you have two options.


await cosmoStore.UpdateAsync(entity);

... and Upsert

await cosmoStore.UpsertAsync(entity);

The main difference is of course in the functionality. Update will only update if the item you are updating exists in the database with this id. Upsert on the other hand will either add the item if there is no item with this id or update it if an item with this id exists.

Removing entities
await cosmoStore.RemoveAsync(x => x.Name == "Nick"); // Removes all the entities that match the criteria
await cosmoStore.RemoveAsync(entity);// Removes the specific entity
await cosmoStore.RemoveByIdAsync("<<anId>>");// Removes an entity with the specified ID

Collection sharing

Cosmonaut is all about making the integration with CosmosDB easy as well as making things like cost optimisation part of the library.

That's why Cosmonaut support collection sharing between different types of entities.

Why would you do that?

Cosmos is charging you based on how many RU/s your individual collection is provisioned at. This means that if you don't need to have one collection per entity because you won't use it that much, even on the minimum 400 RU/s, you will be charged money. That's where the magic of schemaless comes in.

How can you do that?

Well it's actually pretty simple. Just implement the ISharedCosmosEntity interface and decorate your object with the SharedCosmosCollection attribute.

The attribute accepts two properties, SharedCollectionName which is mandatory and EntityName which is optional. The SharedCollectionName property will be used to name the collection that the entity will share with other entities.

The EntityName will be used to make the object identifiable for Cosmosnaut. Be default it will pluralize the name of the class, but you can specify it to override this behavior. You can override this by providing your own name by setting the EntityName value at the attribute level.

Once you set this up you can add individual CosmosStores with shared collections.


Because of the way the internal id property of Cosmosdb works, there is a mandatory restriction made. You cannot have a property named Id or a property with the attribute [JsonProperty("id")] without it being a string. A cosmos id needs to exist somehow on your entity model. For that reason if it isn't part of your entity you can just extend the CosmosEntity class.

It is HIGHLY RECOMMENDED that you decorate your Id property with the [JsonProperty("id")] attribute to prevent any unexpected behaviour.


Cosmonaut has it's own version of a DocumentClient called CosmonautClient. The difference is that the CosmonautClient interface is more user friendly and it looks more like something you would use in a real life scenario. It won't throw not found exceptions if an item is not found but it will return null instead. It will also retry automatically when you get 429s (too many requests).

It also has support for logging and monitoring as you are going to see in the logging section of this page.


There is currently no way to reliably do transactions with the current CosmosDB SDK. Because Cosmonaut is a wrapper around the CosmosDB SDK it doesn't support them either. However there are plans for investigating potential other ways to achieve transactional operations such as server side stored procedures that Cosmonaut could provision and call.

Every operational call (Add, Update, Upsert, Delete) however returns it's status back alongside the reason it failed, if it failed, and the entity so you can add your own retry logic.


By default CosmosDB is created with the following indexing rules

    "indexingMode": "consistent",
    "automatic": true,
    "includedPaths": [
            "path": "/*",
            "indexes": [
                    "kind": "Range",
                    "dataType": "Number",
                    "precision": -1
                    "kind": "Range",
                    "dataType": "String",
                    "precision": -1
                    "kind": "Spatial",
                    "dataType": "Point"
    "excludedPaths": []

Indexing in necessary for things like querying the collections. Keep in mind that when you manage indexing policy, you can make fine-grained trade-offs between index storage overhead, write and query throughput, and query consistency.

For example if the String datatype is Hash then exact matches like the following, cosmoStore.Query().FirstOrDefaultAsync(x => x.SomeProperty.Equals($"Nick Chapsas") will return the item if it exists in CosmosDB but cosmoStore.Query().FirstOrDefaultAsync(x => x.SomeProperty.StartsWith($"Nick Ch") will throw an error. Changing the Hash to Range will work.

However you can get around that by setting the FeedOptions.EnableScanInQuery to true for this Query()

Same goes for ordering. If you use OrderBy on a string property you need to have this property's path set up as Range and precision -1.

More about CosmosDB Indexing here


Cosmonaut supports partitions out of the box. You can specify which property you want to be your Partition Key by adding the [CosmosPartitionKey] attribute above it.

Unless you really know what you're doing, it is recommended make your Id property the Partition Key. This will enable random distribution for your collection.

If you do not set a Partition Key then the collection created will be single partition. Here is a quote from Microsoft about single partition collections:

Single-partition collections have lower price options and the ability to execute queries and perform transactions across all collection data. They have the scalability and storage limits of a single partition (10GB and 10,000 RU/s). You do not have to specify a partition key for these collections. For scenarios that do not need large volumes of storage or throughput, single partition collections are a good fit. link

Known hiccups

Partitions are great but you should these 3 very important things about them and about the way Cosmonaut will react.

  • Once a collection is created with a partition key, it cannot be removed or changed.
  • You cannot add a partition key later to a single partition collection.
  • If you use the the Upsert method to update an entity that had the value of the property that is the partition key changed, then CosmosDB won't update the document but instead it will create a whole different document with the same id but the changed partition key value.

More on the third issue here Unique keys in Azure Cosmos DB

Collection naming

Your collections will automatically be named based on the plural of the object you are using in the generic type. However you can override that by decorating the class with the CosmosCollection attribute.




Event source

Cosmonaut uses the .NET Standard's System.Diagnostics to log it's actions as dependency events. By default, this system is deactivated. In order to activated and actually do something with those events you need to create an EventListener which will activate the logging and give you the option do something with the logs.


By using this package you are able to log the events as dependencies in Application Insights in detail. The logs are batched and send in intervals OR automatically sent when the batch buffer is filled to max.

Just initialise the AppInsightsTelemetryModule in your Startup or setup pipeline like this. Example: AppInsightsTelemetryModule.Instance.Initialize(new TelemetryConfiguration("InstrumentationKey"))

If you already have initialised TelemetryConfiguration for your application then use TelemetryConfiguration.Active instead of new TelemetryConfiguration because if you don't there will be no association between the dependency calls and the parent request.