Object mapping, and more, for Redis and .NET
Redis OM .NET makes it easy to model Redis data in your .NET Applications.
Redis OM .NET | Redis OM Node.js | Redis OM Spring | Redis OM Python
Table of contents
Redis OM provides high-level abstractions for using Redis in .NET, making it easy to model and query your Redis domain objects.
This preview release contains the following features:
- Declarative object mapping for Redis objects
- Declarative secondary-index generation
- Fluent APIs for querying Redis
- Fluent APIs for performing Redis aggregations
Using the dotnet cli, run:
dotnet add package Redis.OM
Before writing any code you'll need a Redis instance with the appropriate Redis modules! The quickest way to get this is with Docker:
docker run -p 6379:6379 -p 8001:8001 redis/redis-stackThis launches the redis-stack an extension of Redis that adds all manner of modern data structures to Redis. You'll also notice that if you open up http://localhost:8001 you'll have access to the redis-insight GUI, a GUI you can use to visualize and work with your data in Redis.
With Redis OM, you can model your data and declare indexes with minimal code. For example, here's how we might model a customer object:
[Document(StorageType = StorageType.Json)]
public class Customer
{
[Indexed] public string FirstName { get; set; }
[Indexed] public string LastName { get; set; }
public string Email { get; set; }
[Indexed(Sortable = true)] public int Age { get; set; }
[Indexed] public string[] NickNames {get; set;}
}Notice that we've applied the Document attribute to this class. We've also specified that certain fields should be Indexed.
Now we need to create the Redis index. So we'll connect to Redis and then call CreateIndex on an IRedisConnection:
var provider = new RedisConnectionProvider("redis://localhost:6379");
provider.Connection.CreateIndex(typeof(Customer));There are two methods for indexing embedded documents with Redis.OM, an embedded document is a complex object, e.g. if our Customer model had an Address property with the following model:
[Document(IndexName = "address-idx", StorageType = StorageType.Json)]
public partial class Address
{
public string StreetName { get; set; }
public string ZipCode { get; set; }
[Indexed] public string City { get; set; }
[Indexed] public string State { get; set; }
[Indexed(CascadeDepth = 1)] public Address ForwardingAddress { get; set; }
[Indexed] public GeoLoc Location { get; set; }
[Indexed] public int HouseNumber { get; set; }
}You can index fields by JSON path, in the top level model, in this case Customer you can decorate the Address property with an Indexed and/or Searchable attribute, specifying the JSON path to the desired field:
[Document(StorageType = StorageType.Json)]
public class Customer
{
[Indexed] public string FirstName { get; set; }
[Indexed] public string LastName { get; set; }
public string Email { get; set; }
[Indexed(Sortable = true)] public int Age { get; set; }
[Indexed] public string[] NickNames {get; set;}
[Indexed(JsonPath = "$.ZipCode")]
[Searchable(JsonPath = "$.StreetAddress")]
public Address Address {get; set;}
}This methodology can also be used for indexing string and string-like value-types within objects within Arrays and Lists, so for example if we had an array of Addresses, and we wanted to index the cities within those addresses we could do so with the following
[Indexed(JsonPath = "$.City")]
public Address[] Addresses { get; set; }Those Cities can then be queried with an Any predicate within the main Where clause.
collection.Where(c=>c.Addresses.Any(a=>a.City == "Satellite Beach"))Alternatively, you can also embedded models by cascading indexes. In this instance you'd simply decorate the property with Indexed and set the CascadeDepth to whatever to however may levels you want the model to cascade for. The default is 0, so if CascadeDepth is not set, indexing an object will be a no-op:
[Document(StorageType = StorageType.Json)]
public class Customer
{
[Indexed] public string FirstName { get; set; }
[Indexed] public string LastName { get; set; }
public string Email { get; set; }
[Indexed(Sortable = true)] public int Age { get; set; }
[Indexed] public string[] NickNames {get; set;}
[Indexed(CascadeDepth = 2)]
public Address Address {get; set;}
}In the above case, all indexed/searchable fields in Address will be indexed down 2 levels, so the ForwardingAddress field in Address will also be indexed.
Once the index is created, we can:
- Insert Customer objects into Redis
- Get a Customer object by ID from Redis
- Query Customers from Redis
- Run aggregations on Customers in Redis
Let's see how!
As of version 0.4.0, all DateTime objects are indexed as numerics, and they are inserted as numerics into JSON documents. Because of this, you can query them as if they were numerics!
Ids are unique per object, and are used as part of key generation for the primary index in Redis. The natively supported Id type in Redis OM is the ULID. You can bind ids to your model, by explicitly decorating your Id field with the RedisIdField attribute:
[Document(StorageType = StorageType.Json)]
public class Customer
{
[RedisIdField] public Ulid Id { get; set; }
[Indexed] public string FirstName { get; set; }
[Indexed] public string LastName { get; set; }
public string Email { get; set; }
[Indexed(Sortable = true)] public int Age { get; set; }
[Indexed] public string[] NickNames { get; set; }
}When you call Set on the RedisConnection or call Insert in the RedisCollection, to insert your object into Redis, Redis OM will automatically set the id for you and you will be able to access it in the object. If the Id type is a string, and there is no explicitly overriding IdGenerationStrategy on the object, the ULID for the object will bind to the string.
Redis OM also supports other types of ids, ids must either be strings or value types (e.g. ints, longs, GUIDs etc. . .), if you want a non-ULID id type, you must either set the id on each object prior to insertion, or you must register an IIdGenerationStrategy with the DocumentAttribute class.
To Register an IIdGenerationStrategy with the DocumentAttribute class, simply call DocumentAttribute.RegisterIdGenerationStrategy passing in the strategy name, and the implementation of IIdGenerationStrategy you want to use. Let's say for example you had the StaticIncrementStrategy, which maintains a static counter in memory, and increments ids based off that counter:
public class StaticIncrementStrategy : IIdGenerationStrategy
{
public static int Current = 0;
public string GenerateId()
{
return (Current++).ToString();
}
}You would then register that strategy with Redis.OM like so:
DocumentAttribute.RegisterIdGenerationStrategy(nameof(StaticIncrementStrategy), new StaticIncrementStrategy());Then, when you want to use that strategy for generating the Ids of a document, you can simply set the IdGenerationStrategy of your document attribute to the name of the strategy.
[Document(IdGenerationStrategyName = nameof(StaticIncrementStrategy))]
public class ObjectWithCustomIdGenerationStrategy
{
[RedisIdField] public string Id { get; set; }
}The key names are, by default, the fully qualified class name of the object, followed by a colon, followed by the Id. For example, there is a Person class in the Unit Test project, an example id of that person class would be Redis.OM.Unit.Tests.RediSearchTests.Person:01FTHAF0D1EKSN0XG67HYG36GZ, because Redis.OM.Unit.Tests.RediSearchTests.Person is the fully qualified class name, and 01FTHAF0D1EKSN0XG67HYG36GZ is the ULID (the default id type). If you want to change the prefix (the fully qualified class name), you can change that in the DocumentAttribute by setting the Prefixes property, which is an array of strings e.g.
[Document(Prefixes = new []{"Person"})]
public class PersonNote: At this time, Redis.OM will only use the first prefix in the prefix list as the prefix when creating a key name. However, when an index is created, it will be created on all prefixes enumerated in the Prefixes property
We can query our domain using expressions in LINQ, like so:
var customers = provider.RedisCollection<Customer>();
// Insert customer
customers.Insert(new Customer()
{
FirstName = "James",
LastName = "Bond",
Age = 68,
Email = "[email protected]"
});
// Find all customers whose last name is "Bond"
customers.Where(x => x.LastName == "Bond");
// Find all customers whose last name is Bond OR whose age is greater than 65
customers.Where(x => x.LastName == "Bond" || x.Age > 65);
// Find all customers whose last name is Bond AND whose first name is James
customers.Where(x => x.LastName == "Bond" && x.FirstName == "James");
// Find all customers with the nickname of Jim
customers.Where(x=>x.NickNames.Contains("Jim"));Redis OM .NET also supports storing and querying Vectors stored in Redis.
A Vector<T> is a representation of an object that can be transformed into a vector by a Vectorizer.
A VectorizerAttribute is the abstract class you use to decorate your Vector fields, it is responsible for defining the logic to convert the object's that Vector<T> is a container for into actual vector embeddings needed. In the package Redis.OM.Vectorizers we provide vectorizers for HuggingFace, OpenAI, and AzureOpenAI to allow you to easily integrate them into your workflows.
To define a vector in your model, simply decorate a Vector<T> field with an Indexed attribute which defines the algorithm and algorithmic parameters and a Vectorizer attribute which defines the shape of the vectors, (in this case we'll use OpenAI):
[Document(StorageType = StorageType.Json)]
public class OpenAICompletionResponse
{
[RedisIdField]
public string Id { get; set; }
[Indexed(DistanceMetric = DistanceMetric.COSINE, Algorithm = VectorAlgorithm.HNSW, M = 16)]
[OpenAIVectorizer]
public Vector<string> Prompt { get; set; }
public string Response { get; set; }
[Indexed]
public string Language { get; set; }
[Indexed]
public DateTime TimeStamp { get; set; }
}With the vector defined in our model, all we need to do is create Vectors of the generic type, and insert them with our model. Using our RedisCollection, you can do this by simply using Insert:
var collection = _provider.RedisCollection<OpenAICompletionResponse>();
var completionResult = new OpenAICompletionResponse
{
Language = "en_us",
Prompt = Vector.Of("What is the Capital of France?"),
Response = "Paris",
TimeStamp = DateTime.Now - TimeSpan.FromHours(3)
};
collection.Insert(completionResult);The Vectorizer will manage the embedding generation for you without you having to intervene.
To query vector fields in Redis, all you need to do is use the VectorRange method on a vector within our normal LINQ queries, and/or use the NearestNeighbors with whatever other filters you want to use, here's some examples:
var prompt = "What really is the Capital of France?";
// simple vector range, find first within .15
var result = collection.First(x => x.Prompt.VectorRange(prompt, .15));
// simple nearest neighbors query, finds first nearest neighbor
result = collection.NearestNeighbors(x => x.Prompt, 1, prompt).First();
// hybrid query, pre-filters result set for english responses, then runs a nearest neighbors search.
result = collection.Where(x=>x.Language == "en_us").NearestNeighbors(x => x.Prompt, 1, prompt).First();
// hybrid query, pre-filters responses newer than 4 hours, and finds first result within .15
var ts = DateTimeOffset.Now - TimeSpan.FromHours(4);
result = collection.First(x=>x.TimeStamp > ts && x.Prompt.VectorRange(prompt, .15));With Redis OM, the embeddings can be completely transparent to you, they are generated and bound to the Vector<T> when you query/insert your vectors. If however you needed your embedding after the insertion/Query, they are available at Vector<T>.Embedding, and be queried either as the raw bytes, as an array of doubles or as an array of floats (depending on your vectorizer).
The Vectorizers provided by the Redis.OM.Vectorizers package have some configuration parameters that it will pull in either from your appsettings.json file, or your environment variables (with your appsettings taking precedence).
| Configuration Parameter | Description |
|---|---|
| REDIS_OM_HF_TOKEN | HuggingFace Authorization token. |
| REDIS_OM_OAI_TOKEN | OpenAI Authorization token |
| REDIS_OM_OAI_API_URL | OpenAI URL |
| REDIS_OM_AZURE_OAI_TOKEN | Azure OpenAI api key |
| REDIS_OM_AZURE_OAI_RESOURCE_NAME | Azure resource name |
| REDIS_OM_AZURE_OAI_DEPLOYMENT_NAME | Azure deployment |
Redis OM also provides the ability to use Semantic Caching, as well as providers for OpenAI, HuggingFace, and Azure OpenAI to perform semantic caching. To use a Semantic Cache, simply pull one out of the RedisConnectionProvider and use Store to insert items, and GetSimilar to retrieve items. For example:
var cache = _provider.OpenAISemanticCache(token, threshold: .15);
cache.Store("What is the capital of France?", "Paris");
var res = cache.GetSimilar("What really is the capital of France?").First();We also provide the packages Redis.OM.Vectorizers.ResNet18 and Redis.OM.Vectorizers.AllMiniLML6V2 which have embedded models / ML Pipelines in them to
allow you to easily Vectorize Images and Sentences respectively without the need to depend on an external API.
We can also run aggregations on the customer object, again using expressions in LINQ:
// Get our average customer age
customerAggregations.Average(x => x.RecordShell.Age);
// Format customer full names
customerAggregations.Apply(x => string.Format("{0} {1}", x.RecordShell.FirstName, x.RecordShell.LastName),
"FullName");
// Get each customer's distance from the Mall of America
customerAggregations.Apply(x => ApplyFunctions.GeoDistance(x.RecordShell.Home, -93.241786, 44.853816),
"DistanceToMall");This README just scratches the surface. You can find a full tutorial on the Redis Developer Site. All the summary docs for this library can be found on the repo's github page.
If you run into trouble or have any questions, we're here to help!
First, check the FAQ. If you don't find the answer there, hit us up on the Redis Discord Server.
Redis OM can be used with regular Redis for Object mapping and getting objects by their IDs. For more advanced features like indexing, querying, and aggregation, Redis OM is dependent on the Redis Stack platform, a collection of modules that extend Redis.
Without Redis Stack, you can still use Redis OM to create declarative models backed by Redis.
We'll store your model data in Redis as Hashes, and you can retrieve models using their primary keys.
So, what won't work without Redis Stack?
- You won't be able to nest models inside each other.
- You won't be able to use our expressive queries to find object -- you'll only be able to query by primary key.
You can use Redis Stack with your self-hosted Redis deployment. Just follow the instructions for Installing Redis Stack.
Don't want to run Redis yourself? Redis Stack is also available on Redis Cloud. Get started here.
We'd love your contributions! If you want to contribute please read our Contributing document.