Using the Redis connector (Preview)

Getting started

Add the Redis Vector Store connector nuget package to your project.

dotnet add package Microsoft.SemanticKernel.Connectors.Redis --prerelease

You can add the vector store to the dependency injection container available on the KernelBuilder or to the IServiceCollection dependency injection container using extension methods provided by Semantic Kernel.

using Microsoft.SemanticKernel;

// Using Kernel Builder.
var kernelBuilder = Kernel
    .CreateBuilder()
    .AddRedisVectorStore("localhost:6379");

using Microsoft.SemanticKernel;

// Using IServiceCollection with ASP.NET Core.
var builder = WebApplication.CreateBuilder(args);
builder.Services.AddRedisVectorStore("localhost:6379");

Extension methods that take no parameters are also provided. These require an instance of the Redis IDatabase to be separately registered with the dependency injection container.

using Microsoft.Extensions.DependencyInjection;
using Microsoft.SemanticKernel;
using StackExchange.Redis;

// Using Kernel Builder.
var kernelBuilder = Kernel.CreateBuilder();
kernelBuilder.Services.AddSingleton<IDatabase>(sp => ConnectionMultiplexer.Connect("localhost:6379").GetDatabase());
kernelBuilder.AddRedisVectorStore();

using Microsoft.Extensions.DependencyInjection;
using Microsoft.SemanticKernel;
using StackExchange.Redis;

// Using IServiceCollection with ASP.NET Core.
var builder = WebApplication.CreateBuilder(args);
builder.Services.AddSingleton<IDatabase>(sp => ConnectionMultiplexer.Connect("localhost:6379").GetDatabase());
builder.Services.AddRedisVectorStore();

You can construct a Redis Vector Store instance directly.

using Microsoft.SemanticKernel.Connectors.Redis;
using StackExchange.Redis;

var vectorStore = new RedisVectorStore(ConnectionMultiplexer.Connect("localhost:6379").GetDatabase());

It is possible to construct a direct reference to a named collection. When doing so, you have to choose between the JSON or Hashes instance depending on how you wish to store data in Redis.

using Microsoft.SemanticKernel.Connectors.Redis;
using StackExchange.Redis;

// Using Hashes.
var hashesCollection = new RedisHashSetVectorStoreRecordCollection<Hotel>(
    ConnectionMultiplexer.Connect("localhost:6379").GetDatabase(),
    "skhotelshashes");

using Microsoft.SemanticKernel.Connectors.Redis;
using StackExchange.Redis;

// Using JSON.
var jsonCollection = new RedisJsonVectorStoreRecordCollection<Hotel>(
    ConnectionMultiplexer.Connect("localhost:6379").GetDatabase(),
    "skhotelsjson");

When constructing a RedisVectorStore or registering it with the dependency injection container, it's possible to pass a RedisVectorStoreOptions instance that configures the preferred storage type / mode used: Hashes or JSON. If not specified, the default is JSON.

using Microsoft.SemanticKernel.Connectors.Redis;
using StackExchange.Redis;

var vectorStore = new RedisVectorStore(
    ConnectionMultiplexer.Connect("localhost:6379").GetDatabase(),
    new() { StorageType = RedisStorageType.HashSet });

Getting started

Install semantic kernel with the redis extras, which includes the redis client.

pip install semantic-kernel[redis]

You can then create a vector store instance using the RedisStore class, this will use the environment variables REDIS_CONNECTION_STRING to connect to a Redis instance, those values can also be supplied directly.

from semantic_kernel.connectors.memory.redis import RedisStore

vector_store = RedisStore()

You can also create the vector store with your own instance of the redis database client.

from redis.asyncio.client import Redis
from semantic_kernel.connectors.memory.redis import RedisStore

redis_database = Redis.from_url(url="https://<your-redis-service-name>")
vector_store = RedisStore(redis_database=redis_database)

You can also create a collection directly, but there are two types of collections, one for Hashes and one for JSON.

from semantic_kernel.connectors.memory.redis import RedisHashsetCollection, RedisJsonCollection

hash_collection = RedisHashsetCollection(collection_name="skhotels", data_model_type=Hotel)
json_collection = RedisJsonCollection(collection_name="skhotels", data_model_type=Hotel)

When creating a collection from the vector store, you can pass in the collection type, as a enum: RedisCollectionTypes, the default is a hash collection.

from semantic_kernel.connectors.memory.redis import RedisStore, RedisCollectionTypes

vector_store = RedisStore()
collection = vector_store.get_collection(
    collection_name="skhotels", 
    data_model_type=Hotel, 
    collection_type=RedisCollectionTypes.JSON,
)

Serialization

The redis collections both use a dict as the data format when upserting, however the structure of the dicts are different between them.

For JSON collections see redis docs for a example.

For Hashset collections, it uses the hset command with the key field as name, data fields as mapping -> metadata and vectors as mapping -> [vector_field_name] , see here for more information.

For more details on this concept see the serialization documentation.

Getting started

Include the latest version of the Semantic Kernel Redis data connector in your Maven project by adding the following dependency to your pom.xml:

<dependency>
    <groupId>com.microsoft.semantic-kernel</groupId>
    <artifactId>semantickernel-data-redis</artifactId>
    <version>[LATEST]</version>
</dependency>

You can then create a vector store instance using the RedisVectorStore class, having the Redis client (JedisPooled) as a parameter.

import com.microsoft.semantickernel.data.redis.RedisJsonVectorStoreRecordCollectionOptions;
import com.microsoft.semantickernel.data.redis.RedisStorageType;
import com.microsoft.semantickernel.data.redis.RedisVectorStore;
import com.microsoft.semantickernel.data.redis.RedisVectorStoreOptions;
import redis.clients.jedis.JedisPooled;

public class Main {
    public static void main(String[] args) {
        JedisPooled jedis = new JedisPooled("<your-redis-url>");

        // Build a Redis Vector Store
        // Available storage types are JSON and HASHSET. Default is JSON.
        var vectorStore = RedisVectorStore.builder()
            .withClient(jedis)
            .withOptions(
                RedisVectorStoreOptions.builder()
                    .withStorageType(RedisStorageType.HASH_SET).build())
            .build();
    }
}

You can also retrieve a collection directly.

var collection = vectorStore.getCollection("skhotels",
    RedisJsonVectorStoreRecordCollectionOptions.<Hotel>builder()
        .withRecordClass(Hotel.class)
        .build());

using Microsoft.SemanticKernel.Connectors.Redis;
using StackExchange.Redis;

var collection = new RedisJsonVectorStoreRecordCollection<Hotel>(
    ConnectionMultiplexer.Connect("localhost:6379").GetDatabase(),
    "skhotelsjson",
    new() { PrefixCollectionNameToKeyNames = false });

await collection.GetAsync("myprefix_h1");

from semantic_kernel.connectors.memory.redis import RedisJsonCollection

collection = RedisJsonCollection(collection_name="skhotels", data_model_type=hotel, prefix_collection_name_to_key_names=False)

await collection.get("myprefix_h1")

var collection = vectorStore.getCollection("skhotels",
    RedisJsonVectorStoreRecordCollectionOptions.<Hotel>builder()
        .withRecordClass(Hotel.class)
        .withPrefixCollectionName(false)
        .build());

collection.getAsync("myprefix_h1", null).block();

Data mapping

Redis supports two modes for storing data: JSON and Hashes. The Redis connector supports both storage types, and mapping differs depending on the chosen storage type.

Data mapping when using the JSON storage type

When using the JSON storage type, the Redis connector will use System.Text.Json.JsonSerializer to do mapping. Since Redis stores records with a separate key and value, the mapper will serialize all properties except for the key to a JSON object and use that as the value.

Usage of the JsonPropertyNameAttribute is supported if a different storage name to the data model property name is required. It is also possible to use a custom JsonSerializerOptions instance with a customized property naming policy. To enable this, the JsonSerializerOptions must be passed to the RedisJsonVectorStoreRecordCollection on construction.

var jsonSerializerOptions = new JsonSerializerOptions { PropertyNamingPolicy = JsonNamingPolicy.SnakeCaseUpper };
var collection = new RedisJsonVectorStoreRecordCollection<Hotel>(
    ConnectionMultiplexer.Connect("localhost:6379").GetDatabase(),
    "skhotelsjson",
    new() { JsonSerializerOptions = jsonSerializerOptions });

Since a naming policy of snake case upper was chosen, here is an example of how this data type will be set in Redis. Also note the use of JsonPropertyNameAttribute on the Description property to further customize the storage naming.

using System.Text.Json.Serialization;
using Microsoft.Extensions.VectorData;

public class Hotel
{
    [VectorStoreRecordKey]
    public ulong HotelId { get; set; }

    [VectorStoreRecordData(IsFilterable = true)]
    public string HotelName { get; set; }

    [JsonPropertyName("HOTEL_DESCRIPTION")]
    [VectorStoreRecordData(IsFullTextSearchable = true)]
    public string Description { get; set; }

    [VectorStoreRecordVector(Dimensions: 4, DistanceFunction.CosineDistance, IndexKind.Hnsw)]
    public ReadOnlyMemory<float>? DescriptionEmbedding { get; set; }
}

JSON.SET skhotelsjson:h1 $ '{ "HOTEL_NAME": "Hotel Happy", "HOTEL_DESCRIPTION": "A place where everyone can be happy.", "DESCRIPTION_EMBEDDING": [0.9, 0.1, 0.1, 0.1] }'

Data mapping when using the Hashes storage type

When using the Hashes storage type, the Redis connector provides its own mapper to do mapping. This mapper will map each property to a field-value pair as supported by the Redis HSET command.

For data properties and vector properties, you can provide override field names to use in storage that is different to the property names on the data model. This is not supported for keys, since keys cannot be named in Redis.

Property name overriding is done by setting the StoragePropertyName option via the data model attributes or record definition.

Here is an example of a data model with StoragePropertyName set on its attributes and how these are set in Redis.

using Microsoft.Extensions.VectorData;

public class Hotel
{
    [VectorStoreRecordKey]
    public ulong HotelId { get; set; }

    [VectorStoreRecordData(IsFilterable = true, StoragePropertyName = "hotel_name")]
    public string HotelName { get; set; }

    [VectorStoreRecordData(IsFullTextSearchable = true, StoragePropertyName = "hotel_description")]
    public string Description { get; set; }

    [VectorStoreRecordVector(Dimensions: 4, DistanceFunction.CosineDistance, IndexKind.Hnsw, StoragePropertyName = "hotel_description_embedding")]
    public ReadOnlyMemory<float>? DescriptionEmbedding { get; set; }
}

HSET skhotelshashes:h1 hotel_name "Hotel Happy" hotel_description 'A place where everyone can be happy.' hotel_description_embedding <vector_bytes>