Vector Library

The AI-native Redis Java client

Home    Documentation    Recipes   

Introduction

Redis Vector Library (RedisVL) is the ultimate Java client designed for AI-native applications harnessing the power of Redis.

RedisVL is your go-to client for:

• Lightning-fast information retrieval & vector similarity search
• Real-time RAG pipelines
• LLM semantic caching
• Hybrid search combining vectors and metadata

💪 Getting Started

Installation

Add RedisVL to your Java (17+) project using Maven or Gradle:

Maven:

<dependency>
    <groupId>com.redis</groupId>
    <artifactId>redisvl</artifactId>
    <version>0.0.1-SNAPSHOT</version>
</dependency>

Gradle:

implementation 'com.redis:redisvl:0.0.1-SNAPSHOT'

Setting up Redis

Choose from multiple Redis deployment options:

1. Redis Cloud: Managed cloud database (free tier available)
2. Redis Stack: Docker image for development

   docker run -d --name redis-stack -p 6379:6379 -p 8001:8001 redis/redis-stack:latest

3. Redis Enterprise: Commercial, self-hosted database
4. Azure Managed Redis: Fully managed Redis Enterprise on Azure

Enhance your experience and observability with the free Redis Insight GUI.

Overview

🗃️ Redis Index Management

1. Design a schema for your use case that models your dataset with built-in Redis and indexable fields (e.g. text, tags, numerics, geo, and vectors). Load a schema from a YAML file:

   index:
     name: user-idx
     prefix: user
     storage_type: json
   
   fields:
     - name: user
       type: tag
     - name: credit_score
       type: tag
     - name: embedding
       type: vector
       attrs:
         algorithm: flat
         dims: 4
         distance_metric: cosine
         datatype: float32

   import com.redis.vl.schema.IndexSchema;
   
   IndexSchema schema = IndexSchema.fromYaml("schemas/schema.yaml");

   Or load directly from a Java Map:

   Map<String, Object> schemaMap = Map.of(
       "index", Map.of(
           "name", "user-idx",
           "prefix", "user",
           "storage_type", "json"
       ),
       "fields", List.of(
           Map.of("name", "user", "type", "tag"),
           Map.of("name", "credit_score", "type", "tag"),
           Map.of(
               "name", "embedding",
               "type", "vector",
               "attrs", Map.of(
                   "algorithm", "flat",
                   "datatype", "float32",
                   "dims", 4,
                   "distance_metric", "cosine"
               )
           )
       )
   );
   
   IndexSchema schema = IndexSchema.fromMap(schemaMap);

2. Create a SearchIndex class with an input schema to perform admin and search operations on your index in Redis:

   import redis.clients.jedis.UnifiedJedis;
   import com.redis.vl.index.SearchIndex;
   
   // Define the index
   UnifiedJedis jedis = new UnifiedJedis("redis://localhost:6379");
   SearchIndex index = new SearchIndex(schema, jedis);
   
   // Create the index in Redis
   index.create(true); // overwrite if exists

3. Load and fetch data to/from your Redis instance:

   Map<String, Object> data = Map.of(
       "user", "john",
       "credit_score", "high",
       "embedding", new float[]{0.23f, 0.49f, -0.18f, 0.95f}
   );
   
   // load list of maps, specify the "id" field
   index.load(List.of(data), "user");
   
   // fetch by "id"
   Map<String, Object> john = index.fetch("john");

🔍 Retrieval

Define queries and perform advanced searches over your indices, including the combination of vectors, metadata filters, and more.

• VectorQuery - Flexible vector queries with customizable filters enabling semantic search:

  import com.redis.vl.query.VectorQuery;
  
  VectorQuery query = VectorQuery.builder()
      .vector(new float[]{0.16f, -0.34f, 0.98f, 0.23f})
      .field("embedding")
      .numResults(3)
      .build();
  
  // run the vector search query against the embedding field
  List<Map<String, Object>> results = index.query(query);

  Incorporate complex metadata filters on your queries:

  import com.redis.vl.query.filter.FilterQuery;
  
  // define a tag match filter
  FilterQuery tagFilter = FilterQuery.tag("user", "john");
  
  // update query definition
  VectorQuery filteredQuery = query.toBuilder()
      .withPreFilter(tagFilter.build())
      .build();
  
  // execute query
  List<Map<String, Object>> results = index.query(filteredQuery);

• VectorRangeQuery - Vector search within a defined range paired with customizable filters

• FilterQuery - Standard search using filters and the full-text search

• CountQuery - Count the number of indexed records given attributes

Read more about building advanced Redis queries.

🔧 Utilities

Vectorizers

Integrate with popular embedding providers to greatly simplify the process of vectorizing unstructured data for your index and queries:

• LangChain4J - Integration with LangChain4J embedding models

  • OpenAI
  • Cohere
  • HuggingFace
  • Ollama
  • Vertex AI
  • Azure OpenAI
  • Mistral AI
  • Voyage AI

• ONNX Models - Local embedding models via ONNX Runtime

  • Sentence Transformers (all-MiniLM-L6-v2, etc.)
  • Custom ONNX models

• DJL (Deep Java Library) - Face recognition and embeddings

import com.redis.vl.utils.vectorize.LangChain4JVectorizer;
import dev.langchain4j.model.cohere.CoherEmbeddingModel;

// Using LangChain4J with Cohere
CoherEmbeddingModel cohereModel = CoherEmbeddingModel.withApiKey(apiKey);
LangChain4JVectorizer vectorizer = new LangChain4JVectorizer(cohereModel);

float[] embedding = vectorizer.embed("What is the capital city of France?");

List<float[]> embeddings = vectorizer.embedBatch(
    List.of(
        "my document chunk content",
        "my other document chunk content"
    )
);

import com.redis.vl.utils.vectorize.SentenceTransformersVectorizer;

// Using local ONNX model
SentenceTransformersVectorizer vectorizer = new SentenceTransformersVectorizer();

float[] embedding = vectorizer.embed("What is the capital city of France?");

Learn more about using vectorizers in your embedding workflows.

Rerankers

Integrate with popular reranking providers to improve the relevancy of the initial search results from Redis:

• Cohere Reranker - Using Cohere's reranking API
• HuggingFace Cross-Encoder - Local reranking with cross-encoder models

import com.redis.vl.utils.rerank.CohereReranker;

CohereReranker reranker = new CohereReranker(apiKey);

List<Map<String, Object>> reranked = reranker.rerank(
    "What is the capital of France?",
    results,
    topK
);

💫 Extensions

RedisVL supports Extensions that implement interfaces exposing best practices and design patterns for working with LLM memory and agents.

LLM Semantic Caching

Increase application throughput and reduce the cost of using LLM models in production by leveraging previously generated knowledge with the SemanticCache.

import com.redis.vl.extensions.cache.llm.SemanticCache;
import com.redis.vl.extensions.cache.llm.CacheConfiguration;

// init cache with TTL and semantic distance threshold
CacheConfiguration config = CacheConfiguration.builder()
    .name("llmcache")
    .ttl(360L)
    .distanceThreshold(0.1)
    .build();

SemanticCache cache = new SemanticCache(config, jedis);

// store user queries and LLM responses in the semantic cache
cache.store(
    "What is the capital city of France?",
    "Paris"
);

// quickly check the cache with a slightly different prompt (before invoking an LLM)
CacheResult result = cache.check("What is France's capital city?");
if (result.isHit()) {
    System.out.println(result.getResponse());
    // Output: Paris
}

Learn more about semantic caching for LLMs.

LLM Semantic Routing

Build fast decision models that run directly in Redis and route user queries to the nearest "route" or "topic".

import com.redis.vl.extensions.router.SemanticRouter;
import com.redis.vl.extensions.router.Route;

List<Route> routes = List.of(
    Route.builder()
        .name("greeting")
        .references(List.of("hello", "hi"))
        .metadata(Map.of("type", "greeting"))
        .distanceThreshold(0.3)
        .build(),
    Route.builder()
        .name("farewell")
        .references(List.of("bye", "goodbye"))
        .metadata(Map.of("type", "farewell"))
        .distanceThreshold(0.3)
        .build()
);

// build semantic router from routes
SemanticRouter router = SemanticRouter.builder()
    .name("topic-router")
    .routes(routes)
    .jedis(jedis)
    .build();

RouteMatch match = router.route("Hi, good morning");
System.out.println(match.getName()); // Output: greeting
System.out.println(match.getDistance()); // Output: 0.273891836405

Learn more about semantic routing.

🚀 Why RedisVL?

In the age of GenAI, vector databases and LLMs are transforming information retrieval systems. With emerging and popular frameworks like LangChain4J and Spring AI, innovation is rapid. Yet, many organizations face the challenge of delivering AI solutions quickly and at scale.

Enter Redis – a cornerstone of the NoSQL world, renowned for its versatile data structures and processing engines. Redis excels in real-time workloads like caching, session management, and search. It's also a powerhouse as a vector database for RAG, an LLM cache, and a chat session memory store for conversational AI.

The Redis Vector Library bridges the gap between the AI-native developer ecosystem and Redis's robust capabilities. With a lightweight, elegant, and intuitive interface, RedisVL makes it easy to leverage Redis's power. Built on the Jedis client, RedisVL transforms Redis's features into a grammar perfectly aligned with the needs of today's AI/ML Engineers and Data Scientists.

📚 Examples and Notebooks

Check out the notebooks directory for interactive Jupyter notebook examples:

• Getting Started - Introduction to RedisVL basics
• Hybrid Queries - Combining vector and metadata search
• LLM Cache - Semantic caching for LLMs
• Hash vs JSON Storage - Storage type comparison
• Vectorizers - Working with embedding models
• Rerankers - Improving search relevance

Running the Notebooks

The notebooks use Java kernel support via JJava. To run them:

1. Start the notebook environment:

   cd notebooks
   docker compose up -d

2. Access Jupyter Lab: Navigate to http://localhost:8888/

3. Stop when done:

   docker compose down

See the notebooks README for more details.

😁 Helpful Links

For additional help, check out the following resources:

• Getting Started Guide
• API Reference (Javadoc)
• Example Notebooks
• Redis AI Recipes
• Official Redis Vector API Docs

🫱🏼‍🫲🏽 Contributing

Please help us by contributing PRs, opening GitHub issues for bugs or new feature ideas, improving documentation, or increasing test coverage. Read more about how to contribute!

Requirements

• Java 17+
• Redis Stack 7.2+ or Redis with RediSearch module

License

This project is licensed under the MIT License - see the LICENSE file for details.

🚧 Maintenance

This project is supported by Redis, Inc on a good faith effort basis. To report bugs, request features, or receive assistance, please file an issue.