Skip to main content

Spring AI EmbeddingModel

Banner Spring AI Icon

🚀 Spring AI Vector Embeddings – A Fun & Engaging Guide​

Welcome to the Spring AI Vector Embedding tutorial! 🎉 Here, we'll unravel the mysteries of vectors, embeddings, and how they supercharge semantic searches. Plus, we’ll generate embeddings using popular LLM models like OpenAI and DeepseekAI using Ollama! 🤖✨

🤔 1. What is a Vector Embedding?​

Imagine you're in a room full of people speaking different languages. How do you find someone who understands you? Vector embeddings work like a universal translator—turning words, sentences, and even images into arrays of numbers that capture meaning! 🔢💡

In AI terms, a vector (or embedding) is a magical array of numbers that represents an object, be it a word, sentence, image, audio snippet, or even a product description. 📦

Each number in the vector is a component, corresponding to a dimension in hyperspace (yes, like sci-fi 🚀). More dimensions mean more details about the object!

🌟 Example Time​

Let’s take the word “happiness” and give it some numbers:

[2.5, 1.3, 0.8]
  • 2.5 → Emotional positivity (yay! 🎉)
  • 1.3 → Emotional intensity (oh wow! 😃)
  • 0.8 → Context usage (poetry or science? 🤔)

Now, here’s the kicker: these numbers don’t mean much alone, but together, they create a powerful representation! 🧠✨

🔍 2. Semantic Search: Finding Meaning, Not Just Words​

Ever wondered how Google knows exactly what you're looking for, even when you type weird search queries? That’s semantic search in action! 🕵️‍♂️💡

Instead of matching exact words, semantic search looks at meaning. It uses vector distances—small distances = high similarity, big distances = low similarity. 🔬

🚀 Real-world applications:

  • Google Search 🧐 – Finds the perfect answer, even if your query is gibberish.
  • Netflix & Spotify 🎶 – Suggests movies & songs you actually like.
  • Chatbots & AI Assistants 🤖 – Understands you better than your best friend!

🏗 3. Meet the EmbeddingModel Interface​

Spring AI makes embedding easy with the EmbeddingModel interface! 🎯

Think of it as your personal AI librarian, transforming text into numerical vectors! 📚🔢

public interface EmbeddingModel extends Model<EmbeddingRequest, EmbeddingResponse> {
  List<Double> embed(String text);
  List<Double> embed(Document document);
  EmbeddingResponse embedForResponse(List<String> texts);
  EmbeddingResponse call(EmbeddingRequest request);
}

spring-ai-embedding-model

💡 The embedForResponse() method gives extra metadata, just in case you want more than just numbers!

⚡ Usage Example​

String text = "...";
List<Double> embeddings = embeddingModel.embed(text);

Or, for more control:

List<Double> embeddings = embeddingModel.call(
  new EmbeddingRequest(List.of(text), OpenAiEmbeddingOptions.builder()
    .withModel("text-embedding-3-small")
    .withUser("jon.snow") // Because he knows nothing, right? 😆
    .build()))
  .getResult().getOutput();

🚀 4. EmbeddingModel Implementations​

Spring AI supports tons of embedding models! 🏆 Each has its own implementation class.

ProviderImplementation ClassSupported Models
OpenAIOpenAiEmbeddingModeltext-embedding-ada-002, text-embedding-3-large, text-embedding-3-small
Azure OpenAIAzureOpenAiEmbeddingModeltext-embedding-ada-002
OllamaOllamaEmbeddingModelmistral
PostgresMLPostgresMlEmbeddingModeldistilbert-base-uncased
Mistral AIMistralAiEmbeddingModelmistral-embed
MiniMaxMiniMaxEmbeddingModelembo-01
QianFanQianFanEmbeddingModelbge_large_zh
Amazon BedrockBedrockTitanEmbeddingModelamazon.titan-embed-image-v1

📜 For the full list, check the official docs!

🎬 5. Let’s Get Our Hands Dirty: DeepseekAI Demo​

We’ll use Ollama’s mxbai-embed-large and nomic-embed-text model to generate vector embeddings. 🛠✨

🛠 5.1. Maven Setup​

First, add the Spring Boot starter dependency:

<dependencies>
  <dependency>
    <groupId>org.springframework.ai</groupId>
    <artifactId>spring-ai-ollama-spring-boot-starter</artifactId>
  </dependency>
</dependencies>

🔑 Don’t forget to add your application properties!

server.port=8084
spring.ai.ollama.base-url=http://localhost:11434
spring.ai.ollama.chat.options.model=deepseek-r1:latest
spring.ai.ollama.chat.options.temperature=0.7
spring.ai.ollama.init.pull-model-strategy=always
spring.ai.ollama.init.pull-model-strategy.embedding.additional-models=mxbai-embed-large

🎛 5.2. DeepseekAiEmbeddingModel Bean​

Spring Boot auto-configures an DeepseekAiEmbeddingModel bean. Just inject it into your class:

@RestController
public class DeepseekAiEmbeddingController {

    private final EmbeddingModel embeddingModel;

    @Autowired
    public DeepseekAiEmbeddingController(EmbeddingModel embeddingModel) {
        this.embeddingModel = embeddingModel;
    }

    @GetMapping("/ai/embedding")
    public Map embed(@RequestParam(value = "message", defaultValue = "tell me a joke") String message) {
        EmbeddingResponse embeddingResponse = embeddingModel.embedForResponse(List.of(message));
        return Map.of("embedding", embeddingResponse);
    }

}

🛠 Test it with a REST tool! 🚀

spring-ai-embedding-result

🎯 6. Wrapping It Up​

🎉 Here’s what we covered:

✅ Vectors (embeddings) represent objects in numerical form. ✅ Semantic search finds meaning instead of just words. ✅ The EmbeddingModel interface is your AI magic wand! 🪄 ✅ Spring AI supports multiple embedding models. ✅ We saw a live demo using Deepseek AI with Ollama!

💡 What’s next? Start building cool AI-powered apps! 🚀💡

👉 Happy Learning & Happy Coding! 🦾🔥