Boost Java AI Apps with DeepSeek4j: Full Chain Support & Reactive Streaming

Why DeepSeek4j?

DeepSeek R1 is praised for its powerful chain‑of‑thought abilities, yet mainstream frameworks like Spring AI lack proper support for chain content retention and streaming output. The newly released DeepSeek4j 1.4 addresses these limitations.

Project repository: https://github.com/pig-mesh/deepseek4j

Documentation: https://javaai.pig4cloud.com/deepseek

Existing Framework Limitations

Chain content loss: The core reasoning process of R1 is completely ignored.

Incompatible response mode: Cannot handle "think first, conclude later" output patterns.

Parameter restrictions: Critical parameters such as temperature and top_p become ineffective.

Streaming handling is incomplete: Leads to poor user experience.

Solution

The author of the Pig project built DeepSeek4j on top of the excellent OpenAI4J architecture, creating a ready‑to‑use solution specifically for DeepSeek.

Core Features

✨ Complete preservation of chain‑of‑thought and billing.

🚀 Reactive streaming support.

🛠 Simple and elegant API design.

📦 Out‑of‑the‑box Spring Boot integration for versions 2.x and 3.x.

💡 Built‑in debugging page.

🔍 Detailed request/response logging.

🔧 Flexible proxy configuration.

⚡️ Full reactive programming support.

Quick Start

Add Dependency :

<dependency>
    <groupId>io.github.pig-mesh.ai</groupId>
    <artifactId>deepseek-spring-boot-starter</artifactId>
    <version>1.1.0</version>
</dependency>

Configuration Parameters (in application.yml):

deepseek:
  api-key: your-api-key-here
  base-url: https://api.deepseek.com/v1  # optional, defaults to official API address

Basic Usage :

@Autowired
private DeepSeekClient deepSeekClient;

// SSE streaming response
@GetMapping(value = "/chat", produces = MediaType.TEXT_EVENT_STREAM_VALUE)
public Flux<ChatCompletionResponse> chat(String prompt) {
    return deepSeekClient.chatFluxCompletion(prompt);
}

Frontend Debugging

Open the sse.html file in the project root, enter the backend SSE endpoint, and click Send to view the inference process and final result in real time. The page includes the complete front‑end implementation code for reference.