openai(ChatGPT)의 completions API 사용 시 stream으로 응답받기

Openai(ChatGPT) completions API를 사용할 때 stream으로 응답받는 방법을 java로 만들어보자.

우선 테스트를 하기 위해 api 키가 필요한데 api 키 발급 받는 방법은 생략하도록 한다.

api key 발급: https://platform.openai.com/account/api-keys

api 테스트

우선 발급받은 api가 잘 동작하는 지 api로만 테스트해보자.

intellij에서 제공하는 http를 통해 테스트하기로 한다.

<인증키> 부분은 각자 api 인증키를 추가로 넣어야 한다.

사용 api는 아래를 참고하도록 하자.

https://platform.openai.com/docs/api-reference/chat/create

[gpt.http]

POST https://api.openai.com/v1/chat/completions
content-Type: application/json
Authorization: Bearer <인증키>

{
  "model": "gpt-3.5-turbo",
  "messages": [{
    "role": "user",
    "content": "안녕"
  }],
  "stream": true
}

위의 코드를 실행하면 아래와 같은 결과가 표시된다.

[응답결과]

data: {"id":"chatcmpl-7JAWKy97ys368Xme5fIRjj1ujwfS5","object":"chat.completion.chunk","created":1684803028,"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"role":"assistant"},"index":0,"finish_reason":null}]}

data: {"id":"chatcmpl-7JAWKy97ys368Xme5fIRjj1ujwfS5","object":"chat.completion.chunk","created":1684803028,"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"하세요"},"index":0,"finish_reason":null}]}

<계속>

data: {"id":"chatcmpl-7JAWKy97ys368Xme5fIRjj1ujwfS5","object":"chat.completion.chunk","created":1684803028,"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"?"},"index":0,"finish_reason":null}]}

data: {"id":"chatcmpl-7JAWKy97ys368Xme5fIRjj1ujwfS5","object":"chat.completion.chunk","created":1684803028,"model":"gpt-3.5-turbo-0301","choices":[{"delta":{},"index":0,"finish_reason":"stop"}]}

data: [DONE]

pom.xml

pom.xml에 필요한 정보를 추가하자.

spring-boot-starter-webflux를 사용하여 stream으로 받아올 예정이다.

<parent>
        <groupId>org.springframework.boot</groupId>
        <artifactId>spring-boot-starter-parent</artifactId>
        <version>2.7.9</version>
        <relativePath/>
    </parent>

    <groupId>org.example</groupId>
    <artifactId>chatgpt-test</artifactId>
    <version>1.0-SNAPSHOT</version>

    <properties>
        <maven.compiler.source>8</maven.compiler.source>
        <maven.compiler.target>8</maven.compiler.target>
        <project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
    </properties>

    <dependencies>
        <dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-starter-webflux</artifactId>
        </dependency>
        <dependency>
            <groupId>io.netty</groupId>
            <artifactId>netty-resolver-dns-native-macos</artifactId>
            <version>4.1.73.Final</version>
            <classifier>osx-aarch_64</classifier>
        </dependency>

        <dependency>
            <groupId>org.projectlombok</groupId>
            <artifactId>lombok</artifactId>
        </dependency>
    </dependencies>

OpenAIService

WebClient로 인증키를 사용하여 https://api.openai.com/v1/chat/completions를 호출하고 응답값을 String으로 리턴한다.

@Service
public class OpenAIService {

    @Value("${openai.url}")
    private String openAiUrl;

    @Value("${openai.key}")
    private String openAiKey;

    private WebClient client;

    private final ObjectMapper objectMapper = new ObjectMapper()
        .configure(DeserializationFeature.FAIL_ON_UNKNOWN_PROPERTIES, false)
        .setPropertyNamingStrategy(PropertyNamingStrategies.SNAKE_CASE );

    @PostConstruct
    public void init() {
        client = WebClient.builder()
            .baseUrl(openAiUrl)
            .defaultHeader(HttpHeaders.CONTENT_TYPE, MediaType.APPLICATION_JSON_VALUE)
            .defaultHeader("Authorization", "Bearer " + openAiKey)
            .build();
    }

    public Flux<String> ask(String question) throws JsonProcessingException {
        CompletionRequest request = new CompletionRequest(question);
        String requestValue = objectMapper.writeValueAsString(request);

        Flux<String> eventStream = client.post()
            .bodyValue(requestValue)
            .accept(MediaType.TEXT_EVENT_STREAM)
            .retrieve()
            .bodyToFlux(String.class);

        return eventStream;
    }

OpenAIResource

RestController를 하나 만들어서 요청을 받게 한다.

@RestController
@RequiredArgsConstructor
@Slf4j
public class OpenAIResource {

    private final OpenAIService openAIService;

    @PostMapping(value = "ask", produces = MediaType.TEXT_EVENT_STREAM_VALUE)
    public Flux<String> ask(@RequestBody Question question) {
        try {
            return openAIService.ask(question.getQuestion());
        } catch (JsonProcessingException e) {
            log.error(e.getMessage());
            return Flux.empty();
        }
    }
}

@Getter
public class Question {

    private String question;
}

Application

SpringBoot Application을 띄워서 API요청을 받도록 한다.

@SpringBootApplication
public class Application {

    public static void main(String[] args) {
        SpringApplication.run(Application.class, args);
    }
}

application.yml

server:
  port: 5001


openai:
  url: https://api.openai.com/v1/chat/completions
  key: <인증키>

SpringBoot Application 실행

Application을 실행한다.

API 실행

아래와 같이 API를 실행한다.

POST http://localhost:5001/ask
content-Type: application/json

{
  "question": "안녕"
}

API 응답

아래와 같이 응답이 표시되는 것을 확인할 수 있다.

data:{"id":"chatcmpl-7JAf0j9pHX1DEst6VNUhYoK5xoeEj","object":"chat.completion.chunk","created":1684803566,"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"role":"assistant"},"index":0,"finish_reason":null}]}

data:{"id":"chatcmpl-7JAf0j9pHX1DEst6VNUhYoK5xoeEj","object":"chat.completion.chunk","created":1684803566,"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"하세요"},"index":0,"finish_reason":null}]}

<계속>

data:{"id":"chatcmpl-7JAf0j9pHX1DEst6VNUhYoK5xoeEj","object":"chat.completion.chunk","created":1684803566,"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"?"},"index":0,"finish_reason":null}]}

data:{"id":"chatcmpl-7JAf0j9pHX1DEst6VNUhYoK5xoeEj","object":"chat.completion.chunk","created":1684803566,"model":"gpt-3.5-turbo-0301","choices":[{"delta":{},"index":0,"finish_reason":"stop"}]}

data:[DONE]