Bob’s constructing a buyer assist system that may:
- Summarize messages
- Auto-reply utilizing a language mannequin
- Classify sentiment
- Generate product descriptions
As an alternative of constructing fashions from scratch, Bob makes use of:
- OpenAI API for LLMs (GPT-like magic)
- TensorFlow Java for native ML inference
[Client]
⬇️
[Spring Boot API]
⬇️⬇️
[OpenAI API] ← LLM duties (chat, summarization)
[TensorFlow] ← In-house fashions (sentiment, picture tags)
Bob’s API acts as a sensible gateway, combining cloud-based AI with native mannequin inference.
@Configuration
public class OpenAIConfig {
@Bean
public WebClient openAIClient() {
return WebClient.builder()
.baseUrl("https://api.openai.com/v1")
.defaultHeader("Authorization", "Bearer " + System.getenv("OPENAI_API_KEY"))
.construct();
}
}
@Service
public class GPTService {
@Autowired
personal WebClient openAIClient;public Mono getChatResponse(String immediate) {
return openAIClient.publish()
.uri("/chat/completions")
.bodyValue(Map.of(
"mannequin", "gpt-3.5-turbo",
"messages", Record.of(Map.of("function", "person", "content material", immediate))
))
.retrieve()
.bodyToMono(String.class);
}
}
Helps:
- Summarization
- Textual content technology
- Translation
- Code technology
