A simple Java library for seamless integration of your Java applications with OpenAI API.
Kindly note that this library is not officially endorsed by OpenAI; rather, its development and maintenance are driven by the community.
<dependency>
<groupId>br.com.rcaneppele</groupId>
<artifactId>simple-openai-client</artifactId>
<version>1.5.2</version>
</dependency>implementation 'br.com.rcaneppele:simple-openai-client:1.5.2'
To utilize the library, your project needs to be built with Java 17 or a later version.
var client = new OpenAIClient("YOUR_API_KEY");
var request = new ChatCompletionRequestBuilder()
.model(OpenAIModel.GPT_4_1106_PREVIEW)
.systemMessage("You are a sustainable product name generator")
.userMessage("Please generate two product names")
.build();
var response = client.chatCompletion(request);
System.out.println(response);The response is an object of type ChatCompletion.
If you only need to get the generated message content:
// First choice
System.out.println(response.firstChoiceMessageContent());
// Last choice
System.out.println(response.lastChoiceMessageContent());
// All choices
response.choices().forEach(c -> System.out.println(c.messageContent()));The ChatCompletionRequestBuilder object has methods to set API parameters:
var request = new ChatCompletionRequestBuilder()
.model(OpenAIModel.GPT_4_1106_PREVIEW)
.systemMessage("You are a sustainable product name generator")
.userMessage("Please generate two product names")
.n(3)
.maxTokens(2048)
.temperature(1.3)
.frequencyPenalty(1.4)
.seed(123)
.user("user-id")
.build();
var response = client.chatCompletion(request);
response.choices().forEach(c -> System.out.println(c.messageContent()));If you require Chat Completion Streaming support, you can use the library as shown below:
var request = new ChatCompletionRequestBuilder()
.model(OpenAIModel.GPT_4_1106_PREVIEW)
.systemMessage("You are a sustainable product name generator")
.userMessage("Please generate two product names")
.build();
client.streamChatCompletion(request).subscribe(response -> {
var message = response.firstChoiceMessageContent();
if (message != null) {
System.out.println(message);
}
});Optionally, you can listen to events such as errors and completion during streaming:
client.streamChatCompletion(request).subscribe(response -> {
var message = response.firstChoiceMessageContent();
if (message != null) {
System.out.println(message);
}
}, error -> {
System.out.println("Error during streaming: " +error.getMessage());
}, () -> {
System.out.println("Streaming completed");
});The default OpenAI API response timeout is 15 seconds. If you need to change it:
// 30 seconds timeout
var client = new OpenAIClient("YOUR_API_KEY", 30);
// No timeout
var client = new OpenAIClient("YOUR_API_KEY", 0);To send requests, generate an API KEY
ATTENTION! The API key is sensitive and confidential information; do not use it directly in your code. Instead, use Environment Variables to protect your key:
// Assuming you have an environment variable named OPENAI_API_KEY
var apiKey = System.getenv("OPENAI_API_KEY");
var client = new OpenAIClient(apiKey);var request = (CreateAssistantRequest) new CreateAssistantRequestBuilder()
.model(OpenAIModel.GPT_4_1106_PREVIEW)
.name("Assistant name")
.description("Assistant description")
.instructions("Assistant instructions")
.retrieval()
.codeInterpreter()
.fileIds("fileId-1", "fileId-2", "fileId-3")
.metadata(Map.of("metadata-key-1", "metadata-value-1", "metadata-key-2", "metadata-value-2"))
.build();
var response = client.createAssistant(request);
System.out.println(response);The response is an object of type Assistant.
If you need to create an Assistant with Function Calling support:
// The third parameter is a Map<String, Object> representing the function parameters
var myFunction = new Function("function-name", "function description", Map.of("name", "string", "age", "number"));
var request = (CreateAssistantRequest) new CreateAssistantRequestBuilder()
.model(OpenAIModel.GPT_4_1106_PREVIEW)
.name("Assistant name")
.description("Assistant description")
.instructions("Assistant instructions")
.function(myFunction)
.build();var response = client.createThread();
System.out.println(response);The response is an object of type Thread.
Optionally, you can create a Thread with metadata and a list of messages:
var messageMetadata = Map.of("message-metadata-key", "message-metadata-value");
var threadMetadata = Map.of("thread-metadata-key", "thread-metadata-value");
var fileIds = Set.of("fileId-1", "fileId-2");
var request = new CreateThreadRequestBuilder()
.addUserMessage("message 1", null, null)
.addUserMessage("message 2", fileIds, messageMetadata)
.metadata(threadMetadata)
.build();
var response = client.createThread(request);
System.out.println(response);var request = new CreateMessageRequestBuilder()
.content("content")
.fileIds("fileId-1", "fileId-2")
.metadata(Map.of("key", "value"))
.build();
var response = client.createMessage("thread_id", request);
System.out.println(response);The response is an object of type Message.
var request = new CreateRunRequestBuilder()
.assistantId("assistant-id")
.additionalInstructions("additional instructions")
.build();
var response = client.createRun("thread_id", request);
System.out.println(response);The response is an object of type Run.
var request = new SubmitToolOutputsToRunRequestBuilder()
.toolOutput("tool_call_id", "output")
.build();
var response = client.submitToolOutputsToRun("thread_id", "run_id", request);
System.out.println(response);For detailed usage scenarios and code snippets, refer to the Simple OpenAI Client Examples repository. The examples are organized by endpoint, making it easy to understand and implement the library in your projects.
Feel free to experiment with these examples to accelerate your integration process and unlock the full potential of the OpenAI API in your Java applications.
We welcome contributions from the community! Whether you want to report a bug or contribute code, your help is greatly appreciated. Please take a moment to review our contribution guidelines before getting started.