OpenAI Java API Library

The OpenAI Java SDK provides convenient access to the OpenAI REST API from applications written in Java.

The REST API documentation can be found on platform.openai.com. Javadocs are available on javadoc.io.

Installation

Try openai-java-spring-boot-starter if you're using Spring Boot!

Gradle

implementation("com.openai:openai-java:4.32.0")

Maven

<dependency>
  <groupId>com.openai</groupId>
  <artifactId>openai-java</artifactId>
  <version>4.32.0</version>
</dependency>

Requirements

This library requires Java 8 or later.

Usage

Tip

See the openai-java-example directory for complete and runnable examples!

The primary API for interacting with OpenAI models is the Responses API. You can generate text from the model with the code below.

import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.models.ChatModel;
import com.openai.models.responses.Response;
import com.openai.models.responses.ResponseCreateParams;

// Configures using the `OPENAI_API_KEY`, `OPENAI_ORG_ID` and `OPENAI_PROJECT_ID` environment variables
OpenAIClient client = OpenAIOkHttpClient.fromEnv();

ResponseCreateParams params = ResponseCreateParams.builder()
        .input("Say this is a test")
        .model(ChatModel.GPT_5_2)
        .build();
Response response = client.responses().create(params);

The previous standard (supported indefinitely) for generating text is the Chat Completions API. You can use that API to generate text from the model with the code below.

import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.models.ChatModel;
import com.openai.models.chat.completions.ChatCompletion;
import com.openai.models.chat.completions.ChatCompletionCreateParams;

// Configures using the `openai.apiKey`, `openai.orgId`, `openai.projectId`, `openai.webhookSecret` and `openai.baseUrl` system properties
// Or configures using the `OPENAI_API_KEY`, `OPENAI_ORG_ID`, `OPENAI_PROJECT_ID`, `OPENAI_WEBHOOK_SECRET` and `OPENAI_BASE_URL` environment variables
OpenAIClient client = OpenAIOkHttpClient.fromEnv();

ChatCompletionCreateParams params = ChatCompletionCreateParams.builder()
    .addUserMessage("Say this is a test")
    .model(ChatModel.GPT_5_2)
    .build();
ChatCompletion chatCompletion = client.chat().completions().create(params);

Client configuration

Configure the client using system properties or environment variables:

import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;

// Configures using the `openai.apiKey`, `openai.orgId`, `openai.projectId`, `openai.webhookSecret` and `openai.baseUrl` system properties
// Or configures using the `OPENAI_API_KEY`, `OPENAI_ORG_ID`, `OPENAI_PROJECT_ID`, `OPENAI_WEBHOOK_SECRET` and `OPENAI_BASE_URL` environment variables
OpenAIClient client = OpenAIOkHttpClient.fromEnv();

Or manually:

import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;

OpenAIClient client = OpenAIOkHttpClient.builder()
    .apiKey("My API Key")
    .build();

Or using a combination of the two approaches:

import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;

OpenAIClient client = OpenAIOkHttpClient.builder()
    // Configures using the `openai.apiKey`, `openai.orgId`, `openai.projectId`, `openai.webhookSecret` and `openai.baseUrl` system properties
    // Or configures using the `OPENAI_API_KEY`, `OPENAI_ORG_ID`, `OPENAI_PROJECT_ID`, `OPENAI_WEBHOOK_SECRET` and `OPENAI_BASE_URL` environment variables
    .fromEnv()
    .apiKey("My API Key")
    .build();

See this table for the available options:

Setter	System property	Environment variable	Required	Default value
`apiKey`	`openai.apiKey`	`OPENAI_API_KEY`	true	-
`organization`	`openai.orgId`	`OPENAI_ORG_ID`	false	-
`project`	`openai.projectId`	`OPENAI_PROJECT_ID`	false	-
`webhookSecret`	`openai.webhookSecret`	`OPENAI_WEBHOOK_SECRET`	false	-
`baseUrl`	`openai.baseUrl`	`OPENAI_BASE_URL`	true	`"https://api.openai.com/v1"`

System properties take precedence over environment variables.

Tip

Don't create more than one client in the same application. Each client has a connection pool and thread pools, which are more efficient to share between requests.

Modifying configuration

To temporarily use a modified client configuration, while reusing the same connection and thread pools, call withOptions() on any client or service:

import com.openai.client.OpenAIClient;

OpenAIClient clientWithOptions = client.withOptions(optionsBuilder -> {
    optionsBuilder.baseUrl("https://example.com");
    optionsBuilder.maxRetries(42);
});

The withOptions() method does not affect the original client or service.

Workload identity authentication

Workload identity authentication allows applications running in cloud environments (Kubernetes, Azure, GCP) to authenticate using short-lived tokens issued by the cloud provider, instead of long-lived API keys.

Basic setup

import com.openai.auth.*;
import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;

SubjectTokenProvider provider = K8sServiceAccountTokenProvider.builder().build();

WorkloadIdentity workloadIdentity = WorkloadIdentity.builder()
    .clientId("your-client-id")
    .identityProviderId("your-identity-provider-id")
    .serviceAccountId("your-service-account-id")
    .provider(provider)
    .build();

OpenAIClient client = OpenAIOkHttpClient.builder()
    .workloadIdentity(workloadIdentity)
    .build();

Kubernetes service account token provider

// Use default token path (/var/run/secrets/kubernetes.io/serviceaccount/token)
SubjectTokenProvider provider = K8sServiceAccountTokenProvider.builder().build();

// Or specify a custom token path
SubjectTokenProvider provider = K8sServiceAccountTokenProvider.builder()
    .tokenPath("/custom/path/to/token")
    .build();

Azure Managed Identity provider

import com.openai.auth.*;

// Use defaults (resource: https://management.azure.com/, api-version: 2018-02-01)
SubjectTokenProvider provider = AzureManagedIdentityTokenProvider.builder()
    .build();

import com.openai.auth.*;

// Or customize
SubjectTokenProvider provider = AzureManagedIdentityTokenProvider.builder()
    .resource("https://management.azure.com/")
    .apiVersion("2018-02-01")
    .build();

GCP ID token provider

import com.openai.auth.*;

SubjectTokenProvider provider = GcpIdTokenProvider.builder()
    .build();

import com.openai.auth.*;

// Or customize the audience
SubjectTokenProvider provider = GcpIdTokenProvider.builder()
    .audience("https://api.openai.com/v1")
    .build();

Requests and responses

To send a request to the OpenAI API, build an instance of some Params class and pass it to the corresponding client method. When the response is received, it will be deserialized into an instance of a Java class.

For example, client.chat().completions().create(...) should be called with an instance of ChatCompletionCreateParams, and it will return an instance of ChatCompletion.

Immutability

Each class in the SDK has an associated builder or factory method for constructing it.

Each class is immutable once constructed. If the class has an associated builder, then it has a toBuilder() method, which can be used to convert it back to a builder for making a modified copy.

Because each class is immutable, builder modification will never affect already built class instances.

Asynchronous execution

The default client is synchronous. To switch to asynchronous execution, call the async() method:

import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.models.ChatModel;
import com.openai.models.chat.completions.ChatCompletion;
import com.openai.models.chat.completions.ChatCompletionCreateParams;
import java.util.concurrent.CompletableFuture;

// Configures using the `openai.apiKey`, `openai.orgId`, `openai.projectId`, `openai.webhookSecret` and `openai.baseUrl` system properties
// Or configures using the `OPENAI_API_KEY`, `OPENAI_ORG_ID`, `OPENAI_PROJECT_ID`, `OPENAI_WEBHOOK_SECRET` and `OPENAI_BASE_URL` environment variables
OpenAIClient client = OpenAIOkHttpClient.fromEnv();

ChatCompletionCreateParams params = ChatCompletionCreateParams.builder()
    .addUserMessage("Say this is a test")
    .model(ChatModel.GPT_5_2)
    .build();
CompletableFuture<ChatCompletion> chatCompletion = client.async().chat().completions().create(params);

Or create an asynchronous client from the beginning:

import com.openai.client.OpenAIClientAsync;
import com.openai.client.okhttp.OpenAIOkHttpClientAsync;
import com.openai.models.ChatModel;
import com.openai.models.chat.completions.ChatCompletion;
import com.openai.models.chat.completions.ChatCompletionCreateParams;
import java.util.concurrent.CompletableFuture;

// Configures using the `openai.apiKey`, `openai.orgId`, `openai.projectId`, `openai.webhookSecret` and `openai.baseUrl` system properties
// Or configures using the `OPENAI_API_KEY`, `OPENAI_ORG_ID`, `OPENAI_PROJECT_ID`, `OPENAI_WEBHOOK_SECRET` and `OPENAI_BASE_URL` environment variables
OpenAIClientAsync client = OpenAIOkHttpClientAsync.fromEnv();

ChatCompletionCreateParams params = ChatCompletionCreateParams.builder()
    .addUserMessage("Say this is a test")
    .model(ChatModel.GPT_5_2)
    .build();
CompletableFuture<ChatCompletion> chatCompletion = client.chat().completions().create(params);

The asynchronous client supports the same options as the synchronous one, except most methods return CompletableFutures.

Streaming

The SDK defines methods that return response "chunk" streams, where each chunk can be individually processed as soon as it arrives instead of waiting on the full response. Streaming methods generally correspond to SSE or JSONL responses.

Some of these methods may have streaming and non-streaming variants, but a streaming method will always have a Streaming suffix in its name, even if it doesn't have a non-streaming variant.

These streaming methods return StreamResponse for synchronous clients:

import com.openai.core.http.StreamResponse;
import com.openai.models.chat.completions.ChatCompletionChunk;

try (StreamResponse<ChatCompletionChunk> streamResponse = client.chat().completions().createStreaming(params)) {
    streamResponse.stream().forEach(chunk -> {
        System.out.println(chunk);
    });
    System.out.println("No more chunks!");
}

Or AsyncStreamResponse for asynchronous clients:

import com.openai.core.http.AsyncStreamResponse;
import com.openai.models.chat.completions.ChatCompletionChunk;
import java.util.Optional;

client.async().chat().completions().createStreaming(params).subscribe(chunk -> {
    System.out.println(chunk);
});

// If you need to handle errors or completion of the stream
client.async().chat().completions().createStreaming(params).subscribe(new AsyncStreamResponse.Handler<>() {
    @Override
    public void onNext(ChatCompletionChunk chunk) {
        System.out.println(chunk);
    }

    @Override
    public void onComplete(Optional<Throwable> error) {
        if (error.isPresent()) {
            System.out.println("Something went wrong!");
            throw new RuntimeException(error.get());
        } else {
            System.out.println("No more chunks!");
        }
    }
});

// Or use futures
client.async().chat().completions().createStreaming(params)
    .subscribe(chunk -> {
        System.out.println(chunk);
    })
    .onCompleteFuture();
    .whenComplete((unused, error) -> {
        if (error != null) {
            System.out.println("Something went wrong!");
            throw new RuntimeException(error);
        } else {
            System.out.println("No more chunks!");
        }
    });

Async streaming uses a dedicated per-client cached thread pool Executor to stream without blocking the current thread. This default is suitable for most purposes.

To use a different Executor, configure the subscription using the executor parameter:

import java.util.concurrent.Executor;
import java.util.concurrent.Executors;

Executor executor = Executors.newFixedThreadPool(4);
client.async().chat().completions().createStreaming(params).subscribe(
    chunk -> System.out.println(chunk), executor
);

Or configure the client globally using the streamHandlerExecutor method:

import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import java.util.concurrent.Executors;

OpenAIClient client = OpenAIOkHttpClient.builder()
    .fromEnv()
    .streamHandlerExecutor(Executors.newFixedThreadPool(4))
    .build();

Streaming helpers

The SDK provides conveniences for streamed chat completions. A ChatCompletionAccumulator can record the stream of chat completion chunks in the response as they are processed and accumulate a ChatCompletion object similar to that which would have been returned by the non-streaming API.

For a synchronous response add a Stream.peek() call to the stream pipeline to accumulate each chunk:

import com.openai.core.http.StreamResponse;
import com.openai.helpers.ChatCompletionAccumulator;
import com.openai.models.chat.completions.ChatCompletion;
import com.openai.models.chat.completions.ChatCompletionChunk;

ChatCompletionAccumulator chatCompletionAccumulator = ChatCompletionAccumulator.create();

try (StreamResponse<ChatCompletionChunk> streamResponse =
        client.chat().completions().createStreaming(createParams)) {
    streamResponse.stream()
            .peek(chatCompletionAccumulator::accumulate)
            .flatMap(completion -> completion.choices().stream())
            .flatMap(choice -> choice.delta().content().stream())
            .forEach(System.out::print);
}

ChatCompletion chatCompletion = chatCompletionAccumulator.chatCompletion();

For an asynchronous response, add the ChatCompletionAccumulator to the subscribe() call:

import com.openai.helpers.ChatCompletionAccumulator;
import com.openai.models.chat.completions.ChatCompletion;

ChatCompletionAccumulator chatCompletionAccumulator = ChatCompletionAccumulator.create();

client.chat()
        .completions()
        .createStreaming(createParams)
        .subscribe(chunk -> chatCompletionAccumulator.accumulate(chunk).choices().stream()
                .flatMap(choice -> choice.delta().content().stream())
                .forEach(System.out::print))
        .onCompleteFuture()
        .join();

ChatCompletion chatCompletion = chatCompletionAccumulator.chatCompletion();

The SDK provides conveniences for streamed responses. A ResponseAccumulator can record the stream of response events as they are processed and accumulate a Response object similar to that which would have been returned by the non-streaming API.

For a synchronous response add a Stream.peek() call to the stream pipeline to accumulate each event:

import com.openai.core.http.StreamResponse;
import com.openai.helpers.ResponseAccumulator;
import com.openai.models.responses.Response;
import com.openai.models.responses.ResponseStreamEvent;

ResponseAccumulator responseAccumulator = ResponseAccumulator.create();

try (StreamResponse<ResponseStreamEvent> streamResponse =
        client.responses().createStreaming(createParams)) {
    streamResponse.stream()
            .peek(responseAccumulator::accumulate)
            .flatMap(event -> event.outputTextDelta().stream())
            .forEach(textEvent -> System.out.print(textEvent.delta()));
}

Response response = responseAccumulator.response();

For an asynchronous response, add the ResponseAccumulator to the subscribe() call:

import com.openai.helpers.ResponseAccumulator;
import com.openai.models.responses.Response;

ResponseAccumulator responseAccumulator = ResponseAccumulator.create();

client.responses()
        .createStreaming(createParams)
        .subscribe(event -> responseAccumulator.accumulate(event)
                .outputTextDelta().ifPresent(textEvent -> System.out.print(textEvent.delta())))
        .onCompleteFuture()
        .join();

Response response = responseAccumulator.response();

Structured outputs with JSON schemas

Open AI Structured Outputs is a feature that ensures that the model will always generate responses that adhere to a supplied JSON schema.

A JSON schema can be defined by creating a ResponseFormatJsonSchema and setting it on the input parameters. However, for greater convenience, a JSON schema can instead be derived automatically from the structure of an arbitrary Java class. The JSON content from the response will then be converted automatically to an instance of that Java class. A full, working example of the use of Structured Outputs with arbitrary Java classes can be seen in StructuredOutputsExample.

Java classes can contain fields declared to be instances of other classes and can use collections (see Defining JSON schema properties for more details):

class Person {
    public String name;
    public int birthYear;
}

class Book {
    public String title;
    public Person author;
    public int publicationYear;
}

class BookList {
    public List<Book> books;
}

Pass the top-level class—BookList in this example—to responseFormat(Class<T>) when building the parameters and then access an instance of BookList from the generated message content in the response:

import com.openai.models.ChatModel;
import com.openai.models.chat.completions.ChatCompletionCreateParams;
import com.openai.models.chat.completions.StructuredChatCompletionCreateParams;

StructuredChatCompletionCreateParams<BookList> params = ChatCompletionCreateParams.builder()
        .addUserMessage("List some famous late twentieth century novels.")
        .model(ChatModel.GPT_5_2)
        .responseFormat(BookList.class)
        .build();

client.chat().completions().create(params).choices().stream()
        .flatMap(choice -> choice.message().content().stream())
        .flatMap(bookList -> bookList.books.stream())
        .forEach(book -> System.out.println(book.title + " by " + book.author.name));

You can start building the parameters with an instance of ChatCompletionCreateParams.Builder or StructuredChatCompletionCreateParams.Builder. If you start with the former (which allows for more compact code) the builder type will change to the latter when ChatCompletionCreateParams.Builder.responseFormat(Class<T>) is called.

If a field in a class is optional and does not require a defined value, you can represent this using the java.util.Optional class. It is up to the AI model to decide whether to provide a value for that field or leave it empty.

import java.util.Optional;

class Book {
    public String title;
    public Person author;
    public int publicationYear;
    public Optional<String> isbn;
}

Generic type information for fields is retained in the class's metadata, but generic type erasure applies in other scopes. While, for example, a JSON schema defining an array of books can be derived from the BookList.books field with type List<Book>, a valid JSON schema cannot be derived from a local variable of that same type, so the following will not work:

List<Book> books = new ArrayList<>();

StructuredChatCompletionCreateParams<List<Book>> params = ChatCompletionCreateParams.builder()
        .responseFormat(books.getClass())
        // ...
        .build();

If an error occurs while converting a JSON response to an instance of a Java class, the error message will include the JSON response to assist in diagnosis. For instance, if the response is truncated, the JSON data will be incomplete and cannot be converted to a class instance. If your JSON response may contain sensitive information, avoid logging it directly, or ensure that you redact any sensitive details from the error message.

Local JSON schema validation

Structured Outputs supports a subset of the JSON Schema language. Schemas are generated automatically from classes to align with this subset. However, due to the inherent structure of the classes, the generated schema may still violate certain OpenAI schema restrictions, such as exceeding the maximum nesting depth or utilizing unsupported data types.

To facilitate compliance, the method responseFormat(Class<T>) performs a validation check on the schema derived from the specified class. This validation ensures that all restrictions are adhered to. If any issues are detected, an exception will be thrown, providing a detailed message outlining the reasons for the validation failure.

Local Validation: The validation process occurs locally, meaning no requests are sent to the remote AI model. If the schema passes local validation, it is likely to pass remote validation as well.
Remote Validation: The remote AI model will conduct its own validation upon receiving the JSON schema in the request.
Version Compatibility: There may be instances where local validation fails while remote validation succeeds. This can occur if the SDK version is outdated compared to the restrictions enforced by the remote AI model.
Disabling Local Validation: If you encounter compatibility issues and wish to bypass local validation, you can disable it by passing JsonSchemaLocalValidation.NO to the responseFormat(Class<T>, JsonSchemaLocalValidation) method when building the parameters. (The default value for this parameter is JsonSchemaLocalValidation.YES.)

import com.openai.core.JsonSchemaLocalValidation;
import com.openai.models.ChatModel;
import com.openai.models.chat.completions.ChatCompletionCreateParams;
import com.openai.models.chat.completions.StructuredChatCompletionCreateParams;

StructuredChatCompletionCreateParams<BookList> params = ChatCompletionCreateParams.builder()
        .addUserMessage("List some famous late twentieth century novels.")
        .model(ChatModel.GPT_5_2)
        .responseFormat(BookList.class, JsonSchemaLocalValidation.NO)
        .build();

By following these guidelines, you can ensure that your structured outputs conform to the necessary schema requirements and minimize the risk of remote validation errors.

Usage with the Responses API

Structured Outputs are also supported for the Responses API. The usage is the same as described except where the Responses API differs slightly from the Chat Completions API. Pass the top-level class to text(Class<T>) when building the parameters and then access an instance of the class from the generated message content in the response.

You can start building the parameters with an instance of ResponseCreateParams.Builder or StructuredResponseCreateParams.Builder. If you start with the former (which allows for more compact code) the builder type will change to the latter when ResponseCreateParams.Builder.text(Class<T>) is called.

For a full example of the usage of Structured Outputs with the Responses API, see ResponsesStructuredOutputsExample.

Instead of using ResponseCreateParams.text(Class<T>), you can build a StructuredResponseTextConfig and set it on the ResponseCreateParams using the text(StructuredResponseTextConfig) method. Similar to using ResponseCreateParams, you can start with a ResponseTextConfig.Builder and its format(Class<T>) method will change it to a StructuredResponseTextConfig.Builder. This also allows you to set the verbosity configuration parameter on the text configuration before adding it to the ResponseCreateParams.

For a full example of the usage of Structured Outputs with the ResponseTextConfig and its verbosity parameter, see ResponsesStructuredOutputsVerbosityExample.

Usage with streaming

Structured Outputs can also be used with Streaming and the Chat Completions API. As responses are returned in "chunks", the full response must first be accumulated to concatenate the JSON strings that can then be converted into instances of the arbitrary Java class. Normal streaming operations can be performed while accumulating the JSON strings.

Use the ChatCompletionAccumulator as described in the section on Streaming helpers to accumulate the JSON strings. Once accumulated, use ChatCompletionAccumulator.chatCompletion(Class<T>) to convert the accumulated ChatCompletion into a StructuredChatCompletion. The StructuredChatCompletion can then automatically deserialize the JSON strings into instances of your Java class.

For a full example of the usage of Structured Outputs with Streaming and the Chat Completions API, see StructuredOutputsStreamingExample.

With the Responses API, accumulate events while streaming using the ResponseAccumulator. Once accumulated, use ResponseAccumulator.response(Class<T>) to convert the accumulated Response into a StructuredResponse. The [StructuredResponse] can then automatically deserialize the JSON strings into instances of your Java class.

For a full example of the usage of Structured Outputs with Streaming and the Responses API, see ResponsesStructuredOutputsStreamingExample.

Defining JSON schema properties

When a JSON schema is derived from your Java classes, all properties represented by public fields or public getter methods are included in the schema by default. Non-public fields and getter methods are not included by default. You can exclude public, or include non-public fields or getter methods, by using the @JsonIgnore or @JsonProperty annotations respectively (see Annotating classes and JSON schemas for details).

If you do not want to define public fields, you can define private fields and corresponding public getter methods. For example, a private field myValue with a public getter method getMyValue() will result in a "myValue" property being included in the JSON schema. If you prefer not to use the conventional Java "get" prefix for the name of the getter method, then you must annotate the getter method with the @JsonProperty annotation and the full method name will be used as the property name. You do not have to define any corresponding setter methods if you do not need them.

Each of your classes must define at least one property to be included in the JSON schema. A validation error will occur if any class contains no fields or getter methods from which schema properties can be derived. This may occur if, for example:

There are no fields or getter methods in the class.
All fields and getter methods are public, but all are annotated with @JsonIgnore.
All fields and getter methods are non-public, but none are annotated with @JsonProperty.
A field or getter method is declared with a Map type. A Map is treated like a separate class with no named properties, so it will result in an empty "properties" field in the JSON schema.

Annotating classes and JSON schemas

You can use annotations to add further information to the JSON schema derived from your Java classes, or to control which fields or getter methods will be included in the schema. Details from annotations captured in the JSON schema may be used by the AI model to improve its response. The SDK supports the use of Jackson Databind annotations.

import com.fasterxml.jackson.annotation.JsonClassDescription;
import com.fasterxml.jackson.annotation.JsonIgnore;
import com.fasterxml.jackson.annotation.JsonPropertyDescription;

class Person {
    @JsonPropertyDescription("The first name and surname of the person")
    public String name;
    public int birthYear;
    @JsonPropertyDescription("The year the person died, or 'present' if the person is living.")
    public String deathYear;
}

@JsonClassDescription("The details of one published book")
class Book {
    public String title;
    public Person author;
    @JsonPropertyDescription("The year in which the book was first published.")
    public int publicationYear;
    @JsonIgnore public String genre;
}

class BookList {
    public List<Book> books;
}

Use @JsonClassDescription to add a detailed description to a class.
Use @JsonPropertyDescription to add a detailed description to a field or getter method of a class.
Use @JsonIgnore to exclude a public field or getter method of a class from the generated JSON schema.
Use @JsonProperty to include a non-public field or getter method of a class in the generated JSON schema.

If you use @JsonProperty(required = false), the false value will be ignored. OpenAI JSON schemas must mark all properties as required, so the schema generated from your Java classes will respect that restriction and ignore any annotation that would violate it.

You can also use OpenAPI Swagger 2 @Schema and @ArraySchema annotations. These allow type-specific constraints to be added to your schema properties. You can learn more about the supported constraints in the OpenAI documentation on Supported properties.

import io.swagger.v3.oas.annotations.media.Schema;
import io.swagger.v3.oas.annotations.media.ArraySchema;

class Article {
    @ArraySchema(minItems = 1, maxItems = 10)
    public List<String> authors;

    @Schema(pattern = "^[A-Za-z ]+$")
    public String title;

    @Schema(format = "date")
    public String publicationDate;

    @Schema(minimum = "1")
    public int pageCount;
}

Local validation will check that you have not used any unsupported constraint keywords. However, the values of the constraints are not validated locally. For example, if you use a value for the "format" constraint of a string property that is not in the list of supported format names, then local validation will pass, but the AI model may report an error.

If you use both Jackson and Swagger annotations to set the same schema field, the Jackson annotation will take precedence. In the following example, the description of myProperty will be set to "Jackson description"; "Swagger description" will be ignored:

import com.fasterxml.jackson.annotation.JsonPropertyDescription;
import io.swagger.v3.oas.annotations.media.Schema;

class MyObject {
    @Schema(description = "Swagger description")
    @JsonPropertyDescription("Jackson description")
    public String myProperty;
}

Function calling with JSON schemas

OpenAI Function Calling lets you integrate external functions directly into the language model's responses. Instead of producing plain text, the model can output instructions (with parameters) for calling a function when appropriate. You define a JSON schema for functions, and the model uses it to decide when and how to trigger these calls, enabling more interactive, data-driven applications.

A JSON schema describing a function's parameters can be defined via the API by building a ChatCompletionTool containing a FunctionDefinition and then using addTool to set it on the input parameters. The response from the AI model may then contain requests to call your functions, detailing the functions' names and their parameter values as JSON data that conforms to the JSON schema from the function definition. You can then parse the parameter values from this JSON, invoke your functions, and pass your functions' results back to the AI model. A full, working example of Function Calling using the low-level API can be seen in FunctionCallingRawExample.

However, for greater convenience, the SDK can derive a function and its parameters automatically from the structure of an arbitrary Java class: the class's name provides the function name, and the class's fields define the function's parameters. When the AI model responds with the parameter values in JSON form, you can then easily convert that JSON to an instance of your Java class and use the parameter values to invoke your custom function. A full, working example of the use of Function Calling with Java classes to define function parameters can be seen in FunctionCallingExample.

Like for Structured Outputs, Java classes can contain fields declared to be instances of other classes and can use collections (see Defining JSON schema properties for more details). Optionally, annotations can be used to set the descriptions of the function (class) and its parameters (fields) to assist the AI model in understanding the purpose of the function and the possible values of its parameters.

import com.fasterxml.jackson.annotation.JsonClassDescription;
import com.fasterxml.jackson.annotation.JsonPropertyDescription;

@JsonClassDescription("Gets the quality of the given SDK.")
static class GetSdkQuality {
    @JsonPropertyDescription("The name of the SDK.")
    public String name;

    public SdkQuality execute() {
        return new SdkQuality(
                name, name.contains("OpenAI") ? "It's robust and polished!" : "*shrug*");
    }
}

static class SdkQuality {
    public String quality;

    public SdkQuality(String name, String evaluation) {
        quality = name + ": " + evaluation;
    }
}

@JsonClassDescription("Gets the review score (out of 10) for the named SDK.")
static class GetSdkScore {
  public String name;

  public int execute() {
    return name.contains("OpenAI") ? 10 : 3;
  }
}

When your functions are defined, add them to the input parameters using addTool(Class<T>) and then call them if requested to do so in the AI model's response. Function.arguments(Class<T>) can be used to parse a function's parameters in JSON form to an instance of your function-defining class. The fields of that instance will be set to the values of the parameters to the function call.

After calling the function, use ChatCompletionToolMessageParam.Builder.contentAsJson(Object) to pass the function's result back to the AI model. The method will convert the result to JSON form for consumption by the model. The Object can be any object, including simple String instances and boxed primitive types.

import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.models.ChatModel;
import com.openai.models.chat.completions.*;
import java.util.Collection;

OpenAIClient client = OpenAIOkHttpClient.fromEnv();

ChatCompletionCreateParams.Builder createParamsBuilder = ChatCompletionCreateParams.builder()
        .model(ChatModel.GPT_3_5_TURBO)
        .maxCompletionTokens(2048)
        .addTool(GetSdkQuality.class)
        .addTool(GetSdkScore.class)
        .addUserMessage("How good are the following SDKs and what do reviewers say: "
                + "OpenAI Java SDK, Unknown Company SDK.");

client.chat().completions().create(createParamsBuilder.build()).choices().stream()
        .map(ChatCompletion.Choice::message)
        // Add each assistant message onto the builder so that we keep track of the
        // conversation for asking a follow-up question later.
        .peek(createParamsBuilder::addMessage)
        .flatMap(message -> {
            message.content().ifPresent(System.out::println);
            return message.toolCalls().stream().flatMap(Collection::stream);
        })
        .forEach(toolCall -> {
            Object result = callFunction(toolCall.function());
            // Add the tool call result to the conversation.
            createParamsBuilder.addMessage(ChatCompletionToolMessageParam.builder()
                    .toolCallId(toolCall.id())
                    .contentAsJson(result)
                    .build());
        });

// Ask a follow-up question about the function call result.
createParamsBuilder.addUserMessage("Why do you say that?");
client.chat().completions().create(createParamsBuilder.build()).choices().stream()
        .flatMap(choice -> choice.message().content().stream())
        .forEach(System.out::println);

static Object callFunction(ChatCompletionMessageToolCall.Function function) {
  switch (function.name()) {
    case "GetSdkQuality":
      return function.arguments(GetSdkQuality.class).execute();
    case "GetSdkScore":
      return function.arguments(GetSdkScore.class).execute();
    default:
      throw new IllegalArgumentException("Unknown function: " + function.name());
  }
}

In the code above, an execute() method encapsulates each function's logic. However, there is no requirement to follow that pattern. You are free to implement your function's logic in any way that best suits your use case. The pattern above is only intended to suggest that a suitable pattern may make the process of function calling simpler to understand and implement.

Usage with the Responses API

Function Calling is also supported for the Responses API. The usage is the same as described except where the Responses API differs slightly from the Chat Completions API. Pass the top-level class to addTool(Class<T>) when building the parameters. In the response, look for RepoonseOutputItem instances that are function calls. Parse the parameters to each function call to an instance of the class using ResponseFunctionToolCall.arguments(Class<T>). Finally, pass the result of each call back to the model.

For a full example of the usage of Function Calling with the Responses API using the low-level API to define and parse function parameters, see ResponsesFunctionCallingRawExample.

For a full example of the usage of Function Calling with the Responses API using Java classes to define and parse function parameters, see ResponsesFunctionCallingExample.

Local function JSON schema validation

Like for Structured Outputs, you can perform local validation to check that the JSON schema derived from your function class respects the restrictions imposed by OpenAI on such schemas. Local validation is enabled by default, but it can be disabled by adding JsonSchemaLocalValidation.NO to the call to addTool.

ChatCompletionCreateParams.Builder createParamsBuilder = ChatCompletionCreateParams.builder()
        .model(ChatModel.GPT_3_5_TURBO)
        .maxCompletionTokens(2048)
        .addTool(GetSdkQuality.class, JsonSchemaLocalValidation.NO)
        .addTool(GetSdkScore.class, JsonSchemaLocalValidation.NO)
        .addUserMessage("How good are the following SDKs and what do reviewers say: "
                + "OpenAI Java SDK, Unknown Company SDK.");

See Local JSON schema validation for more details on local schema validation and under what circumstances you might want to disable it.

Annotating function classes

You can use annotations to add further information about functions to the JSON schemas that are derived from your function classes, or to control which fields or getter methods will be used as parameters to the function. Details from annotations captured in the JSON schema may be used by the AI model to improve its response. The SDK supports the use of Jackson Databind annotations.

Use @JsonClassDescription to add a description to a function class detailing when and how to use that function.
Use @JsonTypeName to set the function name to something other than the simple name of the class, which is used by default.
Use @JsonPropertyDescription to add a detailed description to function parameter (a field or getter method of a function class).
Use @JsonIgnore to exclude a public field or getter method of a class from the generated JSON schema for a function's parameters.
Use @JsonProperty to include a non-public field or getter method of a class in the generated JSON schema for a function's parameters.

OpenAI provides some Best practices for defining functions that may help you to understand how to use the above annotations effectively for your functions.

See also Defining JSON schema properties for more details on how to use fields and getter methods and combine access modifiers and annotations to define the parameters of your functions. The same rules apply to function classes and to the structured output classes described in that section.

File uploads

The SDK defines methods that accept files.

To upload a file, pass a Path:

import com.openai.models.files.FileCreateParams;
import com.openai.models.files.FileObject;
import com.openai.models.files.FilePurpose;
import java.nio.file.Paths;

FileCreateParams params = FileCreateParams.builder()
    .purpose(FilePurpose.FINE_TUNE)
    .file(Paths.get("input.jsonl"))
    .build();
FileObject fileObject = client.files().create(params);

Or an arbitrary InputStream:

import com.openai.models.files.FileCreateParams;
import com.openai.models.files.FileObject;
import com.openai.models.files.FilePurpose;
import java.net.URL;

FileCreateParams params = FileCreateParams.builder()
    .purpose(FilePurpose.FINE_TUNE)
    .file(new URL("https://example.com/input.jsonl").openStream())
    .build();
FileObject fileObject = client.files().create(params);

Or a byte[] array:

import com.openai.models.files.FileCreateParams;
import com.openai.models.files.FileObject;
import com.openai.models.files.FilePurpose;

FileCreateParams params = FileCreateParams.builder()
    .purpose(FilePurpose.FINE_TUNE)
    .file("content".getBytes())
    .build();
FileObject fileObject = client.files().create(params);

Note that when passing a non-Path its filename is unknown so it will not be included in the request. To manually set a filename, pass a MultipartField:

import com.openai.core.MultipartField;
import com.openai.models.files.FileCreateParams;
import com.openai.models.files.FileObject;
import com.openai.models.files.FilePurpose;
import java.io.InputStream;
import java.net.URL;

FileCreateParams params = FileCreateParams.builder()
    .purpose(FilePurpose.FINE_TUNE)
    .file(MultipartField.<InputStream>builder()
        .value(new URL("https://example.com/input.jsonl").openStream())
        .filename("input.jsonl")
        .build())
    .build();
FileObject fileObject = client.files().create(params);

Webhook Verification

Verifying webhook signatures is optional but encouraged.

For more information about webhooks, see the API docs.

Parsing webhook payloads

For most use cases, you will likely want to verify the webhook and parse the payload at the same time. To achieve this, we provide the method client.webhooks().unwrap(), which parses a webhook request and verifies that it was sent by OpenAI. This method will throw an exception if the signature is invalid.

Note that the body parameter must be the raw JSON string sent from the server (do not parse it first). The .unwrap() method will parse this JSON for you into an event object after verifying the webhook was sent from OpenAI.

import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.core.http.Headers;
import com.openai.models.webhooks.UnwrapWebhookEvent;
import java.util.Optional;

OpenAIClient client = OpenAIOkHttpClient.fromEnv(); // OPENAI_WEBHOOK_SECRET env var used by default

public void handleWebhook(String body, Map<String, String> headers) {
    try {
        Headers headersList = Headers.builder()
                .putAll(headers)
                .build();

        UnwrapWebhookEvent event = client.webhooks().unwrap(body, headersList, Optional.empty());

        if (event.isResponseCompletedWebhookEvent()) {
            System.out.println("Response completed: " + event.asResponseCompletedWebhookEvent().data());
        } else if (event.isResponseFailed()) {
            System.out.println("Response failed: " + event.asResponseFailed().data());
        } else {
            System.out.println("Unhandled event type: " + event.getClass().getSimpleName());
        }
    } catch (Exception e) {
        System.err.println("Invalid webhook signature: " + e.getMessage());
        // Handle invalid signature
    }
}

Verifying webhook payloads directly

In some cases, you may want to verify the webhook separately from parsing the payload. If you prefer to handle these steps separately, we provide the method client.webhooks().verifySignature() to only verify the signature of a webhook request. Like .unwrap(), this method will throw an exception if the signature is invalid.

Note that the body parameter must be the raw JSON string sent from the server (do not parse it first). You will then need to parse the body after verifying the signature.

import com.fasterxml.jackson.databind.ObjectMapper;
import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.core.http.Headers;
import com.openai.models.webhooks.WebhookVerificationParams;
import java.util.Optional;

OpenAIClient client = OpenAIOkHttpClient.fromEnv(); // OPENAI_WEBHOOK_SECRET env var used by default
ObjectMapper objectMapper = new ObjectMapper();

public void handleWebhook(String body, Map<String, String> headers) {
    try {
        Headers headersList = Headers.builder()
                .putAll(headers)
                .build();

        client.webhooks().verifySignature(
            WebhookVerificationParams.builder()
                .payload(body)
                .headers(headersList)
                .build()
        );

        // Parse the body after verification
        Map<String, Object> event = objectMapper.readValue(body, Map.class);
        System.out.println("Verified event: " + event);
    } catch (Exception e) {
        System.err.println("Invalid webhook signature: " + e.getMessage());
        // Handle invalid signature
    }
}

Binary responses

The SDK defines methods that return binary responses, which are used for API responses that shouldn't necessarily be parsed, like non-JSON data.

These methods return HttpResponse:

import com.openai.core.http.HttpResponse;
import com.openai.models.files.FileContentParams;

HttpResponse response = client.files().content("file_id");

To save the response content to a file, use the Files.copy(...) method:

import com.openai.core.http.HttpResponse;
import java.nio.file.Files;
import java.nio.file.Paths;
import java.nio.file.StandardCopyOption;

try (HttpResponse response = client.files().content(params)) {
    Files.copy(
        response.body(),
        Paths.get(path),
        StandardCopyOption.REPLACE_EXISTING
    );
} catch (Exception e) {
    System.out.println("Something went wrong!");
    throw new RuntimeException(e);
}

Or transfer the response content to any OutputStream:

import com.openai.core.http.HttpResponse;
import java.nio.file.Files;
import java.nio.file.Paths;

try (HttpResponse response = client.files().content(params)) {
    response.body().transferTo(Files.newOutputStream(Paths.get(path)));
} catch (Exception e) {
    System.out.println("Something went wrong!");
    throw new RuntimeException(e);
}

Raw responses

The SDK defines methods that deserialize responses into instances of Java classes. However, these methods don't provide access to the response headers, status code, or the raw response body.

To access this data, prefix any HTTP method call on a client or service with withRawResponse():

import com.openai.core.http.Headers;
import com.openai.core.http.HttpResponseFor;
import com.openai.models.ChatModel;
import com.openai.models.chat.completions.ChatCompletion;
import com.openai.models.chat.completions.ChatCompletionCreateParams;

ChatCompletionCreateParams params = ChatCompletionCreateParams.builder()
    .addUserMessage("Say this is a test")
    .model(ChatModel.GPT_5_2)
    .build();
HttpResponseFor<ChatCompletion> chatCompletion = client.chat().completions().withRawResponse().create(params);

int statusCode = chatCompletion.statusCode();
Headers headers = chatCompletion.headers();

You can still deserialize the response into an instance of a Java class if needed:

import com.openai.models.chat.completions.ChatCompletion;

ChatCompletion parsedChatCompletion = chatCompletion.parse();

Request IDs

For more information on debugging requests, see the API docs.

When using raw responses, you can access the x-request-id response header using the requestId() method:

import com.openai.core.http.HttpResponseFor;
import com.openai.models.chat.completions.ChatCompletion;
import java.util.Optional;

HttpResponseFor<ChatCompletion> chatCompletion = client.chat().completions().withRawResponse().create(params);
Optional<String> requestId = chatCompletion.requestId();

This can be used to quickly log failing requests and report them back to OpenAI.

Error handling

The SDK throws custom unchecked exception types:

OpenAIServiceException: Base class for HTTP errors. See this table for which exception subclass is thrown for each HTTP status code:

Status	Exception
400	`BadRequestException`
401	`UnauthorizedException`
403	`PermissionDeniedException`
404	`NotFoundException`
422	`UnprocessableEntityException`
429	`RateLimitException`
5xx	`InternalServerException`
others	`UnexpectedStatusCodeException`

SseException is thrown for errors encountered during SSE streaming after a successful initial HTTP response.

OpenAIIoException: I/O networking errors.
OpenAIRetryableException: Generic error indicating a failure that could be retried by the client.
OpenAIInvalidDataException: Failure to interpret successfully parsed data. For example, when accessing a property that's supposed to be required, but the API unexpectedly omitted it from the response.
OpenAIException: Base class for all exceptions. Most errors will result in one of the previously mentioned ones, but completely generic errors may be thrown using the base class.

Pagination

The SDK defines methods that return a paginated lists of results. It provides convenient ways to access the results either one page at a time or item-by-item across all pages.

Auto-pagination

To iterate through all results across all pages, use the autoPager() method, which automatically fetches more pages as needed.

When using the synchronous client, the method returns an Iterable

import com.openai.models.finetuning.jobs.FineTuningJob;
import com.openai.models.finetuning.jobs.JobListPage;

JobListPage page = client.fineTuning().jobs().list();

// Process as an Iterable
for (FineTuningJob job : page.autoPager()) {
    System.out.println(job);
}

// Process as a Stream
page.autoPager()
    .stream()
    .limit(50)
    .forEach(job -> System.out.println(job));

When using the asynchronous client, the method returns an AsyncStreamResponse:

import com.openai.core.http.AsyncStreamResponse;
import com.openai.models.finetuning.jobs.FineTuningJob;
import com.openai.models.finetuning.jobs.JobListPageAsync;
import java.util.Optional;
import java.util.concurrent.CompletableFuture;

CompletableFuture<JobListPageAsync> pageFuture = client.async().fineTuning().jobs().list();

pageFuture.thenRun(page -> page.autoPager().subscribe(job -> {
    System.out.println(job);
}));

// If you need to handle errors or completion of the stream
pageFuture.thenRun(page -> page.autoPager().subscribe(new AsyncStreamResponse.Handler<>() {
    @Override
    public void onNext(FineTuningJob job) {
        System.out.println(job);
    }

    @Override
    public void onComplete(Optional<Throwable> error) {
        if (error.isPresent()) {
            System.out.println("Something went wrong!");
            throw new RuntimeException(error.get());
        } else {
            System.out.println("No more!");
        }
    }
}));

// Or use futures
pageFuture.thenRun(page -> page.autoPager()
    .subscribe(job -> {
        System.out.println(job);
    })
    .onCompleteFuture()
    .whenComplete((unused, error) -> {
        if (error != null) {
            System.out.println("Something went wrong!");
            throw new RuntimeException(error);
        } else {
            System.out.println("No more!");
        }
    }));

Manual pagination

To access individual page items and manually request the next page, use the items(), hasNextPage(), and nextPage() methods:

import com.openai.models.finetuning.jobs.FineTuningJob;
import com.openai.models.finetuning.jobs.JobListPage;

JobListPage page = client.fineTuning().jobs().list();
while (true) {
    for (FineTuningJob job : page.items()) {
        System.out.println(job);
    }

    if (!page.hasNextPage()) {
        break;
    }

    page = page.nextPage();
}

Logging

The SDK uses the standard OkHttp logging interceptor.

Enable logging by setting the OPENAI_LOG environment variable to info:

export OPENAI_LOG=info

Or to debug for more verbose logging:

export OPENAI_LOG=debug

ProGuard and R8

Although the SDK uses reflection, it is still usable with ProGuard and R8 because openai-java-core is published with a configuration file containing keep rules.

ProGuard and R8 should automatically detect and use the published rules, but you can also manually copy the keep rules if necessary.

GraalVM

Although the SDK uses reflection, it is still usable in GraalVM because openai-java-core is published with reachability metadata.

GraalVM should automatically detect and use the published metadata, but manual configuration is also available.

openai/openai-java

README

OpenAI Java API Library

Installation

Gradle

Maven

Requirements

Usage

Client configuration

Modifying configuration

Workload identity authentication

Basic setup

Kubernetes service account token provider

Azure Managed Identity provider

GCP ID token provider

Requests and responses

Immutability

Asynchronous execution

Streaming

Streaming helpers

Structured outputs with JSON schemas

Local JSON schema validation

Usage with the Responses API

Usage with streaming

Defining JSON schema properties

Annotating classes and JSON schemas

Function calling with JSON schemas

Usage with the Responses API

Local function JSON schema validation

Annotating function classes

File uploads

Webhook Verification

Parsing webhook payloads

Verifying webhook payloads directly

Binary responses

Raw responses

Request IDs

Error handling

Pagination

Auto-pagination

Manual pagination

Logging

ProGuard and R8

GraalVM

Spring Boot

리뷰(0)

아직 리뷰가 없습니다

프로젝트 건강도