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.

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

implementation("com.openai:openai-java:4.11.0")

<dependency> <groupId>com.openai</groupId> <artifactId>openai-java</artifactId> <version>4.11.0</version> </dependency>

This library requires Java 8 or later.

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

importcom.openai.client.OpenAIClient; importcom.openai.client.okhttp.OpenAIOkHttpClient; importcom.openai.models.ChatModel; importcom.openai.models.responses.Response; importcom.openai.models.responses.ResponseCreateParams; // Configures using the `OPENAI_API_KEY`, `OPENAI_ORG_ID` and `OPENAI_PROJECT_ID` environment variablesOpenAIClientclient = OpenAIOkHttpClient.fromEnv(); ResponseCreateParamsparams = ResponseCreateParams.builder() .input("Say this is a test") .model(ChatModel.GPT_4_1) .build(); Responseresponse = 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.

importcom.openai.client.OpenAIClient; importcom.openai.client.okhttp.OpenAIOkHttpClient; importcom.openai.models.ChatModel; importcom.openai.models.chat.completions.ChatCompletion; importcom.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 variablesOpenAIClientclient = OpenAIOkHttpClient.fromEnv(); ChatCompletionCreateParamsparams = ChatCompletionCreateParams.builder() .addUserMessage("Say this is a test") .model(ChatModel.GPT_5_2) .build(); ChatCompletionchatCompletion = client.chat().completions().create(params);

Configure the client using system properties or environment variables:

importcom.openai.client.OpenAIClient; importcom.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 variablesOpenAIClientclient = OpenAIOkHttpClient.fromEnv();

Or manually:

importcom.openai.client.OpenAIClient; importcom.openai.client.okhttp.OpenAIOkHttpClient; OpenAIClientclient = OpenAIOkHttpClient.builder() .apiKey("My API Key") .build();

Or using a combination of the two approaches:

importcom.openai.client.OpenAIClient; importcom.openai.client.okhttp.OpenAIOkHttpClient; OpenAIClientclient = 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:

System properties take precedence over environment variables.

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

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

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

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.

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.

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

importcom.openai.client.OpenAIClient; importcom.openai.client.okhttp.OpenAIOkHttpClient; importcom.openai.models.ChatModel; importcom.openai.models.chat.completions.ChatCompletion; importcom.openai.models.chat.completions.ChatCompletionCreateParams; importjava.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 variablesOpenAIClientclient = OpenAIOkHttpClient.fromEnv(); ChatCompletionCreateParamsparams = 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:

importcom.openai.client.OpenAIClientAsync; importcom.openai.client.okhttp.OpenAIOkHttpClientAsync; importcom.openai.models.ChatModel; importcom.openai.models.chat.completions.ChatCompletion; importcom.openai.models.chat.completions.ChatCompletionCreateParams; importjava.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 variablesOpenAIClientAsyncclient = OpenAIOkHttpClientAsync.fromEnv(); ChatCompletionCreateParamsparams = 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.

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:

importcom.openai.core.http.StreamResponse; importcom.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:

importcom.openai.core.http.AsyncStreamResponse; importcom.openai.models.chat.completions.ChatCompletionChunk; importjava.util.Optional; client.async().chat().completions().createStreaming(params).subscribe(chunk ->{System.out.println(chunk)}); // If you need to handle errors or completion of the streamclient.async().chat().completions().createStreaming(params).subscribe(newAsyncStreamResponse.Handler<>(){@OverridepublicvoidonNext(ChatCompletionChunkchunk){System.out.println(chunk)} @OverridepublicvoidonComplete(Optional<Throwable> error){if (error.isPresent()){System.out.println("Something went wrong!"); thrownewRuntimeException(error.get())} else{System.out.println("No more chunks!")} } }); // Or use futuresclient.async().chat().completions().createStreaming(params) .subscribe(chunk ->{System.out.println(chunk)}) .onCompleteFuture(); .whenComplete((unused, error) ->{if (error != null){System.out.println("Something went wrong!"); thrownewRuntimeException(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:

importjava.util.concurrent.Executor; importjava.util.concurrent.Executors; Executorexecutor = 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:

importcom.openai.client.OpenAIClient; importcom.openai.client.okhttp.OpenAIOkHttpClient; importjava.util.concurrent.Executors; OpenAIClientclient = OpenAIOkHttpClient.builder() .fromEnv() .streamHandlerExecutor(Executors.newFixedThreadPool(4)) .build();

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:

importcom.openai.core.http.StreamResponse; importcom.openai.helpers.ChatCompletionAccumulator; importcom.openai.models.chat.completions.ChatCompletion; importcom.openai.models.chat.completions.ChatCompletionChunk; ChatCompletionAccumulatorchatCompletionAccumulator = 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)} ChatCompletionchatCompletion = chatCompletionAccumulator.chatCompletion();

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

importcom.openai.helpers.ChatCompletionAccumulator; importcom.openai.models.chat.completions.ChatCompletion; ChatCompletionAccumulatorchatCompletionAccumulator = 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(); ChatCompletionchatCompletion = 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:

importcom.openai.core.http.StreamResponse; importcom.openai.helpers.ResponseAccumulator; importcom.openai.models.responses.Response; importcom.openai.models.responses.ResponseStreamEvent; ResponseAccumulatorresponseAccumulator = 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()))} Responseresponse = responseAccumulator.response();

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

importcom.openai.helpers.ResponseAccumulator; importcom.openai.models.responses.Response; ResponseAccumulatorresponseAccumulator = ResponseAccumulator.create(); client.responses() .createStreaming(createParams) .subscribe(event -> responseAccumulator.accumulate(event) .outputTextDelta().ifPresent(textEvent -> System.out.print(textEvent.delta()))) .onCompleteFuture() .join(); Responseresponse = responseAccumulator.response();

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):

classPerson{publicStringname; publicintbirthYear} classBook{publicStringtitle; publicPersonauthor; publicintpublicationYear} classBookList{publicList<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:

importcom.openai.models.ChatModel; importcom.openai.models.chat.completions.ChatCompletionCreateParams; importcom.openai.models.chat.completions.StructuredChatCompletionCreateParams; StructuredChatCompletionCreateParams<BookList> params = ChatCompletionCreateParams.builder() .addUserMessage("List some famous late twentieth century novels.") .model(ChatModel.GPT_4_1) .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.

importjava.util.Optional; classBook{publicStringtitle; publicPersonauthor; publicintpublicationYear; publicOptional<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 = newArrayList<>(); 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.

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.)

importcom.openai.core.JsonSchemaLocalValidation; importcom.openai.models.ChatModel; importcom.openai.models.chat.completions.ChatCompletionCreateParams; importcom.openai.models.chat.completions.StructuredChatCompletionCreateParams; StructuredChatCompletionCreateParams<BookList> params = ChatCompletionCreateParams.builder() .addUserMessage("List some famous late twentieth century novels.") .model(ChatModel.GPT_4_1) .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.

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.

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.

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.

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.

importcom.fasterxml.jackson.annotation.JsonClassDescription; importcom.fasterxml.jackson.annotation.JsonIgnore; importcom.fasterxml.jackson.annotation.JsonPropertyDescription; classPerson{@JsonPropertyDescription("The first name and surname of the person") publicStringname; publicintbirthYear; @JsonPropertyDescription("The year the person died, or 'present' if the person is living.") publicStringdeathYear} @JsonClassDescription("The details of one published book") classBook{publicStringtitle; publicPersonauthor; @JsonPropertyDescription("The year in which the book was first published.") publicintpublicationYear; @JsonIgnorepublicStringgenre} classBookList{publicList<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.

importio.swagger.v3.oas.annotations.media.Schema; importio.swagger.v3.oas.annotations.media.ArraySchema; classArticle{@ArraySchema(minItems = 1, maxItems = 10) publicList<String> authors; @Schema(pattern = "^[A-Za-z ]+$") publicStringtitle; @Schema(format = "date") publicStringpublicationDate; @Schema(minimum = "1") publicintpageCount}

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:

importcom.fasterxml.jackson.annotation.JsonPropertyDescription; importio.swagger.v3.oas.annotations.media.Schema; classMyObject{@Schema(description = "Swagger description") @JsonPropertyDescription("Jackson description") publicStringmyProperty}

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.

importcom.fasterxml.jackson.annotation.JsonClassDescription; importcom.fasterxml.jackson.annotation.JsonPropertyDescription; @JsonClassDescription("Gets the quality of the given SDK.") staticclassGetSdkQuality{@JsonPropertyDescription("The name of the SDK.") publicStringname; publicSdkQualityexecute(){returnnewSdkQuality( name, name.contains("OpenAI") ? "It's robust and polished!" : "*shrug*")} } staticclassSdkQuality{publicStringquality; publicSdkQuality(Stringname, Stringevaluation){quality = name + ": " + evaluation} } @JsonClassDescription("Gets the review score (out of 10) for the named SDK.") staticclassGetSdkScore{publicStringname; publicintexecute(){returnname.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.argments(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.

importcom.openai.client.OpenAIClient; importcom.openai.client.okhttp.OpenAIOkHttpClient; importcom.openai.models.ChatModel; importcom.openai.models.chat.completions.*; importjava.util.Collection; OpenAIClientclient = OpenAIOkHttpClient.fromEnv(); ChatCompletionCreateParams.BuildercreateParamsBuilder = 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); returnmessage.toolCalls().stream().flatMap(Collection::stream)}) .forEach(toolCall ->{Objectresult = 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); staticObjectcallFunction(ChatCompletionMessageToolCall.Functionfunction){switch (function.name()){case"GetSdkQuality": returnfunction.arguments(GetSdkQuality.class).execute(); case"GetSdkScore": returnfunction.arguments(GetSdkScore.class).execute(); default: thrownewIllegalArgumentException("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.

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.

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.BuildercreateParamsBuilder = 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.

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.

The SDK defines methods that accept files.

To upload a file, pass a Path:

importcom.openai.models.files.FileCreateParams; importcom.openai.models.files.FileObject; importcom.openai.models.files.FilePurpose; importjava.nio.file.Paths; FileCreateParamsparams = FileCreateParams.builder() .purpose(FilePurpose.FINE_TUNE) .file(Paths.get("input.jsonl")) .build(); FileObjectfileObject = client.files().create(params);

Or an arbitrary InputStream:

importcom.openai.models.files.FileCreateParams; importcom.openai.models.files.FileObject; importcom.openai.models.files.FilePurpose; importjava.net.URL; FileCreateParamsparams = FileCreateParams.builder() .purpose(FilePurpose.FINE_TUNE) .file(newURL("https://example.com/input.jsonl").openStream()) .build(); FileObjectfileObject = client.files().create(params);

Or a byte[] array:

importcom.openai.models.files.FileCreateParams; importcom.openai.models.files.FileObject; importcom.openai.models.files.FilePurpose; FileCreateParamsparams = FileCreateParams.builder() .purpose(FilePurpose.FINE_TUNE) .file("content".getBytes()) .build(); FileObjectfileObject = 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:

importcom.openai.core.MultipartField; importcom.openai.models.files.FileCreateParams; importcom.openai.models.files.FileObject; importcom.openai.models.files.FilePurpose; importjava.io.InputStream; importjava.net.URL; FileCreateParamsparams = FileCreateParams.builder() .purpose(FilePurpose.FINE_TUNE) .file(MultipartField.<InputStream>builder() .value(newURL("https://example.com/input.jsonl").openStream()) .filename("input.jsonl") .build()) .build(); FileObjectfileObject = client.files().create(params);

Verifying webhook signatures is optional but encouraged.

For more information about webhooks, see the API docs.

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.

importcom.openai.client.OpenAIClient; importcom.openai.client.okhttp.OpenAIOkHttpClient; importcom.openai.core.http.Headers; importcom.openai.models.webhooks.UnwrapWebhookEvent; importjava.util.Optional; OpenAIClientclient = OpenAIOkHttpClient.fromEnv(); // OPENAI_WEBHOOK_SECRET env var used by defaultpublicvoidhandleWebhook(Stringbody, Map<String, String> headers){try{HeadersheadersList = Headers.builder() .putAll(headers) .build(); UnwrapWebhookEventevent = client.webhooks().unwrap(body, headersList, Optional.empty()); if (event.isResponseCompletedWebhookEvent()){System.out.println("Response completed: " + event.asResponseCompletedWebhookEvent().data())} elseif (event.isResponseFailed()){System.out.println("Response failed: " + event.asResponseFailed().data())} else{System.out.println("Unhandled event type: " + event.getClass().getSimpleName())} } catch (Exceptione){System.err.println("Invalid webhook signature: " + e.getMessage()); // Handle invalid signature } }

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.

importcom.fasterxml.jackson.databind.ObjectMapper; importcom.openai.client.OpenAIClient; importcom.openai.client.okhttp.OpenAIOkHttpClient; importcom.openai.core.http.Headers; importcom.openai.models.webhooks.WebhookVerificationParams; importjava.util.Optional; OpenAIClientclient = OpenAIOkHttpClient.fromEnv(); // OPENAI_WEBHOOK_SECRET env var used by defaultObjectMapperobjectMapper = newObjectMapper(); publicvoidhandleWebhook(Stringbody, Map<String, String> headers){try{HeadersheadersList = Headers.builder() .putAll(headers) .build(); client.webhooks().verifySignature( WebhookVerificationParams.builder() .payload(body) .headers(headersList) .build() ); // Parse the body after verificationMap<String, Object> event = objectMapper.readValue(body, Map.class); System.out.println("Verified event: " + event)} catch (Exceptione){System.err.println("Invalid webhook signature: " + e.getMessage()); // Handle invalid signature } }

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:

importcom.openai.core.http.HttpResponse; importcom.openai.models.files.FileContentParams; HttpResponseresponse = client.files().content("file_id");

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

importcom.openai.core.http.HttpResponse; importjava.nio.file.Files; importjava.nio.file.Paths; importjava.nio.file.StandardCopyOption; try (HttpResponseresponse = client.files().content(params)){Files.copy( response.body(), Paths.get(path), StandardCopyOption.REPLACE_EXISTING )} catch (Exceptione){System.out.println("Something went wrong!"); thrownewRuntimeException(e)}

Or transfer the response content to any OutputStream:

importcom.openai.core.http.HttpResponse; importjava.nio.file.Files; importjava.nio.file.Paths; try (HttpResponseresponse = client.files().content(params)){response.body().transferTo(Files.newOutputStream(Paths.get(path)))} catch (Exceptione){System.out.println("Something went wrong!"); thrownewRuntimeException(e)}

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():

importcom.openai.core.http.Headers; importcom.openai.core.http.HttpResponseFor; importcom.openai.models.ChatModel; importcom.openai.models.chat.completions.ChatCompletion; importcom.openai.models.chat.completions.ChatCompletionCreateParams; ChatCompletionCreateParamsparams = ChatCompletionCreateParams.builder() .addUserMessage("Say this is a test") .model(ChatModel.GPT_5_2) .build(); HttpResponseFor<ChatCompletion> chatCompletion = client.chat().completions().withRawResponse().create(params); intstatusCode = chatCompletion.statusCode(); Headersheaders = chatCompletion.headers();

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

importcom.openai.models.chat.completions.ChatCompletion; ChatCompletionparsedChatCompletion = chatCompletion.parse();

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:

importcom.openai.core.http.HttpResponseFor; importcom.openai.models.chat.completions.ChatCompletion; importjava.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.

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.

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.

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

importcom.openai.models.finetuning.jobs.FineTuningJob; importcom.openai.models.finetuning.jobs.JobListPage; JobListPagepage = client.fineTuning().jobs().list(); // Process as an Iterablefor (FineTuningJobjob : page.autoPager()){System.out.println(job)} // Process as a Streampage.autoPager() .stream() .limit(50) .forEach(job -> System.out.println(job));

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

importcom.openai.core.http.AsyncStreamResponse; importcom.openai.models.finetuning.jobs.FineTuningJob; importcom.openai.models.finetuning.jobs.JobListPageAsync; importjava.util.Optional; importjava.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 streampageFuture.thenRun(page -> page.autoPager().subscribe(newAsyncStreamResponse.Handler<>(){@OverridepublicvoidonNext(FineTuningJobjob){System.out.println(job)} @OverridepublicvoidonComplete(Optional<Throwable> error){if (error.isPresent()){System.out.println("Something went wrong!"); thrownewRuntimeException(error.get())} else{System.out.println("No more!")} } })); // Or use futurespageFuture.thenRun(page -> page.autoPager() .subscribe(job ->{System.out.println(job)}) .onCompleteFuture() .whenComplete((unused, error) ->{if (error != null){System.out.println("Something went wrong!"); thrownewRuntimeException(error)} else{System.out.println("No more!")} }));

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

importcom.openai.models.finetuning.jobs.FineTuningJob; importcom.openai.models.finetuning.jobs.JobListPage; JobListPagepage = client.fineTuning().jobs().list(); while (true){for (FineTuningJobjob : page.items()){System.out.println(job)} if (!page.hasNextPage()){break} page = page.nextPage()}

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

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.

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.

If you're using Spring Boot, then you can use the SDK's Spring Boot starter to simplify configuration and get set up quickly.

implementation("com.openai:openai-java-spring-boot-starter:4.11.0")

<dependency> <groupId>com.openai</groupId> <artifactId>openai-java-spring-boot-starter</artifactId> <version>4.11.0</version> </dependency>

The client's environment variable options can be configured in application.properties or application.yml.

openai.base-url=https://api.openai.com/v1 openai.api-key=My API Key openai.org-id=My Organization openai.project-id=My Project openai.webhook-secret=My Webhook Secret

openai: base-url: https://api.openai.com/v1api-key: My API Keyorg-id: My Organizationproject-id: My Projectwebhook-secret: My Webhook Secret

Configure any other client option by providing one or more instances of OpenAIClientCustomizer. For example, here's how you'd set maxRetries:

importcom.openai.springboot.OpenAIClientCustomizer; importorg.springframework.context.annotation.Bean; importorg.springframework.context.annotation.Configuration; @ConfigurationpublicclassOpenAIConfig{@BeanpublicOpenAIClientCustomizercustomizer(){returnbuilder -> builder.maxRetries(3)} }

InjectOpenAIClient anywhere and start using it!

The SDK depends on Jackson for JSON serialization/deserialization. It is compatible with version 2.13.4 or higher, but depends on version 2.18.2 by default.

The SDK throws an exception if it detects an incompatible Jackson version at runtime (e.g. if the default version was overridden in your Maven or Gradle config).

If the SDK threw an exception, but you're certain the version is compatible, then disable the version check using the checkJacksonVersionCompatibility on OpenAIOkHttpClient or OpenAIOkHttpClientAsync.

To use this library with Azure OpenAI, use the same OpenAI client builder but with the Azure-specific configuration.

OpenAIClientclient = OpenAIOkHttpClient.builder() // Gets the API key and endpoint from the `AZURE_OPENAI_KEY` and `OPENAI_BASE_URL` environment variables, respectively .fromEnv() // Set the Azure Entra ID .credential(BearerTokenCredential.create(AuthenticationUtil.getBearerTokenSupplier( newDefaultAzureCredentialBuilder().build(), "https://cognitiveservices.azure.com/.default"))) .build();

See the complete Azure OpenAI example in the openai-java-example directory. The other examples in the directory also work with Azure as long as the client is configured to use it.

The ClientOptions can be configured to treat Azure OpenAI endpoint URLs differently, depending on your service setup. The default value is AzureUrlPathMode.AUTO. To customize the SDK behavior, each value does the following:

• AzureUrlPathMode.LEGACY: forces the deployment or model name into the path.
• AzureUrlPathMode.UNIFIED: for newer endpoints ending in /openai/v1 the service behaviour matches OpenAI's, therefore AzureOpenAIServiceVersion becomes optional and the model is passed in the request object.
• AzureUrlPathMode.AUTO: automatically detects the path mode based on the base URL. Default value.

The SDK automatically retries 2 times by default, with a short exponential backoff between requests.

Only the following error types are retried:

• Connection errors (for example, due to a network connectivity problem)
• 408 Request Timeout
• 409 Conflict
• 429 Rate Limit
• 5xx Internal

The API may also explicitly instruct the SDK to retry or not retry a request.

To set a custom number of retries, configure the client using the maxRetries method:

importcom.openai.client.OpenAIClient; importcom.openai.client.okhttp.OpenAIOkHttpClient; OpenAIClientclient = OpenAIOkHttpClient.builder() .fromEnv() .maxRetries(4) .build();

Requests time out after 10 minutes by default.

To set a custom timeout, configure the method call using the timeout method:

importcom.openai.models.chat.completions.ChatCompletion; ChatCompletionchatCompletion = client.chat().completions().create( params, RequestOptions.builder().timeout(Duration.ofSeconds(30)).build() );

Or configure the default for all method calls at the client level:

importcom.openai.client.OpenAIClient; importcom.openai.client.okhttp.OpenAIOkHttpClient; importjava.time.Duration; OpenAIClientclient = OpenAIOkHttpClient.builder() .fromEnv() .timeout(Duration.ofSeconds(30)) .build();

To route requests through a proxy, configure the client using the proxy method:

importcom.openai.client.OpenAIClient; importcom.openai.client.okhttp.OpenAIOkHttpClient; importjava.net.InetSocketAddress; importjava.net.Proxy; OpenAIClientclient = OpenAIOkHttpClient.builder() .fromEnv() .proxy(newProxy( Proxy.Type.HTTP, newInetSocketAddress( "https://example.com", 8080 ) )) .build();

To configure how HTTPS connections are secured, configure the client using the sslSocketFactory, trustManager, and hostnameVerifier methods:

importcom.openai.client.OpenAIClient; importcom.openai.client.okhttp.OpenAIOkHttpClient; OpenAIClientclient = OpenAIOkHttpClient.builder() .fromEnv() // If `sslSocketFactory` is set, then `trustManager` must be set, and vice versa. .sslSocketFactory(yourSSLSocketFactory) .trustManager(yourTrustManager) .hostnameVerifier(yourHostnameVerifier) .build();

The SDK consists of three artifacts:

• openai-java-core
  • Contains core SDK logic
  • Does not depend on OkHttp
  • Exposes OpenAIClient, OpenAIClientAsync, OpenAIClientImpl, and OpenAIClientAsyncImpl, all of which can work with any HTTP client
• openai-java-client-okhttp
  • Depends on OkHttp
  • Exposes OpenAIOkHttpClient and OpenAIOkHttpClientAsync, which provide a way to construct OpenAIClientImpl and OpenAIClientAsyncImpl, respectively, using OkHttp
• openai-java
  • Depends on and exposes the APIs of both openai-java-core and openai-java-client-okhttp
  • Does not have its own logic

This structure allows replacing the SDK's default HTTP client without pulling in unnecessary dependencies.

To use a customized OkHttpClient:

1. Replace your openai-java dependency with openai-java-core
2. Copy openai-java-client-okhttp's OkHttpClient class into your code and customize it
3. Construct OpenAIClientImpl or OpenAIClientAsyncImpl, similarly to OpenAIOkHttpClient or OpenAIOkHttpClientAsync, using your customized client

To use a completely custom HTTP client:

1. Replace your openai-java dependency with openai-java-core
2. Write a class that implements the HttpClient interface
3. Construct OpenAIClientImpl or OpenAIClientAsyncImpl, similarly to OpenAIOkHttpClient or OpenAIOkHttpClientAsync, using your new client class

The SDK is typed for convenient usage of the documented API. However, it also supports working with undocumented or not yet supported parts of the API.

To set undocumented parameters, call the putAdditionalHeader, putAdditionalQueryParam, or putAdditionalBodyProperty methods on any Params class:

importcom.openai.core.JsonValue; importcom.openai.models.chat.completions.ChatCompletionCreateParams; ChatCompletionCreateParamsparams = ChatCompletionCreateParams.builder() .putAdditionalHeader("Secret-Header", "42") .putAdditionalQueryParam("secret_query_param", "42") .putAdditionalBodyProperty("secretProperty", JsonValue.from("42")) .build();

These can be accessed on the built object later using the _additionalHeaders(), _additionalQueryParams(), and _additionalBodyProperties() methods.

To set undocumented parameters on nested headers, query params, or body classes, call the putAdditionalProperty method on the nested class:

importcom.openai.core.JsonValue; importcom.openai.models.chat.completions.ChatCompletionCreateParams; ChatCompletionCreateParamsparams = ChatCompletionCreateParams.builder() .responseFormat(ChatCompletionCreateParams.ResponseFormat.builder() .putAdditionalProperty("secretProperty", JsonValue.from("42")) .build()) .build();

These properties can be accessed on the nested built object later using the _additionalProperties() method.

To set a documented parameter or property to an undocumented or not yet supported value, pass a JsonValue object to its setter:

importcom.openai.core.JsonValue; importcom.openai.models.ChatModel; importcom.openai.models.chat.completions.ChatCompletionCreateParams; ChatCompletionCreateParamsparams = ChatCompletionCreateParams.builder() .messages(JsonValue.from(42)) .model(ChatModel.GPT_5_2) .build();

The most straightforward way to create a JsonValue is using its from(...) method:

importcom.openai.core.JsonValue; importjava.util.List; importjava.util.Map; // Create primitive JSON valuesJsonValuenullValue = JsonValue.from(null); JsonValuebooleanValue = JsonValue.from(true); JsonValuenumberValue = JsonValue.from(42); JsonValuestringValue = JsonValue.from("Hello World!"); // Create a JSON array value equivalent to `["Hello", "World"]`JsonValuearrayValue = JsonValue.from(List.of( "Hello", "World" )); // Create a JSON object value equivalent to `{"a": 1, "b": 2 }`JsonValueobjectValue = JsonValue.from(Map.of( "a", 1, "b", 2 )); // Create an arbitrarily nested JSON equivalent to://{// "a": [1, 2],// "b": [3, 4]// }JsonValuecomplexValue = JsonValue.from(Map.of( "a", List.of( 1, 2 ), "b", List.of( 3, 4 ) ));

Normally a Builder class's build method will throw IllegalStateException if any required parameter or property is unset.

To forcibly omit a required parameter or property, pass JsonMissing:

importcom.openai.core.JsonMissing; importcom.openai.models.ChatModel; importcom.openai.models.chat.completions.ChatCompletionCreateParams; ChatCompletionCreateParamsparams = ChatCompletionCreateParams.builder() .model(ChatModel.GPT_5_2) .messages(JsonMissing.of()) .build();

To access undocumented response properties, call the _additionalProperties() method:

importcom.openai.core.JsonValue; importjava.util.Map; Map<String, JsonValue> additionalProperties = client.chat().completions().create(params)._additionalProperties(); JsonValuesecretPropertyValue = additionalProperties.get("secretProperty"); Stringresult = secretPropertyValue.accept(newJsonValue.Visitor<>(){@OverridepublicStringvisitNull(){return"It's null!"} @OverridepublicStringvisitBoolean(booleanvalue){return"It's a boolean!"} @OverridepublicStringvisitNumber(Numbervalue){return"It's a number!"} // Other methods include `visitMissing`, `visitString`, `visitArray`, and `visitObject`// The default implementation of each unimplemented method delegates to `visitDefault`, which throws by default, but can also be overridden });

To access a property's raw JSON value, which may be undocumented, call its _ prefixed method:

importcom.openai.core.JsonField; importcom.openai.models.chat.completions.ChatCompletionMessageParam; importjava.util.Optional; JsonField<List<ChatCompletionMessageParam>> messages = client.chat().completions().create(params)._messages(); if (messages.isMissing()){// The property is absent from the JSON response } elseif (messages.isNull()){// The property was set to literal null } else{// Check if value was provided as a string// Other methods include `asNumber()`, `asBoolean()`, etc.Optional<String> jsonString = messages.asString(); // Try to deserialize into a custom typeMyClassmyObject = messages.asUnknown().orElseThrow().convert(MyClass.class)}

In rare cases, the API may return a response that doesn't match the expected type. For example, the SDK may expect a property to contain a String, but the API could return something else.

By default, the SDK will not throw an exception in this case. It will throw OpenAIInvalidDataException only if you directly access the property.

If you would prefer to check that the response is completely well-typed upfront, then either call validate():

importcom.openai.models.chat.completions.ChatCompletion; ChatCompletionchatCompletion = client.chat().completions().create(params).validate();

Or configure the method call to validate the response using the responseValidation method:

importcom.openai.models.chat.completions.ChatCompletion; ChatCompletionchatCompletion = client.chat().completions().create( params, RequestOptions.builder().responseValidation(true).build() );

Or configure the default for all method calls at the client level:

importcom.openai.client.OpenAIClient; importcom.openai.client.okhttp.OpenAIOkHttpClient; OpenAIClientclient = OpenAIOkHttpClient.builder() .fromEnv() .responseValidation(true) .build();

Java enum classes are not trivially forwards compatible. Using them in the SDK could cause runtime exceptions if the API is updated to respond with a new enum value.

Using JsonField<T> enables a few features:

• Allowing usage of undocumented API functionality
• Lazily validating the API response against the expected shape
• Representing absent vs explicitly null values

It is not backwards compatible to add new fields to a data class and we don't want to introduce a breaking change every time we add a field to a class.

Checked exceptions are widely considered a mistake in the Java programming language. In fact, they were omitted from Kotlin for this reason.

Checked exceptions:

• Are verbose to handle
• Encourage error handling at the wrong level of abstraction, where nothing can be done about the error
• Are tedious to propagate due to the function coloring problem
• Don't play well with lambdas (also due to the function coloring problem)

This package generally follows SemVer conventions, though certain backwards-incompatible changes may be released as minor versions:

1. Changes to library internals which are technically public but not intended or documented for external use. (Please open a GitHub issue to let us know if you are relying on such internals.)
2. Changes that we do not expect to impact the vast majority of users in practice.

We take backwards-compatibility seriously and work hard to ensure you can rely on a smooth upgrade experience.

We are keen for your feedback; please open an issue with questions, bugs, or suggestions.