Response Streaming

note

This page describes response streaming with a low-level LLM API. See AI Services for a high-level LLM API.

LLMs generate text one token at a time, so many LLM providers offer a way to stream the response token-by-token instead of waiting for the entire text to be generated. This significantly improves the user experience, as the user does not need to wait an unknown amount of time and can start reading the response almost immediately.

For the ChatLanguageModel and LanguageModel interfaces, there are corresponding StreamingChatLanguageModel and StreamingLanguageModel interfaces. These have a similar API but can stream the responses. They accept an implementation of the StreamingResponseHandler interface as an argument.

public interface StreamingResponseHandler<T> {

    void onNext(String token);
 
    default void onComplete(Response<T> response) {}

    void onError(Throwable error);
}

By implementing StreamingResponseHandler, you can define actions for the following events:

• When the next token is generated: onNext(String token) is invoked. For instance, you can send the token directly to the UI as soon as it becomes available.
• When the LLM has completed generation: onComplete(Response<T> response) is invoked. Here, T stands for AiMessage in the case of StreamingChatLanguageModel, and String for StreamingLanguageModel. The Response object contains the complete response.
• When an error occurs: onError(Throwable error) is invoked.

Below is an example of how to implement streaming with StreamingChatLanguageModel:

StreamingChatLanguageModel model = OpenAiStreamingChatModel.builder()
    .apiKey(System.getenv("OPENAI_API_KEY"))
    .modelName(GPT_4_O_MINI)
    .build();

String userMessage = "Tell me a joke";

model.generate(userMessage, new StreamingResponseHandler<AiMessage>() {

    @Override
    public void onNext(String token) {
        System.out.println("onNext: " + token);
    }

    @Override
    public void onComplete(Response<AiMessage> response) {
        System.out.println("onComplete: " + response);
    }

    @Override
    public void onError(Throwable error) {
        error.printStackTrace();
    }
});

A more compact way to stream the response is to use the LambdaStreamingResponseHandler class. This utility class provides static methods to create a StreamingResponseHandler using lambda expressions. The way to use lambdas to stream the response is quite simple. You just call the onNext() static method with a lambda expression that defines what to do with the token:

import static dev.langchain4j.model.LambdaStreamingResponseHandler.onNext;

model.generate("Tell me a joke", onNext(System.out::print));

The onNextAndError() method allows you to define actions for both the onNext() and onError() events:

import static dev.langchain4j.model.LambdaStreamingResponseHandler.onNextAndError;

model.generate("Tell me a joke", onNextAndError(System.out::print, Throwable::printStackTrace));