simple-openai has been updated to support:
simple-openai has been updated to support Realtime API, so you can enable sync speech-to-speech communication in your Java applications, as well as multimodality (audio, text) and function calling.
Here you can see a demo code of Realtime:
simple-openai has been updated for:
reasoningEffort field and developer role to ChatCompletions.PS: Thanks to OpenAI Staff for including simple-openai in the list of Java community libraries.
Take a look at simple-openai.
Take a look at simple-openai.
Hey @sashirestela, do you have an example of using a WebRTC connection with the realtime API in java? Does the library enable this?
do you have an example of using a WebRTC connection with the realtime API in java?
I can prepare an example for WebRTC supported by simple-openai.
Does the library enable this?
Yes, simple-openai enables this via the Session Token API to provide ephemeral tokens to your web application.
Great thank you :). I saw the session API but didnât know if there was an easier way to use webrtc completely in java, or if it was meant more for use in a client app. I was looking to do Java directly so an example would be amazing!
Itâs correct, WebRTC is for use in client applications, but it needs a backend endpoint to request for an ephemeral token, because you donât want to expose your regular OpenAI api key.
Iâve created a minimalist Java application which serves a frontend component (web files) and a backend component (Java classes). The backend is exposing an endpoint for the frontend to ask for an ephemeral token that is used to start a WebRTC session from the frontend. That backend is supported by simple-openai.
Here is the OpenAI WebRTC Java Demo:
Deepseek API has been included to the simple-openai list of supported providers with compatible OpenAI API.
For Deepseek the supported services are:
Other compatible OpenAI providers supported by simple-openai are:
Mistral, Azure OpenAI, Anyscale.
Whatâs new in simple-openai?
For http client communications it uses Javaâs HttpClient by default, now you can choose to use OkHttp. See how to use.
For websocket communications it uses Javaâs HttpClient by default, now you can choose to use OkHttp. See how to use.
Support for response interceptors to modify the response body after itâs received.
Updated the RealtimeDemo to test the user transcription events in addition to other events.
simple-openai has been updated for:
Just to let you know that simple-openai is ready to work with the latest o3-mini model:
reasoning_effort in the request.usage.completion_tokens_details.reasoning_tokens in the response.Thanks for the update.
Keep up the great work!
Instructions for AndroidFollow the next instructions to run simple-openai in Android devices:
android {
//...
defaultConfig {
//...
minSdk 24
//...
}
//...
compileOptions {
sourceCompatibility JavaVersion.VERSION_11
targetCompatibility JavaVersion.VERSION_11
}
kotlinOptions {
jvmTarget = '11'
}
}
dependencies {
//...
implementation 'io.github.sashirestela:simple-openai:[simple-openai_version]'
implementation 'com.squareup.okhttp3:okhttp:[okhttp_version]'
}
In Java:
SimpleOpenAI openAI = SimpleOpenAI.builder()
.apiKey(API_KEY)
.clientAdapter(new OkHttpClientAdapter()) // Optionally you could add a custom OkHttpClient
.build();
In Kotlin:
val openAI = SimpleOpenAI.builder()
.apiKey(API_KEY)
.clientAdapter(OkHttpClientAdapter()) // Optionally you could add a custom OkHttpClient
.build()
Thanks to simple-openaiâs versatile design, you can work with the Gemini 2.0 Flash model in Java:
simple-openai has been updated for:
code_interpreter, file_search, and function, OpenAI is sending unknown tools such as browser sometimes. Now simple-openai wonât fail if they send any unknown tool.simple-openai has been updated for:
This project is really great, thank you.
@CodeTheory Thank you for using simple-openai and for your kind words.
What do you like most about simple-openai? Is there anything you miss?