Facilitates seamless connection to multiple AI providers through a user-friendly API client. Includes customizable settings, asynchronous execution, and provider-specific configurations for enhanced flexibility.
Easy-to-use generic Kotlin API client for connecting to various AI providers with multiplatform support. Supported AI providers:
repositories {
mavenCentral()
}
dependencies {
implementation("io.github.bay73:generic-ai:0.6.6")
}
Generic-AI uses ktor library to work with http requests so you need to include ktor client corresponding to your platform.
In multiplatform projects, include the generic-AI client dependency in commonMain and select a specific ktor engine for each target.
import com.bay.aiclient.AiClient
fun getResponse() {
// Create a client for specified AI provider.
val client = AiClient.get(AiClient.Type.OPEN_AI) { // Choose provider
apiKey =
defaultModel =
}
job = client.generateText { prompt = }
println(job.await().getOrThrow().response)
models = client.models()
models.models.forEach { println(it.id) }
}
The library created to use in Kotlin code and so usage in Java is cumbersome while possible.
Special AiClientJava wrapper is created to use CompletableFuture for asynchronous execution.
Here is a code example:
There are set of generic settings which can be used for any AI provider to customize model behavior for a specific request.
responseFormat allows to specify returning text, generic JSON or validated JSON schema. The parameter has limited support by different providers:
Some AI providers have additional settings which can be adjusted. To use this you need to request client of a specific class:
fun getResponseWithSpecificParameters() {
// Create a client for specified AI provider.
val client = AiClient.get(CohereClient::class) { // Choose provider
apiKey = "put your API key here"
defaultModel =
}
response = client.generateText {
prompt =
seed =
frequencyPenalty =
}
}
AWS bedrock doesn't support token based authentication. To use it you need to provide special credentials object to the client constructor:
fun getBedrockClient() {
// You can use either [DefaultCredentialsProvider](https://docs.aws.amazon.com/sdk-for-java/latest/developer-guide/credentials-chain.html) or specify access key and token directly.
val credentials = BedrockClient.Credentials("region", true)
// or
val credentials = BedrockClient.Credentials(, , , , )
client = BedrockClient.Builder(creadentals).build()
}
Azure OpenAI requires the resource name which is used to access services as a part of endpoint. To specify resource name use dedicated client constructor which allow to set the resource name as a string:
fun getAzureOpenAiClient() {
// Create a client using specific builder
val client = AzureOpenAiClient.Builder().apply {
resourceName = "your-resource-name"
apiKey = "your-api-key"
}.build()
}
Further usage of this client is the same as for all other AI clients.
Yandex requires to specify a folder - a space where Yandex Cloud resources are created and grouped. It is used as a part of foundation model URI. To specify the folder use dedicated client builder which allow to set the resource folder as a string:
fun getYandexOpenAiClient() {
// Create a client using specific builder
val client = YandexOpenAiClient.Builder().apply {
resourceFolder = "your-resource-name"
apiKey = "your-api-key"
}.build()
}
Further usage of this client is the same as for all other AI clients.
Sample multiplatform application using the library is available at GitHub
Generic AI Kotlin API Client is an open-sourced software licensed under the MIT license. Please note that this is an unofficial library and is not affiliated with or endorsed by any AI provider. Contributions are always welcome!
import com.bay.aiclient.AiClient;
import com.bay.aiclient.AiClientJava;
import com.bay.aiclient.domain.GenerateTextRequest;
import com.bay.aiclient.domain.GenerateTextResponse;
getResponse main() throws ExecutionException, InterruptedException {
AiClient.Builder clientBuilder = AiClient.Companion.getBuilder(AiClient.Type.OPEN_AI); // Choose provider
clientBuilder.setApiKey("put your API key here");
clientBuilder.setDefaultModel("gpt-4o-mini"); // Choose model
AiClient client = clientBuilder.build();
AiClientJava javaClient = new AiClientJava(client); // Java client uses CompletableFuture
// Start request to the AI model.
GenerateTextRequest.Builder requestBuilder = client.textGenerationRequestBuilder();
requestBuilder.setPrompt("When the first LLM were released?");
CompletableFuture<Result<GenerateTextResponse>> response = javaClient.generateText(requestBuilder.build());
// Wait for execution and get response.
response.join();
if (response.isDone()) {
System.out.println(response.get());
}
}
fun customizeRequest() {
val response = cleint.generateText {
model = "model_id" // Model id to use for a specific request.
prompt = "" // User prompt which initiates generation
systemInstructions = "" // Additional system instruction to adjust AI behavior.
responseFormat = ResponseFormat.JSON_OBJECT // Allows to specify response format. See details below.
chatHistory = listOf<TextMessage>() // A list of chat messages in chronological order, representing a conversation between the user and the model.
maxOutputTokens = 100 // The maximum number of tokens that can be generated as part of the response.
stopSequences = listOf<String>() // A list of strings that the model uses to stop generation.
temperature = 0.1 // A non-negative float that tunes the degree of randomness in generation. Lower temperatures mean less random generations.
topP = 0.5 // An alternative way controlling the diversity of the model's responses. It's recommended to use either temperature or topP.
}
}
| Provider | Text | Generic JSON | JSON schema |
|---|
| AI21 Lab | ✅ | ✅ | |
| Anthropic | ✅ | ||
| AWS Bedrock | ✅ | ||
| Azure OpenAI | ✅ | ✅ | ✅ |
| Cerebras | ✅ | ✅ | |
| Cohere | ✅ | ✅ | ✅ |
| DeepSeek | ✅ | ✅ | |
| Google Gemini | ✅ | ✅ | ✅ |
| Grok | ✅ | ✅ | ✅ |
| Inception Labs | ✅ | ||
| Mistral | ✅ | ✅ | |
| Novita | ✅ | ✅ | ✅ |
| OpenAI | ✅ | ✅ | ✅ depending on model |
| SambaNova | ✅ | ||
| Together AI | ✅ | ✅ | ✅ depending on model |
| Yandex AI | ✅ | ✅ | ✅ |
Surfaced from shared tags and platforms — no rankings paid for.