Smallest AI Java SDK

Official Java client for the Smallest AI API — Waves (TTS / STT) and Atoms (voice agents).

Generated with Fern from the same API definition as the Python, Node, and Go SDKs.

Requirements

• Java 17+

Install

Available via JitPack. OkHttp and Jackson are pulled in transitively.

Gradle

repositories {
    mavenCentral()
    maven { url 'https://jitpack.io' }
}
dependencies {
    implementation 'com.github.smallest-inc:smallest-java-sdk:0.1.4'
}

Maven

<repositories>
  <repository>
    <id>jitpack.io</id>
    <url>https://jitpack.io</url>
  </repository>
</repositories>

<dependency>
  <groupId>com.github.smallest-inc</groupId>
  <artifactId>smallest-java-sdk</artifactId>
  <version>0.1.4</version>
</dependency>

Client

import ai.smallest.SmallestAI;

SmallestAI client = SmallestAI.builder()
    .apiKey(System.getenv("SMALLEST_API_KEY")) // defaults to the env var
    .build();

Get an API key at app.smallest.ai.

Text-to-speech

import ai.smallest.resources.waves.types.TtsRequest;
import ai.smallest.resources.waves.types.TtsRequestOutputFormat;

var audio = client.waves().synthesizeTts(
    TtsRequest.builder()
        .text("The quick brown fox jumps over the lazy dog.")
        .voiceId("avery")
        .outputFormat(TtsRequestOutputFormat.WAV)
        .sampleRate(16000)
        .build()); // InputStream of audio

Speech-to-text (file / batch)

import ai.smallest.resources.waves.requests.TranscribePulseWavesRequest;
import ai.smallest.resources.waves.types.TranscribePulseWavesRequestLanguage;

var resp = client.waves().transcribePulse(
    TranscribePulseWavesRequest.builder()
        .body(audioBytes)
        .language(TranscribePulseWavesRequestLanguage.EN) // set explicitly
        .build());
System.out.println(resp.getTranscription().orElse(""));

Streaming speech-to-text (real-time / telephony)

Push audio as it arrives and receive interim + final transcripts over a WebSocket. Match encoding and sampleRate to your audio — the server does not resample.

import ai.smallest.resources.waves.pulsesttstreaming.types.PulseStreamEncoding;
import ai.smallest.resources.waves.pulsesttstreaming.types.PulseFinalizeSignalMessage;
import ai.smallest.resources.waves.pulsesttstreaming.types.PulseFinalizeSignalMessageType;
import ai.smallest.resources.waves.pulsesttstreaming.websocket.PulseSttStreamingConnectOptions;
import okio.ByteString;

var stt = client.waves().pulseSttStreaming().pulseSttStreamingWebSocket();

stt.receiveTranscribeStreamingPulse(e -> {
    if (e.getIsFinal().orElse(false)) {
        System.out.println("final: " + e.getTranscript().orElse(""));
    }
});
stt.onError(ex -> System.err.println(ex.getMessage()));

stt.connect(PulseSttStreamingConnectOptions.builder()
    .language("en")
    .encoding(PulseStreamEncoding.MULAW) // G.711 u-law telephony; LINEAR16 for raw PCM16
    .sampleRate(8000)                    // 8000 for telephony
    .build()).get();

// feed audio as it arrives (e.g. per RTP packet), then finalize
stt.transcribeStreamingPulseSendAudio(ByteString.of(audioChunk)).get();
stt.transcribeStreamingPulseSendFinalize(
    PulseFinalizeSignalMessage.builder()
        .type(PulseFinalizeSignalMessageType.FINALIZE).build()).get();

A full runnable telephony (8 kHz μ-law) example is in examples/StreamingSttTelephony.java.

Status

Initial 0.1.4 snapshot. Real-time streaming STT/TTS WebSocket clients are included and evolving.