Quickstart

import WebSocket from 'ws';
import { randomUUID } from 'node:crypto';

const socket = new WebSocket('wss://router.speko.dev/v1/stt/stream', {
  headers: {
    Authorization: `Bearer ${process.env.SPEKO_API_KEY}`,
    'Idempotency-Key': randomUUID(),
  },
});

socket.on('open', () => {
  // The first frame, within 10 seconds of connecting.
  socket.send(JSON.stringify({
    type: 'session.configure',
    routing: { mode: 'auto', objective: 'latency' },
    audio: { encoding: 'pcm_s16le', sample_rate_hz: 16000, channels: 1 },
    language: 'en',
  }));

  // Then the live audio itself, as binary frames.
  callAudio.on('data', (chunk) => socket.send(chunk));
  callAudio.on('end', () => {
    // Ask for the remaining finals, then end the session.
    socket.send(JSON.stringify({ type: 'input.commit' }));
    socket.send(JSON.stringify({ type: 'session.close' }));
  });
});

socket.on('message', (raw, isBinary) => {
  if (isBinary) return;
  const frame = JSON.parse(raw.toString());
  if (frame.type === 'transcript.delta') process.stdout.write(frame.text);
  // One per finalized utterance, so a long call emits many of these.
  if (frame.type === 'transcript.final') console.log('\n✓', frame.text);
  // Exactly one terminal frame ends the session: session.closed or error.
  if (frame.type === 'session.closed' || frame.type === 'error') socket.close();
});

import asyncio, json, os, uuid
import websockets

async def main():
    async with websockets.connect(
        "wss://router.speko.dev/v1/stt/stream",
        additional_headers={
            "Authorization": f"Bearer {os.environ['SPEKO_API_KEY']}",
            "Idempotency-Key": str(uuid.uuid4()),
        },
    ) as socket:
        # The first frame, within 10 seconds of connecting.
        await socket.send(json.dumps({
            "type": "session.configure",
            "routing": {"mode": "auto", "objective": "latency"},
            "audio": {"encoding": "pcm_s16le", "sample_rate_hz": 16000, "channels": 1},
            "language": "en",
        }))

        async def send_audio():
            async for chunk in call_audio():   # live audio, as binary frames
                await socket.send(chunk)
            # Ask for the remaining finals, then end the session.
            await socket.send(json.dumps({"type": "input.commit"}))
            await socket.send(json.dumps({"type": "session.close"}))

        async def read_frames():
            async for raw in socket:
                frame = json.loads(raw)
                if frame["type"] == "transcript.delta":
                    print(frame["text"], end="", flush=True)
                elif frame["type"] == "transcript.final":
                    # One per finalized utterance, not one per session.
                    print("\n✓", frame["text"])
                elif frame["type"] in ("session.closed", "error"):
                    break   # exactly one terminal frame ends the session

        await asyncio.gather(send_audio(), read_frames())

asyncio.run(main())