60db (HTTP)

@saraudio/sixtydb supports HTTP batch transcription and the existing WAV segment aggregator. It does not implement WebSocket streaming or partial transcripts.

pnpm add @saraudio/sixtydb

On your server, load credentials at runtime:

import { readFile } from 'node:fs/promises';
import { sixtydb } from '@saraudio/sixtydb';

const provider = sixtydb({ auth: { apiKey: process.env.SIXTYDB_API_KEY } });
const result = await provider.transcribe(await readFile('recording.wav'));
console.log(result.text);

For the HTTP quickstart, use this provider, transport: 'http', a mono 16 kHz recorder, and flushOnSegmentEnd: true. The runtime encodes PCM to WAV before upload. Direct byte inputs must contain WAV; other supported audio formats can be supplied as a Blob with its MIME type. Keep workspace API keys server-side; browser applications should use their own backend proxy, configured through baseUrl.

Optional provider settings: language, diarization, plain-string context, and CSV keywords. Per-call language/diarization override provider defaults. Requests enable word timestamps and confidence. Results use milliseconds for word timing and audio-duration span; raw segments, speaker arrays, duration, and warnings remain in result.metadata. No word speaker labels or aggregate confidence are fabricated. Empty transcripts for silence are successful results.

Upload limit: 10,000,000 bytes, following the stricter current endpoint reference. The service validates the one-hour duration limit. Only JSON responses are supported. Cancellation uses the supplied AbortSignal and produces AbortedError. The adapter surfaces authentication, rate-limit (Retry-After), network, malformed response, and provider errors; it does not automatically retry uploads.

See the 60db REST reference for supported languages, server limits, diarization surcharge, and context-plan requirements. No live API key is bundled or needed for the contract tests.