Voice input infrastructure for product interfaces.

import { checkCompatibility, createSession } from "dictum-sdk";

const compatibility = checkCompatibility();
if (!compatibility.supported) {
  showUnsupportedBrowser(compatibility.missing);
} else {
  const session = createSession({
    projectKey: "dpk_live_xxx",
    identityToken: await getDictumIdentityToken(),
    sourceLabel: "editor-main",
    target: document.querySelector("#message")
  });

  session.on("state_changed", event => renderVoiceState(event.state));
  session.on("transcript_delta", event => renderFinalSegment(event.data.text));
  session.on("sdk_error", event => showVoiceError(event.data));

  startButton.onclick = () => session.start().catch(showVoiceError);
  pauseButton.onclick = () => session.pause().catch(showVoiceError);
  resumeButton.onclick = () => session.resume().catch(showVoiceError);
  stopButton.onclick = () => session.stop().catch(showVoiceError);
  retryButton.onclick = () => session.retry().catch(showVoiceError);

  const result = await session.result;
  renderFinalTranscript(result.text, result.transport);
}

const report = checkCompatibility();

if (!report.supported) {
  voiceButton.disabled = true;
  showMessage("Please update your browser to use voice input.");
  console.debug({ capture: report.capture, missing: report.missing });
}

import { createHmac } from "node:crypto";

const MAX_IDENTITY_LIFETIME_SECONDS = 7 * 24 * 60 * 60;
const encodeJson = value => Buffer.from(JSON.stringify(value)).toString("base64url");

function createDictumIdentity({ keyId, signingSecret, projectKey, userId }) {
  const iat = Math.floor(Date.now() / 1000);
  const header = { alg: "HS256", typ: "JWT", kid: keyId };
  const payload = {
    sub: String(userId),
    aud: projectKey,
    iat,
    exp: iat + MAX_IDENTITY_LIFETIME_SECONDS
  };
  const unsignedToken = [encodeJson(header), encodeJson(payload)].join(".");
  const signature = createHmac("sha256", signingSecret)
    .update(unsignedToken)
    .digest("base64url");

  return [unsignedToken, signature].join(".");
}

app.post("/api/dictum/identity", requireUser, (request, response) => {
  const identityToken = createDictumIdentity({
    keyId: process.env.DICTUM_SIGNING_KEY_ID,
    signingSecret: process.env.DICTUM_SIGNING_SECRET,
    projectKey: process.env.DICTUM_PROJECT_KEY,
    userId: request.user.id
  });

  response.json({ identityToken });
});

{
  "product": "Dictum",
  "capability": "headless voice-to-text for desktop and mobile web browsers",
  "browserOnly": true,
  "nonBrowserIntegration": "Use the Dictum API directly",
  "package": "dictum-sdk",
  "packagePublished": false,
  "browserSecrets": [
    "project public key",
    "signed customer identity under the global non-configurable seven-day maximum"
  ],
  "serverOnlySecrets": [
    "Dictum signing key",
    "speech provider API keys"
  ],
  "states": [
    "idle", "requesting_mic", "listening", "paused", "retrying", "transcribing", "failed"
  ],
  "events": [
    "state_changed", "volume_changed", "transcript_delta", "text_inserted",
    "mic_denied", "no_transcript", "network_error", "audio_invalid",
    "compat_failed", "limit_reached", "fallback_started", "quota_exceeded",
    "auth_required", "identity_expired", "auth_invalid", "retry_started", "sdk_error"
  ],
  "transport": "WebSocket first; explicit bounded recovery to POST",
  "transcriptRule": "Every transcript_delta segment is provider-confirmed, final, immutable, and append-only.",
  "usageRule": "Usage is measured by the Worker, never submitted by the browser SDK.",
  "privacyRule": "Never log raw audio, transcript text, signed identity, or provider secrets."
}