@d-id/client-sdk 1.2.7-staging.363 → 1.2.8-staging.365

package/dist/livekit-manager-CgGPcchb.js

@@ -0,0 +1,471 @@
+import { c as Ue, C as z, S as f, a as Oe, b as je, A as g, d as Ne, l as le, e as ze, t as Be, f as A, g as O, h as Je, i as Qe, j as R, n as We } from "./index-BXfxP3Yk.js";
+function qe() {
+  try {
+    return Intl.DateTimeFormat().resolvedOptions().timeZone ?? "";
+  } catch {
+    return "";
+  }
+}
+function He() {
+  if (typeof navigator > "u")
+    return "";
+  const l = navigator.userAgent ?? "", C = (navigator.platform ?? "").toLowerCase();
+  let u = "Unknown";
+  /android/i.test(l) ? u = "Android" : /iphone|ipad|ipod/i.test(l) ? u = "iOS" : C.includes("win") ? u = "Windows" : C.includes("mac") ? u = "Mac OS X" : C.includes("linux") && (u = "Linux");
+  const i = u === "Android" || u === "iOS" || /Mobi/i.test(l);
+  return `${u}, ${i ? "Mobile" : "Desktop"}`;
+}
+function Xe() {
+  const l = {}, C = qe();
+  C && (l.timezone = C);
+  const u = He();
+  return u && (l.device = u), l;
+}
+function Ze(l, C, u, i) {
+  const M = Ue(l, `${C}/v2/agents/${u}`, i);
+  return {
+    async createStream(h) {
+      return M.post("/sessions", h);
+    }
+  };
+}
+const Ge = {
+  [f.ChatAnswer]: z.Answer,
+  [f.ChatPartial]: z.Partial
+}, pe = 2e4;
+async function Ye() {
+  try {
+    return await import("./livekit-client.esm-DypJ_fOm.js");
+  } catch {
+    throw new Error(
+      "LiveKit client is required for this streaming manager. Please install it using: npm install livekit-client"
+    );
+  }
+}
+const en = {
+  excellent: O.Strong,
+  good: O.Strong,
+  poor: O.Weak,
+  lost: O.Unknown,
+  unknown: O.Unknown
+}, D = (l = "Stream Error") => new ze(l);
+function fe(l, C, u) {
+  var i, M;
+  throw C("Failed to connect to LiveKit room:", l), (i = u.onConnectionStateChange) == null || i.call(u, A.Fail, "internal:init-error"), (M = u.onError) == null || M.call(u, l, { sessionId: "" }), l;
+}
+async function tn(l, C, u) {
+  var ae;
+  const i = Oe(u.debug || !1, "LiveKitStreamingManager"), { Room: M, RoomEvent: h, ConnectionState: k, Track: X } = await Ye(), { callbacks: n, auth: ge, baseURL: he, analytics: Z } = u;
+  let a = null, E = !1;
+  const G = je.Fluent;
+  let L = null;
+  const b = { isPublishing: !1, publication: null }, B = { isPublishing: !1, publication: null };
+  let m = null, S = null, j = null, V = !1;
+  a = new M({
+    adaptiveStream: !1,
+    // Must be false to use mediaStreamTrack directly
+    dynacast: !0
+  });
+  let P = null, v = g.Idle, I = !0;
+  const F = /* @__PURE__ */ new Set(), me = Ze(ge, he || Ne, l, n.onError);
+  let y, $, K, Y = !0;
+  try {
+    const e = await me.createStream({
+      transport: C.transport,
+      chat_persist: C.chat_persist ?? !0,
+      verbose: u.verbose ?? !1
+    }), { id: t, session_token: o, session_url: r, interrupt_enabled: s } = e;
+    (ae = n.onStreamCreated) == null || ae.call(n, { session_id: t, stream_id: t, agent_id: l }), y = t, $ = o, K = r, Y = s ?? !0, await a.prepareConnection(K, $);
+  } catch (e) {
+    fe(e, i, n);
+  }
+  if (!K || !$ || !y)
+    return Promise.reject(new Error("Failed to initialize LiveKit stream"));
+  a.on(h.ConnectionStateChanged, ye).on(h.ConnectionQualityChanged, Te).on(h.ParticipantConnected, ve).on(h.ParticipantDisconnected, we).on(h.TrackSubscribed, Ee).on(h.TrackUnsubscribed, Pe).on(h.DataReceived, Me).on(h.MediaDevicesError, ke).on(h.TranscriptionReceived, Ce).on(h.EncryptionError, Ie).on(h.TrackSubscriptionFailed, $e);
+  function Ce(e, t) {
+    t != null && t.isLocal && (le.update(), v === g.Talking && (v = g.Idle));
+  }
+  async function Se() {
+    const e = Xe();
+    if (!(!a || Object.keys(e).length === 0))
+      try {
+        await a.localParticipant.setAttributes(e);
+      } catch (t) {
+        i("Failed to set user context attributes", t);
+      }
+  }
+  try {
+    await a.connect(K, $), i("LiveKit room joined successfully"), Se(), P = setTimeout(() => {
+      var t;
+      i(
+        `Track subscription timeout - no track subscribed within ${pe / 1e3} seconds after connect`
+      ), P = null;
+      const e = D("Track subscription timeout");
+      Z.track("connectivity-error", {
+        error: Be(e),
+        sessionId: y
+      }), (t = n.onError) == null || t.call(n, e, { sessionId: y }), H("internal:track-subscription-timeout");
+    }, pe);
+  } catch (e) {
+    fe(e, i, n);
+  }
+  Z.enrich({
+    "stream-type": G
+  });
+  function ye(e) {
+    var t, o, r, s;
+    switch (i("Connection state changed:", e), e) {
+      case k.Connecting:
+        i("CALLBACK: onConnectionStateChange(Connecting)"), (t = n.onConnectionStateChange) == null || t.call(n, A.Connecting, "livekit:connecting");
+        break;
+      case k.Connected:
+        i("LiveKit room connected successfully"), E = !0;
+        break;
+      case k.Disconnected:
+        i("LiveKit room disconnected"), E = !1, V = !1, b.publication = null, B.publication = null, (o = n.onConnectionStateChange) == null || o.call(n, A.Disconnected, "livekit:disconnected");
+        break;
+      case k.Reconnecting:
+        i("LiveKit room reconnecting..."), (r = n.onConnectionStateChange) == null || r.call(n, A.Connecting, "livekit:reconnecting");
+        break;
+      case k.SignalReconnecting:
+        i("LiveKit room signal reconnecting..."), (s = n.onConnectionStateChange) == null || s.call(n, A.Connecting, "livekit:signal-reconnecting");
+        break;
+    }
+  }
+  function Te(e, t) {
+    var o;
+    i("Connection quality:", e), t != null && t.isLocal && ((o = n.onConnectivityStateChange) == null || o.call(n, en[e]));
+  }
+  function ve(e) {
+    i("Participant connected:", e.identity);
+  }
+  function we(e) {
+    i("Participant disconnected:", e.identity), H("livekit:participant-disconnected");
+  }
+  function Ae() {
+    var e;
+    j !== R.Start && (i("CALLBACK: onVideoStateChange(Start)"), j = R.Start, (e = n.onVideoStateChange) == null || e.call(n, R.Start));
+  }
+  function ee(e) {
+    var t;
+    j !== R.Stop && (i("CALLBACK: onVideoStateChange(Stop)"), j = R.Stop, (t = n.onVideoStateChange) == null || t.call(n, R.Stop, e));
+  }
+  function Ee(e, t, o) {
+    var s, d, p;
+    i(`Track subscribed: ${e.kind} from ${o.identity}`);
+    const r = e.mediaStreamTrack;
+    if (!r) {
+      i(`No mediaStreamTrack available for ${e.kind}`);
+      return;
+    }
+    L ? (L.addTrack(r), i(`Added ${e.kind} track to shared MediaStream`)) : (L = new MediaStream([r]), i(`Created shared MediaStream with ${e.kind} track`)), e.kind === "audio" && (S = Qe(
+      () => e.getRTCStatsReport(),
+      ({ sttLatency: c, serviceLatency: T }) => {
+        var ce, se, de;
+        const x = le.get(!0);
+        let U = 0;
+        if (c) {
+          const ue = ((se = (ce = m == null ? void 0 : m.getReport()) == null ? void 0 : ce.webRTCStats) == null ? void 0 : se.avgRtt) ?? 0;
+          U = ue > 0 ? Math.round(ue * 1e3) : 0;
+        }
+        const _ = x > 0 ? x + (c ?? 0) + U : void 0, w = _ !== void 0 && T !== void 0 ? _ - T : void 0;
+        (de = n.onFirstAudioDetected) == null || de.call(n, { latency: _, networkLatency: w });
+      }
+    )), e.kind === "video" && ((s = n.onStreamReady) == null || s.call(n), i("CALLBACK: onSrcObjectReady"), (d = n.onSrcObjectReady) == null || d.call(n, L), V || (V = !0, i("CALLBACK: onConnectionStateChange(Connected)"), (p = n.onConnectionStateChange) == null || p.call(n, A.Connected, "livekit:track-subscribed")), m = Je(
+      () => e.getRTCStatsReport(),
+      () => E,
+      We,
+      (c, T) => {
+        i(`Video state change: ${c}`), c === R.Start ? (P && (clearTimeout(P), P = null, i("Track subscription timeout cleared")), Ae()) : c === R.Stop && ee(T);
+      }
+    ), m.start());
+  }
+  function Pe(e, t, o) {
+    i(`Track unsubscribed: ${e.kind} from ${o.identity}`), e.kind === "audio" && (S == null || S.destroy(), S = null), e.kind === "video" && (ee(m == null ? void 0 : m.getReport()), m == null || m.stop(), m = null);
+  }
+  function ne(e, t) {
+    var r;
+    const o = Ge[e];
+    o && ((r = n.onMessage) == null || r.call(n, o, { event: o, ...t }));
+  }
+  function J(e, t) {
+    var o, r, s, d, p;
+    if (e === f.ToolCallStarted) {
+      const c = t;
+      I = c.interruptible !== !1, (o = n.onInterruptibleChange) == null || o.call(n, I), F.add(c.call_id), v = g.ToolActive, (r = n.onAgentActivityStateChange) == null || r.call(n, g.ToolActive), (s = n.onToolEvent) == null || s.call(n, f.ToolCallStarted, c);
+      return;
+    }
+    if (e === f.ToolCallDone) {
+      const c = t;
+      te(c.call_id), (d = n.onToolEvent) == null || d.call(n, f.ToolCallDone, c);
+      return;
+    }
+    if (e === f.ToolCallError) {
+      const c = t;
+      te(c.call_id), (p = n.onToolEvent) == null || p.call(n, f.ToolCallError, c);
+    }
+  }
+  function te(e) {
+    var t;
+    F.delete(e), F.size === 0 && v === g.ToolActive && (v = g.Idle, (t = n.onAgentActivityStateChange) == null || t.call(n, g.Idle));
+  }
+  function Re(e, t) {
+    var o, r, s, d, p, c;
+    if (I = ((o = t.metadata) == null ? void 0 : o.interruptible) !== !1, (r = n.onInterruptibleChange) == null || r.call(n, I), e === f.StreamVideoCreated) {
+      v = g.Talking, (s = n.onAgentActivityStateChange) == null || s.call(n, g.Talking), S == null || S.arm({
+        sttLatency: (d = t == null ? void 0 : t.stt) == null ? void 0 : d.latency,
+        serviceLatency: t == null ? void 0 : t.serviceLatency
+      });
+      return;
+    }
+    if (F.size > 0) {
+      v !== g.ToolActive && (v = g.ToolActive, (p = n.onAgentActivityStateChange) == null || p.call(n, g.ToolActive));
+      return;
+    }
+    I && (v = g.Idle, (c = n.onAgentActivityStateChange) == null || c.call(n, g.Idle));
+  }
+  function N(e, t) {
+    var d, p, c, T;
+    const o = ((p = (d = m == null ? void 0 : m.getReport()) == null ? void 0 : d.webRTCStats) == null ? void 0 : p.avgRtt) ?? 0, r = o > 0 ? Math.round(o / 2 * 1e3) : 0, s = { ...t, downstreamNetworkLatency: r };
+    u.debug && ((c = t == null ? void 0 : t.metadata) != null && c.sentiment) && (s.sentiment = {
+      id: t.metadata.sentiment.id,
+      name: t.metadata.sentiment.sentiment
+    }), (T = n.onMessage) == null || T.call(n, e, s), Re(e, t);
+  }
+  function Le(e, t) {
+    var o;
+    (o = n.onMessage) == null || o.call(n, z.Transcribe, { event: z.Transcribe, ...t }), queueMicrotask(() => {
+      var r;
+      (r = n.onAgentActivityStateChange) == null || r.call(n, g.Loading);
+    });
+  }
+  const be = {
+    [f.ChatAnswer]: ne,
+    [f.ChatPartial]: ne,
+    [f.ToolCallStarted]: J,
+    [f.ToolCallDone]: J,
+    [f.ToolCallError]: J,
+    [f.StreamVideoCreated]: N,
+    [f.StreamVideoDone]: N,
+    [f.StreamVideoError]: N,
+    [f.StreamVideoRejected]: N,
+    [f.ChatAudioTranscribed]: Le
+  };
+  function Me(e, t, o, r) {
+    const s = new TextDecoder().decode(e);
+    try {
+      const d = JSON.parse(s), p = r || d.subject;
+      if (i("Data received:", { subject: p, data: d }), !p) return;
+      const c = be[p];
+      c == null || c(p, d);
+    } catch (d) {
+      i("Failed to parse data channel message:", d);
+    }
+  }
+  function ke(e) {
+    var t;
+    i("Media devices error:", e), (t = n.onError) == null || t.call(n, D(), { sessionId: y });
+  }
+  function Ie(e) {
+    var t;
+    i("Encryption error:", e), (t = n.onError) == null || t.call(n, D(), { sessionId: y });
+  }
+  function $e(e, t, o) {
+    i("Track subscription failed:", { trackSid: e, participant: t, reason: o });
+  }
+  function Ke(e, t, o) {
+    for (const [r, s] of o)
+      if (s.source === t && s.track) {
+        const d = s.track.mediaStreamTrack;
+        if (d === e || (d == null ? void 0 : d.id) === e.id)
+          return s;
+      }
+    return null;
+  }
+  async function ie(e, t, o, r, s, d) {
+    var x, U, _;
+    if (!E || !a)
+      throw i(`Room is not connected, cannot publish ${r} stream`), new Error("Room is not connected");
+    if (e.isPublishing) {
+      i(`${r} publish already in progress, skipping`);
+      return;
+    }
+    const p = o(t);
+    if (p.length === 0)
+      throw new Error(`No ${r} track found in the provided MediaStream`);
+    const c = p[0], T = Ke(c, r, s());
+    if (T) {
+      i(`${r} track is already published, skipping`, {
+        trackId: c.id,
+        publishedTrackId: (U = (x = T.track) == null ? void 0 : x.mediaStreamTrack) == null ? void 0 : U.id
+      }), e.publication = T;
+      return;
+    }
+    if ((_ = e.publication) != null && _.track) {
+      const w = e.publication.track.mediaStreamTrack;
+      w !== c && (w == null ? void 0 : w.id) !== c.id && (i(`Unpublishing existing ${r} track before publishing new one`), await d());
+    }
+    i(`Publishing ${r} track from provided MediaStream`, { trackId: c.id }), e.isPublishing = !0;
+    try {
+      e.publication = await a.localParticipant.publishTrack(c, { source: r }), i(`${r} track published successfully`, { trackSid: e.publication.trackSid });
+    } catch (w) {
+      throw i(`Failed to publish ${r} track:`, w), w;
+    } finally {
+      e.isPublishing = !1;
+    }
+  }
+  async function oe(e, t) {
+    if (!(!e.publication || !e.publication.track))
+      try {
+        a && (await a.localParticipant.unpublishTrack(e.publication.track, !1), i(`${t} track unpublished`));
+      } catch (o) {
+        i(`Error unpublishing ${t} track:`, o);
+      } finally {
+        e.publication = null;
+      }
+  }
+  async function _e(e) {
+    return ie(
+      b,
+      e,
+      (t) => t.getAudioTracks(),
+      X.Source.Microphone,
+      () => a.localParticipant.audioTrackPublications,
+      Q
+    );
+  }
+  async function Q() {
+    return oe(b, "Microphone");
+  }
+  async function De(e) {
+    if (!E || !a)
+      throw i("Cannot replace microphone track: room is not connected"), new Error("Room is not connected");
+    if (e.kind !== "audio")
+      throw i("Cannot replace microphone track: not an audio track", { kind: e.kind }), new Error("Microphone track must be an audio track");
+    if (b.isPublishing)
+      throw i("Cannot replace microphone track: publish in progress"), new Error("Microphone publish in progress");
+    const t = b.publication;
+    if (!t || !t.track)
+      throw i("Cannot replace microphone track: no publication to replace"), new Error("No microphone publication to replace");
+    try {
+      b.isPublishing = !0, await t.track.replaceTrack(e), i("Microphone track replaced", { trackId: e.id, trackSid: t.trackSid });
+    } finally {
+      b.isPublishing = !1;
+    }
+  }
+  async function Ve(e) {
+    return ie(
+      B,
+      e,
+      (t) => t.getVideoTracks(),
+      X.Source.Camera,
+      () => a.localParticipant.videoTrackPublications,
+      W
+    );
+  }
+  async function W() {
+    return oe(B, "Camera");
+  }
+  function Fe() {
+    L && (L.getTracks().forEach((e) => e.stop()), L = null);
+  }
+  async function q(e, t) {
+    var o, r;
+    if (!E || !a) {
+      i("Room is not connected for sending messages"), (o = n.onError) == null || o.call(n, D(), {
+        sessionId: y
+      });
+      return;
+    }
+    try {
+      await a.localParticipant.sendText(e, { topic: t }), i("Message sent successfully:", e);
+    } catch (s) {
+      i("Failed to send message:", s), (r = n.onError) == null || r.call(n, D(), { sessionId: y });
+    }
+  }
+  async function re(e) {
+    var t;
+    try {
+      const r = JSON.parse(e).topic;
+      return q("", r);
+    } catch (o) {
+      i("Failed to send data channel message:", o), (t = n.onError) == null || t.call(n, D(), { sessionId: y });
+    }
+  }
+  function xe(e) {
+    return q(
+      e,
+      "lk.chat"
+      /* Chat */
+    );
+  }
+  async function H(e) {
+    var t, o;
+    P && (clearTimeout(P), P = null), S == null || S.destroy(), S = null, a && ((t = n.onConnectionStateChange) == null || t.call(n, A.Disconnecting, e), await Promise.all([Q(), W()]), await a.disconnect()), Fe(), E = !1, V = !1, F.clear(), (o = n.onAgentActivityStateChange) == null || o.call(n, g.Idle), v = g.Idle;
+  }
+  return {
+    speak(e) {
+      const t = typeof e == "string" ? e : JSON.stringify(e);
+      return q(
+        t,
+        "did.speak"
+        /* Speak */
+      );
+    },
+    disconnect: () => H("user:disconnect"),
+    async reconnect() {
+      var e, t;
+      if ((a == null ? void 0 : a.state) === k.Connected) {
+        i("Room is already connected");
+        return;
+      }
+      if (!a || !K || !$)
+        throw i("Cannot reconnect: missing room, URL or token"), new Error("Cannot reconnect: session not available");
+      i("Reconnecting to LiveKit room, state:", a.state), V = !1, (e = n.onConnectionStateChange) == null || e.call(n, A.Connecting, "user:reconnect");
+      try {
+        if (await a.connect(K, $), i("Room reconnected"), E = !0, a.remoteParticipants.size === 0) {
+          if (i("Waiting for agent to join..."), !await new Promise((r) => {
+            const s = setTimeout(() => {
+              a == null || a.off(h.ParticipantConnected, d), r(!1);
+            }, 5e3), d = () => {
+              clearTimeout(s), a == null || a.off(h.ParticipantConnected, d), r(!0);
+            };
+            a == null || a.on(h.ParticipantConnected, d);
+          }))
+            throw i("Agent did not join within timeout"), await a.disconnect(), new Error("Agent did not rejoin the room");
+          i("Agent joined, reconnection successful");
+        }
+      } catch (o) {
+        throw i("Failed to reconnect:", o), (t = n.onConnectionStateChange) == null || t.call(n, A.Fail, "user:reconnect-failed"), o;
+      }
+    },
+    sendDataChannelMessage: re,
+    sendTextMessage: xe,
+    publishMicrophoneStream: _e,
+    unpublishMicrophoneStream: Q,
+    replaceMicrophoneTrack: De,
+    publishCameraStream: Ve,
+    unpublishCameraStream: W,
+    interrupt(e) {
+      e !== "text" && re(JSON.stringify({
+        topic: "did.interrupt"
+        /* Interrupt */
+      }));
+    },
+    registerRpcMethod(e, t) {
+      a == null || a.registerRpcMethod(e, t);
+    },
+    unregisterRpcMethod(e) {
+      a == null || a.unregisterRpcMethod(e);
+    },
+    sessionId: y,
+    streamId: y,
+    streamType: G,
+    interruptAvailable: Y,
+    isInterruptible: I
+  };
+}
+export {
+  tn as createLiveKitStreamingManager,
+  fe as handleInitError
+};

package/dist/src/utils/user-context.d.ts

@@ -0,0 +1,15 @@
+/**
+ * Best-effort user context (location + device) resolved entirely on the client
+ * and attached to the LiveKit participant as attributes on connect. The agent
+ * backend reads these to enrich the LLM context (see `get_user_context`).
+ *
+ * `location` is the IANA timezone (e.g. `America/New_York`) — a coarse but
+ * network-free location proxy. `device` is a short human-readable descriptor.
+ */
+/**
+ * Non-empty `{ timezone, device }` attributes to attach to the participant.
+ * Keys with no resolvable value are omitted. The backend infers the user's
+ * country from the timezone (a timezone identifies a country reliably, but its
+ * reference city is not the user's actual city).
+ */
+export declare function getUserContextAttributes(): Record<string, string>;

package/dist/src/utils/user-context.test.d.ts

@@ -0,0 +1 @@
+export {};

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@d-id/client-sdk",
   "private": false,
-  "version": "1.2.7-staging.363",
+  "version": "1.2.8-staging.365",
   "type": "module",
   "description": "d-id client sdk",
   "repository": {