@voicethere/agent 0.7.1 → 0.7.3

package/dist/templates/voice-starter/agent.js

@@ -130,6 +130,24 @@ var SessionSerialQueue = class {
 var SESSION_START_INIT_DELAY_ENABLED_ENV = "AGENT_SESSION_START_INIT_DELAY_ENABLED";
 var SESSION_START_INIT_DELAY_MS_ENV = "AGENT_SESSION_START_INIT_DELAY_MS";
 var DEFAULT_SESSION_START_INIT_DELAY_MS = 500;
+function isPlayAckMessage(value) {
+  if (!value || typeof value !== "object")
+    return false;
+  const msg = value;
+  return msg.type === "play_ack" && typeof msg.requestId === "string";
+}
+function isPlayStatusAckMessage(value) {
+  if (!value || typeof value !== "object")
+    return false;
+  const msg = value;
+  return msg.type === "play_status_ack" && typeof msg.requestId === "string";
+}
+function isPlayStopAckMessage(value) {
+  if (!value || typeof value !== "object")
+    return false;
+  const msg = value;
+  return msg.type === "play_stop_ack" && typeof msg.requestId === "string";
+}
 function isRecordingControlAckMessage(value) {
   if (!value || typeof value !== "object")
     return false;
@@ -190,7 +208,7 @@ function isParentMessage(value) {
   if (!value || typeof value !== "object")
     return false;
   const msg = value;
-  return msg.type === "session_start" || msg.type === "speech_event" || msg.type === "session_end" || msg.type === "data_channel_message" || msg.type === "data_channel_binary" || msg.type === "idle_timeout" || msg.type === "recording_control_ack" || msg.type === "mix_control_ack" || msg.type === "stt_control_ack" || msg.type === "webhook";
+  return msg.type === "session_start" || msg.type === "speech_event" || msg.type === "session_end" || msg.type === "data_channel_message" || msg.type === "data_channel_binary" || msg.type === "idle_timeout" || msg.type === "recording_control_ack" || msg.type === "play_ack" || msg.type === "play_status_ack" || msg.type === "play_stop_ack" || msg.type === "mix_control_ack" || msg.type === "stt_control_ack" || msg.type === "webhook";
 }
 function isSessionScopedParentMessage(value) {
   return isParentMessage(value) && !isWebhookMessage(value);
@@ -207,7 +225,11 @@ var recordingAvailableBySessionId = /* @__PURE__ */ new Map();
 var mixAvailableBySessionId = /* @__PURE__ */ new Map();
 var ttsPoseAvailableBySessionId = /* @__PURE__ */ new Map();
 var endedSessionIds = /* @__PURE__ */ new Set();
+var PLAY_BYTES_MAX_DECODED_LENGTH = 64 * 1024;
 var pendingRecordingAcks = /* @__PURE__ */ new Map();
+var pendingPlayAcks = /* @__PURE__ */ new Map();
+var pendingPlayStatusAcks = /* @__PURE__ */ new Map();
+var pendingPlayStopAcks = /* @__PURE__ */ new Map();
 var pendingMixAcks = /* @__PURE__ */ new Map();
 var pendingSttAcks = /* @__PURE__ */ new Map();
 function handleRecordingControlAck(message) {
@@ -222,6 +244,46 @@ function handleRecordingControlAck(message) {
     requestId: message.requestId
   });
 }
+function handlePlayAck(message) {
+  const pending = pendingPlayAcks.get(message.requestId);
+  if (!pending)
+    return;
+  clearTimeout(pending.timer);
+  pendingPlayAcks.delete(message.requestId);
+  pending.resolve({
+    ok: message.ok,
+    ...message.playId ? { playId: message.playId } : {},
+    reason: message.reason,
+    requestId: message.requestId
+  });
+}
+function handlePlayStatusAck(message) {
+  const pending = pendingPlayStatusAcks.get(message.requestId);
+  if (!pending)
+    return;
+  clearTimeout(pending.timer);
+  pendingPlayStatusAcks.delete(message.requestId);
+  pending.resolve({
+    ok: message.ok,
+    playId: message.playId,
+    ...message.status ? { status: message.status } : {},
+    reason: message.reason,
+    requestId: message.requestId
+  });
+}
+function handlePlayStopAck(message) {
+  const pending = pendingPlayStopAcks.get(message.requestId);
+  if (!pending)
+    return;
+  clearTimeout(pending.timer);
+  pendingPlayStopAcks.delete(message.requestId);
+  pending.resolve({
+    ok: message.ok,
+    playId: message.playId,
+    reason: message.reason,
+    requestId: message.requestId
+  });
+}
 function handleMixControlAck(message) {
   const pending = pendingMixAcks.get(message.requestId);
   if (!pending)
@@ -307,6 +369,10 @@ function sendParentMessage(message) {
     process.send?.(message);
     return;
   }
+  if (msgType === "play" || msgType === "play_status" || msgType === "play_stop") {
+    process.send?.(message);
+    return;
+  }
   const sessionId = message && typeof message === "object" && "sessionId" in message && typeof message.sessionId === "string" ? message.sessionId : void 0;
   if (!allowOutboundForSession(sessionId))
     return;
@@ -458,6 +524,18 @@ function defineAgent(handlers) {
       handleRecordingControlAck(message);
       return;
     }
+    if (isPlayAckMessage(message)) {
+      handlePlayAck(message);
+      return;
+    }
+    if (isPlayStatusAckMessage(message)) {
+      handlePlayStatusAck(message);
+      return;
+    }
+    if (isPlayStopAckMessage(message)) {
+      handlePlayStopAck(message);
+      return;
+    }
     if (isMixControlAckMessage(message)) {
       handleMixControlAck(message);
       return;

package/dist/templates/webhooks/agent.js

@@ -133,6 +133,24 @@ var SessionSerialQueue = class {
 var SESSION_START_INIT_DELAY_ENABLED_ENV = "AGENT_SESSION_START_INIT_DELAY_ENABLED";
 var SESSION_START_INIT_DELAY_MS_ENV = "AGENT_SESSION_START_INIT_DELAY_MS";
 var DEFAULT_SESSION_START_INIT_DELAY_MS = 500;
+function isPlayAckMessage(value) {
+  if (!value || typeof value !== "object")
+    return false;
+  const msg = value;
+  return msg.type === "play_ack" && typeof msg.requestId === "string";
+}
+function isPlayStatusAckMessage(value) {
+  if (!value || typeof value !== "object")
+    return false;
+  const msg = value;
+  return msg.type === "play_status_ack" && typeof msg.requestId === "string";
+}
+function isPlayStopAckMessage(value) {
+  if (!value || typeof value !== "object")
+    return false;
+  const msg = value;
+  return msg.type === "play_stop_ack" && typeof msg.requestId === "string";
+}
 function isRecordingControlAckMessage(value) {
   if (!value || typeof value !== "object")
     return false;
@@ -193,7 +211,7 @@ function isParentMessage(value) {
   if (!value || typeof value !== "object")
     return false;
   const msg = value;
-  return msg.type === "session_start" || msg.type === "speech_event" || msg.type === "session_end" || msg.type === "data_channel_message" || msg.type === "data_channel_binary" || msg.type === "idle_timeout" || msg.type === "recording_control_ack" || msg.type === "mix_control_ack" || msg.type === "stt_control_ack" || msg.type === "webhook";
+  return msg.type === "session_start" || msg.type === "speech_event" || msg.type === "session_end" || msg.type === "data_channel_message" || msg.type === "data_channel_binary" || msg.type === "idle_timeout" || msg.type === "recording_control_ack" || msg.type === "play_ack" || msg.type === "play_status_ack" || msg.type === "play_stop_ack" || msg.type === "mix_control_ack" || msg.type === "stt_control_ack" || msg.type === "webhook";
 }
 function isSessionScopedParentMessage(value) {
   return isParentMessage(value) && !isWebhookMessage(value);
@@ -210,7 +228,11 @@ var recordingAvailableBySessionId = /* @__PURE__ */ new Map();
 var mixAvailableBySessionId = /* @__PURE__ */ new Map();
 var ttsPoseAvailableBySessionId = /* @__PURE__ */ new Map();
 var endedSessionIds = /* @__PURE__ */ new Set();
+var PLAY_BYTES_MAX_DECODED_LENGTH = 64 * 1024;
 var pendingRecordingAcks = /* @__PURE__ */ new Map();
+var pendingPlayAcks = /* @__PURE__ */ new Map();
+var pendingPlayStatusAcks = /* @__PURE__ */ new Map();
+var pendingPlayStopAcks = /* @__PURE__ */ new Map();
 var pendingMixAcks = /* @__PURE__ */ new Map();
 var pendingSttAcks = /* @__PURE__ */ new Map();
 function handleRecordingControlAck(message) {
@@ -225,6 +247,46 @@ function handleRecordingControlAck(message) {
     requestId: message.requestId
   });
 }
+function handlePlayAck(message) {
+  const pending = pendingPlayAcks.get(message.requestId);
+  if (!pending)
+    return;
+  clearTimeout(pending.timer);
+  pendingPlayAcks.delete(message.requestId);
+  pending.resolve({
+    ok: message.ok,
+    ...message.playId ? { playId: message.playId } : {},
+    reason: message.reason,
+    requestId: message.requestId
+  });
+}
+function handlePlayStatusAck(message) {
+  const pending = pendingPlayStatusAcks.get(message.requestId);
+  if (!pending)
+    return;
+  clearTimeout(pending.timer);
+  pendingPlayStatusAcks.delete(message.requestId);
+  pending.resolve({
+    ok: message.ok,
+    playId: message.playId,
+    ...message.status ? { status: message.status } : {},
+    reason: message.reason,
+    requestId: message.requestId
+  });
+}
+function handlePlayStopAck(message) {
+  const pending = pendingPlayStopAcks.get(message.requestId);
+  if (!pending)
+    return;
+  clearTimeout(pending.timer);
+  pendingPlayStopAcks.delete(message.requestId);
+  pending.resolve({
+    ok: message.ok,
+    playId: message.playId,
+    reason: message.reason,
+    requestId: message.requestId
+  });
+}
 function handleMixControlAck(message) {
   const pending = pendingMixAcks.get(message.requestId);
   if (!pending)
@@ -310,6 +372,10 @@ function sendParentMessage(message) {
     process.send?.(message);
     return;
   }
+  if (msgType === "play" || msgType === "play_status" || msgType === "play_stop") {
+    process.send?.(message);
+    return;
+  }
   const sessionId = message && typeof message === "object" && "sessionId" in message && typeof message.sessionId === "string" ? message.sessionId : void 0;
   if (!allowOutboundForSession(sessionId))
     return;
@@ -461,6 +527,18 @@ function defineAgent(handlers) {
       handleRecordingControlAck(message);
       return;
     }
+    if (isPlayAckMessage(message)) {
+      handlePlayAck(message);
+      return;
+    }
+    if (isPlayStatusAckMessage(message)) {
+      handlePlayStatusAck(message);
+      return;
+    }
+    if (isPlayStopAckMessage(message)) {
+      handlePlayStopAck(message);
+      return;
+    }
     if (isMixControlAckMessage(message)) {
       handleMixControlAck(message);
       return;

package/dist/templates/webhooks-redis/agent.js

@@ -10885,6 +10885,24 @@ var SessionSerialQueue = class {
 var SESSION_START_INIT_DELAY_ENABLED_ENV = "AGENT_SESSION_START_INIT_DELAY_ENABLED";
 var SESSION_START_INIT_DELAY_MS_ENV = "AGENT_SESSION_START_INIT_DELAY_MS";
 var DEFAULT_SESSION_START_INIT_DELAY_MS = 500;
+function isPlayAckMessage(value) {
+  if (!value || typeof value !== "object")
+    return false;
+  const msg = value;
+  return msg.type === "play_ack" && typeof msg.requestId === "string";
+}
+function isPlayStatusAckMessage(value) {
+  if (!value || typeof value !== "object")
+    return false;
+  const msg = value;
+  return msg.type === "play_status_ack" && typeof msg.requestId === "string";
+}
+function isPlayStopAckMessage(value) {
+  if (!value || typeof value !== "object")
+    return false;
+  const msg = value;
+  return msg.type === "play_stop_ack" && typeof msg.requestId === "string";
+}
 function isRecordingControlAckMessage(value) {
   if (!value || typeof value !== "object")
     return false;
@@ -10945,7 +10963,7 @@ function isParentMessage(value) {
   if (!value || typeof value !== "object")
     return false;
   const msg = value;
-  return msg.type === "session_start" || msg.type === "speech_event" || msg.type === "session_end" || msg.type === "data_channel_message" || msg.type === "data_channel_binary" || msg.type === "idle_timeout" || msg.type === "recording_control_ack" || msg.type === "mix_control_ack" || msg.type === "stt_control_ack" || msg.type === "webhook";
+  return msg.type === "session_start" || msg.type === "speech_event" || msg.type === "session_end" || msg.type === "data_channel_message" || msg.type === "data_channel_binary" || msg.type === "idle_timeout" || msg.type === "recording_control_ack" || msg.type === "play_ack" || msg.type === "play_status_ack" || msg.type === "play_stop_ack" || msg.type === "mix_control_ack" || msg.type === "stt_control_ack" || msg.type === "webhook";
 }
 function isSessionScopedParentMessage(value) {
   return isParentMessage(value) && !isWebhookMessage(value);
@@ -10962,7 +10980,11 @@ var recordingAvailableBySessionId = /* @__PURE__ */ new Map();
 var mixAvailableBySessionId = /* @__PURE__ */ new Map();
 var ttsPoseAvailableBySessionId = /* @__PURE__ */ new Map();
 var endedSessionIds = /* @__PURE__ */ new Set();
+var PLAY_BYTES_MAX_DECODED_LENGTH = 64 * 1024;
 var pendingRecordingAcks = /* @__PURE__ */ new Map();
+var pendingPlayAcks = /* @__PURE__ */ new Map();
+var pendingPlayStatusAcks = /* @__PURE__ */ new Map();
+var pendingPlayStopAcks = /* @__PURE__ */ new Map();
 var pendingMixAcks = /* @__PURE__ */ new Map();
 var pendingSttAcks = /* @__PURE__ */ new Map();
 function handleRecordingControlAck(message) {
@@ -10977,6 +10999,46 @@ function handleRecordingControlAck(message) {
     requestId: message.requestId
   });
 }
+function handlePlayAck(message) {
+  const pending = pendingPlayAcks.get(message.requestId);
+  if (!pending)
+    return;
+  clearTimeout(pending.timer);
+  pendingPlayAcks.delete(message.requestId);
+  pending.resolve({
+    ok: message.ok,
+    ...message.playId ? { playId: message.playId } : {},
+    reason: message.reason,
+    requestId: message.requestId
+  });
+}
+function handlePlayStatusAck(message) {
+  const pending = pendingPlayStatusAcks.get(message.requestId);
+  if (!pending)
+    return;
+  clearTimeout(pending.timer);
+  pendingPlayStatusAcks.delete(message.requestId);
+  pending.resolve({
+    ok: message.ok,
+    playId: message.playId,
+    ...message.status ? { status: message.status } : {},
+    reason: message.reason,
+    requestId: message.requestId
+  });
+}
+function handlePlayStopAck(message) {
+  const pending = pendingPlayStopAcks.get(message.requestId);
+  if (!pending)
+    return;
+  clearTimeout(pending.timer);
+  pendingPlayStopAcks.delete(message.requestId);
+  pending.resolve({
+    ok: message.ok,
+    playId: message.playId,
+    reason: message.reason,
+    requestId: message.requestId
+  });
+}
 function handleMixControlAck(message) {
   const pending = pendingMixAcks.get(message.requestId);
   if (!pending)
@@ -11062,6 +11124,10 @@ function sendParentMessage(message) {
     process.send?.(message);
     return;
   }
+  if (msgType === "play" || msgType === "play_status" || msgType === "play_stop") {
+    process.send?.(message);
+    return;
+  }
   const sessionId = message && typeof message === "object" && "sessionId" in message && typeof message.sessionId === "string" ? message.sessionId : void 0;
   if (!allowOutboundForSession(sessionId))
     return;
@@ -11213,6 +11279,18 @@ function defineAgent(handlers) {
       handleRecordingControlAck(message);
       return;
     }
+    if (isPlayAckMessage(message)) {
+      handlePlayAck(message);
+      return;
+    }
+    if (isPlayStatusAckMessage(message)) {
+      handlePlayStatusAck(message);
+      return;
+    }
+    if (isPlayStopAckMessage(message)) {
+      handlePlayStopAck(message);
+      return;
+    }
     if (isMixControlAckMessage(message)) {
       handleMixControlAck(message);
       return;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@voicethere/agent",
-  "version": "0.7.1",
+  "version": "0.7.3",
   "description": "VoiceThere customer agent SDK — IPC types and runtime helpers for sandboxed child bundles",
   "type": "module",
   "exports": {
@@ -95,9 +95,9 @@
     "@node-webrtc-rust/sdk": ">=0.6.5"
   },
   "devDependencies": {
-    "@node-webrtc-rust/helpers": "0.8.0",
-    "@node-webrtc-rust/sdk": "0.8.0",
-    "@node-webrtc-rust/signaling": "0.8.0",
+    "@node-webrtc-rust/helpers": "0.9.0",
+    "@node-webrtc-rust/sdk": "0.9.0",
+    "@node-webrtc-rust/signaling": "0.9.0",
     "@types/node": "22.20.1",
     "http-server": "14.1.1",
     "ioredis": "5.11.1",

package/templates/mix-smoke.ts

@@ -10,20 +10,28 @@
  * - `{ type: "mix", action: "set_global_mute", clientId, muted, sttEnabled? }`
  * - `{ type: "mix", action: "set_listener_mute", listenerId, targetId, muted }`
  * - `{ type: "mix", action: "get_status", clientId? }`
+ * - `{ type: "mix", action: "set_tts_pose", clientId, pose }`
+ * - `{ type: "mix", action: "speak", clientId, text }`
+ * - `{ type: "mix", action: "clear_tts_pose", clientId }`
+ * - `{ type: "mix", action: "play", sessionIds?, bytes?, url?, volume? }`
  *
  * Acks: `{ type: "mix_ack", action, ok: true, statuses?, ... }` or `{ ok: false, error }`.
- * Ignores `ping` / chat strings (voice-control readiness). No TTS during smoke.
+ * Ignores `ping` / chat strings (voice-control readiness).
  */
 import {
   MIX_REQUIRES_VOICE_PLUS_DATA,
+  clearTtsPose,
   createMixGroup,
   defineAgent,
   getClientMixStatus,
+  play,
   sendToClient,
   setClientPose,
   setGlobalMute,
   setListenerMute,
   setPositionalMixing,
+  setTtsPose,
+  speak,
 } from "@voicethere/agent";
 
 /** E2E fixture marker — rewritten before each fixture upload when needed. */
@@ -59,7 +67,18 @@ export type MixCommand =
       targetId: string;
       muted: boolean;
     }
-  | { type: "mix"; action: "get_status"; clientId?: string };
+  | { type: "mix"; action: "get_status"; clientId?: string }
+  | { type: "mix"; action: "set_tts_pose"; clientId: string; pose: MixPose }
+  | { type: "mix"; action: "speak"; clientId: string; text: string }
+  | { type: "mix"; action: "clear_tts_pose"; clientId: string }
+  | {
+      type: "mix";
+      action: "play";
+      sessionIds?: string[];
+      bytes?: string;
+      url?: string;
+      volume?: number;
+    };
 
 export type MixClientStatus = {
   clientId?: string;
@@ -77,6 +96,7 @@ export type MixAck =
       sessionId?: string;
       clientIds?: string[];
       statuses?: MixClientStatus[];
+      playId?: string;
     }
   | { type: "mix_ack"; action: string; ok: false; error: string };
 
@@ -161,6 +181,71 @@ export function isMixCommand(message: unknown): message is MixCommand {
         typeof status.clientId === "string" && status.clientId.trim().length > 0
       );
     }
+    case "set_tts_pose": {
+      const ttsPose = message as { clientId?: unknown; pose?: unknown };
+      return (
+        typeof ttsPose.clientId === "string" &&
+        ttsPose.clientId.trim().length > 0 &&
+        isMixPose(ttsPose.pose)
+      );
+    }
+    case "speak": {
+      const speakMsg = message as { clientId?: unknown; text?: unknown };
+      return (
+        typeof speakMsg.clientId === "string" &&
+        speakMsg.clientId.trim().length > 0 &&
+        typeof speakMsg.text === "string" &&
+        speakMsg.text.trim().length > 0
+      );
+    }
+    case "clear_tts_pose": {
+      const clearMsg = message as { clientId?: unknown };
+      return (
+        typeof clearMsg.clientId === "string" &&
+        clearMsg.clientId.trim().length > 0
+      );
+    }
+    case "play": {
+      const playMsg = message as {
+        sessionIds?: unknown;
+        bytes?: unknown;
+        url?: unknown;
+        volume?: unknown;
+      };
+      const hasBytes =
+        typeof playMsg.bytes === "string" && playMsg.bytes.trim().length > 0;
+      const hasUrl =
+        typeof playMsg.url === "string" && playMsg.url.trim().length > 0;
+      if (!hasBytes && !hasUrl) {
+        return false;
+      }
+      if (
+        playMsg.bytes !== undefined &&
+        (typeof playMsg.bytes !== "string" || playMsg.bytes.trim().length === 0)
+      ) {
+        return false;
+      }
+      if (
+        playMsg.url !== undefined &&
+        (typeof playMsg.url !== "string" || playMsg.url.trim().length === 0)
+      ) {
+        return false;
+      }
+      if (playMsg.sessionIds !== undefined) {
+        if (
+          !Array.isArray(playMsg.sessionIds) ||
+          !playMsg.sessionIds.every(
+            (id) => typeof id === "string" && id.trim().length > 0,
+          )
+        ) {
+          return false;
+        }
+      }
+      if (playMsg.volume !== undefined && typeof playMsg.volume !== "number") {
+        return false;
+      }
+      return true;
+    }
     default:
       return false;
   }
@@ -202,6 +287,7 @@ function ackOk(
     sessionId?: string;
     clientIds?: string[];
     statuses?: MixClientStatus[];
+    playId?: string;
   },
 ): void {
   sendToClient(sessionId, {
@@ -327,6 +413,50 @@ async function handleMixCommand(
       });
       return;
     }
+    case "set_tts_pose": {
+      if (!requireMix(sessionId, action)) return;
+      const result = await setTtsPose(command.clientId, command.pose);
+      if (!result.ok) {
+        ackError(sessionId, action, result.reason ?? "set_tts_pose failed");
+        return;
+      }
+      ackOk(sessionId, action);
+      return;
+    }
+    case "speak": {
+      if (!requireMix(sessionId, action)) return;
+      speak(command.clientId, command.text);
+      ackOk(sessionId, action);
+      return;
+    }
+    case "clear_tts_pose": {
+      if (!requireMix(sessionId, action)) return;
+      const result = await clearTtsPose(command.clientId);
+      if (!result.ok) {
+        ackError(sessionId, action, result.reason ?? "clear_tts_pose failed");
+        return;
+      }
+      ackOk(sessionId, action);
+      return;
+    }
+    case "play": {
+      const inlineUrl =
+        command.url?.trim() || "https://example.com/inline-clip.wav";
+      const result = await play({
+        url: inlineUrl,
+        ...(command.sessionIds?.length
+          ? { sessionIds: command.sessionIds }
+          : {}),
+        ...(command.bytes ? { bytes: command.bytes } : {}),
+        ...(command.volume !== undefined ? { volume: command.volume } : {}),
+      });
+      if (!result.ok || !result.playId) {
+        ackError(sessionId, action, result.reason ?? "play failed");
+        return;
+      }
+      ackOk(sessionId, action, { playId: result.playId });
+      return;
+    }
     default:
       ackError(sessionId, "unknown", "unsupported mix action");
   }