@estuary-ai/sdk 0.1.3 → 0.1.5

package/dist/index.mjs

@@ -3738,7 +3738,7 @@ var SocketManager = class extends TypedEventEmitter {
 // src/voice/voice-manager.ts
 async function createVoiceManager(transport, socketManager, sampleRate, logger) {
   if (transport === "websocket") {
-    const { WebSocketVoiceManager } = await import('./websocket-voice-A4CK3UTM.mjs');
+    const { WebSocketVoiceManager } = await import('./websocket-voice-HYHCIYEW.mjs');
     return new WebSocketVoiceManager(socketManager, sampleRate, logger);
   }
   if (transport === "livekit") {
@@ -3747,7 +3747,7 @@ async function createVoiceManager(transport, socketManager, sampleRate, logger)
       return new LiveKitVoiceManager(socketManager, logger);
     } catch {
       logger.warn("livekit-client not installed, falling back to WebSocket voice");
-      const { WebSocketVoiceManager } = await import('./websocket-voice-A4CK3UTM.mjs');
+      const { WebSocketVoiceManager } = await import('./websocket-voice-HYHCIYEW.mjs');
       return new WebSocketVoiceManager(socketManager, sampleRate, logger);
     }
   }
@@ -3756,7 +3756,7 @@ async function createVoiceManager(transport, socketManager, sampleRate, logger)
       const { LiveKitVoiceManager } = await import('./livekit-voice-H6WZ44LM.mjs');
       return new LiveKitVoiceManager(socketManager, logger);
     } catch {
-      const { WebSocketVoiceManager } = await import('./websocket-voice-A4CK3UTM.mjs');
+      const { WebSocketVoiceManager } = await import('./websocket-voice-HYHCIYEW.mjs');
       return new WebSocketVoiceManager(socketManager, sampleRate, logger);
     }
   }
@@ -3989,6 +3989,51 @@ var Logger = class {
   }
 };
 
+// src/utils/action-parser.ts
+var ACTION_TAG_RE = /<action\s+([^>]*?)\/>/gi;
+var ATTR_RE = /(\w+)\s*=\s*"([^"]*)"|(\w+)\s*=\s*'([^']*)'/g;
+function parseAttributes(attrString) {
+  const attrs = {};
+  let match;
+  while ((match = ATTR_RE.exec(attrString)) !== null) {
+    const key = match[1] ?? match[3];
+    const value2 = match[2] ?? match[4];
+    attrs[key] = value2;
+  }
+  return attrs;
+}
+var StreamingActionParser = class {
+  emittedCount = 0;
+  /**
+   * Parse the accumulated response text and return any new actions found
+   * since the last call. Also returns the text with all action tags stripped.
+   */
+  parse(accumulatedText) {
+    const allActions = [];
+    let match;
+    ACTION_TAG_RE.lastIndex = 0;
+    while ((match = ACTION_TAG_RE.exec(accumulatedText)) !== null) {
+      const attrs = parseAttributes(match[1]);
+      const name = attrs.name;
+      if (name) {
+        delete attrs.name;
+        allActions.push({ name, params: attrs });
+      }
+    }
+    const newActions = allActions.slice(this.emittedCount);
+    this.emittedCount = allActions.length;
+    const cleanText = accumulatedText.replace(ACTION_TAG_RE, "").replace(/\s{2,}/g, " ").trim();
+    return { actions: newActions, cleanText };
+  }
+  reset() {
+    this.emittedCount = 0;
+  }
+};
+function parseActions(text) {
+  const parser = new StreamingActionParser();
+  return parser.parse(text);
+}
+
 // src/client.ts
 var DEFAULT_SAMPLE_RATE = 16e3;
 var EstuaryClient = class extends TypedEventEmitter {
@@ -3999,6 +4044,7 @@ var EstuaryClient = class extends TypedEventEmitter {
   audioPlayer = null;
   _memory;
   _sessionInfo = null;
+  actionParsers = /* @__PURE__ */ new Map();
   constructor(config) {
     super();
     this.config = config;
@@ -4050,6 +4096,9 @@ var EstuaryClient = class extends TypedEventEmitter {
     this.ensureConnected();
     this.socketManager.emitEvent("client_interrupt", { message_id: messageId });
     this.audioPlayer?.clear();
+    if (this.config.suppressMicDuringPlayback) {
+      this.voiceManager?.setSuppressed?.(false);
+    }
   }
   /** Send a camera image for vision processing */
   sendCameraImage(imageBase64, mimeType, requestId, text) {
@@ -4088,9 +4137,15 @@ var EstuaryClient = class extends TypedEventEmitter {
       this.audioPlayer = new AudioPlayer(sampleRate, (event) => {
         if (event.type === "started") {
           this.emit("audioPlaybackStarted", event.messageId);
+          if (this.config.suppressMicDuringPlayback) {
+            this.voiceManager?.setSuppressed?.(true);
+          }
         } else if (event.type === "complete") {
           this.emit("audioPlaybackComplete", event.messageId);
           this.notifyAudioPlaybackComplete(event.messageId);
+          if (this.config.suppressMicDuringPlayback) {
+            this.voiceManager?.setSuppressed?.(false);
+          }
         }
       });
     }
@@ -4134,15 +4189,20 @@ var EstuaryClient = class extends TypedEventEmitter {
     });
     this.socketManager.on("disconnected", (reason) => {
       this._sessionInfo = null;
+      this.actionParsers.clear();
       this.emit("disconnected", reason);
     });
     this.socketManager.on("reconnecting", (attempt) => this.emit("reconnecting", attempt));
     this.socketManager.on("connectionStateChanged", (state) => this.emit("connectionStateChanged", state));
-    this.socketManager.on("botResponse", (response) => this.emit("botResponse", response));
+    this.socketManager.on("botResponse", (response) => this.handleBotResponse(response));
     this.socketManager.on("botVoice", (voice) => this.handleBotVoice(voice));
     this.socketManager.on("sttResponse", (response) => this.emit("sttResponse", response));
     this.socketManager.on("interrupt", (data) => {
       this.audioPlayer?.clear();
+      this.actionParsers.clear();
+      if (this.config.suppressMicDuringPlayback) {
+        this.voiceManager?.setSuppressed?.(false);
+      }
       this.emit("interrupt", data);
     });
     this.socketManager.on("error", (error) => this.emit("error", error));
@@ -4153,12 +4213,34 @@ var EstuaryClient = class extends TypedEventEmitter {
     this.socketManager.on("livekitDisconnected", () => this.emit("livekitDisconnected"));
     this.socketManager.on("memoryUpdated", (event) => this.emit("memoryUpdated", event));
   }
+  handleBotResponse(response) {
+    const { messageId } = response;
+    if (!this.actionParsers.has(messageId)) {
+      this.actionParsers.set(messageId, new StreamingActionParser());
+    }
+    const parser = this.actionParsers.get(messageId);
+    const { actions, cleanText } = parser.parse(response.text);
+    for (const action of actions) {
+      this.emit("characterAction", {
+        name: action.name,
+        params: action.params,
+        messageId
+      });
+    }
+    this.emit("botResponse", {
+      ...response,
+      text: cleanText
+    });
+    if (response.isFinal) {
+      this.actionParsers.delete(messageId);
+    }
+  }
   handleBotVoice(voice) {
     this.emit("botVoice", voice);
     this.audioPlayer?.enqueue(voice);
   }
 };
 
-export { ConnectionState, EstuaryClient };
+export { ConnectionState, EstuaryClient, parseActions };
 //# sourceMappingURL=index.mjs.map
 //# sourceMappingURL=index.mjs.map