@estuary-ai/sdk 0.1.3 → 0.1.4

package/dist/index.d.mts

@@ -96,6 +96,14 @@ interface MemoryUpdatedEvent {
     newMemories: MemoryData[];
     timestamp: string;
 }
+interface CharacterAction {
+    /** Action name (e.g., "follow_user", "sit", "look_at") */
+    name: string;
+    /** Action parameters as key-value pairs */
+    params: Record<string, string>;
+    /** Message ID of the bot response that contained this action */
+    messageId: string;
+}
 type EstuaryEventMap = {
     connected: (session: SessionInfo) => void;
     disconnected: (reason: string) => void;
@@ -109,6 +117,7 @@ type EstuaryEventMap = {
     authError: (error: string) => void;
     quotaExceeded: (data: QuotaExceededData) => void;
     cameraCaptureRequest: (request: CameraCaptureRequest) => void;
+    characterAction: (action: CharacterAction) => void;
     voiceStarted: () => void;
     voiceStopped: () => void;
     livekitConnected: (room: string) => void;
@@ -229,6 +238,7 @@ declare class EstuaryClient extends TypedEventEmitter<EstuaryEventMap> {
     private audioPlayer;
     private _memory;
     private _sessionInfo;
+    private actionParsers;
     constructor(config: EstuaryConfig);
     /** Memory API client for querying memories, graphs, and facts */
     get memory(): MemoryClient;
@@ -266,6 +276,7 @@ declare class EstuaryClient extends TypedEventEmitter<EstuaryEventMap> {
     get isVoiceActive(): boolean;
     private ensureConnected;
     private forwardSocketEvents;
+    private handleBotResponse;
     private handleBotVoice;
 }
 
@@ -289,4 +300,23 @@ declare class EstuaryError extends Error {
     constructor(code: ErrorCode, message: string, details?: unknown);
 }
 
-export { type BotResponse, type BotVoice, type CameraCaptureRequest, ConnectionState, type CoreFactsResponse, ErrorCode, EstuaryClient, type EstuaryConfig, EstuaryError, type EstuaryEventMap, type InterruptData, type LiveKitTokenResponse, MemoryClient, type MemoryData, type MemoryGraphOptions, type MemoryGraphResponse, type MemoryListOptions, type MemoryListResponse, type MemorySearchOptions, type MemorySearchResponse, type MemoryStatsResponse, type MemoryTimelineOptions, type MemoryTimelineResponse, type MemoryUpdatedEvent, type QuotaExceededData, type SessionInfo, type SttResponse, type VoiceManager, type VoiceTransport };
+/**
+ * Parses `<action name="..." .../>` XML tags from bot response text.
+ *
+ * Designed for streaming: call `parse()` with the accumulated text on each
+ * chunk and it returns only newly-discovered actions since the last call.
+ */
+interface ParsedAction {
+    name: string;
+    params: Record<string, string>;
+}
+/**
+ * One-shot parse: extract all actions and return clean text.
+ * Useful for non-streaming contexts.
+ */
+declare function parseActions(text: string): {
+    actions: ParsedAction[];
+    cleanText: string;
+};
+
+export { type BotResponse, type BotVoice, type CameraCaptureRequest, type CharacterAction, ConnectionState, type CoreFactsResponse, ErrorCode, EstuaryClient, type EstuaryConfig, EstuaryError, type EstuaryEventMap, type InterruptData, type LiveKitTokenResponse, MemoryClient, type MemoryData, type MemoryGraphOptions, type MemoryGraphResponse, type MemoryListOptions, type MemoryListResponse, type MemorySearchOptions, type MemorySearchResponse, type MemoryStatsResponse, type MemoryTimelineOptions, type MemoryTimelineResponse, type MemoryUpdatedEvent, type ParsedAction, type QuotaExceededData, type SessionInfo, type SttResponse, type VoiceManager, type VoiceTransport, parseActions };

package/dist/index.js

@@ -9310,6 +9310,53 @@ var Logger = class {
 
 // src/client.ts
 init_errors();
+
+// 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 {
   config;
@@ -9319,6 +9366,7 @@ var EstuaryClient = class extends TypedEventEmitter {
   audioPlayer = null;
   _memory;
   _sessionInfo = null;
+  actionParsers = /* @__PURE__ */ new Map();
   constructor(config) {
     super();
     this.config = config;
@@ -9454,15 +9502,17 @@ 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();
       this.emit("interrupt", data);
     });
     this.socketManager.on("error", (error) => this.emit("error", error));
@@ -9473,6 +9523,28 @@ 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);
@@ -9500,5 +9572,6 @@ xmlhttprequest-ssl/lib/XMLHttpRequest.js:
 
 exports.ConnectionState = ConnectionState;
 exports.EstuaryClient = EstuaryClient;
+exports.parseActions = parseActions;
 //# sourceMappingURL=index.js.map
 //# sourceMappingURL=index.js.map