phonic 0.15.0 → 0.16.0

package/README.md

@@ -96,7 +96,7 @@ if (error === null) {
 To start a conversation, open a WebSocket connection:
 
 ```ts
-const { data, error } = await phonic.sts.websocket({
+const phonicWebSocket = phonic.sts.websocket({
   input_format: "mulaw_8000",
 
   // Optional fields
@@ -106,12 +106,6 @@ const { data, error } = await phonic.sts.websocket({
   voice_id: "meredith",
   output_format: "mulaw_8000"
 });
-
-if (error !== null) {
-  throw new Error(`Failed to start conversation: ${error.message}`);
-}
-
-const { phonicWebSocket } = data;
 ```
 
 Stream input (user) audio chunks:

package/dist/index.d.mts

@@ -96,6 +96,9 @@ type PhonicSTSWebSocketResponseMessage = {
 } | {
     type: "is_user_speaking";
     isUserSpeaking: boolean;
+} | {
+    type: "interrupted_response";
+    interruptedResponse: string;
 } | {
     type: "error";
     error: {
@@ -116,10 +119,13 @@ type OnErrorCallback = (event: WebSocket.ErrorEvent) => void;
 
 declare class PhonicSTSWebSocket {
     private readonly ws;
+    private readonly config;
     private onMessageCallback;
     private onCloseCallback;
     private onErrorCallback;
-    constructor(ws: WebSocket);
+    private buffer;
+    private isOpen;
+    constructor(ws: WebSocket, config: PhonicSTSConfig);
     onMessage(callback: OnMessageCallback): void;
     onClose(callback: OnCloseCallback): void;
     onError(callback: OnErrorCallback): void;
@@ -138,10 +144,7 @@ declare class PhonicSTSWebSocket {
 declare class SpeechToSpeech {
     private readonly phonic;
     constructor(phonic: Phonic);
-    private connectToPhonicAPI;
-    websocket(config: PhonicSTSConfig): DataOrError<{
-        phonicWebSocket: PhonicSTSWebSocket;
-    }>;
+    websocket(config: PhonicSTSConfig): PhonicSTSWebSocket;
 }
 
 type Voice = {

package/dist/index.d.ts

@@ -96,6 +96,9 @@ type PhonicSTSWebSocketResponseMessage = {
 } | {
     type: "is_user_speaking";
     isUserSpeaking: boolean;
+} | {
+    type: "interrupted_response";
+    interruptedResponse: string;
 } | {
     type: "error";
     error: {
@@ -116,10 +119,13 @@ type OnErrorCallback = (event: WebSocket.ErrorEvent) => void;
 
 declare class PhonicSTSWebSocket {
     private readonly ws;
+    private readonly config;
     private onMessageCallback;
     private onCloseCallback;
     private onErrorCallback;
-    constructor(ws: WebSocket);
+    private buffer;
+    private isOpen;
+    constructor(ws: WebSocket, config: PhonicSTSConfig);
     onMessage(callback: OnMessageCallback): void;
     onClose(callback: OnCloseCallback): void;
     onError(callback: OnErrorCallback): void;
@@ -138,10 +144,7 @@ declare class PhonicSTSWebSocket {
 declare class SpeechToSpeech {
     private readonly phonic;
     constructor(phonic: Phonic);
-    private connectToPhonicAPI;
-    websocket(config: PhonicSTSConfig): DataOrError<{
-        phonicWebSocket: PhonicSTSWebSocket;
-    }>;
+    websocket(config: PhonicSTSConfig): PhonicSTSWebSocket;
 }
 
 type Voice = {

package/dist/index.js

@@ -35,7 +35,7 @@ __export(index_exports, {
 module.exports = __toCommonJS(index_exports);
 
 // package.json
-var version = "0.15.0";
+var version = "0.16.0";
 
 // src/conversations/index.ts
 var Conversations = class {
@@ -87,8 +87,21 @@ var import_ws = __toESM(require("ws"));
 
 // src/sts/websocket.ts
 var PhonicSTSWebSocket = class {
-  constructor(ws) {
+  constructor(ws, config) {
     this.ws = ws;
+    this.config = config;
+    this.buffer.push(
+      JSON.stringify({
+        type: "config",
+        ...this.config
+      })
+    );
+    this.ws.onopen = () => {
+      for (const message of this.buffer) {
+        this.ws.send(message);
+      }
+      this.isOpen = true;
+    };
     this.ws.onmessage = (event) => {
       if (this.onMessageCallback === null) {
         return;
@@ -122,6 +135,8 @@ var PhonicSTSWebSocket = class {
   onMessageCallback = null;
   onCloseCallback = null;
   onErrorCallback = null;
+  buffer = [];
+  isOpen = false;
   onMessage(callback) {
     this.onMessageCallback = callback;
   }
@@ -132,28 +147,37 @@ var PhonicSTSWebSocket = class {
     this.onErrorCallback = callback;
   }
   audioChunk({ audio }) {
-    this.ws.send(
-      JSON.stringify({
-        type: "audio_chunk",
-        audio
-      })
-    );
+    const audiochunkMessage = JSON.stringify({
+      type: "audio_chunk",
+      audio
+    });
+    if (this.isOpen) {
+      this.ws.send(audiochunkMessage);
+    } else {
+      this.buffer.push(audiochunkMessage);
+    }
   }
   updateSystemPrompt({ systemPrompt }) {
-    this.ws.send(
-      JSON.stringify({
-        type: "update_system_prompt",
-        system_prompt: systemPrompt
-      })
-    );
+    const updateSystemPromptMessage = JSON.stringify({
+      type: "update_system_prompt",
+      system_prompt: systemPrompt
+    });
+    if (this.isOpen) {
+      this.ws.send(updateSystemPromptMessage);
+    } else {
+      this.buffer.push(updateSystemPromptMessage);
+    }
   }
   setExternalId({ externalId }) {
-    this.ws.send(
-      JSON.stringify({
-        type: "set_external_id",
-        external_id: externalId
-      })
-    );
+    const setExternalIdMessage = JSON.stringify({
+      type: "set_external_id",
+      external_id: externalId
+    });
+    if (this.isOpen) {
+      this.ws.send(setExternalIdMessage);
+    } else {
+      this.buffer.push(setExternalIdMessage);
+    }
   }
   close(code) {
     this.ws.close(code ?? 1e3);
@@ -161,64 +185,11 @@ var PhonicSTSWebSocket = class {
 };
 
 // src/sts/index.ts
-var phonicApiCloseCodes = {
-  insuffucientCapacityAvailable: 4004
-};
 var SpeechToSpeech = class {
   constructor(phonic) {
     this.phonic = phonic;
   }
-  async connectToPhonicAPI(phonicApiWsUrl, config) {
-    return new Promise((resolve) => {
-      const ws = new import_ws.default(phonicApiWsUrl, {
-        headers: this.phonic.headers
-      });
-      ws.onopen = () => {
-        ws.send(
-          JSON.stringify({
-            type: "config",
-            ...config
-          })
-        );
-      };
-      ws.onmessage = (event) => {
-        if (typeof event.data !== "string") {
-          throw new Error("Received non-string message");
-        }
-        const dataObj = JSON.parse(
-          event.data
-        );
-        if (dataObj.type === "ready_to_start_conversation") {
-          resolve({
-            data: {
-              phonicWebSocket: new PhonicSTSWebSocket(ws)
-            },
-            error: null
-          });
-        }
-      };
-      ws.onerror = (error) => {
-        resolve({
-          data: null,
-          error: {
-            message: error.message
-          }
-        });
-      };
-      ws.onclose = (event) => {
-        if (event.code === phonicApiCloseCodes.insuffucientCapacityAvailable) {
-          resolve({
-            data: null,
-            error: {
-              message: event.reason,
-              code: "insuffucient_capacity_available"
-            }
-          });
-        }
-      };
-    });
-  }
-  async websocket(config) {
+  websocket(config) {
     const wsBaseUrl = this.phonic.baseUrl.replace(/^http/, "ws");
     const queryString = new URLSearchParams({
       ...this.phonic.__downstreamWebSocketUrl !== null && {
@@ -226,31 +197,10 @@ var SpeechToSpeech = class {
       }
     }).toString();
     const phonicApiWsUrl = `${wsBaseUrl}/v1/sts/ws?${queryString}`;
-    let retryNumber = 0;
-    const maxRetries = 14;
-    const retryDelay = 15e3;
-    while (true) {
-      const connectResult = await this.connectToPhonicAPI(
-        phonicApiWsUrl,
-        config
-      );
-      if (connectResult.data !== null) {
-        return connectResult;
-      }
-      if (connectResult.error.code === "insuffucient_capacity_available") {
-        if (retryNumber >= maxRetries) {
-          return connectResult;
-        }
-        console.info(
-          `${connectResult.error.message}, will retry in ${retryDelay / 1e3}sec`
-        );
-        await new Promise((resolve) => setTimeout(resolve, retryDelay));
-        retryNumber += 1;
-        console.info(`Retrying... ${retryNumber}/${maxRetries}`);
-        continue;
-      }
-      return connectResult;
-    }
+    const ws = new import_ws.default(phonicApiWsUrl, {
+      headers: this.phonic.headers
+    });
+    return new PhonicSTSWebSocket(ws, config);
   }
 };
 

package/dist/index.mjs

@@ -1,5 +1,5 @@
 // package.json
-var version = "0.15.0";
+var version = "0.16.0";
 
 // src/conversations/index.ts
 var Conversations = class {
@@ -51,8 +51,21 @@ import WebSocket from "ws";
 
 // src/sts/websocket.ts
 var PhonicSTSWebSocket = class {
-  constructor(ws) {
+  constructor(ws, config) {
     this.ws = ws;
+    this.config = config;
+    this.buffer.push(
+      JSON.stringify({
+        type: "config",
+        ...this.config
+      })
+    );
+    this.ws.onopen = () => {
+      for (const message of this.buffer) {
+        this.ws.send(message);
+      }
+      this.isOpen = true;
+    };
     this.ws.onmessage = (event) => {
       if (this.onMessageCallback === null) {
         return;
@@ -86,6 +99,8 @@ var PhonicSTSWebSocket = class {
   onMessageCallback = null;
   onCloseCallback = null;
   onErrorCallback = null;
+  buffer = [];
+  isOpen = false;
   onMessage(callback) {
     this.onMessageCallback = callback;
   }
@@ -96,28 +111,37 @@ var PhonicSTSWebSocket = class {
     this.onErrorCallback = callback;
   }
   audioChunk({ audio }) {
-    this.ws.send(
-      JSON.stringify({
-        type: "audio_chunk",
-        audio
-      })
-    );
+    const audiochunkMessage = JSON.stringify({
+      type: "audio_chunk",
+      audio
+    });
+    if (this.isOpen) {
+      this.ws.send(audiochunkMessage);
+    } else {
+      this.buffer.push(audiochunkMessage);
+    }
   }
   updateSystemPrompt({ systemPrompt }) {
-    this.ws.send(
-      JSON.stringify({
-        type: "update_system_prompt",
-        system_prompt: systemPrompt
-      })
-    );
+    const updateSystemPromptMessage = JSON.stringify({
+      type: "update_system_prompt",
+      system_prompt: systemPrompt
+    });
+    if (this.isOpen) {
+      this.ws.send(updateSystemPromptMessage);
+    } else {
+      this.buffer.push(updateSystemPromptMessage);
+    }
   }
   setExternalId({ externalId }) {
-    this.ws.send(
-      JSON.stringify({
-        type: "set_external_id",
-        external_id: externalId
-      })
-    );
+    const setExternalIdMessage = JSON.stringify({
+      type: "set_external_id",
+      external_id: externalId
+    });
+    if (this.isOpen) {
+      this.ws.send(setExternalIdMessage);
+    } else {
+      this.buffer.push(setExternalIdMessage);
+    }
   }
   close(code) {
     this.ws.close(code ?? 1e3);
@@ -125,64 +149,11 @@ var PhonicSTSWebSocket = class {
 };
 
 // src/sts/index.ts
-var phonicApiCloseCodes = {
-  insuffucientCapacityAvailable: 4004
-};
 var SpeechToSpeech = class {
   constructor(phonic) {
     this.phonic = phonic;
   }
-  async connectToPhonicAPI(phonicApiWsUrl, config) {
-    return new Promise((resolve) => {
-      const ws = new WebSocket(phonicApiWsUrl, {
-        headers: this.phonic.headers
-      });
-      ws.onopen = () => {
-        ws.send(
-          JSON.stringify({
-            type: "config",
-            ...config
-          })
-        );
-      };
-      ws.onmessage = (event) => {
-        if (typeof event.data !== "string") {
-          throw new Error("Received non-string message");
-        }
-        const dataObj = JSON.parse(
-          event.data
-        );
-        if (dataObj.type === "ready_to_start_conversation") {
-          resolve({
-            data: {
-              phonicWebSocket: new PhonicSTSWebSocket(ws)
-            },
-            error: null
-          });
-        }
-      };
-      ws.onerror = (error) => {
-        resolve({
-          data: null,
-          error: {
-            message: error.message
-          }
-        });
-      };
-      ws.onclose = (event) => {
-        if (event.code === phonicApiCloseCodes.insuffucientCapacityAvailable) {
-          resolve({
-            data: null,
-            error: {
-              message: event.reason,
-              code: "insuffucient_capacity_available"
-            }
-          });
-        }
-      };
-    });
-  }
-  async websocket(config) {
+  websocket(config) {
     const wsBaseUrl = this.phonic.baseUrl.replace(/^http/, "ws");
     const queryString = new URLSearchParams({
       ...this.phonic.__downstreamWebSocketUrl !== null && {
@@ -190,31 +161,10 @@ var SpeechToSpeech = class {
       }
     }).toString();
     const phonicApiWsUrl = `${wsBaseUrl}/v1/sts/ws?${queryString}`;
-    let retryNumber = 0;
-    const maxRetries = 14;
-    const retryDelay = 15e3;
-    while (true) {
-      const connectResult = await this.connectToPhonicAPI(
-        phonicApiWsUrl,
-        config
-      );
-      if (connectResult.data !== null) {
-        return connectResult;
-      }
-      if (connectResult.error.code === "insuffucient_capacity_available") {
-        if (retryNumber >= maxRetries) {
-          return connectResult;
-        }
-        console.info(
-          `${connectResult.error.message}, will retry in ${retryDelay / 1e3}sec`
-        );
-        await new Promise((resolve) => setTimeout(resolve, retryDelay));
-        retryNumber += 1;
-        console.info(`Retrying... ${retryNumber}/${maxRetries}`);
-        continue;
-      }
-      return connectResult;
-    }
+    const ws = new WebSocket(phonicApiWsUrl, {
+      headers: this.phonic.headers
+    });
+    return new PhonicSTSWebSocket(ws, config);
   }
 };
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "phonic",
-  "version": "0.15.0",
+  "version": "0.16.0",
   "description": "Phonic Node.js SDK",
   "scripts": {
     "build": "tsup",