Custom realtime model selection

Author: AnasMations

lib/api.js

@@ -63,6 +63,7 @@ export class RealtimeAPI extends RealtimeEventHandler {
     if (this.isConnected()) {
       throw new Error(`Already connected`);
     }
+    console.log("Connecting to model: ", model);
     if (globalThis.WebSocket) {
       /**
        * Web browser

lib/client.js

@@ -189,15 +189,14 @@ import { RealtimeUtils } from './utils.js';
 export class RealtimeClient extends RealtimeEventHandler {
   /**
    * Create a new RealtimeClient instance
-   * @param {{url?: string, apiKey?: string, dangerouslyAllowAPIKeyInBrowser?: boolean, debug?: boolean}} [settings]
+   * @param {{url?: string, apiKey?: string, model?: string, dangerouslyAllowAPIKeyInBrowser?: boolean, debug?: boolean}} [settings]
    */
-  constructor({ url, apiKey, dangerouslyAllowAPIKeyInBrowser, debug } = {}) {
+  constructor({ url, apiKey, model='gpt-4o-realtime-preview-2024-10-01', dangerouslyAllowAPIKeyInBrowser, debug } = {}) {
     super();
     this.defaultSessionConfig = {
       modalities: ['text', 'audio'],
       instructions: '',
       voice: 'verse',
-      model: 'gpt-4o-realtime-preview-2024-10-01',
       input_audio_format: 'pcm16',
       output_audio_format: 'pcm16',
       input_audio_transcription: null,
@@ -207,6 +206,7 @@ export class RealtimeClient extends RealtimeEventHandler {
       temperature: 0.8,
       max_response_output_tokens: 4096,
     };
+    this.realtimeModel = model;
     this.sessionConfig = {};
     this.transcriptionModels = [
       {
@@ -394,8 +394,8 @@ export class RealtimeClient extends RealtimeEventHandler {
     if (this.isConnected()) {
       throw new Error(`Already connected, use .disconnect() first`);
     }
+    await this.realtime.connect(this.realtimeModel);
     this.updateSession();
-    await this.realtime.connect(this.sessionConfig?.model);
     return true;
   }
 
@@ -486,7 +486,6 @@ export class RealtimeClient extends RealtimeEventHandler {
     modalities = void 0,
     instructions = void 0,
     voice = void 0,
-    model = void 0,
     input_audio_format = void 0,
     output_audio_format = void 0,
     input_audio_transcription = void 0,
@@ -499,7 +498,6 @@ export class RealtimeClient extends RealtimeEventHandler {
     modalities !== void 0 && (this.sessionConfig.modalities = modalities);
     instructions !== void 0 && (this.sessionConfig.instructions = instructions);
     voice !== void 0 && (this.sessionConfig.voice = voice);
-    model !== void 0 && (this.sessionConfig.model = model);
     input_audio_format !== void 0 &&
       (this.sessionConfig.input_audio_format = input_audio_format);
     output_audio_format !== void 0 &&