@keyframelabs/elements 0.5.0 → 0.5.1

package/dist/agents/index.d.ts

@@ -1,5 +1,5 @@
 import { ElevenLabsAgent, ElevenLabsConfig } from './elevenlabs';
-import { OpenAIRealtimeAgent, OpenAIRealtimeConfig, TurnDetection } from './openai-realtime';
+import { OpenAIRealtimeAgent, OpenAIRealtimeConfig } from './openai-realtime';
 import { KflAgent, KflAgentConfig } from './kfl';
 /**
  * Agent implementations for voice AI platforms.
@@ -10,7 +10,7 @@ import { KflAgent, KflAgentConfig } from './kfl';
 export { BaseAgent, DEFAULT_INPUT_SAMPLE_RATE } from './base';
 export type { Agent, AgentConfig, AgentEventMap, AgentState, Emotion } from './types';
 export { ElevenLabsAgent, type ElevenLabsConfig };
-export { OpenAIRealtimeAgent, type OpenAIRealtimeConfig, type TurnDetection };
+export { OpenAIRealtimeAgent, type OpenAIRealtimeConfig };
 export { KflAgent, type KflAgentConfig };
 export { SAMPLE_RATE, base64ToBytes, bytesToBase64, resamplePcm, createEventEmitter, floatTo16BitPCM, createPcmWorkletNode } from './audio-utils';
 /** Supported agent types */

package/dist/agents/openai-realtime.d.ts

@@ -1,28 +1,9 @@
 import { AgentConfig } from './types';
 import { BaseAgent } from './base';
-/**
- * Turn detection configuration for OpenAI Realtime.
- * @see https://developers.openai.com/api/docs/guides/realtime-vad
- */
-export type TurnDetection = {
-    type: 'server_vad';
-    /** Activation threshold 0-1. Higher = requires louder audio. */
-    threshold?: number;
-    /** Audio (ms) to include before detected speech. */
-    prefix_padding_ms?: number;
-    /** Silence duration (ms) before speech stop is detected. */
-    silence_duration_ms?: number;
-} | {
-    type: 'semantic_vad';
-    /** How eager the model is to consider a turn finished. Default: 'auto'. */
-    eagerness?: 'low' | 'medium' | 'high' | 'auto';
-};
 /** OpenAI Realtime specific configuration */
 export interface OpenAIRealtimeConfig extends AgentConfig {
     /** Model to use (defaults to gpt-realtime) */
     model?: string;
-    /** Turn detection / VAD settings. Defaults to semantic_vad with eagerness 'high'. */
-    turnDetection?: TurnDetection;
 }
 /**
  * OpenAI Realtime agent implementation.

package/dist/index.d.ts

@@ -4,7 +4,7 @@ export { PersonaView } from './PersonaView';
 export type { PersonaViewOptions } from './PersonaView';
 export type { EmbedStatus, VideoFit, VoiceAgentDetails, SessionDetails, BaseCallbacks, } from './types';
 export { createAgent, ElevenLabsAgent, OpenAIRealtimeAgent, KflAgent, BaseAgent, AGENT_REGISTRY, getAgentInfo, } from './agents';
-export type { AgentType, AgentConfig, AgentEventMap, Agent, AnyAgent, AgentTypeInfo, ElevenLabsConfig, OpenAIRealtimeConfig, KflAgentConfig, TurnDetection, } from './agents';
+export type { AgentType, AgentConfig, AgentEventMap, Agent, AnyAgent, AgentTypeInfo, ElevenLabsConfig, OpenAIRealtimeConfig, KflAgentConfig, } from './agents';
 export type { AgentState } from '@keyframelabs/sdk';
 export { floatTo16BitPCM, resamplePcm, base64ToBytes, bytesToBase64, SAMPLE_RATE, createEventEmitter, } from './agents';
 export { KflEmbedElement } from './KflEmbedElement';

package/dist/index.js

@@ -344,7 +344,7 @@ class F extends v {
     this.initialized = !1, this.lastInterruptId = 0, this.resetTurnState(), super.close();
   }
 }
-const D = ["neutral", "angry", "sad", "happy"], j = "wss://api.openai.com/v1/realtime", U = "gpt-realtime", m = 24e3, z = {
+const j = ["neutral", "angry", "sad", "happy"], U = "wss://api.openai.com/v1/realtime", D = "gpt-realtime", m = 24e3, z = {
   type: "function",
   name: "set_emotion",
   description: "Set the emotional expression of the avatar. Call this on every turn to reflect the tone of your response.",
@@ -378,12 +378,12 @@ class $ extends v {
     if (!e.apiKey)
       throw new Error("OpenAI Realtime token is required");
     e.inputSampleRate && (this.sourceInputSampleRate = e.inputSampleRate);
-    const t = e.model ?? U;
-    return this.initialSessionUpdate = this.buildSessionUpdate(e, t), new Promise((i, n) => {
+    const t = e.model ?? D;
+    return this.initialSessionUpdate = this.buildSessionUpdate(t), new Promise((i, n) => {
       this.connectResolve = i, this.connectReject = n, this.connectTimeout = setTimeout(() => {
         this.rejectPendingConnect(new Error("Timed out waiting for OpenAI Realtime session setup")), this.close();
       }, 1e4), this.ws = new WebSocket(
-        `${j}?model=${encodeURIComponent(t)}`,
+        `${U}?model=${encodeURIComponent(t)}`,
         ["realtime", `openai-insecure-api-key.${e.apiKey}`]
       ), this.ws.onopen = () => {
       }, this.ws.onerror = () => {
@@ -469,29 +469,25 @@ class $ extends v {
   close() {
     this.rejectPendingConnect(new Error("Connection closed")), this.clearConnectTimeout(), this.resetTurnState(), this.initialSessionUpdate = null, this.handledFunctionCallIds.clear(), super.close();
   }
-  buildSessionUpdate(e, t) {
-    const i = e.turnDetection ?? { type: "semantic_vad", eagerness: "high" };
+  buildSessionUpdate(e) {
     return {
       type: "session.update",
       session: {
         type: "realtime",
-        model: t,
+        model: e,
         output_modalities: ["audio"],
-        instructions: e.systemPrompt,
         audio: {
           input: {
             format: {
               type: "audio/pcm",
               rate: m
-            },
-            turn_detection: i
+            }
           },
           output: {
             format: {
               type: "audio/pcm",
               rate: h
-            },
-            ...e.voice ? { voice: e.voice } : {}
+            }
           }
         },
         tools: [z],
@@ -540,7 +536,7 @@ class $ extends v {
       return { error: `Unsupported function: ${e.name}` };
     try {
       const i = (e.arguments ? JSON.parse(e.arguments) : {}).emotion?.toLowerCase();
-      return i && D.includes(i) ? (this.events.emit("emotion", i), { result: "ok" }) : { error: "Invalid emotion" };
+      return i && j.includes(i) ? (this.events.emit("emotion", i), { result: "ok" }) : { error: "Invalid emotion" };
     } catch {
       return { error: "Invalid function arguments" };
     }
@@ -882,10 +878,7 @@ class Y {
     else if (e.type === "openai")
       await this.agent.connect({
         ...t,
-        apiKey: e.token,
-        systemPrompt: e.system_prompt,
-        voice: e.voice,
-        turnDetection: e.turn_detection
+        apiKey: e.token
       });
     else if (e.type === "kfl") {
       if (!e.signed_url)
@@ -1028,10 +1021,7 @@ class be {
     else if (e.type === "openai")
       await this.agent.connect({
         ...t,
-        apiKey: e.token,
-        systemPrompt: e.system_prompt,
-        voice: e.voice,
-        turnDetection: e.turn_detection
+        apiKey: e.token
       });
     else if (e.type === "kfl") {
       if (!e.signed_url)

package/dist/kfl-embed.js

@@ -17896,7 +17896,7 @@ class Wu extends ts {
       throw new Error("OpenAI Realtime token is required");
     e.inputSampleRate && (this.sourceInputSampleRate = e.inputSampleRate);
     const t = e.model ?? Vu;
-    return this.initialSessionUpdate = this.buildSessionUpdate(e, t), new Promise((i, s) => {
+    return this.initialSessionUpdate = this.buildSessionUpdate(t), new Promise((i, s) => {
       this.connectResolve = i, this.connectReject = s, this.connectTimeout = setTimeout(() => {
         this.rejectPendingConnect(new Error("Timed out waiting for OpenAI Realtime session setup")), this.close();
       }, 1e4), this.ws = new WebSocket(
@@ -17986,29 +17986,25 @@ class Wu extends ts {
   close() {
     this.rejectPendingConnect(new Error("Connection closed")), this.clearConnectTimeout(), this.resetTurnState(), this.initialSessionUpdate = null, this.handledFunctionCallIds.clear(), super.close();
   }
-  buildSessionUpdate(e, t) {
-    const i = e.turnDetection ?? { type: "semantic_vad", eagerness: "high" };
+  buildSessionUpdate(e) {
     return {
       type: "session.update",
       session: {
         type: "realtime",
-        model: t,
+        model: e,
         output_modalities: ["audio"],
-        instructions: e.systemPrompt,
         audio: {
           input: {
             format: {
               type: "audio/pcm",
               rate: Zi
-            },
-            turn_detection: i
+            }
           },
           output: {
             format: {
               type: "audio/pcm",
               rate: si
-            },
-            ...e.voice ? { voice: e.voice } : {}
+            }
           }
         },
         tools: [qu],
@@ -18391,10 +18387,7 @@ class $u {
     else if (e.type === "openai")
       await this.agent.connect({
         ...t,
-        apiKey: e.token,
-        systemPrompt: e.system_prompt,
-        voice: e.voice,
-        turnDetection: e.turn_detection
+        apiKey: e.token
       });
     else if (e.type === "kfl") {
       if (!e.signed_url)

package/dist/types.d.ts

@@ -1,4 +1,4 @@
-import { AgentType, TurnDetection } from './agents';
+import { AgentType } from './agents';
 import { AgentState } from '@keyframelabs/sdk';
 export type EmbedStatus = 'connecting' | 'connected' | 'error' | 'disconnected';
 export type VideoFit = 'cover' | 'contain';
@@ -7,9 +7,6 @@ export type VoiceAgentDetails = {
     token?: string;
     agent_id?: string;
     signed_url?: string;
-    system_prompt?: string;
-    voice?: string;
-    turn_detection?: TurnDetection;
 };
 export type SessionDetails = {
     server_url: string;

package/package.json

@@ -4,7 +4,7 @@
   "publishConfig": {
     "access": "public"
   },
-  "version": "0.5.0",
+  "version": "0.5.1",
   "type": "module",
   "main": "./dist/index.js",
   "types": "./dist/index.d.ts",