@livekit/agents-plugin-openai - npm Package Compare versions

Comparing version 0.3.0 to 0.3.1

CHANGELOG.md

 # @livekit/agents-plugin-openai
 
+## 0.3.1
+
+### Patch Changes
+
+- implement session close - [#79](https://github.com/livekit/agents-js/pull/79) ([@nbsp](https://github.com/nbsp))
+
+- make sure rejects return errors, not string - [#84](https://github.com/livekit/agents-js/pull/84) ([@nbsp](https://github.com/nbsp))
+
+- Updated dependencies [[`d3db7cf19c696f611b5717ff8d510b2f910da712`](https://github.com/livekit/agents-js/commit/d3db7cf19c696f611b5717ff8d510b2f910da712), [`c0cce8a0f71cd8def7052917d8a6479e06178447`](https://github.com/livekit/agents-js/commit/c0cce8a0f71cd8def7052917d8a6479e06178447), [`e748aa4f7be76361c5fcafb03bdb760314b29a9f`](https://github.com/livekit/agents-js/commit/e748aa4f7be76361c5fcafb03bdb760314b29a9f), [`b35952ca243fecb087c898b670f5db0eaa1949bf`](https://github.com/livekit/agents-js/commit/b35952ca243fecb087c898b670f5db0eaa1949bf), [`4edacb8ba7dbbdd060dfedffe3116f1af4739b52`](https://github.com/livekit/agents-js/commit/4edacb8ba7dbbdd060dfedffe3116f1af4739b52)]:
+  - @livekit/agents@0.3.1
+
 ## 0.3.0
@@ -4,0 +15,0 @@

dist/realtime/api_proto.d.ts

@@ -38,4 +38,2 @@ export declare const SAMPLE_RATE = 24000;
     silence_duration_ms?: number;
-} | {
-    type: 'none';
 };
@@ -105,12 +103,13 @@ export type InputAudioTranscription = {
     modalities: ['text', 'audio'] | ['text'];
-    instructions?: string;
+    instructions: string;
     voice: Voice;
     input_audio_format: AudioFormat;
     output_audio_format: AudioFormat;
-    input_audio_transcription?: InputAudioTranscription;
-    turn_detection: TurnDetectionType;
+    input_audio_transcription: InputAudioTranscription | null;
+    turn_detection: TurnDetectionType | null;
     tools: Tool[];
     tool_choice: ToolChoice;
     temperature: number;
-    max_response_output_tokens: number | null;
+    max_response_output_tokens: number | 'inf';
+    expires_at: number;
 }
@@ -158,8 +157,8 @@ export interface ConversationResource {
         output_audio_format: AudioFormat;
-        input_audio_transcription?: InputAudioTranscription;
-        turn_detection: TurnDetectionType;
+        input_audio_transcription: InputAudioTranscription | null;
+        turn_detection: TurnDetectionType | null;
         tools: Tool[];
         tool_choice: ToolChoice;
         temperature: number;
-        max_response_output_tokens: number;
+        max_response_output_tokens: number | 'inf';
     }>;
@@ -223,3 +222,3 @@ }
         temperature: number;
-        max_response_output_tokens: number;
+        max_output_tokens: number | 'inf';
     }>;
@@ -226,0 +225,0 @@ }

dist/realtime/realtime_model.d.ts

@@ -7,10 +7,10 @@ import { AsyncIterableQueue, Future } from '@livekit/agents';
     modalities: ['text', 'audio'] | ['text'];
-    instructions?: string;
+    instructions: string;
     voice: api_proto.Voice;
     inputAudioFormat: api_proto.AudioFormat;
     outputAudioFormat: api_proto.AudioFormat;
-    inputAudioTranscription?: api_proto.InputAudioTranscription;
-    turnDetection: api_proto.TurnDetectionType;
+    inputAudioTranscription: api_proto.InputAudioTranscription | null;
+    turnDetection: api_proto.TurnDetectionType | null;
     temperature: number;
-    maxResponseOutputTokens?: number;
+    maxResponseOutputTokens: number;
     model: api_proto.Model;
@@ -118,4 +118,4 @@ apiKey: string;
         outputAudioFormat?: api_proto.AudioFormat;
-        inputAudioTranscription?: api_proto.InputAudioTranscription;
-        turnDetection?: api_proto.TurnDetectionType;
+        inputAudioTranscription?: api_proto.InputAudioTranscription | null;
+        turnDetection?: api_proto.TurnDetectionType | null;
         temperature?: number;
@@ -134,2 +134,3 @@ maxResponseOutputTokens?: number;
     get response(): Response;
+    get expiration(): number;
     queueMsg(command: api_proto.ClientEvent): void;
@@ -142,4 +143,4 @@ sessionUpdate({ modalities, instructions, voice, inputAudioFormat, outputAudioFormat, inputAudioTranscription, turnDetection, temperature, maxResponseOutputTokens, toolChoice, }: {
         outputAudioFormat?: api_proto.AudioFormat;
-        inputAudioTranscription?: api_proto.InputAudioTranscription;
-        turnDetection?: api_proto.TurnDetectionType;
+        inputAudioTranscription?: api_proto.InputAudioTranscription | null;
+        turnDetection?: api_proto.TurnDetectionType | null;
         temperature?: number;
@@ -146,0 +147,0 @@ maxResponseOutputTokens?: number;

dist/realtime/realtime_model.js

@@ -12,3 +12,3 @@ var __classPrivateFieldSet = (this && this.__classPrivateFieldSet) || function (receiver, state, value, kind, f) {
 };
-var _InputAudioBuffer_session, _ConversationItem_session, _Conversation_session, _Response_session, _RealtimeModel_defaultOpts, _RealtimeModel_sessions, _RealtimeSession_instances, _RealtimeSession_fncCtx, _RealtimeSession_opts, _RealtimeSession_pendingResponses, _RealtimeSession_sessionId, _RealtimeSession_ws, _RealtimeSession_logger, _RealtimeSession_task, _RealtimeSession_closing, _RealtimeSession_sendQueue, _RealtimeSession_loggableEvent, _RealtimeSession_start, _RealtimeSession_getContent, _RealtimeSession_handleError, _RealtimeSession_handleSessionCreated, _RealtimeSession_handleSessionUpdated, _RealtimeSession_handleConversationCreated, _RealtimeSession_handleInputAudioBufferCommitted, _RealtimeSession_handleInputAudioBufferCleared, _RealtimeSession_handleInputAudioBufferSpeechStarted, _RealtimeSession_handleInputAudioBufferSpeechStopped, _RealtimeSession_handleConversationItemCreated, _RealtimeSession_handleConversationItemInputAudioTranscriptionCompleted, _RealtimeSession_handleConversationItemInputAudioTranscriptionFailed, _RealtimeSession_handleConversationItemTruncated, _RealtimeSession_handleConversationItemDeleted, _RealtimeSession_handleResponseCreated, _RealtimeSession_handleResponseDone, _RealtimeSession_handleResponseOutputItemAdded, _RealtimeSession_handleResponseOutputItemDone, _RealtimeSession_handleResponseContentPartAdded, _RealtimeSession_handleResponseContentPartDone, _RealtimeSession_handleResponseTextDelta, _RealtimeSession_handleResponseTextDone, _RealtimeSession_handleResponseAudioTranscriptDelta, _RealtimeSession_handleResponseAudioTranscriptDone, _RealtimeSession_handleResponseAudioDelta, _RealtimeSession_handleResponseAudioDone, _RealtimeSession_handleResponseFunctionCallArgumentsDelta, _RealtimeSession_handleResponseFunctionCallArgumentsDone, _RealtimeSession_handleRateLimitsUpdated;
+var _InputAudioBuffer_session, _ConversationItem_session, _Conversation_session, _Response_session, _RealtimeModel_defaultOpts, _RealtimeModel_sessions, _RealtimeSession_instances, _RealtimeSession_fncCtx, _RealtimeSession_opts, _RealtimeSession_pendingResponses, _RealtimeSession_sessionId, _RealtimeSession_ws, _RealtimeSession_expiresAt, _RealtimeSession_logger, _RealtimeSession_task, _RealtimeSession_closing, _RealtimeSession_sendQueue, _RealtimeSession_loggableEvent, _RealtimeSession_start, _RealtimeSession_getContent, _RealtimeSession_handleError, _RealtimeSession_handleSessionCreated, _RealtimeSession_handleSessionUpdated, _RealtimeSession_handleConversationCreated, _RealtimeSession_handleInputAudioBufferCommitted, _RealtimeSession_handleInputAudioBufferCleared, _RealtimeSession_handleInputAudioBufferSpeechStarted, _RealtimeSession_handleInputAudioBufferSpeechStopped, _RealtimeSession_handleConversationItemCreated, _RealtimeSession_handleConversationItemInputAudioTranscriptionCompleted, _RealtimeSession_handleConversationItemInputAudioTranscriptionFailed, _RealtimeSession_handleConversationItemTruncated, _RealtimeSession_handleConversationItemDeleted, _RealtimeSession_handleResponseCreated, _RealtimeSession_handleResponseDone, _RealtimeSession_handleResponseOutputItemAdded, _RealtimeSession_handleResponseOutputItemDone, _RealtimeSession_handleResponseContentPartAdded, _RealtimeSession_handleResponseContentPartDone, _RealtimeSession_handleResponseTextDelta, _RealtimeSession_handleResponseTextDone, _RealtimeSession_handleResponseAudioTranscriptDelta, _RealtimeSession_handleResponseAudioTranscriptDone, _RealtimeSession_handleResponseAudioDelta, _RealtimeSession_handleResponseAudioDone, _RealtimeSession_handleResponseFunctionCallArgumentsDelta, _RealtimeSession_handleResponseFunctionCallArgumentsDone, _RealtimeSession_handleRateLimitsUpdated;
 // SPDX-FileCopyrightText: 2024 LiveKit, Inc.
@@ -102,3 +102,3 @@ //
 export class RealtimeModel extends multimodal.RealtimeModel {
-    constructor({ modalities = ['text', 'audio'], instructions = undefined, voice = 'alloy', inputAudioFormat = 'pcm16', outputAudioFormat = 'pcm16', inputAudioTranscription = { model: 'whisper-1' }, turnDetection = { type: 'server_vad' }, temperature = 0.8, maxResponseOutputTokens = undefined, model = 'gpt-4o-realtime-preview-2024-10-01', apiKey = process.env.OPENAI_API_KEY || '', baseURL = api_proto.API_URL, }) {
+    constructor({ modalities = ['text', 'audio'], instructions = '', voice = 'alloy', inputAudioFormat = 'pcm16', outputAudioFormat = 'pcm16', inputAudioTranscription = { model: 'whisper-1' }, turnDetection = { type: 'server_vad' }, temperature = 0.8, maxResponseOutputTokens = Infinity, model = 'gpt-4o-realtime-preview-2024-10-01', apiKey = process.env.OPENAI_API_KEY || '', baseURL = api_proto.API_URL, }) {
         super();
@@ -152,4 +152,3 @@ this.sampleRate = api_proto.SAMPLE_RATE;
     async close() {
-        // TODO: Implement close method
-        throw new Error('Not implemented');
+        await Promise.allSettled(__classPrivateFieldGet(this, _RealtimeModel_sessions, "f").map((session) => session.close()));
     }
@@ -167,2 +166,3 @@ }
         _RealtimeSession_ws.set(this, null);
+        _RealtimeSession_expiresAt.set(this, null);
         _RealtimeSession_logger.set(this, log());
@@ -203,2 +203,8 @@ _RealtimeSession_task.set(this, void 0);
     }
+    get expiration() {
+        if (!__classPrivateFieldGet(this, _RealtimeSession_expiresAt, "f")) {
+            throw new Error('session not started');
+        }
+        return __classPrivateFieldGet(this, _RealtimeSession_expiresAt, "f") * 1000;
+    }
     queueMsg(command) {
@@ -241,3 +247,5 @@ __classPrivateFieldGet(this, _RealtimeSession_sendQueue, "f").put(command);
                 temperature: __classPrivateFieldGet(this, _RealtimeSession_opts, "f").temperature,
-                max_response_output_tokens: __classPrivateFieldGet(this, _RealtimeSession_opts, "f").maxResponseOutputTokens,
+                max_response_output_tokens: __classPrivateFieldGet(this, _RealtimeSession_opts, "f").maxResponseOutputTokens === Infinity
+                    ? 'inf'
+                    : __classPrivateFieldGet(this, _RealtimeSession_opts, "f").maxResponseOutputTokens,
                 tools,
@@ -249,7 +257,10 @@ tool_choice: toolChoice,
     async close() {
-        // TODO: Implement close method
-        throw new Error('Not implemented');
+        if (!__classPrivateFieldGet(this, _RealtimeSession_ws, "f"))
+            return;
+        __classPrivateFieldSet(this, _RealtimeSession_closing, true, "f");
+        __classPrivateFieldGet(this, _RealtimeSession_ws, "f").close();
+        await __classPrivateFieldGet(this, _RealtimeSession_task, "f");
     }
 }
-_RealtimeSession_fncCtx = new WeakMap(), _RealtimeSession_opts = new WeakMap(), _RealtimeSession_pendingResponses = new WeakMap(), _RealtimeSession_sessionId = new WeakMap(), _RealtimeSession_ws = new WeakMap(), _RealtimeSession_logger = new WeakMap(), _RealtimeSession_task = new WeakMap(), _RealtimeSession_closing = new WeakMap(), _RealtimeSession_sendQueue = new WeakMap(), _RealtimeSession_instances = new WeakSet(), _RealtimeSession_loggableEvent = function _RealtimeSession_loggableEvent(event, maxLength = 30) {
+_RealtimeSession_fncCtx = new WeakMap(), _RealtimeSession_opts = new WeakMap(), _RealtimeSession_pendingResponses = new WeakMap(), _RealtimeSession_sessionId = new WeakMap(), _RealtimeSession_ws = new WeakMap(), _RealtimeSession_expiresAt = new WeakMap(), _RealtimeSession_logger = new WeakMap(), _RealtimeSession_task = new WeakMap(), _RealtimeSession_closing = new WeakMap(), _RealtimeSession_sendQueue = new WeakMap(), _RealtimeSession_instances = new WeakSet(), _RealtimeSession_loggableEvent = function _RealtimeSession_loggableEvent(event, maxLength = 30) {
     const untypedEvent = {};
@@ -281,3 +292,3 @@ for (const [key, value] of Object.entries(event)) {
         __classPrivateFieldGet(this, _RealtimeSession_ws, "f").onerror = (error) => {
-            reject(error.message);
+            reject(error);
         };
@@ -392,4 +403,7 @@ await once(__classPrivateFieldGet(this, _RealtimeSession_ws, "f"), 'open');
         __classPrivateFieldGet(this, _RealtimeSession_ws, "f").onclose = () => {
+            if (__classPrivateFieldGet(this, _RealtimeSession_expiresAt, "f") && Date.now() >= __classPrivateFieldGet(this, _RealtimeSession_expiresAt, "f") * 1000) {
+                __classPrivateFieldSet(this, _RealtimeSession_closing, true, "f");
+            }
             if (!__classPrivateFieldGet(this, _RealtimeSession_closing, "f")) {
-                reject('OpenAI Realtime connection closed unexpectedly');
+                reject(new Error('OpenAI Realtime connection closed unexpectedly'));
             }
@@ -409,2 +423,4 @@ __classPrivateFieldSet(this, _RealtimeSession_ws, null, "f");
     __classPrivateFieldSet(this, _RealtimeSession_sessionId, event.session.id, "f");
+    __classPrivateFieldSet(this, _RealtimeSession_expiresAt, event.session.expires_at, "f");
+    __classPrivateFieldSet(this, _RealtimeSession_logger, __classPrivateFieldGet(this, _RealtimeSession_logger, "f").child({ sessionId: __classPrivateFieldGet(this, _RealtimeSession_sessionId, "f") }), "f");
 }, _RealtimeSession_handleSessionUpdated = function _RealtimeSession_handleSessionUpdated(event) { }, _RealtimeSession_handleConversationCreated = function _RealtimeSession_handleConversationCreated(event) { }, _RealtimeSession_handleInputAudioBufferCommitted = function _RealtimeSession_handleInputAudioBufferCommitted(event) {
@@ -411,0 +427,0 @@ this.emit('input_speech_committed', {

package.json

 {
   "name": "@livekit/agents-plugin-openai",
-  "version": "0.3.0",
+  "version": "0.3.1",
   "description": "OpenAI plugin for LiveKit Node Agents",
@@ -17,3 +17,3 @@ "main": "dist/index.js",
     "ws": "^8.16.0",
-    "@livekit/agents": "0.3.0"
+    "@livekit/agents": "0.3.1"
   },
@@ -20,0 +20,0 @@ "scripts": {

src/realtime/api_proto.ts

@@ -86,12 +86,8 @@ // SPDX-FileCopyrightText: 2024 LiveKit, Inc.
 
-export type TurnDetectionType =
-  | {
-      type: 'server_vad';
-      threshold?: number; // 0.0 to 1.0, default: 0.5
-      prefix_padding_ms?: number; // default: 300
-      silence_duration_ms?: number; // default: 200
-    }
-  | {
-      type: 'none';
-    };
+export type TurnDetectionType = {
+  type: 'server_vad';
+  threshold?: number; // 0.0 to 1.0, default: 0.5
+  prefix_padding_ms?: number; // default: 300
+  silence_duration_ms?: number; // default: 200
+};
 
@@ -180,12 +176,13 @@ export type InputAudioTranscription = {
   modalities: ['text', 'audio'] | ['text']; // default: ["text", "audio"]
-  instructions?: string; // default: null
+  instructions: string;
   voice: Voice; // default: "alloy"
   input_audio_format: AudioFormat; // default: "pcm16"
   output_audio_format: AudioFormat; // default: "pcm16"
-  input_audio_transcription?: InputAudioTranscription; // default: null
-  turn_detection: TurnDetectionType;
+  input_audio_transcription: InputAudioTranscription | null;
+  turn_detection: TurnDetectionType | null;
   tools: Tool[];
   tool_choice: ToolChoice; // default: "auto"
   temperature: number; // default: 0.8
-  max_response_output_tokens: number | null;
+  max_response_output_tokens: number | 'inf';
+  expires_at: number;
 }
@@ -243,8 +240,8 @@
     output_audio_format: AudioFormat;
-    input_audio_transcription?: InputAudioTranscription;
-    turn_detection: TurnDetectionType;
+    input_audio_transcription: InputAudioTranscription | null;
+    turn_detection: TurnDetectionType | null;
     tools: Tool[];
     tool_choice: ToolChoice;
     temperature: number;
-    max_response_output_tokens: number;
+    max_response_output_tokens: number | 'inf';
   }>;
@@ -324,3 +321,3 @@ }
     temperature: number;
-    max_response_output_tokens: number;
+    max_output_tokens: number | 'inf';
   }>;
@@ -327,0 +324,0 @@ }

src/realtime/realtime_model.ts

@@ -13,10 +13,10 @@ // SPDX-FileCopyrightText: 2024 LiveKit, Inc.
   modalities: ['text', 'audio'] | ['text'];
-  instructions?: string;
+  instructions: string;
   voice: api_proto.Voice;
   inputAudioFormat: api_proto.AudioFormat;
   outputAudioFormat: api_proto.AudioFormat;
-  inputAudioTranscription?: api_proto.InputAudioTranscription;
-  turnDetection: api_proto.TurnDetectionType;
+  inputAudioTranscription: api_proto.InputAudioTranscription | null;
+  turnDetection: api_proto.TurnDetectionType | null;
   temperature: number;
-  maxResponseOutputTokens?: number;
+  maxResponseOutputTokens: number;
   model: api_proto.Model;
@@ -188,3 +188,3 @@ apiKey: string;
     modalities = ['text', 'audio'],
-    instructions = undefined,
+    instructions = '',
     voice = 'alloy',
@@ -196,3 +196,3 @@ inputAudioFormat = 'pcm16',
     temperature = 0.8,
-    maxResponseOutputTokens = undefined,
+    maxResponseOutputTokens = Infinity,
     model = 'gpt-4o-realtime-preview-2024-10-01',
@@ -261,4 +261,4 @@ apiKey = process.env.OPENAI_API_KEY || '',
     outputAudioFormat?: api_proto.AudioFormat;
-    inputAudioTranscription?: api_proto.InputAudioTranscription;
-    turnDetection?: api_proto.TurnDetectionType;
+    inputAudioTranscription?: api_proto.InputAudioTranscription | null;
+    turnDetection?: api_proto.TurnDetectionType | null;
     temperature?: number;
@@ -287,5 +287,4 @@ maxResponseOutputTokens?: number;
 
-  async close(): Promise<void> {
-    // TODO: Implement close method
-    throw new Error('Not implemented');
+  async close() {
+    await Promise.allSettled(this.#sessions.map((session) => session.close()));
   }
@@ -300,2 +299,3 @@ }
   #ws: WebSocket | null = null;
+  #expiresAt: number | null = null;
   #logger = log();
@@ -348,2 +348,9 @@ #task: Promise<void>;
 
+  get expiration(): number {
+    if (!this.#expiresAt) {
+      throw new Error('session not started');
+    }
+    return this.#expiresAt * 1000;
+  }
+
   queueMsg(command: api_proto.ClientEvent): void {
@@ -400,4 +407,4 @@ this.#sendQueue.put(command);
     outputAudioFormat?: api_proto.AudioFormat;
-    inputAudioTranscription?: api_proto.InputAudioTranscription;
-    turnDetection?: api_proto.TurnDetectionType;
+    inputAudioTranscription?: api_proto.InputAudioTranscription | null;
+    turnDetection?: api_proto.TurnDetectionType | null;
     temperature?: number;
@@ -442,3 +449,6 @@ maxResponseOutputTokens?: number;
         temperature: this.#opts.temperature,
-        max_response_output_tokens: this.#opts.maxResponseOutputTokens,
+        max_response_output_tokens:
+          this.#opts.maxResponseOutputTokens === Infinity
+            ? 'inf'
+            : this.#opts.maxResponseOutputTokens,
         tools,
@@ -460,3 +470,3 @@ tool_choice: toolChoice,
       this.#ws.onerror = (error) => {
-        reject(error.message);
+        reject(error);
       };
@@ -575,4 +585,7 @@
       this.#ws.onclose = () => {
+        if (this.#expiresAt && Date.now() >= this.#expiresAt * 1000) {
+          this.#closing = true;
+        }
         if (!this.#closing) {
-          reject('OpenAI Realtime connection closed unexpectedly');
+          reject(new Error('OpenAI Realtime connection closed unexpectedly'));
         }
@@ -585,5 +598,7 @@ this.#ws = null;
 
-  async close(): Promise<void> {
-    // TODO: Implement close method
-    throw new Error('Not implemented');
+  async close() {
+    if (!this.#ws) return;
+    this.#closing = true;
+    this.#ws.close();
+    await this.#task;
   }
@@ -604,2 +619,4 @@
     this.#sessionId = event.session.id;
+    this.#expiresAt = event.session.expires_at;
+    this.#logger = this.#logger.child({ sessionId: this.#sessionId });
   }
@@ -606,0 +623,0 @@

Fixed alerts

Major refactor

Supply chain risk

Package has recently undergone a major refactor. It may be unstable or indicate significant internal changes. Use caution when updating to versions that include significant changes.

Found 1 instance in 1 package

Improved metrics

Total package byte prevSize

152722

2.52%

Lines of code

2455

1.2%

Number of medium supply chain risk alerts

0

-100%

Dependency changes

• + Added

  @livekit/agents@0.3.1
  (transitive)

• - Removed

  @livekit/agents@0.3.0
  (transitive)

• Updated

  @livekit/agents@0.3.1