New Case Study:See how Anthropic automated 95% of dependency reviews with Socket.Learn More →
Socket
Sign inDemoInstall
Socket
Sign inDemoInstall

@livekit/agents

Package Overview
Dependencies
Maintainers
0
Versions
23
Alerts
File Explorer

Advanced tools

License
Socket logo

Install Socket

Detect and block malicious and high-risk dependencies

Install

@livekit/agents - npm Package Compare versions

Comparing version 0.4.3 to 0.4.4

dist/stt/stream_adapter.d.ts
dist/stt/stream_adapter.d.ts.map
dist/stt/stream_adapter.js
dist/stt/stream_adapter.js.map
dist/tts/stream_adapter.d.ts
dist/tts/stream_adapter.d.ts.map
dist/tts/stream_adapter.js
dist/tts/stream_adapter.js.map
src/stt/stream_adapter.ts
src/tts/stream_adapter.ts

12

CHANGELOG.md
# @livekit/agents
## 0.4.4
### Patch Changes
- add ChunkedStream, openai.TTS - [#155](https://github.com/livekit/agents-js/pull/155) ([@nbsp](https://github.com/nbsp))
- feat(stt): implement StreamAdapter - [#156](https://github.com/livekit/agents-js/pull/156) ([@nbsp](https://github.com/nbsp))
- export VPAEvent not as type - [#161](https://github.com/livekit/agents-js/pull/161) ([@nbsp](https://github.com/nbsp))
- add tts.StreamAdapter - [#156](https://github.com/livekit/agents-js/pull/156) ([@nbsp](https://github.com/nbsp))
## 0.4.3

@@ -4,0 +16,0 @@

12

dist/pipeline/agent_output.js

@@ -110,4 +110,5 @@ import { log } from '../log.js';

for await (const audio of ttsStream) {
if (cancelled || audio === SynthesizeStream.END_OF_STREAM)
if (cancelled || audio === SynthesizeStream.END_OF_STREAM) {
break;
}
handle.queue.put(audio.frame);

@@ -128,3 +129,2 @@ }

const readGeneratedAudio = async () => {
let started = false;
for await (const audio of ttsStream) {

@@ -134,11 +134,5 @@ if (cancelled)

if (audio === SynthesizeStream.END_OF_STREAM) {
if (started) {
break;
}
else {
continue;
}
break;
}
handle.queue.put(audio.frame);
started = true;
}

@@ -145,0 +139,0 @@ handle.queue.put(SynthesisHandle.FLUSH_SENTINEL);

3

dist/pipeline/agent_playout.js

@@ -107,4 +107,5 @@ import EventEmitter from 'node:events';

for await (const frame of handle.playoutSource) {
if (cancelled || frame === SynthesisHandle.FLUSH_SENTINEL)
if (cancelled || frame === SynthesisHandle.FLUSH_SENTINEL) {
break;
}
if (firstFrame) {

@@ -111,0 +112,0 @@ this.#logger

2

dist/pipeline/index.d.ts

@@ -1,2 +0,2 @@

export { type AgentState, type BeforeTTSCallback, type BeforeLLMCallback, type VPAEvent, type VPACallbacks, type AgentCallContext, type AgentTranscriptionOptions, type VPAOptions, VoicePipelineAgent, } from './pipeline_agent.js';
export { type AgentState, type BeforeTTSCallback, type BeforeLLMCallback, type VPACallbacks, type AgentCallContext, type AgentTranscriptionOptions, type VPAOptions, VPAEvent, VoicePipelineAgent, } from './pipeline_agent.js';
//# sourceMappingURL=index.d.ts.map

2

dist/pipeline/index.js
// SPDX-FileCopyrightText: 2024 LiveKit, Inc.
//
// SPDX-License-Identifier: Apache-2.0
export { VoicePipelineAgent, } from './pipeline_agent.js';
export { VPAEvent, VoicePipelineAgent, } from './pipeline_agent.js';
//# sourceMappingURL=index.js.map

2

dist/pipeline/pipeline_agent.d.ts

@@ -6,3 +6,3 @@ import type { RemoteParticipant, Room } from '@livekit/rtc-node';

import { ChatContext, ChatMessage } from '../llm/index.js';
import type { STT } from '../stt/index.js';
import { type STT } from '../stt/index.js';
import type { SentenceTokenizer, WordTokenizer } from '../tokenize/tokenizer.js';

@@ -9,0 +9,0 @@ import type { TTS } from '../tts/index.js';

13

dist/pipeline/pipeline_agent.js

@@ -7,3 +7,5 @@ var _a;

import { log } from '../log.js';
import { StreamAdapter as STTStreamAdapter } from '../stt/index.js';
import { SentenceTokenizer as BasicSentenceTokenizer, WordTokenizer as BasicWordTokenizer, hyphenateWord, } from '../tokenize/basic/index.js';
import { StreamAdapter as TTSStreamAdapter } from '../tts/index.js';
import { AsyncIterableQueue, CancellablePromise, Future, gracefullyCancel } from '../utils.js';

@@ -121,2 +123,8 @@ import { AgentOutput } from './agent_output.js';

this.#opts = { ...defaultVPAOptions, ...opts };
if (!stt.capabilities.streaming) {
stt = new STTStreamAdapter(stt, vad);
}
if (!tts.capabilities.streaming) {
tts = new TTSStreamAdapter(tts, new BasicSentenceTokenizer());
}
this.#vad = vad;

@@ -394,2 +402,3 @@ this.#stt = stt;

this.#transcribedText = this.#transcribedText.slice(userQuestion.length);
handle.markUserCommitted();
};

@@ -417,3 +426,3 @@ // wait for the playHandle to finish and check every 1s if user question should be committed

if (isUsingTools && !interrupted) {
if (!userQuestion || handle.userCommitted) {
if (!userQuestion || !handle.userCommitted) {
throw new Error('user speech should have been committed before using tools');

@@ -590,3 +599,3 @@ }

log()
.child({ speechId, elapsed: Math.round(Date.now() * 1000 - startTime) / 1000 })
.child({ speechId, elapsed: Math.round(Date.now() - startTime) })
.debug('received first LLM token');

@@ -593,0 +602,0 @@ }

1

dist/stt/index.d.ts
export { type SpeechEvent, type SpeechData, type STTCapabilities, SpeechEventType, STT, SpeechStream, } from './stt.js';
export { StreamAdapter, StreamAdapterWrapper } from './stream_adapter.js';
//# sourceMappingURL=index.d.ts.map

1

dist/stt/index.js

@@ -5,2 +5,3 @@ // SPDX-FileCopyrightText: 2024 LiveKit, Inc.

export { SpeechEventType, STT, SpeechStream, } from './stt.js';
export { StreamAdapter, StreamAdapterWrapper } from './stream_adapter.js';
//# sourceMappingURL=index.js.map

3

dist/stt/stt.d.ts
import type { AudioFrame } from '@livekit/rtc-node';
import type { AudioBuffer } from '../utils.js';
import { AsyncIterableQueue } from '../utils.js';

@@ -61,2 +62,4 @@ /** Indicates start/middle/end of speech */

get capabilities(): STTCapabilities;
/** Receives an audio buffer and returns transcription in the form of a {@link SpeechEvent} */
abstract recognize(frame: AudioBuffer): Promise<SpeechEvent>;
/**

@@ -63,0 +66,0 @@ * Returns a {@link SpeechStream} that can be used to push audio frames and receive

1

dist/tokenize/token_stream.d.ts

@@ -25,2 +25,3 @@ import { AsyncIterableQueue } from '../utils.js';

pushText(text: string): void;
flush(): void;
close(): void;

@@ -27,0 +28,0 @@ next(): Promise<IteratorResult<TokenData>>;

5

dist/tokenize/token_stream.js

@@ -33,3 +33,3 @@ // SPDX-FileCopyrightText: 2024 LiveKit, Inc.

const tokens = this.#func(this.#inBuf);
if (tokens.length === 0)
if (tokens.length <= 1)
break;

@@ -112,2 +112,5 @@ if (this.#outBuf)

}
flush() {
this.#stream.flush();
}
close() {

@@ -114,0 +117,0 @@ super.close();

3

dist/tts/index.d.ts

@@ -1,2 +0,3 @@

export { type SynthesizedAudio, type TTSCapabilities, TTS, SynthesizeStream } from './tts.js';
export { type SynthesizedAudio, type TTSCapabilities, TTS, SynthesizeStream, ChunkedStream, } from './tts.js';
export { StreamAdapter, StreamAdapterWrapper } from './stream_adapter.js';
//# sourceMappingURL=index.d.ts.map

3

dist/tts/index.js
// SPDX-FileCopyrightText: 2024 LiveKit, Inc.
//
// SPDX-License-Identifier: Apache-2.0
export { TTS, SynthesizeStream } from './tts.js';
export { TTS, SynthesizeStream, ChunkedStream, } from './tts.js';
export { StreamAdapter, StreamAdapterWrapper } from './stream_adapter.js';
//# sourceMappingURL=index.js.map

28

dist/tts/tts.d.ts

@@ -41,2 +41,6 @@ import type { AudioFrame } from '@livekit/rtc-node';

/**
* Receives text and returns synthesis in the form of a {@link ChunkedStream}
*/
abstract synthesize(text: string): ChunkedStream;
/**
* Returns a {@link SynthesizeStream} that can be used to push text and receive audio data

@@ -77,2 +81,26 @@ */

}
/**
* An instance of a text-to-speech response, as an asynchronous iterable iterator.
*
* @example Looping through frames
* ```ts
* for await (const event of stream) {
* await source.captureFrame(event.frame);
* }
* ```
*
* @remarks
* This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that
* exports its own child ChunkedStream class, which inherits this class's methods.
*/
export declare abstract class ChunkedStream implements AsyncIterableIterator<SynthesizedAudio> {
protected queue: AsyncIterableQueue<SynthesizedAudio>;
protected closed: boolean;
/** Collect every frame into one in a single call */
collect(): Promise<AudioFrame>;
next(): Promise<IteratorResult<SynthesizedAudio>>;
/** Close both the input and output of the TTS stream */
close(): void;
[Symbol.asyncIterator](): ChunkedStream;
}
//# sourceMappingURL=tts.d.ts.map

39

dist/tts/tts.js

@@ -1,2 +0,2 @@

import { AsyncIterableQueue } from '../utils.js';
import { AsyncIterableQueue, mergeFrames } from '../utils.js';
/**

@@ -94,2 +94,39 @@ * An instance of a text-to-speech adapter.

}
/**
* An instance of a text-to-speech response, as an asynchronous iterable iterator.
*
* @example Looping through frames
* ```ts
* for await (const event of stream) {
* await source.captureFrame(event.frame);
* }
* ```
*
* @remarks
* This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that
* exports its own child ChunkedStream class, which inherits this class's methods.
*/
export class ChunkedStream {
queue = new AsyncIterableQueue();
closed = false;
/** Collect every frame into one in a single call */
async collect() {
const frames = [];
for await (const event of this) {
frames.push(event.frame);
}
return mergeFrames(frames);
}
next() {
return this.queue.next();
}
/** Close both the input and output of the TTS stream */
close() {
this.queue.close();
this.closed = true;
}
[Symbol.asyncIterator]() {
return this;
}
}
//# sourceMappingURL=tts.js.map

2

package.json
{
"name": "@livekit/agents",
"version": "0.4.3",
"version": "0.4.4",
"description": "LiveKit Agents - Node.js",

@@ -5,0 +5,0 @@ "main": "dist/index.js",

12

src/pipeline/agent_output.ts

@@ -137,3 +137,5 @@ // SPDX-FileCopyrightText: 2024 LiveKit, Inc.

for await (const audio of ttsStream) {
if (cancelled || audio === SynthesizeStream.END_OF_STREAM) break;
if (cancelled || audio === SynthesizeStream.END_OF_STREAM) {
break;
}
handle.queue.put(audio.frame);

@@ -160,14 +162,8 @@ }

const readGeneratedAudio = async () => {
let started = false;
for await (const audio of ttsStream) {
if (cancelled) break;
if (audio === SynthesizeStream.END_OF_STREAM) {
if (started) {
break;
} else {
continue;
}
break;
}
handle.queue.put(audio.frame);
started = true;
}

@@ -174,0 +170,0 @@ handle.queue.put(SynthesisHandle.FLUSH_SENTINEL);

4

src/pipeline/agent_playout.ts

@@ -151,3 +151,5 @@ // SPDX-FileCopyrightText: 2024 LiveKit, Inc.

for await (const frame of handle.playoutSource) {
if (cancelled || frame === SynthesisHandle.FLUSH_SENTINEL) break;
if (cancelled || frame === SynthesisHandle.FLUSH_SENTINEL) {
break;
}
if (firstFrame) {

@@ -154,0 +156,0 @@ this.#logger

2

src/pipeline/index.ts

@@ -9,3 +9,2 @@ // SPDX-FileCopyrightText: 2024 LiveKit, Inc.

type BeforeLLMCallback,
type VPAEvent,
type VPACallbacks,

@@ -15,3 +14,4 @@ type AgentCallContext,

type VPAOptions,
VPAEvent,
VoicePipelineAgent,
} from './pipeline_agent.js';

16

src/pipeline/pipeline_agent.ts

@@ -23,3 +23,3 @@ // SPDX-FileCopyrightText: 2024 LiveKit, Inc.

import { log } from '../log.js';
import type { STT } from '../stt/index.js';
import { type STT, StreamAdapter as STTStreamAdapter } from '../stt/index.js';
import {

@@ -32,2 +32,3 @@ SentenceTokenizer as BasicSentenceTokenizer,

import type { TTS } from '../tts/index.js';
import { StreamAdapter as TTSStreamAdapter } from '../tts/index.js';
import { AsyncIterableQueue, CancellablePromise, Future, gracefullyCancel } from '../utils.js';

@@ -257,2 +258,10 @@ import type { VAD, VADEvent } from '../vad.js';

if (!stt.capabilities.streaming) {
stt = new STTStreamAdapter(stt, vad);
}
if (!tts.capabilities.streaming) {
tts = new TTSStreamAdapter(tts, new BasicSentenceTokenizer());
}
this.#vad = vad;

@@ -599,2 +608,3 @@ this.#stt = stt;

this.#transcribedText = this.#transcribedText.slice(userQuestion.length);
handle.markUserCommitted();
};

@@ -625,3 +635,3 @@

if (isUsingTools && !interrupted) {
if (!userQuestion || handle.userCommitted) {
if (!userQuestion || !handle.userCommitted) {
throw new Error('user speech should have been committed before using tools');

@@ -831,3 +841,3 @@ }

log()
.child({ speechId, elapsed: Math.round(Date.now() * 1000 - startTime) / 1000 })
.child({ speechId, elapsed: Math.round(Date.now() - startTime) })
.debug('received first LLM token');

@@ -834,0 +844,0 @@ }

1

src/stt/index.ts

@@ -13,1 +13,2 @@ // SPDX-FileCopyrightText: 2024 LiveKit, Inc.

} from './stt.js';
export { StreamAdapter, StreamAdapterWrapper } from './stream_adapter.js';

4

src/stt/stt.ts

@@ -5,2 +5,3 @@ // SPDX-FileCopyrightText: 2024 LiveKit, Inc.

import type { AudioFrame } from '@livekit/rtc-node';
import type { AudioBuffer } from '../utils.js';
import { AsyncIterableQueue } from '../utils.js';

@@ -77,2 +78,5 @@

/** Receives an audio buffer and returns transcription in the form of a {@link SpeechEvent} */
abstract recognize(frame: AudioBuffer): Promise<SpeechEvent>;
/**

@@ -79,0 +83,0 @@ * Returns a {@link SpeechStream} that can be used to push audio frames and receive

6

src/tokenize/token_stream.ts

@@ -42,3 +42,3 @@ // SPDX-FileCopyrightText: 2024 LiveKit, Inc.

const tokens = this.#func(this.#inBuf);
if (tokens.length === 0) break;
if (tokens.length <= 1) break;

@@ -134,2 +134,6 @@ if (this.#outBuf) this.#outBuf += ' ';

flush() {
this.#stream.flush();
}
close() {

@@ -136,0 +140,0 @@ super.close();

9

src/tts/index.ts
// SPDX-FileCopyrightText: 2024 LiveKit, Inc.
//
// SPDX-License-Identifier: Apache-2.0
export { type SynthesizedAudio, type TTSCapabilities, TTS, SynthesizeStream } from './tts.js';
export {
type SynthesizedAudio,
type TTSCapabilities,
TTS,
SynthesizeStream,
ChunkedStream,
} from './tts.js';
export { StreamAdapter, StreamAdapterWrapper } from './stream_adapter.js';

49

src/tts/tts.ts

@@ -5,3 +5,3 @@ // SPDX-FileCopyrightText: 2024 LiveKit, Inc.

import type { AudioFrame } from '@livekit/rtc-node';
import { AsyncIterableQueue } from '../utils.js';
import { AsyncIterableQueue, mergeFrames } from '../utils.js';

@@ -65,2 +65,7 @@ /** SynthesizedAudio is a packet of speech synthesis as returned by the TTS. */

/**
* Receives text and returns synthesis in the form of a {@link ChunkedStream}
*/
abstract synthesize(text: string): ChunkedStream;
/**
* Returns a {@link SynthesizeStream} that can be used to push text and receive audio data

@@ -144,1 +149,43 @@ */

}
/**
* An instance of a text-to-speech response, as an asynchronous iterable iterator.
*
* @example Looping through frames
* ```ts
* for await (const event of stream) {
* await source.captureFrame(event.frame);
* }
* ```
*
* @remarks
* This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that
* exports its own child ChunkedStream class, which inherits this class's methods.
*/
export abstract class ChunkedStream implements AsyncIterableIterator<SynthesizedAudio> {
protected queue = new AsyncIterableQueue<SynthesizedAudio>();
protected closed = false;
/** Collect every frame into one in a single call */
async collect(): Promise<AudioFrame> {
const frames = [];
for await (const event of this) {
frames.push(event.frame);
}
return mergeFrames(frames);
}
next(): Promise<IteratorResult<SynthesizedAudio>> {
return this.queue.next();
}
/** Close both the input and output of the TTS stream */
close() {
this.queue.close();
this.closed = true;
}
[Symbol.asyncIterator](): ChunkedStream {
return this;
}
}
.turbo/turbo-build.log

Sorry, the diff of this file is not supported yet

dist/pipeline/agent_output.js.map

Sorry, the diff of this file is not supported yet

dist/pipeline/agent_playout.d.ts.map

Sorry, the diff of this file is not supported yet

dist/pipeline/agent_playout.js.map

Sorry, the diff of this file is not supported yet

dist/pipeline/index.d.ts.map

Sorry, the diff of this file is not supported yet

dist/pipeline/index.js.map

Sorry, the diff of this file is not supported yet

dist/pipeline/pipeline_agent.d.ts.map

Sorry, the diff of this file is not supported yet

dist/pipeline/pipeline_agent.js.map

Sorry, the diff of this file is not supported yet

dist/stt/index.d.ts.map

Sorry, the diff of this file is not supported yet

dist/stt/index.js.map

Sorry, the diff of this file is not supported yet

dist/stt/stt.d.ts.map

Sorry, the diff of this file is not supported yet

dist/stt/stt.js.map

Sorry, the diff of this file is not supported yet

dist/tokenize/token_stream.d.ts.map

Sorry, the diff of this file is not supported yet

dist/tokenize/token_stream.js.map

Sorry, the diff of this file is not supported yet

dist/tts/index.d.ts.map

Sorry, the diff of this file is not supported yet

dist/tts/index.js.map

Sorry, the diff of this file is not supported yet

dist/tts/tts.d.ts.map

Sorry, the diff of this file is not supported yet

dist/tts/tts.js.map

Sorry, the diff of this file is not supported yet

tsconfig.tsbuildinfo

Sorry, the diff of this file is not supported yet

SocketSocket SOC 2 Logo

Product

  • Package Alerts
  • Integrations
  • Docs
  • Pricing
  • FAQ
  • Roadmap
  • Changelog

About

Packages

npm

Go

Maven

PyPI

Rubygems

Stay in touch

Get open source security insights delivered straight into your inbox.

  • Terms
  • Privacy
  • Security

Made with ⚡️ by Socket Inc