@micdrop/server 1.0.2 → 1.1.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +18 -6
- package/dist/index.d.mts +3 -2
- package/dist/index.d.ts +3 -2
- package/dist/index.js +23 -3
- package/dist/index.js.map +1 -1
- package/dist/index.mjs +23 -3
- package/dist/index.mjs.map +1 -1
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -7,11 +7,11 @@ For browser implementation, see [@micdrop/client](../client/README.md) package.
|
|
|
7
7
|
## Features
|
|
8
8
|
|
|
9
9
|
- 🌐 WebSocket server for real-time audio streaming
|
|
10
|
-
- 🔊
|
|
10
|
+
- 🔊 Advanced audio processing:
|
|
11
|
+
- Streaming TTS support
|
|
12
|
+
- Efficient audio chunk delivery
|
|
13
|
+
- Interrupt handling
|
|
11
14
|
- 💬 Conversation state management
|
|
12
|
-
- ⚡ Event-based architecture
|
|
13
|
-
- 🔄 Bi-directional communication
|
|
14
|
-
- 🛡️ Built-in error handling
|
|
15
15
|
- 🎙️ Speech-to-text and text-to-speech integration
|
|
16
16
|
- 🤖 AI conversation generation support
|
|
17
17
|
- 💾 Debug mode with optional audio saving
|
|
@@ -57,7 +57,10 @@ const config: CallConfig = {
|
|
|
57
57
|
},
|
|
58
58
|
|
|
59
59
|
// Function to convert text to speech
|
|
60
|
-
|
|
60
|
+
// Can return either a complete ArrayBuffer or a ReadableStream for streaming
|
|
61
|
+
async text2Speech(
|
|
62
|
+
text: string
|
|
63
|
+
): Promise<ArrayBuffer | NodeJS.ReadableStream> {
|
|
61
64
|
// Implement your TTS logic
|
|
62
65
|
return new ArrayBuffer(0) // Audio data
|
|
63
66
|
},
|
|
@@ -110,6 +113,12 @@ The main class for managing WebSocket connections and audio streaming.
|
|
|
110
113
|
```typescript
|
|
111
114
|
class CallSocket {
|
|
112
115
|
constructor(socket: WebSocket, config: CallConfig)
|
|
116
|
+
|
|
117
|
+
// Add assistant message and send to client with audio (TTS)
|
|
118
|
+
answer(message: string): Promise<void>
|
|
119
|
+
|
|
120
|
+
// Reset conversation (including system prompt)
|
|
121
|
+
resetConversation(conversation: Conversation): void
|
|
113
122
|
}
|
|
114
123
|
```
|
|
115
124
|
|
|
@@ -141,7 +150,8 @@ interface CallConfig {
|
|
|
141
150
|
speech2Text(blob: Blob, prompt?: string): Promise<string>
|
|
142
151
|
|
|
143
152
|
// Convert text to audio
|
|
144
|
-
|
|
153
|
+
// Can return either a complete ArrayBuffer or a ReadableStream for streaming
|
|
154
|
+
text2Speech(text: string): Promise<ArrayBuffer | NodeJS.ReadableStream>
|
|
145
155
|
|
|
146
156
|
// Optional callbacks
|
|
147
157
|
onMessage?(message: ConversationMessage): void
|
|
@@ -259,6 +269,8 @@ server.get('/call', { websocket: true }, (socket) => {
|
|
|
259
269
|
server.listen({ port: 8080 })
|
|
260
270
|
```
|
|
261
271
|
|
|
272
|
+
See [@micdrop/demo-server](../demo-server/src/ai/index.ts) for a complete example using OpenAI and ElevenLabs.
|
|
273
|
+
|
|
262
274
|
## Debug Mode
|
|
263
275
|
|
|
264
276
|
The server includes a debug mode that can:
|
package/dist/index.d.mts
CHANGED
|
@@ -19,7 +19,7 @@ interface CallConfig {
|
|
|
19
19
|
disableTTS?: boolean;
|
|
20
20
|
generateAnswer(conversation: Conversation): Promise<string>;
|
|
21
21
|
speech2Text(blob: Blob, prompt?: string): Promise<string>;
|
|
22
|
-
text2Speech(text: string): Promise<ArrayBuffer>;
|
|
22
|
+
text2Speech(text: string): Promise<ArrayBuffer | NodeJS.ReadableStream>;
|
|
23
23
|
onMessage?(message: ConversationMessage): void;
|
|
24
24
|
onEnd?(call: CallSummary): void;
|
|
25
25
|
}
|
|
@@ -44,11 +44,12 @@ declare class CallSocket {
|
|
|
44
44
|
private chunks;
|
|
45
45
|
private conversation;
|
|
46
46
|
constructor(socket: WebSocket, config: CallConfig);
|
|
47
|
+
resetConversation(conversation: Conversation): void;
|
|
47
48
|
private addMessage;
|
|
48
49
|
private onClose;
|
|
49
50
|
private onMessage;
|
|
50
51
|
private onStopSpeaking;
|
|
51
|
-
|
|
52
|
+
answer(message: string): Promise<void>;
|
|
52
53
|
private log;
|
|
53
54
|
}
|
|
54
55
|
|
package/dist/index.d.ts
CHANGED
|
@@ -19,7 +19,7 @@ interface CallConfig {
|
|
|
19
19
|
disableTTS?: boolean;
|
|
20
20
|
generateAnswer(conversation: Conversation): Promise<string>;
|
|
21
21
|
speech2Text(blob: Blob, prompt?: string): Promise<string>;
|
|
22
|
-
text2Speech(text: string): Promise<ArrayBuffer>;
|
|
22
|
+
text2Speech(text: string): Promise<ArrayBuffer | NodeJS.ReadableStream>;
|
|
23
23
|
onMessage?(message: ConversationMessage): void;
|
|
24
24
|
onEnd?(call: CallSummary): void;
|
|
25
25
|
}
|
|
@@ -44,11 +44,12 @@ declare class CallSocket {
|
|
|
44
44
|
private chunks;
|
|
45
45
|
private conversation;
|
|
46
46
|
constructor(socket: WebSocket, config: CallConfig);
|
|
47
|
+
resetConversation(conversation: Conversation): void;
|
|
47
48
|
private addMessage;
|
|
48
49
|
private onClose;
|
|
49
50
|
private onMessage;
|
|
50
51
|
private onStopSpeaking;
|
|
51
|
-
|
|
52
|
+
answer(message: string): Promise<void>;
|
|
52
53
|
private log;
|
|
53
54
|
}
|
|
54
55
|
|
package/dist/index.js
CHANGED
|
@@ -87,6 +87,11 @@ var CallSocket = class {
|
|
|
87
87
|
this.socket.on("close", this.onClose.bind(this));
|
|
88
88
|
this.socket.on("message", this.onMessage.bind(this));
|
|
89
89
|
}
|
|
90
|
+
// Reset conversation
|
|
91
|
+
resetConversation(conversation) {
|
|
92
|
+
this.log("Reset conversation");
|
|
93
|
+
this.conversation = conversation;
|
|
94
|
+
}
|
|
90
95
|
addMessage(message) {
|
|
91
96
|
this.conversation.push(message);
|
|
92
97
|
this.socket.send(
|
|
@@ -164,6 +169,7 @@ var CallSocket = class {
|
|
|
164
169
|
this.socket.close();
|
|
165
170
|
}
|
|
166
171
|
}
|
|
172
|
+
// Add assistant message and send to client with audio (TTS)
|
|
167
173
|
async answer(message) {
|
|
168
174
|
let isEnd = false;
|
|
169
175
|
if (message.includes(END_INTERVIEW)) {
|
|
@@ -175,17 +181,31 @@ var CallSocket = class {
|
|
|
175
181
|
this.addMessage({ role: "assistant", content: message });
|
|
176
182
|
if (!this.config.disableTTS) {
|
|
177
183
|
const audio = await this.config.text2Speech(message);
|
|
178
|
-
|
|
184
|
+
const abort = () => {
|
|
179
185
|
this.log("Answer aborted, removing last assistant message");
|
|
180
186
|
const lastMessage = this.conversation[this.conversation.length - 1];
|
|
181
187
|
if (lastMessage?.role === "assistant") {
|
|
182
188
|
this.conversation.pop();
|
|
183
189
|
this.socket.send("cancelLastAssistantMessage" /* CancelLastAssistantMessage */);
|
|
184
190
|
}
|
|
191
|
+
};
|
|
192
|
+
if (this.abortAnswer) {
|
|
193
|
+
abort();
|
|
185
194
|
return;
|
|
186
195
|
}
|
|
187
|
-
|
|
188
|
-
|
|
196
|
+
if (audio instanceof ArrayBuffer) {
|
|
197
|
+
this.log(`Send audio: (${audio.byteLength} bytes)`);
|
|
198
|
+
this.socket.send(audio);
|
|
199
|
+
} else {
|
|
200
|
+
for await (const chunk of audio) {
|
|
201
|
+
if (this.abortAnswer) {
|
|
202
|
+
abort();
|
|
203
|
+
return;
|
|
204
|
+
}
|
|
205
|
+
this.log(`Send audio: (${chunk.length} bytes)`);
|
|
206
|
+
this.socket.send(chunk);
|
|
207
|
+
}
|
|
208
|
+
}
|
|
189
209
|
}
|
|
190
210
|
}
|
|
191
211
|
if (isEnd) {
|
package/dist/index.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/index.ts","../src/CallSocket.ts","../src/types.ts","../src/errors.ts","../src/waitForParams.ts"],"sourcesContent":["export * from './CallSocket'\nexport * from './errors'\nexport * from './types'\nexport * from './waitForParams'\n","import * as fs from 'fs'\nimport * as path from 'path'\nimport { WebSocket } from 'ws'\nimport {\n CallClientCommands,\n CallConfig,\n CallServerCommands,\n Conversation,\n ConversationMessage,\n} from './types'\n\nexport const END_INTERVIEW = 'END_INTERVIEW'\n\nexport class CallSocket {\n private startTime = Date.now()\n private lastDebug = Date.now()\n\n // An answer can be aborted if user is speaking\n private abortAnswer = false\n\n // When user is speaking, we're waiting to chunks or to stop\n private isSpeaking = false\n\n // Chunks of user speech since user started speaking\n private chunks: Buffer[] = []\n\n // Conversation history\n private conversation: Conversation\n\n constructor(\n public socket: WebSocket,\n public config: CallConfig\n ) {\n this.conversation = [{ role: 'system', content: config.systemPrompt }]\n this.log(`Call started`)\n\n // Assistant speaks first\n\n // LLM: Generate answer\n if (config.firstMessage) {\n this.answer(config.firstMessage)\n } else {\n this.config\n .generateAnswer(this.conversation)\n .then((answer) => this.answer(answer))\n .catch((error) => {\n console.error('[WS]', error)\n this.socket.close()\n })\n }\n\n // Listen to events\n this.socket.on('close', this.onClose.bind(this))\n this.socket.on('message', this.onMessage.bind(this))\n }\n\n private addMessage(message: ConversationMessage) {\n this.conversation.push(message)\n this.socket.send(\n `${\n message.role === 'user'\n ? CallServerCommands.UserMessage\n : CallServerCommands.AssistantMessage\n } ${message.content}`\n )\n this.config.onMessage?.(message)\n }\n\n private onClose() {\n this.log('Connection closed')\n this.abortAnswer = true\n const duration = Math.round((Date.now() - this.startTime) / 1000)\n\n // End call callback\n this.config.onEnd?.({\n conversation: this.conversation.slice(1), // Remove system message\n duration,\n })\n\n // Unset params\n // @ts-ignore\n this.socket = null\n // @ts-ignore\n this.config = null\n }\n\n private async onMessage(message: Buffer) {\n if (!Buffer.isBuffer(message)) {\n console.warn(`[WS] Message is not a buffer`)\n return\n }\n\n // Commands\n if (message.byteLength < 15) {\n const cmd = message.toString()\n this.log(`Command: ${cmd}`)\n\n if (cmd === CallClientCommands.StartSpeaking) {\n // User started speaking\n this.isSpeaking = true\n // Abort answer if there is generation in progress\n this.abortAnswer = true\n } else if (cmd === CallClientCommands.Mute) {\n // User muted the call\n // Abort answer if there is generation in progress\n this.abortAnswer = true\n } else if (cmd === CallClientCommands.StopSpeaking) {\n // User stopped speaking\n this.isSpeaking = false\n await this.onStopSpeaking()\n }\n }\n\n // Audio chunk\n else if (Buffer.isBuffer(message) && this.isSpeaking) {\n this.log(`Received chunk (${message.byteLength} bytes)`)\n this.chunks.push(message)\n }\n }\n\n private async onStopSpeaking() {\n // Do nothing if there is no chunk\n if (this.chunks.length === 0) return\n\n this.abortAnswer = false\n\n // Combine audio blob\n const blob = new Blob(this.chunks, { type: 'audio/ogg' })\n\n // Reset chunks for next user speech\n this.chunks.length = 0\n\n try {\n // Save file to disk\n if (this.config.debugSaveSpeech) {\n const filePath = path.join(__dirname, 'speech.ogg')\n fs.writeFileSync(filePath, Buffer.from(await blob.arrayBuffer()))\n return\n }\n\n // STT: Get transcript and send to client\n const transcript = await this.config.speech2Text(\n blob,\n this.conversation[this.conversation.length - 1]?.content\n )\n if (!transcript) {\n this.log('Ignoring empty transcript')\n return\n }\n\n this.log('User transcript:', transcript)\n\n // Send transcript to client\n this.addMessage({ role: 'user', content: transcript })\n\n if (this.abortAnswer) {\n this.log('Answer aborted, no answer generated')\n return\n }\n\n // LLM: Generate answer\n const answer = await this.config.generateAnswer(this.conversation)\n if (this.abortAnswer) {\n this.log('Answer aborted, ignoring answer')\n return\n }\n\n await this.answer(answer)\n } catch (error) {\n console.error('[WS]', error)\n this.socket.close()\n }\n }\n\n private async answer(message: string) {\n let isEnd = false\n\n // Detect end of interview\n if (message.includes(END_INTERVIEW)) {\n message = message.replace(END_INTERVIEW, '').trim()\n isEnd = true\n }\n\n if (message.length) {\n // Send answer to client\n this.log('Assistant message:', message)\n this.addMessage({ role: 'assistant', content: message })\n\n // TTS: Generate answer audio\n if (!this.config.disableTTS) {\n const audio = await this.config.text2Speech(message)\n if (this.abortAnswer) {\n // Remove last assistant message if aborted\n this.log('Answer aborted, removing last assistant message')\n const lastMessage = this.conversation[this.conversation.length - 1]\n if (lastMessage?.role === 'assistant') {\n this.conversation.pop()\n this.socket.send(CallServerCommands.CancelLastAssistantMessage)\n }\n return\n }\n\n // Send audio to client\n this.log(`Send audio: (${audio.byteLength} bytes)`)\n this.socket.send(audio)\n }\n }\n\n // End of call\n if (isEnd) {\n this.log('Interview ended')\n this.socket.send(CallServerCommands.EndInterview)\n }\n }\n\n private log(...message: any[]) {\n if (!this.config.debugLog) return\n const now = Date.now()\n const delta = now - this.lastDebug\n this.lastDebug = now\n console.log(`[Debug +${delta}ms]`, ...message)\n }\n}\n","export enum CallClientCommands {\n StartSpeaking = 'startSpeaking',\n StopSpeaking = 'stopSpeaking',\n Mute = 'mute',\n}\n\nexport enum CallServerCommands {\n UserMessage = 'userMessage',\n AssistantMessage = 'assistantMessage',\n CancelLastAssistantMessage = 'cancelLastAssistantMessage',\n EndInterview = 'endInterview',\n}\n\nexport interface CallConfig {\n systemPrompt: string\n firstMessage?: string\n debugLog?: boolean\n debugSaveSpeech?: boolean\n disableTTS?: boolean\n generateAnswer(conversation: Conversation): Promise<string>\n speech2Text(blob: Blob, prompt?: string): Promise<string>\n text2Speech(text: string): Promise<ArrayBuffer>\n onMessage?(message: ConversationMessage): void\n onEnd?(call: CallSummary): void\n}\n\nexport interface CallSummary {\n conversation: Conversation\n duration: number\n}\n\nexport type Conversation = ConversationMessage[]\n\nexport interface ConversationMessage {\n role: 'system' | 'user' | 'assistant'\n content: string\n}\n","import WebSocket from 'ws'\n\nexport enum CallErrorCode {\n BadRequest = 4400,\n Unauthorized = 4401,\n NotFound = 4404,\n}\n\nexport class CallError extends Error {\n code: number\n\n constructor(code: number, message: string) {\n super(message)\n this.code = code\n }\n}\n\nexport function handleError(socket: WebSocket, error: unknown) {\n if (error instanceof CallError) {\n socket.close(error.code, error.message)\n } else {\n console.error(error)\n socket.close(1011)\n }\n socket.terminate()\n}\n","import { WebSocket } from 'ws'\nimport { CallError, CallErrorCode } from './errors'\n\nexport async function waitForParams<CallParams>(\n socket: WebSocket,\n validate: (params: any) => CallParams\n): Promise<CallParams> {\n return new Promise<CallParams>((resolve, reject) => {\n // Handle timeout\n const timeout = setTimeout(() => {\n reject(new CallError(CallErrorCode.BadRequest, 'Missing params'))\n }, 3000)\n\n const onParams = (payload: string) => {\n // Clear timeout and listener\n clearTimeout(timeout)\n socket.off('message', onParams)\n\n try {\n // Parse JSON payload\n const params = validate(JSON.parse(payload))\n resolve(params)\n } catch (error) {\n reject(new CallError(CallErrorCode.BadRequest, 'Invalid params'))\n }\n }\n\n // Listen for params\n socket.on('message', onParams)\n })\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;;;ACAA,SAAoB;AACpB,WAAsB;;;ACDf,IAAK,qBAAL,kBAAKA,wBAAL;AACL,EAAAA,oBAAA,mBAAgB;AAChB,EAAAA,oBAAA,kBAAe;AACf,EAAAA,oBAAA,UAAO;AAHG,SAAAA;AAAA,GAAA;AAML,IAAK,qBAAL,kBAAKC,wBAAL;AACL,EAAAA,oBAAA,iBAAc;AACd,EAAAA,oBAAA,sBAAmB;AACnB,EAAAA,oBAAA,gCAA6B;AAC7B,EAAAA,oBAAA,kBAAe;AAJL,SAAAA;AAAA,GAAA;;;ADKL,IAAM,gBAAgB;AAEtB,IAAM,aAAN,MAAiB;AAAA,EAgBtB,YACS,QACA,QACP;AAFO;AACA;AAjBT,SAAQ,YAAY,KAAK,IAAI;AAC7B,SAAQ,YAAY,KAAK,IAAI;AAG7B;AAAA,SAAQ,cAAc;AAGtB;AAAA,SAAQ,aAAa;AAGrB;AAAA,SAAQ,SAAmB,CAAC;AAS1B,SAAK,eAAe,CAAC,EAAE,MAAM,UAAU,SAAS,OAAO,aAAa,CAAC;AACrE,SAAK,IAAI,cAAc;AAKvB,QAAI,OAAO,cAAc;AACvB,WAAK,OAAO,OAAO,YAAY;AAAA,IACjC,OAAO;AACL,WAAK,OACF,eAAe,KAAK,YAAY,EAChC,KAAK,CAAC,WAAW,KAAK,OAAO,MAAM,CAAC,EACpC,MAAM,CAAC,UAAU;AAChB,gBAAQ,MAAM,QAAQ,KAAK;AAC3B,aAAK,OAAO,MAAM;AAAA,MACpB,CAAC;AAAA,IACL;AAGA,SAAK,OAAO,GAAG,SAAS,KAAK,QAAQ,KAAK,IAAI,CAAC;AAC/C,SAAK,OAAO,GAAG,WAAW,KAAK,UAAU,KAAK,IAAI,CAAC;AAAA,EACrD;AAAA,EAEQ,WAAW,SAA8B;AAC/C,SAAK,aAAa,KAAK,OAAO;AAC9B,SAAK,OAAO;AAAA,MACV,GACE,QAAQ,SAAS,oFAGnB,IAAI,QAAQ,OAAO;AAAA,IACrB;AACA,SAAK,OAAO,YAAY,OAAO;AAAA,EACjC;AAAA,EAEQ,UAAU;AAChB,SAAK,IAAI,mBAAmB;AAC5B,SAAK,cAAc;AACnB,UAAM,WAAW,KAAK,OAAO,KAAK,IAAI,IAAI,KAAK,aAAa,GAAI;AAGhE,SAAK,OAAO,QAAQ;AAAA,MAClB,cAAc,KAAK,aAAa,MAAM,CAAC;AAAA;AAAA,MACvC;AAAA,IACF,CAAC;AAID,SAAK,SAAS;AAEd,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAc,UAAU,SAAiB;AACvC,QAAI,CAAC,OAAO,SAAS,OAAO,GAAG;AAC7B,cAAQ,KAAK,8BAA8B;AAC3C;AAAA,IACF;AAGA,QAAI,QAAQ,aAAa,IAAI;AAC3B,YAAM,MAAM,QAAQ,SAAS;AAC7B,WAAK,IAAI,YAAY,GAAG,EAAE;AAE1B,UAAI,6CAA0C;AAE5C,aAAK,aAAa;AAElB,aAAK,cAAc;AAAA,MACrB,WAAW,2BAAiC;AAG1C,aAAK,cAAc;AAAA,MACrB,WAAW,2CAAyC;AAElD,aAAK,aAAa;AAClB,cAAM,KAAK,eAAe;AAAA,MAC5B;AAAA,IACF,WAGS,OAAO,SAAS,OAAO,KAAK,KAAK,YAAY;AACpD,WAAK,IAAI,mBAAmB,QAAQ,UAAU,SAAS;AACvD,WAAK,OAAO,KAAK,OAAO;AAAA,IAC1B;AAAA,EACF;AAAA,EAEA,MAAc,iBAAiB;AAE7B,QAAI,KAAK,OAAO,WAAW,EAAG;AAE9B,SAAK,cAAc;AAGnB,UAAM,OAAO,IAAI,KAAK,KAAK,QAAQ,EAAE,MAAM,YAAY,CAAC;AAGxD,SAAK,OAAO,SAAS;AAErB,QAAI;AAEF,UAAI,KAAK,OAAO,iBAAiB;AAC/B,cAAM,WAAgB,UAAK,WAAW,YAAY;AAClD,QAAG,iBAAc,UAAU,OAAO,KAAK,MAAM,KAAK,YAAY,CAAC,CAAC;AAChE;AAAA,MACF;AAGA,YAAM,aAAa,MAAM,KAAK,OAAO;AAAA,QACnC;AAAA,QACA,KAAK,aAAa,KAAK,aAAa,SAAS,CAAC,GAAG;AAAA,MACnD;AACA,UAAI,CAAC,YAAY;AACf,aAAK,IAAI,2BAA2B;AACpC;AAAA,MACF;AAEA,WAAK,IAAI,oBAAoB,UAAU;AAGvC,WAAK,WAAW,EAAE,MAAM,QAAQ,SAAS,WAAW,CAAC;AAErD,UAAI,KAAK,aAAa;AACpB,aAAK,IAAI,qCAAqC;AAC9C;AAAA,MACF;AAGA,YAAM,SAAS,MAAM,KAAK,OAAO,eAAe,KAAK,YAAY;AACjE,UAAI,KAAK,aAAa;AACpB,aAAK,IAAI,iCAAiC;AAC1C;AAAA,MACF;AAEA,YAAM,KAAK,OAAO,MAAM;AAAA,IAC1B,SAAS,OAAO;AACd,cAAQ,MAAM,QAAQ,KAAK;AAC3B,WAAK,OAAO,MAAM;AAAA,IACpB;AAAA,EACF;AAAA,EAEA,MAAc,OAAO,SAAiB;AACpC,QAAI,QAAQ;AAGZ,QAAI,QAAQ,SAAS,aAAa,GAAG;AACnC,gBAAU,QAAQ,QAAQ,eAAe,EAAE,EAAE,KAAK;AAClD,cAAQ;AAAA,IACV;AAEA,QAAI,QAAQ,QAAQ;AAElB,WAAK,IAAI,sBAAsB,OAAO;AACtC,WAAK,WAAW,EAAE,MAAM,aAAa,SAAS,QAAQ,CAAC;AAGvD,UAAI,CAAC,KAAK,OAAO,YAAY;AAC3B,cAAM,QAAQ,MAAM,KAAK,OAAO,YAAY,OAAO;AACnD,YAAI,KAAK,aAAa;AAEpB,eAAK,IAAI,iDAAiD;AAC1D,gBAAM,cAAc,KAAK,aAAa,KAAK,aAAa,SAAS,CAAC;AAClE,cAAI,aAAa,SAAS,aAAa;AACrC,iBAAK,aAAa,IAAI;AACtB,iBAAK,OAAO,kEAAkD;AAAA,UAChE;AACA;AAAA,QACF;AAGA,aAAK,IAAI,gBAAgB,MAAM,UAAU,SAAS;AAClD,aAAK,OAAO,KAAK,KAAK;AAAA,MACxB;AAAA,IACF;AAGA,QAAI,OAAO;AACT,WAAK,IAAI,iBAAiB;AAC1B,WAAK,OAAO,sCAAoC;AAAA,IAClD;AAAA,EACF;AAAA,EAEQ,OAAO,SAAgB;AAC7B,QAAI,CAAC,KAAK,OAAO,SAAU;AAC3B,UAAM,MAAM,KAAK,IAAI;AACrB,UAAM,QAAQ,MAAM,KAAK;AACzB,SAAK,YAAY;AACjB,YAAQ,IAAI,WAAW,KAAK,OAAO,GAAG,OAAO;AAAA,EAC/C;AACF;;;AE5NO,IAAK,gBAAL,kBAAKC,mBAAL;AACL,EAAAA,8BAAA,gBAAa,QAAb;AACA,EAAAA,8BAAA,kBAAe,QAAf;AACA,EAAAA,8BAAA,cAAW,QAAX;AAHU,SAAAA;AAAA,GAAA;AAML,IAAM,YAAN,cAAwB,MAAM;AAAA,EAGnC,YAAY,MAAc,SAAiB;AACzC,UAAM,OAAO;AACb,SAAK,OAAO;AAAA,EACd;AACF;AAEO,SAAS,YAAY,QAAmB,OAAgB;AAC7D,MAAI,iBAAiB,WAAW;AAC9B,WAAO,MAAM,MAAM,MAAM,MAAM,OAAO;AAAA,EACxC,OAAO;AACL,YAAQ,MAAM,KAAK;AACnB,WAAO,MAAM,IAAI;AAAA,EACnB;AACA,SAAO,UAAU;AACnB;;;ACtBA,eAAsB,cACpB,QACA,UACqB;AACrB,SAAO,IAAI,QAAoB,CAAC,SAAS,WAAW;AAElD,UAAM,UAAU,WAAW,MAAM;AAC/B,aAAO,IAAI,iCAAoC,gBAAgB,CAAC;AAAA,IAClE,GAAG,GAAI;AAEP,UAAM,WAAW,CAAC,YAAoB;AAEpC,mBAAa,OAAO;AACpB,aAAO,IAAI,WAAW,QAAQ;AAE9B,UAAI;AAEF,cAAM,SAAS,SAAS,KAAK,MAAM,OAAO,CAAC;AAC3C,gBAAQ,MAAM;AAAA,MAChB,SAAS,OAAO;AACd,eAAO,IAAI,iCAAoC,gBAAgB,CAAC;AAAA,MAClE;AAAA,IACF;AAGA,WAAO,GAAG,WAAW,QAAQ;AAAA,EAC/B,CAAC;AACH;","names":["CallClientCommands","CallServerCommands","CallErrorCode"]}
|
|
1
|
+
{"version":3,"sources":["../src/index.ts","../src/CallSocket.ts","../src/types.ts","../src/errors.ts","../src/waitForParams.ts"],"sourcesContent":["export * from './CallSocket'\nexport * from './errors'\nexport * from './types'\nexport * from './waitForParams'\n","import * as fs from 'fs'\nimport * as path from 'path'\nimport { WebSocket } from 'ws'\nimport {\n CallClientCommands,\n CallConfig,\n CallServerCommands,\n Conversation,\n ConversationMessage,\n} from './types'\n\nexport const END_INTERVIEW = 'END_INTERVIEW'\n\nexport class CallSocket {\n private startTime = Date.now()\n private lastDebug = Date.now()\n\n // An answer can be aborted if user is speaking\n private abortAnswer = false\n\n // When user is speaking, we're waiting to chunks or to stop\n private isSpeaking = false\n\n // Chunks of user speech since user started speaking\n private chunks: Buffer[] = []\n\n // Conversation history\n private conversation: Conversation\n\n constructor(\n public socket: WebSocket,\n public config: CallConfig\n ) {\n this.conversation = [{ role: 'system', content: config.systemPrompt }]\n this.log(`Call started`)\n\n // Assistant speaks first\n\n // LLM: Generate answer\n if (config.firstMessage) {\n this.answer(config.firstMessage)\n } else {\n this.config\n .generateAnswer(this.conversation)\n .then((answer) => this.answer(answer))\n .catch((error) => {\n console.error('[WS]', error)\n this.socket.close()\n })\n }\n\n // Listen to events\n this.socket.on('close', this.onClose.bind(this))\n this.socket.on('message', this.onMessage.bind(this))\n }\n\n // Reset conversation\n public resetConversation(conversation: Conversation) {\n this.log('Reset conversation')\n this.conversation = conversation\n }\n\n private addMessage(message: ConversationMessage) {\n this.conversation.push(message)\n this.socket.send(\n `${\n message.role === 'user'\n ? CallServerCommands.UserMessage\n : CallServerCommands.AssistantMessage\n } ${message.content}`\n )\n this.config.onMessage?.(message)\n }\n\n private onClose() {\n this.log('Connection closed')\n this.abortAnswer = true\n const duration = Math.round((Date.now() - this.startTime) / 1000)\n\n // End call callback\n this.config.onEnd?.({\n conversation: this.conversation.slice(1), // Remove system message\n duration,\n })\n\n // Unset params\n // @ts-ignore\n this.socket = null\n // @ts-ignore\n this.config = null\n }\n\n private async onMessage(message: Buffer) {\n if (!Buffer.isBuffer(message)) {\n console.warn(`[WS] Message is not a buffer`)\n return\n }\n\n // Commands\n if (message.byteLength < 15) {\n const cmd = message.toString()\n this.log(`Command: ${cmd}`)\n\n if (cmd === CallClientCommands.StartSpeaking) {\n // User started speaking\n this.isSpeaking = true\n // Abort answer if there is generation in progress\n this.abortAnswer = true\n } else if (cmd === CallClientCommands.Mute) {\n // User muted the call\n // Abort answer if there is generation in progress\n this.abortAnswer = true\n } else if (cmd === CallClientCommands.StopSpeaking) {\n // User stopped speaking\n this.isSpeaking = false\n await this.onStopSpeaking()\n }\n }\n\n // Audio chunk\n else if (Buffer.isBuffer(message) && this.isSpeaking) {\n this.log(`Received chunk (${message.byteLength} bytes)`)\n this.chunks.push(message)\n }\n }\n\n private async onStopSpeaking() {\n // Do nothing if there is no chunk\n if (this.chunks.length === 0) return\n\n this.abortAnswer = false\n\n // Combine audio blob\n const blob = new Blob(this.chunks, { type: 'audio/ogg' })\n\n // Reset chunks for next user speech\n this.chunks.length = 0\n\n try {\n // Save file to disk\n if (this.config.debugSaveSpeech) {\n const filePath = path.join(__dirname, 'speech.ogg')\n fs.writeFileSync(filePath, Buffer.from(await blob.arrayBuffer()))\n return\n }\n\n // STT: Get transcript and send to client\n const transcript = await this.config.speech2Text(\n blob,\n this.conversation[this.conversation.length - 1]?.content\n )\n if (!transcript) {\n this.log('Ignoring empty transcript')\n return\n }\n\n this.log('User transcript:', transcript)\n\n // Send transcript to client\n this.addMessage({ role: 'user', content: transcript })\n\n if (this.abortAnswer) {\n this.log('Answer aborted, no answer generated')\n return\n }\n\n // LLM: Generate answer\n const answer = await this.config.generateAnswer(this.conversation)\n if (this.abortAnswer) {\n this.log('Answer aborted, ignoring answer')\n return\n }\n\n await this.answer(answer)\n } catch (error) {\n console.error('[WS]', error)\n this.socket.close()\n }\n }\n\n // Add assistant message and send to client with audio (TTS)\n public async answer(message: string) {\n let isEnd = false\n\n // Detect end of interview\n if (message.includes(END_INTERVIEW)) {\n message = message.replace(END_INTERVIEW, '').trim()\n isEnd = true\n }\n\n if (message.length) {\n // Send answer to client\n this.log('Assistant message:', message)\n this.addMessage({ role: 'assistant', content: message })\n\n // TTS: Generate answer audio\n if (!this.config.disableTTS) {\n const audio = await this.config.text2Speech(message)\n\n // Remove last assistant message if aborted\n const abort = () => {\n this.log('Answer aborted, removing last assistant message')\n const lastMessage = this.conversation[this.conversation.length - 1]\n if (lastMessage?.role === 'assistant') {\n this.conversation.pop()\n this.socket.send(CallServerCommands.CancelLastAssistantMessage)\n }\n }\n\n if (this.abortAnswer) {\n abort()\n return\n }\n\n // Send audio to client\n if (audio instanceof ArrayBuffer) {\n this.log(`Send audio: (${audio.byteLength} bytes)`)\n this.socket.send(audio)\n } else {\n for await (const chunk of audio) {\n if (this.abortAnswer) {\n abort()\n return\n }\n this.log(`Send audio: (${chunk.length} bytes)`)\n this.socket.send(chunk)\n }\n }\n }\n }\n\n // End of call\n if (isEnd) {\n this.log('Interview ended')\n this.socket.send(CallServerCommands.EndInterview)\n }\n }\n\n private log(...message: any[]) {\n if (!this.config.debugLog) return\n const now = Date.now()\n const delta = now - this.lastDebug\n this.lastDebug = now\n console.log(`[Debug +${delta}ms]`, ...message)\n }\n}\n","export enum CallClientCommands {\n StartSpeaking = 'startSpeaking',\n StopSpeaking = 'stopSpeaking',\n Mute = 'mute',\n}\n\nexport enum CallServerCommands {\n UserMessage = 'userMessage',\n AssistantMessage = 'assistantMessage',\n CancelLastAssistantMessage = 'cancelLastAssistantMessage',\n EndInterview = 'endInterview',\n}\n\nexport interface CallConfig {\n systemPrompt: string\n firstMessage?: string\n debugLog?: boolean\n debugSaveSpeech?: boolean\n disableTTS?: boolean\n generateAnswer(conversation: Conversation): Promise<string>\n speech2Text(blob: Blob, prompt?: string): Promise<string>\n text2Speech(text: string): Promise<ArrayBuffer | NodeJS.ReadableStream>\n onMessage?(message: ConversationMessage): void\n onEnd?(call: CallSummary): void\n}\n\nexport interface CallSummary {\n conversation: Conversation\n duration: number\n}\n\nexport type Conversation = ConversationMessage[]\n\nexport interface ConversationMessage {\n role: 'system' | 'user' | 'assistant'\n content: string\n}\n","import WebSocket from 'ws'\n\nexport enum CallErrorCode {\n BadRequest = 4400,\n Unauthorized = 4401,\n NotFound = 4404,\n}\n\nexport class CallError extends Error {\n code: number\n\n constructor(code: number, message: string) {\n super(message)\n this.code = code\n }\n}\n\nexport function handleError(socket: WebSocket, error: unknown) {\n if (error instanceof CallError) {\n socket.close(error.code, error.message)\n } else {\n console.error(error)\n socket.close(1011)\n }\n socket.terminate()\n}\n","import { WebSocket } from 'ws'\nimport { CallError, CallErrorCode } from './errors'\n\nexport async function waitForParams<CallParams>(\n socket: WebSocket,\n validate: (params: any) => CallParams\n): Promise<CallParams> {\n return new Promise<CallParams>((resolve, reject) => {\n // Handle timeout\n const timeout = setTimeout(() => {\n reject(new CallError(CallErrorCode.BadRequest, 'Missing params'))\n }, 3000)\n\n const onParams = (payload: string) => {\n // Clear timeout and listener\n clearTimeout(timeout)\n socket.off('message', onParams)\n\n try {\n // Parse JSON payload\n const params = validate(JSON.parse(payload))\n resolve(params)\n } catch (error) {\n reject(new CallError(CallErrorCode.BadRequest, 'Invalid params'))\n }\n }\n\n // Listen for params\n socket.on('message', onParams)\n })\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;;;ACAA,SAAoB;AACpB,WAAsB;;;ACDf,IAAK,qBAAL,kBAAKA,wBAAL;AACL,EAAAA,oBAAA,mBAAgB;AAChB,EAAAA,oBAAA,kBAAe;AACf,EAAAA,oBAAA,UAAO;AAHG,SAAAA;AAAA,GAAA;AAML,IAAK,qBAAL,kBAAKC,wBAAL;AACL,EAAAA,oBAAA,iBAAc;AACd,EAAAA,oBAAA,sBAAmB;AACnB,EAAAA,oBAAA,gCAA6B;AAC7B,EAAAA,oBAAA,kBAAe;AAJL,SAAAA;AAAA,GAAA;;;ADKL,IAAM,gBAAgB;AAEtB,IAAM,aAAN,MAAiB;AAAA,EAgBtB,YACS,QACA,QACP;AAFO;AACA;AAjBT,SAAQ,YAAY,KAAK,IAAI;AAC7B,SAAQ,YAAY,KAAK,IAAI;AAG7B;AAAA,SAAQ,cAAc;AAGtB;AAAA,SAAQ,aAAa;AAGrB;AAAA,SAAQ,SAAmB,CAAC;AAS1B,SAAK,eAAe,CAAC,EAAE,MAAM,UAAU,SAAS,OAAO,aAAa,CAAC;AACrE,SAAK,IAAI,cAAc;AAKvB,QAAI,OAAO,cAAc;AACvB,WAAK,OAAO,OAAO,YAAY;AAAA,IACjC,OAAO;AACL,WAAK,OACF,eAAe,KAAK,YAAY,EAChC,KAAK,CAAC,WAAW,KAAK,OAAO,MAAM,CAAC,EACpC,MAAM,CAAC,UAAU;AAChB,gBAAQ,MAAM,QAAQ,KAAK;AAC3B,aAAK,OAAO,MAAM;AAAA,MACpB,CAAC;AAAA,IACL;AAGA,SAAK,OAAO,GAAG,SAAS,KAAK,QAAQ,KAAK,IAAI,CAAC;AAC/C,SAAK,OAAO,GAAG,WAAW,KAAK,UAAU,KAAK,IAAI,CAAC;AAAA,EACrD;AAAA;AAAA,EAGO,kBAAkB,cAA4B;AACnD,SAAK,IAAI,oBAAoB;AAC7B,SAAK,eAAe;AAAA,EACtB;AAAA,EAEQ,WAAW,SAA8B;AAC/C,SAAK,aAAa,KAAK,OAAO;AAC9B,SAAK,OAAO;AAAA,MACV,GACE,QAAQ,SAAS,oFAGnB,IAAI,QAAQ,OAAO;AAAA,IACrB;AACA,SAAK,OAAO,YAAY,OAAO;AAAA,EACjC;AAAA,EAEQ,UAAU;AAChB,SAAK,IAAI,mBAAmB;AAC5B,SAAK,cAAc;AACnB,UAAM,WAAW,KAAK,OAAO,KAAK,IAAI,IAAI,KAAK,aAAa,GAAI;AAGhE,SAAK,OAAO,QAAQ;AAAA,MAClB,cAAc,KAAK,aAAa,MAAM,CAAC;AAAA;AAAA,MACvC;AAAA,IACF,CAAC;AAID,SAAK,SAAS;AAEd,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAc,UAAU,SAAiB;AACvC,QAAI,CAAC,OAAO,SAAS,OAAO,GAAG;AAC7B,cAAQ,KAAK,8BAA8B;AAC3C;AAAA,IACF;AAGA,QAAI,QAAQ,aAAa,IAAI;AAC3B,YAAM,MAAM,QAAQ,SAAS;AAC7B,WAAK,IAAI,YAAY,GAAG,EAAE;AAE1B,UAAI,6CAA0C;AAE5C,aAAK,aAAa;AAElB,aAAK,cAAc;AAAA,MACrB,WAAW,2BAAiC;AAG1C,aAAK,cAAc;AAAA,MACrB,WAAW,2CAAyC;AAElD,aAAK,aAAa;AAClB,cAAM,KAAK,eAAe;AAAA,MAC5B;AAAA,IACF,WAGS,OAAO,SAAS,OAAO,KAAK,KAAK,YAAY;AACpD,WAAK,IAAI,mBAAmB,QAAQ,UAAU,SAAS;AACvD,WAAK,OAAO,KAAK,OAAO;AAAA,IAC1B;AAAA,EACF;AAAA,EAEA,MAAc,iBAAiB;AAE7B,QAAI,KAAK,OAAO,WAAW,EAAG;AAE9B,SAAK,cAAc;AAGnB,UAAM,OAAO,IAAI,KAAK,KAAK,QAAQ,EAAE,MAAM,YAAY,CAAC;AAGxD,SAAK,OAAO,SAAS;AAErB,QAAI;AAEF,UAAI,KAAK,OAAO,iBAAiB;AAC/B,cAAM,WAAgB,UAAK,WAAW,YAAY;AAClD,QAAG,iBAAc,UAAU,OAAO,KAAK,MAAM,KAAK,YAAY,CAAC,CAAC;AAChE;AAAA,MACF;AAGA,YAAM,aAAa,MAAM,KAAK,OAAO;AAAA,QACnC;AAAA,QACA,KAAK,aAAa,KAAK,aAAa,SAAS,CAAC,GAAG;AAAA,MACnD;AACA,UAAI,CAAC,YAAY;AACf,aAAK,IAAI,2BAA2B;AACpC;AAAA,MACF;AAEA,WAAK,IAAI,oBAAoB,UAAU;AAGvC,WAAK,WAAW,EAAE,MAAM,QAAQ,SAAS,WAAW,CAAC;AAErD,UAAI,KAAK,aAAa;AACpB,aAAK,IAAI,qCAAqC;AAC9C;AAAA,MACF;AAGA,YAAM,SAAS,MAAM,KAAK,OAAO,eAAe,KAAK,YAAY;AACjE,UAAI,KAAK,aAAa;AACpB,aAAK,IAAI,iCAAiC;AAC1C;AAAA,MACF;AAEA,YAAM,KAAK,OAAO,MAAM;AAAA,IAC1B,SAAS,OAAO;AACd,cAAQ,MAAM,QAAQ,KAAK;AAC3B,WAAK,OAAO,MAAM;AAAA,IACpB;AAAA,EACF;AAAA;AAAA,EAGA,MAAa,OAAO,SAAiB;AACnC,QAAI,QAAQ;AAGZ,QAAI,QAAQ,SAAS,aAAa,GAAG;AACnC,gBAAU,QAAQ,QAAQ,eAAe,EAAE,EAAE,KAAK;AAClD,cAAQ;AAAA,IACV;AAEA,QAAI,QAAQ,QAAQ;AAElB,WAAK,IAAI,sBAAsB,OAAO;AACtC,WAAK,WAAW,EAAE,MAAM,aAAa,SAAS,QAAQ,CAAC;AAGvD,UAAI,CAAC,KAAK,OAAO,YAAY;AAC3B,cAAM,QAAQ,MAAM,KAAK,OAAO,YAAY,OAAO;AAGnD,cAAM,QAAQ,MAAM;AAClB,eAAK,IAAI,iDAAiD;AAC1D,gBAAM,cAAc,KAAK,aAAa,KAAK,aAAa,SAAS,CAAC;AAClE,cAAI,aAAa,SAAS,aAAa;AACrC,iBAAK,aAAa,IAAI;AACtB,iBAAK,OAAO,kEAAkD;AAAA,UAChE;AAAA,QACF;AAEA,YAAI,KAAK,aAAa;AACpB,gBAAM;AACN;AAAA,QACF;AAGA,YAAI,iBAAiB,aAAa;AAChC,eAAK,IAAI,gBAAgB,MAAM,UAAU,SAAS;AAClD,eAAK,OAAO,KAAK,KAAK;AAAA,QACxB,OAAO;AACL,2BAAiB,SAAS,OAAO;AAC/B,gBAAI,KAAK,aAAa;AACpB,oBAAM;AACN;AAAA,YACF;AACA,iBAAK,IAAI,gBAAgB,MAAM,MAAM,SAAS;AAC9C,iBAAK,OAAO,KAAK,KAAK;AAAA,UACxB;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAGA,QAAI,OAAO;AACT,WAAK,IAAI,iBAAiB;AAC1B,WAAK,OAAO,sCAAoC;AAAA,IAClD;AAAA,EACF;AAAA,EAEQ,OAAO,SAAgB;AAC7B,QAAI,CAAC,KAAK,OAAO,SAAU;AAC3B,UAAM,MAAM,KAAK,IAAI;AACrB,UAAM,QAAQ,MAAM,KAAK;AACzB,SAAK,YAAY;AACjB,YAAQ,IAAI,WAAW,KAAK,OAAO,GAAG,OAAO;AAAA,EAC/C;AACF;;;AEnPO,IAAK,gBAAL,kBAAKC,mBAAL;AACL,EAAAA,8BAAA,gBAAa,QAAb;AACA,EAAAA,8BAAA,kBAAe,QAAf;AACA,EAAAA,8BAAA,cAAW,QAAX;AAHU,SAAAA;AAAA,GAAA;AAML,IAAM,YAAN,cAAwB,MAAM;AAAA,EAGnC,YAAY,MAAc,SAAiB;AACzC,UAAM,OAAO;AACb,SAAK,OAAO;AAAA,EACd;AACF;AAEO,SAAS,YAAY,QAAmB,OAAgB;AAC7D,MAAI,iBAAiB,WAAW;AAC9B,WAAO,MAAM,MAAM,MAAM,MAAM,OAAO;AAAA,EACxC,OAAO;AACL,YAAQ,MAAM,KAAK;AACnB,WAAO,MAAM,IAAI;AAAA,EACnB;AACA,SAAO,UAAU;AACnB;;;ACtBA,eAAsB,cACpB,QACA,UACqB;AACrB,SAAO,IAAI,QAAoB,CAAC,SAAS,WAAW;AAElD,UAAM,UAAU,WAAW,MAAM;AAC/B,aAAO,IAAI,iCAAoC,gBAAgB,CAAC;AAAA,IAClE,GAAG,GAAI;AAEP,UAAM,WAAW,CAAC,YAAoB;AAEpC,mBAAa,OAAO;AACpB,aAAO,IAAI,WAAW,QAAQ;AAE9B,UAAI;AAEF,cAAM,SAAS,SAAS,KAAK,MAAM,OAAO,CAAC;AAC3C,gBAAQ,MAAM;AAAA,MAChB,SAAS,OAAO;AACd,eAAO,IAAI,iCAAoC,gBAAgB,CAAC;AAAA,MAClE;AAAA,IACF;AAGA,WAAO,GAAG,WAAW,QAAQ;AAAA,EAC/B,CAAC;AACH;","names":["CallClientCommands","CallServerCommands","CallErrorCode"]}
|
package/dist/index.mjs
CHANGED
|
@@ -44,6 +44,11 @@ var CallSocket = class {
|
|
|
44
44
|
this.socket.on("close", this.onClose.bind(this));
|
|
45
45
|
this.socket.on("message", this.onMessage.bind(this));
|
|
46
46
|
}
|
|
47
|
+
// Reset conversation
|
|
48
|
+
resetConversation(conversation) {
|
|
49
|
+
this.log("Reset conversation");
|
|
50
|
+
this.conversation = conversation;
|
|
51
|
+
}
|
|
47
52
|
addMessage(message) {
|
|
48
53
|
this.conversation.push(message);
|
|
49
54
|
this.socket.send(
|
|
@@ -121,6 +126,7 @@ var CallSocket = class {
|
|
|
121
126
|
this.socket.close();
|
|
122
127
|
}
|
|
123
128
|
}
|
|
129
|
+
// Add assistant message and send to client with audio (TTS)
|
|
124
130
|
async answer(message) {
|
|
125
131
|
let isEnd = false;
|
|
126
132
|
if (message.includes(END_INTERVIEW)) {
|
|
@@ -132,17 +138,31 @@ var CallSocket = class {
|
|
|
132
138
|
this.addMessage({ role: "assistant", content: message });
|
|
133
139
|
if (!this.config.disableTTS) {
|
|
134
140
|
const audio = await this.config.text2Speech(message);
|
|
135
|
-
|
|
141
|
+
const abort = () => {
|
|
136
142
|
this.log("Answer aborted, removing last assistant message");
|
|
137
143
|
const lastMessage = this.conversation[this.conversation.length - 1];
|
|
138
144
|
if (lastMessage?.role === "assistant") {
|
|
139
145
|
this.conversation.pop();
|
|
140
146
|
this.socket.send("cancelLastAssistantMessage" /* CancelLastAssistantMessage */);
|
|
141
147
|
}
|
|
148
|
+
};
|
|
149
|
+
if (this.abortAnswer) {
|
|
150
|
+
abort();
|
|
142
151
|
return;
|
|
143
152
|
}
|
|
144
|
-
|
|
145
|
-
|
|
153
|
+
if (audio instanceof ArrayBuffer) {
|
|
154
|
+
this.log(`Send audio: (${audio.byteLength} bytes)`);
|
|
155
|
+
this.socket.send(audio);
|
|
156
|
+
} else {
|
|
157
|
+
for await (const chunk of audio) {
|
|
158
|
+
if (this.abortAnswer) {
|
|
159
|
+
abort();
|
|
160
|
+
return;
|
|
161
|
+
}
|
|
162
|
+
this.log(`Send audio: (${chunk.length} bytes)`);
|
|
163
|
+
this.socket.send(chunk);
|
|
164
|
+
}
|
|
165
|
+
}
|
|
146
166
|
}
|
|
147
167
|
}
|
|
148
168
|
if (isEnd) {
|
package/dist/index.mjs.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"sources":["../src/CallSocket.ts","../src/types.ts","../src/errors.ts","../src/waitForParams.ts"],"sourcesContent":["import * as fs from 'fs'\nimport * as path from 'path'\nimport { WebSocket } from 'ws'\nimport {\n CallClientCommands,\n CallConfig,\n CallServerCommands,\n Conversation,\n ConversationMessage,\n} from './types'\n\nexport const END_INTERVIEW = 'END_INTERVIEW'\n\nexport class CallSocket {\n private startTime = Date.now()\n private lastDebug = Date.now()\n\n // An answer can be aborted if user is speaking\n private abortAnswer = false\n\n // When user is speaking, we're waiting to chunks or to stop\n private isSpeaking = false\n\n // Chunks of user speech since user started speaking\n private chunks: Buffer[] = []\n\n // Conversation history\n private conversation: Conversation\n\n constructor(\n public socket: WebSocket,\n public config: CallConfig\n ) {\n this.conversation = [{ role: 'system', content: config.systemPrompt }]\n this.log(`Call started`)\n\n // Assistant speaks first\n\n // LLM: Generate answer\n if (config.firstMessage) {\n this.answer(config.firstMessage)\n } else {\n this.config\n .generateAnswer(this.conversation)\n .then((answer) => this.answer(answer))\n .catch((error) => {\n console.error('[WS]', error)\n this.socket.close()\n })\n }\n\n // Listen to events\n this.socket.on('close', this.onClose.bind(this))\n this.socket.on('message', this.onMessage.bind(this))\n }\n\n private addMessage(message: ConversationMessage) {\n this.conversation.push(message)\n this.socket.send(\n `${\n message.role === 'user'\n ? CallServerCommands.UserMessage\n : CallServerCommands.AssistantMessage\n } ${message.content}`\n )\n this.config.onMessage?.(message)\n }\n\n private onClose() {\n this.log('Connection closed')\n this.abortAnswer = true\n const duration = Math.round((Date.now() - this.startTime) / 1000)\n\n // End call callback\n this.config.onEnd?.({\n conversation: this.conversation.slice(1), // Remove system message\n duration,\n })\n\n // Unset params\n // @ts-ignore\n this.socket = null\n // @ts-ignore\n this.config = null\n }\n\n private async onMessage(message: Buffer) {\n if (!Buffer.isBuffer(message)) {\n console.warn(`[WS] Message is not a buffer`)\n return\n }\n\n // Commands\n if (message.byteLength < 15) {\n const cmd = message.toString()\n this.log(`Command: ${cmd}`)\n\n if (cmd === CallClientCommands.StartSpeaking) {\n // User started speaking\n this.isSpeaking = true\n // Abort answer if there is generation in progress\n this.abortAnswer = true\n } else if (cmd === CallClientCommands.Mute) {\n // User muted the call\n // Abort answer if there is generation in progress\n this.abortAnswer = true\n } else if (cmd === CallClientCommands.StopSpeaking) {\n // User stopped speaking\n this.isSpeaking = false\n await this.onStopSpeaking()\n }\n }\n\n // Audio chunk\n else if (Buffer.isBuffer(message) && this.isSpeaking) {\n this.log(`Received chunk (${message.byteLength} bytes)`)\n this.chunks.push(message)\n }\n }\n\n private async onStopSpeaking() {\n // Do nothing if there is no chunk\n if (this.chunks.length === 0) return\n\n this.abortAnswer = false\n\n // Combine audio blob\n const blob = new Blob(this.chunks, { type: 'audio/ogg' })\n\n // Reset chunks for next user speech\n this.chunks.length = 0\n\n try {\n // Save file to disk\n if (this.config.debugSaveSpeech) {\n const filePath = path.join(__dirname, 'speech.ogg')\n fs.writeFileSync(filePath, Buffer.from(await blob.arrayBuffer()))\n return\n }\n\n // STT: Get transcript and send to client\n const transcript = await this.config.speech2Text(\n blob,\n this.conversation[this.conversation.length - 1]?.content\n )\n if (!transcript) {\n this.log('Ignoring empty transcript')\n return\n }\n\n this.log('User transcript:', transcript)\n\n // Send transcript to client\n this.addMessage({ role: 'user', content: transcript })\n\n if (this.abortAnswer) {\n this.log('Answer aborted, no answer generated')\n return\n }\n\n // LLM: Generate answer\n const answer = await this.config.generateAnswer(this.conversation)\n if (this.abortAnswer) {\n this.log('Answer aborted, ignoring answer')\n return\n }\n\n await this.answer(answer)\n } catch (error) {\n console.error('[WS]', error)\n this.socket.close()\n }\n }\n\n private async answer(message: string) {\n let isEnd = false\n\n // Detect end of interview\n if (message.includes(END_INTERVIEW)) {\n message = message.replace(END_INTERVIEW, '').trim()\n isEnd = true\n }\n\n if (message.length) {\n // Send answer to client\n this.log('Assistant message:', message)\n this.addMessage({ role: 'assistant', content: message })\n\n // TTS: Generate answer audio\n if (!this.config.disableTTS) {\n const audio = await this.config.text2Speech(message)\n if (this.abortAnswer) {\n // Remove last assistant message if aborted\n this.log('Answer aborted, removing last assistant message')\n const lastMessage = this.conversation[this.conversation.length - 1]\n if (lastMessage?.role === 'assistant') {\n this.conversation.pop()\n this.socket.send(CallServerCommands.CancelLastAssistantMessage)\n }\n return\n }\n\n // Send audio to client\n this.log(`Send audio: (${audio.byteLength} bytes)`)\n this.socket.send(audio)\n }\n }\n\n // End of call\n if (isEnd) {\n this.log('Interview ended')\n this.socket.send(CallServerCommands.EndInterview)\n }\n }\n\n private log(...message: any[]) {\n if (!this.config.debugLog) return\n const now = Date.now()\n const delta = now - this.lastDebug\n this.lastDebug = now\n console.log(`[Debug +${delta}ms]`, ...message)\n }\n}\n","export enum CallClientCommands {\n StartSpeaking = 'startSpeaking',\n StopSpeaking = 'stopSpeaking',\n Mute = 'mute',\n}\n\nexport enum CallServerCommands {\n UserMessage = 'userMessage',\n AssistantMessage = 'assistantMessage',\n CancelLastAssistantMessage = 'cancelLastAssistantMessage',\n EndInterview = 'endInterview',\n}\n\nexport interface CallConfig {\n systemPrompt: string\n firstMessage?: string\n debugLog?: boolean\n debugSaveSpeech?: boolean\n disableTTS?: boolean\n generateAnswer(conversation: Conversation): Promise<string>\n speech2Text(blob: Blob, prompt?: string): Promise<string>\n text2Speech(text: string): Promise<ArrayBuffer>\n onMessage?(message: ConversationMessage): void\n onEnd?(call: CallSummary): void\n}\n\nexport interface CallSummary {\n conversation: Conversation\n duration: number\n}\n\nexport type Conversation = ConversationMessage[]\n\nexport interface ConversationMessage {\n role: 'system' | 'user' | 'assistant'\n content: string\n}\n","import WebSocket from 'ws'\n\nexport enum CallErrorCode {\n BadRequest = 4400,\n Unauthorized = 4401,\n NotFound = 4404,\n}\n\nexport class CallError extends Error {\n code: number\n\n constructor(code: number, message: string) {\n super(message)\n this.code = code\n }\n}\n\nexport function handleError(socket: WebSocket, error: unknown) {\n if (error instanceof CallError) {\n socket.close(error.code, error.message)\n } else {\n console.error(error)\n socket.close(1011)\n }\n socket.terminate()\n}\n","import { WebSocket } from 'ws'\nimport { CallError, CallErrorCode } from './errors'\n\nexport async function waitForParams<CallParams>(\n socket: WebSocket,\n validate: (params: any) => CallParams\n): Promise<CallParams> {\n return new Promise<CallParams>((resolve, reject) => {\n // Handle timeout\n const timeout = setTimeout(() => {\n reject(new CallError(CallErrorCode.BadRequest, 'Missing params'))\n }, 3000)\n\n const onParams = (payload: string) => {\n // Clear timeout and listener\n clearTimeout(timeout)\n socket.off('message', onParams)\n\n try {\n // Parse JSON payload\n const params = validate(JSON.parse(payload))\n resolve(params)\n } catch (error) {\n reject(new CallError(CallErrorCode.BadRequest, 'Invalid params'))\n }\n }\n\n // Listen for params\n socket.on('message', onParams)\n })\n}\n"],"mappings":";AAAA,YAAY,QAAQ;AACpB,YAAY,UAAU;;;ACDf,IAAK,qBAAL,kBAAKA,wBAAL;AACL,EAAAA,oBAAA,mBAAgB;AAChB,EAAAA,oBAAA,kBAAe;AACf,EAAAA,oBAAA,UAAO;AAHG,SAAAA;AAAA,GAAA;AAML,IAAK,qBAAL,kBAAKC,wBAAL;AACL,EAAAA,oBAAA,iBAAc;AACd,EAAAA,oBAAA,sBAAmB;AACnB,EAAAA,oBAAA,gCAA6B;AAC7B,EAAAA,oBAAA,kBAAe;AAJL,SAAAA;AAAA,GAAA;;;ADKL,IAAM,gBAAgB;AAEtB,IAAM,aAAN,MAAiB;AAAA,EAgBtB,YACS,QACA,QACP;AAFO;AACA;AAjBT,SAAQ,YAAY,KAAK,IAAI;AAC7B,SAAQ,YAAY,KAAK,IAAI;AAG7B;AAAA,SAAQ,cAAc;AAGtB;AAAA,SAAQ,aAAa;AAGrB;AAAA,SAAQ,SAAmB,CAAC;AAS1B,SAAK,eAAe,CAAC,EAAE,MAAM,UAAU,SAAS,OAAO,aAAa,CAAC;AACrE,SAAK,IAAI,cAAc;AAKvB,QAAI,OAAO,cAAc;AACvB,WAAK,OAAO,OAAO,YAAY;AAAA,IACjC,OAAO;AACL,WAAK,OACF,eAAe,KAAK,YAAY,EAChC,KAAK,CAAC,WAAW,KAAK,OAAO,MAAM,CAAC,EACpC,MAAM,CAAC,UAAU;AAChB,gBAAQ,MAAM,QAAQ,KAAK;AAC3B,aAAK,OAAO,MAAM;AAAA,MACpB,CAAC;AAAA,IACL;AAGA,SAAK,OAAO,GAAG,SAAS,KAAK,QAAQ,KAAK,IAAI,CAAC;AAC/C,SAAK,OAAO,GAAG,WAAW,KAAK,UAAU,KAAK,IAAI,CAAC;AAAA,EACrD;AAAA,EAEQ,WAAW,SAA8B;AAC/C,SAAK,aAAa,KAAK,OAAO;AAC9B,SAAK,OAAO;AAAA,MACV,GACE,QAAQ,SAAS,oFAGnB,IAAI,QAAQ,OAAO;AAAA,IACrB;AACA,SAAK,OAAO,YAAY,OAAO;AAAA,EACjC;AAAA,EAEQ,UAAU;AAChB,SAAK,IAAI,mBAAmB;AAC5B,SAAK,cAAc;AACnB,UAAM,WAAW,KAAK,OAAO,KAAK,IAAI,IAAI,KAAK,aAAa,GAAI;AAGhE,SAAK,OAAO,QAAQ;AAAA,MAClB,cAAc,KAAK,aAAa,MAAM,CAAC;AAAA;AAAA,MACvC;AAAA,IACF,CAAC;AAID,SAAK,SAAS;AAEd,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAc,UAAU,SAAiB;AACvC,QAAI,CAAC,OAAO,SAAS,OAAO,GAAG;AAC7B,cAAQ,KAAK,8BAA8B;AAC3C;AAAA,IACF;AAGA,QAAI,QAAQ,aAAa,IAAI;AAC3B,YAAM,MAAM,QAAQ,SAAS;AAC7B,WAAK,IAAI,YAAY,GAAG,EAAE;AAE1B,UAAI,6CAA0C;AAE5C,aAAK,aAAa;AAElB,aAAK,cAAc;AAAA,MACrB,WAAW,2BAAiC;AAG1C,aAAK,cAAc;AAAA,MACrB,WAAW,2CAAyC;AAElD,aAAK,aAAa;AAClB,cAAM,KAAK,eAAe;AAAA,MAC5B;AAAA,IACF,WAGS,OAAO,SAAS,OAAO,KAAK,KAAK,YAAY;AACpD,WAAK,IAAI,mBAAmB,QAAQ,UAAU,SAAS;AACvD,WAAK,OAAO,KAAK,OAAO;AAAA,IAC1B;AAAA,EACF;AAAA,EAEA,MAAc,iBAAiB;AAE7B,QAAI,KAAK,OAAO,WAAW,EAAG;AAE9B,SAAK,cAAc;AAGnB,UAAM,OAAO,IAAI,KAAK,KAAK,QAAQ,EAAE,MAAM,YAAY,CAAC;AAGxD,SAAK,OAAO,SAAS;AAErB,QAAI;AAEF,UAAI,KAAK,OAAO,iBAAiB;AAC/B,cAAM,WAAgB,UAAK,WAAW,YAAY;AAClD,QAAG,iBAAc,UAAU,OAAO,KAAK,MAAM,KAAK,YAAY,CAAC,CAAC;AAChE;AAAA,MACF;AAGA,YAAM,aAAa,MAAM,KAAK,OAAO;AAAA,QACnC;AAAA,QACA,KAAK,aAAa,KAAK,aAAa,SAAS,CAAC,GAAG;AAAA,MACnD;AACA,UAAI,CAAC,YAAY;AACf,aAAK,IAAI,2BAA2B;AACpC;AAAA,MACF;AAEA,WAAK,IAAI,oBAAoB,UAAU;AAGvC,WAAK,WAAW,EAAE,MAAM,QAAQ,SAAS,WAAW,CAAC;AAErD,UAAI,KAAK,aAAa;AACpB,aAAK,IAAI,qCAAqC;AAC9C;AAAA,MACF;AAGA,YAAM,SAAS,MAAM,KAAK,OAAO,eAAe,KAAK,YAAY;AACjE,UAAI,KAAK,aAAa;AACpB,aAAK,IAAI,iCAAiC;AAC1C;AAAA,MACF;AAEA,YAAM,KAAK,OAAO,MAAM;AAAA,IAC1B,SAAS,OAAO;AACd,cAAQ,MAAM,QAAQ,KAAK;AAC3B,WAAK,OAAO,MAAM;AAAA,IACpB;AAAA,EACF;AAAA,EAEA,MAAc,OAAO,SAAiB;AACpC,QAAI,QAAQ;AAGZ,QAAI,QAAQ,SAAS,aAAa,GAAG;AACnC,gBAAU,QAAQ,QAAQ,eAAe,EAAE,EAAE,KAAK;AAClD,cAAQ;AAAA,IACV;AAEA,QAAI,QAAQ,QAAQ;AAElB,WAAK,IAAI,sBAAsB,OAAO;AACtC,WAAK,WAAW,EAAE,MAAM,aAAa,SAAS,QAAQ,CAAC;AAGvD,UAAI,CAAC,KAAK,OAAO,YAAY;AAC3B,cAAM,QAAQ,MAAM,KAAK,OAAO,YAAY,OAAO;AACnD,YAAI,KAAK,aAAa;AAEpB,eAAK,IAAI,iDAAiD;AAC1D,gBAAM,cAAc,KAAK,aAAa,KAAK,aAAa,SAAS,CAAC;AAClE,cAAI,aAAa,SAAS,aAAa;AACrC,iBAAK,aAAa,IAAI;AACtB,iBAAK,OAAO,kEAAkD;AAAA,UAChE;AACA;AAAA,QACF;AAGA,aAAK,IAAI,gBAAgB,MAAM,UAAU,SAAS;AAClD,aAAK,OAAO,KAAK,KAAK;AAAA,MACxB;AAAA,IACF;AAGA,QAAI,OAAO;AACT,WAAK,IAAI,iBAAiB;AAC1B,WAAK,OAAO,sCAAoC;AAAA,IAClD;AAAA,EACF;AAAA,EAEQ,OAAO,SAAgB;AAC7B,QAAI,CAAC,KAAK,OAAO,SAAU;AAC3B,UAAM,MAAM,KAAK,IAAI;AACrB,UAAM,QAAQ,MAAM,KAAK;AACzB,SAAK,YAAY;AACjB,YAAQ,IAAI,WAAW,KAAK,OAAO,GAAG,OAAO;AAAA,EAC/C;AACF;;;AE5NO,IAAK,gBAAL,kBAAKC,mBAAL;AACL,EAAAA,8BAAA,gBAAa,QAAb;AACA,EAAAA,8BAAA,kBAAe,QAAf;AACA,EAAAA,8BAAA,cAAW,QAAX;AAHU,SAAAA;AAAA,GAAA;AAML,IAAM,YAAN,cAAwB,MAAM;AAAA,EAGnC,YAAY,MAAc,SAAiB;AACzC,UAAM,OAAO;AACb,SAAK,OAAO;AAAA,EACd;AACF;AAEO,SAAS,YAAY,QAAmB,OAAgB;AAC7D,MAAI,iBAAiB,WAAW;AAC9B,WAAO,MAAM,MAAM,MAAM,MAAM,OAAO;AAAA,EACxC,OAAO;AACL,YAAQ,MAAM,KAAK;AACnB,WAAO,MAAM,IAAI;AAAA,EACnB;AACA,SAAO,UAAU;AACnB;;;ACtBA,eAAsB,cACpB,QACA,UACqB;AACrB,SAAO,IAAI,QAAoB,CAAC,SAAS,WAAW;AAElD,UAAM,UAAU,WAAW,MAAM;AAC/B,aAAO,IAAI,iCAAoC,gBAAgB,CAAC;AAAA,IAClE,GAAG,GAAI;AAEP,UAAM,WAAW,CAAC,YAAoB;AAEpC,mBAAa,OAAO;AACpB,aAAO,IAAI,WAAW,QAAQ;AAE9B,UAAI;AAEF,cAAM,SAAS,SAAS,KAAK,MAAM,OAAO,CAAC;AAC3C,gBAAQ,MAAM;AAAA,MAChB,SAAS,OAAO;AACd,eAAO,IAAI,iCAAoC,gBAAgB,CAAC;AAAA,MAClE;AAAA,IACF;AAGA,WAAO,GAAG,WAAW,QAAQ;AAAA,EAC/B,CAAC;AACH;","names":["CallClientCommands","CallServerCommands","CallErrorCode"]}
|
|
1
|
+
{"version":3,"sources":["../src/CallSocket.ts","../src/types.ts","../src/errors.ts","../src/waitForParams.ts"],"sourcesContent":["import * as fs from 'fs'\nimport * as path from 'path'\nimport { WebSocket } from 'ws'\nimport {\n CallClientCommands,\n CallConfig,\n CallServerCommands,\n Conversation,\n ConversationMessage,\n} from './types'\n\nexport const END_INTERVIEW = 'END_INTERVIEW'\n\nexport class CallSocket {\n private startTime = Date.now()\n private lastDebug = Date.now()\n\n // An answer can be aborted if user is speaking\n private abortAnswer = false\n\n // When user is speaking, we're waiting to chunks or to stop\n private isSpeaking = false\n\n // Chunks of user speech since user started speaking\n private chunks: Buffer[] = []\n\n // Conversation history\n private conversation: Conversation\n\n constructor(\n public socket: WebSocket,\n public config: CallConfig\n ) {\n this.conversation = [{ role: 'system', content: config.systemPrompt }]\n this.log(`Call started`)\n\n // Assistant speaks first\n\n // LLM: Generate answer\n if (config.firstMessage) {\n this.answer(config.firstMessage)\n } else {\n this.config\n .generateAnswer(this.conversation)\n .then((answer) => this.answer(answer))\n .catch((error) => {\n console.error('[WS]', error)\n this.socket.close()\n })\n }\n\n // Listen to events\n this.socket.on('close', this.onClose.bind(this))\n this.socket.on('message', this.onMessage.bind(this))\n }\n\n // Reset conversation\n public resetConversation(conversation: Conversation) {\n this.log('Reset conversation')\n this.conversation = conversation\n }\n\n private addMessage(message: ConversationMessage) {\n this.conversation.push(message)\n this.socket.send(\n `${\n message.role === 'user'\n ? CallServerCommands.UserMessage\n : CallServerCommands.AssistantMessage\n } ${message.content}`\n )\n this.config.onMessage?.(message)\n }\n\n private onClose() {\n this.log('Connection closed')\n this.abortAnswer = true\n const duration = Math.round((Date.now() - this.startTime) / 1000)\n\n // End call callback\n this.config.onEnd?.({\n conversation: this.conversation.slice(1), // Remove system message\n duration,\n })\n\n // Unset params\n // @ts-ignore\n this.socket = null\n // @ts-ignore\n this.config = null\n }\n\n private async onMessage(message: Buffer) {\n if (!Buffer.isBuffer(message)) {\n console.warn(`[WS] Message is not a buffer`)\n return\n }\n\n // Commands\n if (message.byteLength < 15) {\n const cmd = message.toString()\n this.log(`Command: ${cmd}`)\n\n if (cmd === CallClientCommands.StartSpeaking) {\n // User started speaking\n this.isSpeaking = true\n // Abort answer if there is generation in progress\n this.abortAnswer = true\n } else if (cmd === CallClientCommands.Mute) {\n // User muted the call\n // Abort answer if there is generation in progress\n this.abortAnswer = true\n } else if (cmd === CallClientCommands.StopSpeaking) {\n // User stopped speaking\n this.isSpeaking = false\n await this.onStopSpeaking()\n }\n }\n\n // Audio chunk\n else if (Buffer.isBuffer(message) && this.isSpeaking) {\n this.log(`Received chunk (${message.byteLength} bytes)`)\n this.chunks.push(message)\n }\n }\n\n private async onStopSpeaking() {\n // Do nothing if there is no chunk\n if (this.chunks.length === 0) return\n\n this.abortAnswer = false\n\n // Combine audio blob\n const blob = new Blob(this.chunks, { type: 'audio/ogg' })\n\n // Reset chunks for next user speech\n this.chunks.length = 0\n\n try {\n // Save file to disk\n if (this.config.debugSaveSpeech) {\n const filePath = path.join(__dirname, 'speech.ogg')\n fs.writeFileSync(filePath, Buffer.from(await blob.arrayBuffer()))\n return\n }\n\n // STT: Get transcript and send to client\n const transcript = await this.config.speech2Text(\n blob,\n this.conversation[this.conversation.length - 1]?.content\n )\n if (!transcript) {\n this.log('Ignoring empty transcript')\n return\n }\n\n this.log('User transcript:', transcript)\n\n // Send transcript to client\n this.addMessage({ role: 'user', content: transcript })\n\n if (this.abortAnswer) {\n this.log('Answer aborted, no answer generated')\n return\n }\n\n // LLM: Generate answer\n const answer = await this.config.generateAnswer(this.conversation)\n if (this.abortAnswer) {\n this.log('Answer aborted, ignoring answer')\n return\n }\n\n await this.answer(answer)\n } catch (error) {\n console.error('[WS]', error)\n this.socket.close()\n }\n }\n\n // Add assistant message and send to client with audio (TTS)\n public async answer(message: string) {\n let isEnd = false\n\n // Detect end of interview\n if (message.includes(END_INTERVIEW)) {\n message = message.replace(END_INTERVIEW, '').trim()\n isEnd = true\n }\n\n if (message.length) {\n // Send answer to client\n this.log('Assistant message:', message)\n this.addMessage({ role: 'assistant', content: message })\n\n // TTS: Generate answer audio\n if (!this.config.disableTTS) {\n const audio = await this.config.text2Speech(message)\n\n // Remove last assistant message if aborted\n const abort = () => {\n this.log('Answer aborted, removing last assistant message')\n const lastMessage = this.conversation[this.conversation.length - 1]\n if (lastMessage?.role === 'assistant') {\n this.conversation.pop()\n this.socket.send(CallServerCommands.CancelLastAssistantMessage)\n }\n }\n\n if (this.abortAnswer) {\n abort()\n return\n }\n\n // Send audio to client\n if (audio instanceof ArrayBuffer) {\n this.log(`Send audio: (${audio.byteLength} bytes)`)\n this.socket.send(audio)\n } else {\n for await (const chunk of audio) {\n if (this.abortAnswer) {\n abort()\n return\n }\n this.log(`Send audio: (${chunk.length} bytes)`)\n this.socket.send(chunk)\n }\n }\n }\n }\n\n // End of call\n if (isEnd) {\n this.log('Interview ended')\n this.socket.send(CallServerCommands.EndInterview)\n }\n }\n\n private log(...message: any[]) {\n if (!this.config.debugLog) return\n const now = Date.now()\n const delta = now - this.lastDebug\n this.lastDebug = now\n console.log(`[Debug +${delta}ms]`, ...message)\n }\n}\n","export enum CallClientCommands {\n StartSpeaking = 'startSpeaking',\n StopSpeaking = 'stopSpeaking',\n Mute = 'mute',\n}\n\nexport enum CallServerCommands {\n UserMessage = 'userMessage',\n AssistantMessage = 'assistantMessage',\n CancelLastAssistantMessage = 'cancelLastAssistantMessage',\n EndInterview = 'endInterview',\n}\n\nexport interface CallConfig {\n systemPrompt: string\n firstMessage?: string\n debugLog?: boolean\n debugSaveSpeech?: boolean\n disableTTS?: boolean\n generateAnswer(conversation: Conversation): Promise<string>\n speech2Text(blob: Blob, prompt?: string): Promise<string>\n text2Speech(text: string): Promise<ArrayBuffer | NodeJS.ReadableStream>\n onMessage?(message: ConversationMessage): void\n onEnd?(call: CallSummary): void\n}\n\nexport interface CallSummary {\n conversation: Conversation\n duration: number\n}\n\nexport type Conversation = ConversationMessage[]\n\nexport interface ConversationMessage {\n role: 'system' | 'user' | 'assistant'\n content: string\n}\n","import WebSocket from 'ws'\n\nexport enum CallErrorCode {\n BadRequest = 4400,\n Unauthorized = 4401,\n NotFound = 4404,\n}\n\nexport class CallError extends Error {\n code: number\n\n constructor(code: number, message: string) {\n super(message)\n this.code = code\n }\n}\n\nexport function handleError(socket: WebSocket, error: unknown) {\n if (error instanceof CallError) {\n socket.close(error.code, error.message)\n } else {\n console.error(error)\n socket.close(1011)\n }\n socket.terminate()\n}\n","import { WebSocket } from 'ws'\nimport { CallError, CallErrorCode } from './errors'\n\nexport async function waitForParams<CallParams>(\n socket: WebSocket,\n validate: (params: any) => CallParams\n): Promise<CallParams> {\n return new Promise<CallParams>((resolve, reject) => {\n // Handle timeout\n const timeout = setTimeout(() => {\n reject(new CallError(CallErrorCode.BadRequest, 'Missing params'))\n }, 3000)\n\n const onParams = (payload: string) => {\n // Clear timeout and listener\n clearTimeout(timeout)\n socket.off('message', onParams)\n\n try {\n // Parse JSON payload\n const params = validate(JSON.parse(payload))\n resolve(params)\n } catch (error) {\n reject(new CallError(CallErrorCode.BadRequest, 'Invalid params'))\n }\n }\n\n // Listen for params\n socket.on('message', onParams)\n })\n}\n"],"mappings":";AAAA,YAAY,QAAQ;AACpB,YAAY,UAAU;;;ACDf,IAAK,qBAAL,kBAAKA,wBAAL;AACL,EAAAA,oBAAA,mBAAgB;AAChB,EAAAA,oBAAA,kBAAe;AACf,EAAAA,oBAAA,UAAO;AAHG,SAAAA;AAAA,GAAA;AAML,IAAK,qBAAL,kBAAKC,wBAAL;AACL,EAAAA,oBAAA,iBAAc;AACd,EAAAA,oBAAA,sBAAmB;AACnB,EAAAA,oBAAA,gCAA6B;AAC7B,EAAAA,oBAAA,kBAAe;AAJL,SAAAA;AAAA,GAAA;;;ADKL,IAAM,gBAAgB;AAEtB,IAAM,aAAN,MAAiB;AAAA,EAgBtB,YACS,QACA,QACP;AAFO;AACA;AAjBT,SAAQ,YAAY,KAAK,IAAI;AAC7B,SAAQ,YAAY,KAAK,IAAI;AAG7B;AAAA,SAAQ,cAAc;AAGtB;AAAA,SAAQ,aAAa;AAGrB;AAAA,SAAQ,SAAmB,CAAC;AAS1B,SAAK,eAAe,CAAC,EAAE,MAAM,UAAU,SAAS,OAAO,aAAa,CAAC;AACrE,SAAK,IAAI,cAAc;AAKvB,QAAI,OAAO,cAAc;AACvB,WAAK,OAAO,OAAO,YAAY;AAAA,IACjC,OAAO;AACL,WAAK,OACF,eAAe,KAAK,YAAY,EAChC,KAAK,CAAC,WAAW,KAAK,OAAO,MAAM,CAAC,EACpC,MAAM,CAAC,UAAU;AAChB,gBAAQ,MAAM,QAAQ,KAAK;AAC3B,aAAK,OAAO,MAAM;AAAA,MACpB,CAAC;AAAA,IACL;AAGA,SAAK,OAAO,GAAG,SAAS,KAAK,QAAQ,KAAK,IAAI,CAAC;AAC/C,SAAK,OAAO,GAAG,WAAW,KAAK,UAAU,KAAK,IAAI,CAAC;AAAA,EACrD;AAAA;AAAA,EAGO,kBAAkB,cAA4B;AACnD,SAAK,IAAI,oBAAoB;AAC7B,SAAK,eAAe;AAAA,EACtB;AAAA,EAEQ,WAAW,SAA8B;AAC/C,SAAK,aAAa,KAAK,OAAO;AAC9B,SAAK,OAAO;AAAA,MACV,GACE,QAAQ,SAAS,oFAGnB,IAAI,QAAQ,OAAO;AAAA,IACrB;AACA,SAAK,OAAO,YAAY,OAAO;AAAA,EACjC;AAAA,EAEQ,UAAU;AAChB,SAAK,IAAI,mBAAmB;AAC5B,SAAK,cAAc;AACnB,UAAM,WAAW,KAAK,OAAO,KAAK,IAAI,IAAI,KAAK,aAAa,GAAI;AAGhE,SAAK,OAAO,QAAQ;AAAA,MAClB,cAAc,KAAK,aAAa,MAAM,CAAC;AAAA;AAAA,MACvC;AAAA,IACF,CAAC;AAID,SAAK,SAAS;AAEd,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAc,UAAU,SAAiB;AACvC,QAAI,CAAC,OAAO,SAAS,OAAO,GAAG;AAC7B,cAAQ,KAAK,8BAA8B;AAC3C;AAAA,IACF;AAGA,QAAI,QAAQ,aAAa,IAAI;AAC3B,YAAM,MAAM,QAAQ,SAAS;AAC7B,WAAK,IAAI,YAAY,GAAG,EAAE;AAE1B,UAAI,6CAA0C;AAE5C,aAAK,aAAa;AAElB,aAAK,cAAc;AAAA,MACrB,WAAW,2BAAiC;AAG1C,aAAK,cAAc;AAAA,MACrB,WAAW,2CAAyC;AAElD,aAAK,aAAa;AAClB,cAAM,KAAK,eAAe;AAAA,MAC5B;AAAA,IACF,WAGS,OAAO,SAAS,OAAO,KAAK,KAAK,YAAY;AACpD,WAAK,IAAI,mBAAmB,QAAQ,UAAU,SAAS;AACvD,WAAK,OAAO,KAAK,OAAO;AAAA,IAC1B;AAAA,EACF;AAAA,EAEA,MAAc,iBAAiB;AAE7B,QAAI,KAAK,OAAO,WAAW,EAAG;AAE9B,SAAK,cAAc;AAGnB,UAAM,OAAO,IAAI,KAAK,KAAK,QAAQ,EAAE,MAAM,YAAY,CAAC;AAGxD,SAAK,OAAO,SAAS;AAErB,QAAI;AAEF,UAAI,KAAK,OAAO,iBAAiB;AAC/B,cAAM,WAAgB,UAAK,WAAW,YAAY;AAClD,QAAG,iBAAc,UAAU,OAAO,KAAK,MAAM,KAAK,YAAY,CAAC,CAAC;AAChE;AAAA,MACF;AAGA,YAAM,aAAa,MAAM,KAAK,OAAO;AAAA,QACnC;AAAA,QACA,KAAK,aAAa,KAAK,aAAa,SAAS,CAAC,GAAG;AAAA,MACnD;AACA,UAAI,CAAC,YAAY;AACf,aAAK,IAAI,2BAA2B;AACpC;AAAA,MACF;AAEA,WAAK,IAAI,oBAAoB,UAAU;AAGvC,WAAK,WAAW,EAAE,MAAM,QAAQ,SAAS,WAAW,CAAC;AAErD,UAAI,KAAK,aAAa;AACpB,aAAK,IAAI,qCAAqC;AAC9C;AAAA,MACF;AAGA,YAAM,SAAS,MAAM,KAAK,OAAO,eAAe,KAAK,YAAY;AACjE,UAAI,KAAK,aAAa;AACpB,aAAK,IAAI,iCAAiC;AAC1C;AAAA,MACF;AAEA,YAAM,KAAK,OAAO,MAAM;AAAA,IAC1B,SAAS,OAAO;AACd,cAAQ,MAAM,QAAQ,KAAK;AAC3B,WAAK,OAAO,MAAM;AAAA,IACpB;AAAA,EACF;AAAA;AAAA,EAGA,MAAa,OAAO,SAAiB;AACnC,QAAI,QAAQ;AAGZ,QAAI,QAAQ,SAAS,aAAa,GAAG;AACnC,gBAAU,QAAQ,QAAQ,eAAe,EAAE,EAAE,KAAK;AAClD,cAAQ;AAAA,IACV;AAEA,QAAI,QAAQ,QAAQ;AAElB,WAAK,IAAI,sBAAsB,OAAO;AACtC,WAAK,WAAW,EAAE,MAAM,aAAa,SAAS,QAAQ,CAAC;AAGvD,UAAI,CAAC,KAAK,OAAO,YAAY;AAC3B,cAAM,QAAQ,MAAM,KAAK,OAAO,YAAY,OAAO;AAGnD,cAAM,QAAQ,MAAM;AAClB,eAAK,IAAI,iDAAiD;AAC1D,gBAAM,cAAc,KAAK,aAAa,KAAK,aAAa,SAAS,CAAC;AAClE,cAAI,aAAa,SAAS,aAAa;AACrC,iBAAK,aAAa,IAAI;AACtB,iBAAK,OAAO,kEAAkD;AAAA,UAChE;AAAA,QACF;AAEA,YAAI,KAAK,aAAa;AACpB,gBAAM;AACN;AAAA,QACF;AAGA,YAAI,iBAAiB,aAAa;AAChC,eAAK,IAAI,gBAAgB,MAAM,UAAU,SAAS;AAClD,eAAK,OAAO,KAAK,KAAK;AAAA,QACxB,OAAO;AACL,2BAAiB,SAAS,OAAO;AAC/B,gBAAI,KAAK,aAAa;AACpB,oBAAM;AACN;AAAA,YACF;AACA,iBAAK,IAAI,gBAAgB,MAAM,MAAM,SAAS;AAC9C,iBAAK,OAAO,KAAK,KAAK;AAAA,UACxB;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAGA,QAAI,OAAO;AACT,WAAK,IAAI,iBAAiB;AAC1B,WAAK,OAAO,sCAAoC;AAAA,IAClD;AAAA,EACF;AAAA,EAEQ,OAAO,SAAgB;AAC7B,QAAI,CAAC,KAAK,OAAO,SAAU;AAC3B,UAAM,MAAM,KAAK,IAAI;AACrB,UAAM,QAAQ,MAAM,KAAK;AACzB,SAAK,YAAY;AACjB,YAAQ,IAAI,WAAW,KAAK,OAAO,GAAG,OAAO;AAAA,EAC/C;AACF;;;AEnPO,IAAK,gBAAL,kBAAKC,mBAAL;AACL,EAAAA,8BAAA,gBAAa,QAAb;AACA,EAAAA,8BAAA,kBAAe,QAAf;AACA,EAAAA,8BAAA,cAAW,QAAX;AAHU,SAAAA;AAAA,GAAA;AAML,IAAM,YAAN,cAAwB,MAAM;AAAA,EAGnC,YAAY,MAAc,SAAiB;AACzC,UAAM,OAAO;AACb,SAAK,OAAO;AAAA,EACd;AACF;AAEO,SAAS,YAAY,QAAmB,OAAgB;AAC7D,MAAI,iBAAiB,WAAW;AAC9B,WAAO,MAAM,MAAM,MAAM,MAAM,OAAO;AAAA,EACxC,OAAO;AACL,YAAQ,MAAM,KAAK;AACnB,WAAO,MAAM,IAAI;AAAA,EACnB;AACA,SAAO,UAAU;AACnB;;;ACtBA,eAAsB,cACpB,QACA,UACqB;AACrB,SAAO,IAAI,QAAoB,CAAC,SAAS,WAAW;AAElD,UAAM,UAAU,WAAW,MAAM;AAC/B,aAAO,IAAI,iCAAoC,gBAAgB,CAAC;AAAA,IAClE,GAAG,GAAI;AAEP,UAAM,WAAW,CAAC,YAAoB;AAEpC,mBAAa,OAAO;AACpB,aAAO,IAAI,WAAW,QAAQ;AAE9B,UAAI;AAEF,cAAM,SAAS,SAAS,KAAK,MAAM,OAAO,CAAC;AAC3C,gBAAQ,MAAM;AAAA,MAChB,SAAS,OAAO;AACd,eAAO,IAAI,iCAAoC,gBAAgB,CAAC;AAAA,MAClE;AAAA,IACF;AAGA,WAAO,GAAG,WAAW,QAAQ;AAAA,EAC/B,CAAC;AACH;","names":["CallClientCommands","CallServerCommands","CallErrorCode"]}
|