@micdrop/server 1.2.1 → 1.2.2

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 CHANGED
@@ -144,10 +144,12 @@ interface CallConfig {
144
144
  disableTTS?: boolean
145
145
 
146
146
  // Generate assistant's response
147
- generateAnswer(conversation: Conversation): Promise<string>
147
+ generateAnswer(
148
+ conversation: Conversation
149
+ ): Promise<string | ConversationMessage>
148
150
 
149
151
  // Convert audio to text
150
- speech2Text(blob: Blob, prevMessage?: string): Promise<string>
152
+ speech2Text(audioBlob: Blob, prevMessage?: string): Promise<string>
151
153
 
152
154
  // Convert text to audio
153
155
  // Can return either a complete ArrayBuffer or a ReadableStream for streaming
@@ -165,6 +167,7 @@ interface CallConfig {
165
167
  interface ConversationMessage {
166
168
  role: 'system' | 'user' | 'assistant'
167
169
  content: string
170
+ metadata?: any
168
171
  }
169
172
 
170
173
  type Conversation = ConversationMessage[]
@@ -193,8 +196,7 @@ enum CallClientCommands {
193
196
 
194
197
  ```typescript
195
198
  enum CallServerCommands {
196
- UserMessage = 'userMessage',
197
- AssistantMessage = 'assistantMessage',
199
+ Message = 'message',
198
200
  CancelLastAssistantMessage = 'cancelLastAssistantMessage',
199
201
  EndInterview = 'endInterview',
200
202
  }
package/dist/index.d.mts CHANGED
@@ -6,8 +6,7 @@ declare enum CallClientCommands {
6
6
  Mute = "mute"
7
7
  }
8
8
  declare enum CallServerCommands {
9
- UserMessage = "userMessage",
10
- AssistantMessage = "assistantMessage",
9
+ Message = "message",
11
10
  CancelLastAssistantMessage = "cancelLastAssistantMessage",
12
11
  EndInterview = "endInterview"
13
12
  }
@@ -17,8 +16,8 @@ interface CallConfig {
17
16
  debugLog?: boolean;
18
17
  debugSaveSpeech?: boolean;
19
18
  disableTTS?: boolean;
20
- generateAnswer(conversation: Conversation): Promise<string>;
21
- speech2Text(blob: Blob, prevMessage?: string): Promise<string>;
19
+ generateAnswer(conversation: Conversation): Promise<string | ConversationMessage>;
20
+ speech2Text(audioBlob: Blob, prevMessage?: string): Promise<string>;
22
21
  text2Speech(text: string): Promise<ArrayBuffer | NodeJS.ReadableStream>;
23
22
  onMessage?(message: ConversationMessage): void;
24
23
  onEnd?(call: CallSummary): void;
@@ -28,9 +27,10 @@ interface CallSummary {
28
27
  duration: number;
29
28
  }
30
29
  type Conversation = ConversationMessage[];
31
- interface ConversationMessage {
30
+ interface ConversationMessage<Data = any> {
32
31
  role: 'system' | 'user' | 'assistant';
33
32
  content: string;
33
+ metadata?: Data;
34
34
  }
35
35
 
36
36
  declare const END_INTERVIEW = "END_INTERVIEW";
@@ -50,7 +50,7 @@ declare class CallSocket {
50
50
  private onClose;
51
51
  private onMessage;
52
52
  private onStopSpeaking;
53
- answer(message: string): Promise<void>;
53
+ answer(message: string | ConversationMessage): Promise<void>;
54
54
  private log;
55
55
  }
56
56
 
package/dist/index.d.ts CHANGED
@@ -6,8 +6,7 @@ declare enum CallClientCommands {
6
6
  Mute = "mute"
7
7
  }
8
8
  declare enum CallServerCommands {
9
- UserMessage = "userMessage",
10
- AssistantMessage = "assistantMessage",
9
+ Message = "message",
11
10
  CancelLastAssistantMessage = "cancelLastAssistantMessage",
12
11
  EndInterview = "endInterview"
13
12
  }
@@ -17,8 +16,8 @@ interface CallConfig {
17
16
  debugLog?: boolean;
18
17
  debugSaveSpeech?: boolean;
19
18
  disableTTS?: boolean;
20
- generateAnswer(conversation: Conversation): Promise<string>;
21
- speech2Text(blob: Blob, prevMessage?: string): Promise<string>;
19
+ generateAnswer(conversation: Conversation): Promise<string | ConversationMessage>;
20
+ speech2Text(audioBlob: Blob, prevMessage?: string): Promise<string>;
22
21
  text2Speech(text: string): Promise<ArrayBuffer | NodeJS.ReadableStream>;
23
22
  onMessage?(message: ConversationMessage): void;
24
23
  onEnd?(call: CallSummary): void;
@@ -28,9 +27,10 @@ interface CallSummary {
28
27
  duration: number;
29
28
  }
30
29
  type Conversation = ConversationMessage[];
31
- interface ConversationMessage {
30
+ interface ConversationMessage<Data = any> {
32
31
  role: 'system' | 'user' | 'assistant';
33
32
  content: string;
33
+ metadata?: Data;
34
34
  }
35
35
 
36
36
  declare const END_INTERVIEW = "END_INTERVIEW";
@@ -50,7 +50,7 @@ declare class CallSocket {
50
50
  private onClose;
51
51
  private onMessage;
52
52
  private onStopSpeaking;
53
- answer(message: string): Promise<void>;
53
+ answer(message: string | ConversationMessage): Promise<void>;
54
54
  private log;
55
55
  }
56
56
 
package/dist/index.js CHANGED
@@ -52,8 +52,7 @@ var CallClientCommands = /* @__PURE__ */ ((CallClientCommands2) => {
52
52
  return CallClientCommands2;
53
53
  })(CallClientCommands || {});
54
54
  var CallServerCommands = /* @__PURE__ */ ((CallServerCommands2) => {
55
- CallServerCommands2["UserMessage"] = "userMessage";
56
- CallServerCommands2["AssistantMessage"] = "assistantMessage";
55
+ CallServerCommands2["Message"] = "message";
57
56
  CallServerCommands2["CancelLastAssistantMessage"] = "cancelLastAssistantMessage";
58
57
  CallServerCommands2["EndInterview"] = "endInterview";
59
58
  return CallServerCommands2;
@@ -96,9 +95,7 @@ var CallSocket = class {
96
95
  addMessage(message) {
97
96
  if (!this.socket || !this.config) return;
98
97
  this.conversation.push(message);
99
- this.socket.send(
100
- `${message.role === "user" ? "userMessage" /* UserMessage */ : "assistantMessage" /* AssistantMessage */} ${message.content}`
101
- );
98
+ this.socket.send(`${"message" /* Message */} ${JSON.stringify(message)}`);
102
99
  this.config.onMessage?.(message);
103
100
  }
104
101
  async sendAudio(audio, abort) {
@@ -198,16 +195,18 @@ var CallSocket = class {
198
195
  async answer(message) {
199
196
  if (!this.socket || !this.config) return;
200
197
  let isEnd = false;
201
- if (message.includes(END_INTERVIEW)) {
202
- message = message.replace(END_INTERVIEW, "").trim();
198
+ let content = typeof message === "string" ? message : message.content;
199
+ const metadata = typeof message === "string" ? void 0 : message.metadata;
200
+ if (content.includes(END_INTERVIEW)) {
201
+ content = content.replace(END_INTERVIEW, "").trim();
203
202
  isEnd = true;
204
203
  }
205
- if (message.length) {
204
+ if (content.length) {
206
205
  this.log("Assistant message:", message);
207
- this.addMessage({ role: "assistant", content: message });
206
+ this.addMessage({ role: "assistant", content, metadata });
208
207
  if (!this.config.disableTTS) {
209
208
  try {
210
- const audio = await this.config.text2Speech(message);
209
+ const audio = await this.config.text2Speech(content);
211
210
  const abort = () => {
212
211
  this.log("Answer aborted, removing last assistant message");
213
212
  const lastMessage = this.conversation[this.conversation.length - 1];
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 { 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 public socket: WebSocket | null = null\n public config: CallConfig | null = null\n\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(socket: WebSocket, config: CallConfig) {\n this.socket = socket\n this.config = config\n this.conversation = [{ role: 'system', content: config.systemPrompt }]\n this.log(`Call started`)\n\n // Assistant speaks first\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('[CallSocket]', error)\n socket?.close()\n // TODO: Implement retry\n })\n }\n\n // Listen to events\n socket.on('close', this.onClose.bind(this))\n 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 if (!this.socket || !this.config) return\n this.conversation.push(message)\n this.socket.send(\n `${\n message.role === 'user'\n ? CallServerCommands.UserMessage\n : CallServerCommands.AssistantMessage\n } ${message.content}`\n )\n this.config.onMessage?.(message)\n }\n\n private async sendAudio(\n audio: ArrayBuffer | NodeJS.ReadableStream,\n abort?: () => void\n ) {\n if (!this.socket) return\n if (this.abortAnswer) {\n abort?.()\n return\n }\n\n if (Buffer.isBuffer(audio) || audio instanceof ArrayBuffer) {\n // Whole audio as a single buffer\n this.log(`Send audio: (${audio.byteLength} bytes)`)\n this.socket.send(audio)\n } else if ('paused' in audio) {\n // Audio as a stream\n for await (const chunk of audio) {\n if (this.abortAnswer) {\n abort?.()\n return\n }\n this.log(`Send audio chunk (${chunk.length} bytes)`)\n this.socket.send(chunk)\n }\n } else {\n this.log(`Unknown audio type: ${audio}`)\n }\n }\n\n private onClose() {\n if (!this.config) return\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 this.socket = null\n this.config = null\n }\n\n private async onMessage(message: Buffer) {\n if (!Buffer.isBuffer(message)) {\n console.warn(`[CallSocket] 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 if (!this.config) return\n\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 filename = `speech-${Date.now()}.ogg`\n fs.writeFileSync(filename, Buffer.from(await blob.arrayBuffer()))\n this.log(`Saved speech: ${filename}`)\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('[CallSocket]', error)\n // TODO: Implement retry\n }\n }\n\n // Add assistant message and send to client with audio (TTS)\n public async answer(message: string) {\n if (!this.socket || !this.config) return\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 try {\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 // Send audio to client\n await this.sendAudio(audio, abort)\n } catch (error) {\n console.error('[CallSocket]', error)\n // TODO: Implement retry\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, prevMessage?: 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;;;ACAb,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;;;ADIL,IAAM,gBAAgB;AAEtB,IAAM,aAAN,MAAiB;AAAA,EAmBtB,YAAY,QAAmB,QAAoB;AAlBnD,SAAO,SAA2B;AAClC,SAAO,SAA4B;AAEnC,SAAQ,YAAY,KAAK,IAAI;AAC7B,SAAQ,YAAY,KAAK,IAAI;AAG7B;AAAA,SAAQ,cAAc;AAGtB;AAAA,SAAQ,aAAa;AAGrB;AAAA,SAAQ,SAAmB,CAAC;AAM1B,SAAK,SAAS;AACd,SAAK,SAAS;AACd,SAAK,eAAe,CAAC,EAAE,MAAM,UAAU,SAAS,OAAO,aAAa,CAAC;AACrE,SAAK,IAAI,cAAc;AAGvB,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,gBAAgB,KAAK;AACnC,gBAAQ,MAAM;AAAA,MAEhB,CAAC;AAAA,IACL;AAGA,WAAO,GAAG,SAAS,KAAK,QAAQ,KAAK,IAAI,CAAC;AAC1C,WAAO,GAAG,WAAW,KAAK,UAAU,KAAK,IAAI,CAAC;AAAA,EAChD;AAAA;AAAA,EAGO,kBAAkB,cAA4B;AACnD,SAAK,IAAI,oBAAoB;AAC7B,SAAK,eAAe;AAAA,EACtB;AAAA,EAEQ,WAAW,SAA8B;AAC/C,QAAI,CAAC,KAAK,UAAU,CAAC,KAAK,OAAQ;AAClC,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,EAEA,MAAc,UACZ,OACA,OACA;AACA,QAAI,CAAC,KAAK,OAAQ;AAClB,QAAI,KAAK,aAAa;AACpB,cAAQ;AACR;AAAA,IACF;AAEA,QAAI,OAAO,SAAS,KAAK,KAAK,iBAAiB,aAAa;AAE1D,WAAK,IAAI,gBAAgB,MAAM,UAAU,SAAS;AAClD,WAAK,OAAO,KAAK,KAAK;AAAA,IACxB,WAAW,YAAY,OAAO;AAE5B,uBAAiB,SAAS,OAAO;AAC/B,YAAI,KAAK,aAAa;AACpB,kBAAQ;AACR;AAAA,QACF;AACA,aAAK,IAAI,qBAAqB,MAAM,MAAM,SAAS;AACnD,aAAK,OAAO,KAAK,KAAK;AAAA,MACxB;AAAA,IACF,OAAO;AACL,WAAK,IAAI,uBAAuB,KAAK,EAAE;AAAA,IACzC;AAAA,EACF;AAAA,EAEQ,UAAU;AAChB,QAAI,CAAC,KAAK,OAAQ;AAClB,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;AAGD,SAAK,SAAS;AACd,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAc,UAAU,SAAiB;AACvC,QAAI,CAAC,OAAO,SAAS,OAAO,GAAG;AAC7B,cAAQ,KAAK,sCAAsC;AACnD;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;AAC7B,QAAI,CAAC,KAAK,OAAQ;AAGlB,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,WAAW,UAAU,KAAK,IAAI,CAAC;AACrC,QAAG,iBAAc,UAAU,OAAO,KAAK,MAAM,KAAK,YAAY,CAAC,CAAC;AAChE,aAAK,IAAI,iBAAiB,QAAQ,EAAE;AAAA,MACtC;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,gBAAgB,KAAK;AAAA,IAErC;AAAA,EACF;AAAA;AAAA,EAGA,MAAa,OAAO,SAAiB;AACnC,QAAI,CAAC,KAAK,UAAU,CAAC,KAAK,OAAQ;AAClC,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,YAAI;AACF,gBAAM,QAAQ,MAAM,KAAK,OAAO,YAAY,OAAO;AAGnD,gBAAM,QAAQ,MAAM;AAClB,iBAAK,IAAI,iDAAiD;AAC1D,kBAAM,cAAc,KAAK,aAAa,KAAK,aAAa,SAAS,CAAC;AAClE,gBAAI,aAAa,SAAS,aAAa;AACrC,mBAAK,aAAa,IAAI;AACtB,mBAAK,QAAQ,kEAAkD;AAAA,YACjE;AAAA,UACF;AAGA,gBAAM,KAAK,UAAU,OAAO,KAAK;AAAA,QACnC,SAAS,OAAO;AACd,kBAAQ,MAAM,gBAAgB,KAAK;AAAA,QAErC;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,QAAQ,SAAU;AAC5B,UAAM,MAAM,KAAK,IAAI;AACrB,UAAM,QAAQ,MAAM,KAAK;AACzB,SAAK,YAAY;AACjB,YAAQ,IAAI,WAAW,KAAK,OAAO,GAAG,OAAO;AAAA,EAC/C;AACF;;;AEvQO,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 { 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 public socket: WebSocket | null = null\n public config: CallConfig | null = null\n\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(socket: WebSocket, config: CallConfig) {\n this.socket = socket\n this.config = config\n this.conversation = [{ role: 'system', content: config.systemPrompt }]\n this.log(`Call started`)\n\n // Assistant speaks first\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('[CallSocket]', error)\n socket?.close()\n // TODO: Implement retry\n })\n }\n\n // Listen to events\n socket.on('close', this.onClose.bind(this))\n 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 if (!this.socket || !this.config) return\n this.conversation.push(message)\n this.socket.send(`${CallServerCommands.Message} ${JSON.stringify(message)}`)\n this.config.onMessage?.(message)\n }\n\n private async sendAudio(\n audio: ArrayBuffer | NodeJS.ReadableStream,\n abort?: () => void\n ) {\n if (!this.socket) return\n if (this.abortAnswer) {\n abort?.()\n return\n }\n\n if (Buffer.isBuffer(audio) || audio instanceof ArrayBuffer) {\n // Whole audio as a single buffer\n this.log(`Send audio: (${audio.byteLength} bytes)`)\n this.socket.send(audio)\n } else if ('paused' in audio) {\n // Audio as a stream\n for await (const chunk of audio) {\n if (this.abortAnswer) {\n abort?.()\n return\n }\n this.log(`Send audio chunk (${chunk.length} bytes)`)\n this.socket.send(chunk)\n }\n } else {\n this.log(`Unknown audio type: ${audio}`)\n }\n }\n\n private onClose() {\n if (!this.config) return\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 this.socket = null\n this.config = null\n }\n\n private async onMessage(message: Buffer) {\n if (!Buffer.isBuffer(message)) {\n console.warn(`[CallSocket] 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 if (!this.config) return\n\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 filename = `speech-${Date.now()}.ogg`\n fs.writeFileSync(filename, Buffer.from(await blob.arrayBuffer()))\n this.log(`Saved speech: ${filename}`)\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('[CallSocket]', error)\n // TODO: Implement retry\n }\n }\n\n // Add assistant message and send to client with audio (TTS)\n public async answer(message: string | ConversationMessage) {\n if (!this.socket || !this.config) return\n let isEnd = false\n\n let content = typeof message === 'string' ? message : message.content\n const metadata = typeof message === 'string' ? undefined : message.metadata\n\n // Detect end of interview\n if (content.includes(END_INTERVIEW)) {\n content = content.replace(END_INTERVIEW, '').trim()\n isEnd = true\n }\n\n if (content.length) {\n // Send answer to client\n this.log('Assistant message:', message)\n this.addMessage({ role: 'assistant', content, metadata })\n\n // TTS: Generate answer audio\n if (!this.config.disableTTS) {\n try {\n const audio = await this.config.text2Speech(content)\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 // Send audio to client\n await this.sendAudio(audio, abort)\n } catch (error) {\n console.error('[CallSocket]', error)\n // TODO: Implement retry\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 Message = 'message',\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(\n conversation: Conversation\n ): Promise<string | ConversationMessage>\n speech2Text(audioBlob: Blob, prevMessage?: 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<Data = any> {\n role: 'system' | 'user' | 'assistant'\n content: string\n metadata?: Data\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;;;ACAb,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,aAAU;AACV,EAAAA,oBAAA,gCAA6B;AAC7B,EAAAA,oBAAA,kBAAe;AAHL,SAAAA;AAAA,GAAA;;;ADIL,IAAM,gBAAgB;AAEtB,IAAM,aAAN,MAAiB;AAAA,EAmBtB,YAAY,QAAmB,QAAoB;AAlBnD,SAAO,SAA2B;AAClC,SAAO,SAA4B;AAEnC,SAAQ,YAAY,KAAK,IAAI;AAC7B,SAAQ,YAAY,KAAK,IAAI;AAG7B;AAAA,SAAQ,cAAc;AAGtB;AAAA,SAAQ,aAAa;AAGrB;AAAA,SAAQ,SAAmB,CAAC;AAM1B,SAAK,SAAS;AACd,SAAK,SAAS;AACd,SAAK,eAAe,CAAC,EAAE,MAAM,UAAU,SAAS,OAAO,aAAa,CAAC;AACrE,SAAK,IAAI,cAAc;AAGvB,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,gBAAgB,KAAK;AACnC,gBAAQ,MAAM;AAAA,MAEhB,CAAC;AAAA,IACL;AAGA,WAAO,GAAG,SAAS,KAAK,QAAQ,KAAK,IAAI,CAAC;AAC1C,WAAO,GAAG,WAAW,KAAK,UAAU,KAAK,IAAI,CAAC;AAAA,EAChD;AAAA;AAAA,EAGO,kBAAkB,cAA4B;AACnD,SAAK,IAAI,oBAAoB;AAC7B,SAAK,eAAe;AAAA,EACtB;AAAA,EAEQ,WAAW,SAA8B;AAC/C,QAAI,CAAC,KAAK,UAAU,CAAC,KAAK,OAAQ;AAClC,SAAK,aAAa,KAAK,OAAO;AAC9B,SAAK,OAAO,KAAK,0BAA6B,IAAI,KAAK,UAAU,OAAO,CAAC,EAAE;AAC3E,SAAK,OAAO,YAAY,OAAO;AAAA,EACjC;AAAA,EAEA,MAAc,UACZ,OACA,OACA;AACA,QAAI,CAAC,KAAK,OAAQ;AAClB,QAAI,KAAK,aAAa;AACpB,cAAQ;AACR;AAAA,IACF;AAEA,QAAI,OAAO,SAAS,KAAK,KAAK,iBAAiB,aAAa;AAE1D,WAAK,IAAI,gBAAgB,MAAM,UAAU,SAAS;AAClD,WAAK,OAAO,KAAK,KAAK;AAAA,IACxB,WAAW,YAAY,OAAO;AAE5B,uBAAiB,SAAS,OAAO;AAC/B,YAAI,KAAK,aAAa;AACpB,kBAAQ;AACR;AAAA,QACF;AACA,aAAK,IAAI,qBAAqB,MAAM,MAAM,SAAS;AACnD,aAAK,OAAO,KAAK,KAAK;AAAA,MACxB;AAAA,IACF,OAAO;AACL,WAAK,IAAI,uBAAuB,KAAK,EAAE;AAAA,IACzC;AAAA,EACF;AAAA,EAEQ,UAAU;AAChB,QAAI,CAAC,KAAK,OAAQ;AAClB,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;AAGD,SAAK,SAAS;AACd,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAc,UAAU,SAAiB;AACvC,QAAI,CAAC,OAAO,SAAS,OAAO,GAAG;AAC7B,cAAQ,KAAK,sCAAsC;AACnD;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;AAC7B,QAAI,CAAC,KAAK,OAAQ;AAGlB,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,WAAW,UAAU,KAAK,IAAI,CAAC;AACrC,QAAG,iBAAc,UAAU,OAAO,KAAK,MAAM,KAAK,YAAY,CAAC,CAAC;AAChE,aAAK,IAAI,iBAAiB,QAAQ,EAAE;AAAA,MACtC;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,gBAAgB,KAAK;AAAA,IAErC;AAAA,EACF;AAAA;AAAA,EAGA,MAAa,OAAO,SAAuC;AACzD,QAAI,CAAC,KAAK,UAAU,CAAC,KAAK,OAAQ;AAClC,QAAI,QAAQ;AAEZ,QAAI,UAAU,OAAO,YAAY,WAAW,UAAU,QAAQ;AAC9D,UAAM,WAAW,OAAO,YAAY,WAAW,SAAY,QAAQ;AAGnE,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,SAAS,CAAC;AAGxD,UAAI,CAAC,KAAK,OAAO,YAAY;AAC3B,YAAI;AACF,gBAAM,QAAQ,MAAM,KAAK,OAAO,YAAY,OAAO;AAGnD,gBAAM,QAAQ,MAAM;AAClB,iBAAK,IAAI,iDAAiD;AAC1D,kBAAM,cAAc,KAAK,aAAa,KAAK,aAAa,SAAS,CAAC;AAClE,gBAAI,aAAa,SAAS,aAAa;AACrC,mBAAK,aAAa,IAAI;AACtB,mBAAK,QAAQ,kEAAkD;AAAA,YACjE;AAAA,UACF;AAGA,gBAAM,KAAK,UAAU,OAAO,KAAK;AAAA,QACnC,SAAS,OAAO;AACd,kBAAQ,MAAM,gBAAgB,KAAK;AAAA,QAErC;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,QAAQ,SAAU;AAC5B,UAAM,MAAM,KAAK,IAAI;AACrB,UAAM,QAAQ,MAAM,KAAK;AACzB,SAAK,YAAY;AACjB,YAAQ,IAAI,WAAW,KAAK,OAAO,GAAG,OAAO;AAAA,EAC/C;AACF;;;AEpQO,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
@@ -9,8 +9,7 @@ var CallClientCommands = /* @__PURE__ */ ((CallClientCommands2) => {
9
9
  return CallClientCommands2;
10
10
  })(CallClientCommands || {});
11
11
  var CallServerCommands = /* @__PURE__ */ ((CallServerCommands2) => {
12
- CallServerCommands2["UserMessage"] = "userMessage";
13
- CallServerCommands2["AssistantMessage"] = "assistantMessage";
12
+ CallServerCommands2["Message"] = "message";
14
13
  CallServerCommands2["CancelLastAssistantMessage"] = "cancelLastAssistantMessage";
15
14
  CallServerCommands2["EndInterview"] = "endInterview";
16
15
  return CallServerCommands2;
@@ -53,9 +52,7 @@ var CallSocket = class {
53
52
  addMessage(message) {
54
53
  if (!this.socket || !this.config) return;
55
54
  this.conversation.push(message);
56
- this.socket.send(
57
- `${message.role === "user" ? "userMessage" /* UserMessage */ : "assistantMessage" /* AssistantMessage */} ${message.content}`
58
- );
55
+ this.socket.send(`${"message" /* Message */} ${JSON.stringify(message)}`);
59
56
  this.config.onMessage?.(message);
60
57
  }
61
58
  async sendAudio(audio, abort) {
@@ -155,16 +152,18 @@ var CallSocket = class {
155
152
  async answer(message) {
156
153
  if (!this.socket || !this.config) return;
157
154
  let isEnd = false;
158
- if (message.includes(END_INTERVIEW)) {
159
- message = message.replace(END_INTERVIEW, "").trim();
155
+ let content = typeof message === "string" ? message : message.content;
156
+ const metadata = typeof message === "string" ? void 0 : message.metadata;
157
+ if (content.includes(END_INTERVIEW)) {
158
+ content = content.replace(END_INTERVIEW, "").trim();
160
159
  isEnd = true;
161
160
  }
162
- if (message.length) {
161
+ if (content.length) {
163
162
  this.log("Assistant message:", message);
164
- this.addMessage({ role: "assistant", content: message });
163
+ this.addMessage({ role: "assistant", content, metadata });
165
164
  if (!this.config.disableTTS) {
166
165
  try {
167
- const audio = await this.config.text2Speech(message);
166
+ const audio = await this.config.text2Speech(content);
168
167
  const abort = () => {
169
168
  this.log("Answer aborted, removing last assistant message");
170
169
  const lastMessage = this.conversation[this.conversation.length - 1];
@@ -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 { 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 public socket: WebSocket | null = null\n public config: CallConfig | null = null\n\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(socket: WebSocket, config: CallConfig) {\n this.socket = socket\n this.config = config\n this.conversation = [{ role: 'system', content: config.systemPrompt }]\n this.log(`Call started`)\n\n // Assistant speaks first\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('[CallSocket]', error)\n socket?.close()\n // TODO: Implement retry\n })\n }\n\n // Listen to events\n socket.on('close', this.onClose.bind(this))\n 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 if (!this.socket || !this.config) return\n this.conversation.push(message)\n this.socket.send(\n `${\n message.role === 'user'\n ? CallServerCommands.UserMessage\n : CallServerCommands.AssistantMessage\n } ${message.content}`\n )\n this.config.onMessage?.(message)\n }\n\n private async sendAudio(\n audio: ArrayBuffer | NodeJS.ReadableStream,\n abort?: () => void\n ) {\n if (!this.socket) return\n if (this.abortAnswer) {\n abort?.()\n return\n }\n\n if (Buffer.isBuffer(audio) || audio instanceof ArrayBuffer) {\n // Whole audio as a single buffer\n this.log(`Send audio: (${audio.byteLength} bytes)`)\n this.socket.send(audio)\n } else if ('paused' in audio) {\n // Audio as a stream\n for await (const chunk of audio) {\n if (this.abortAnswer) {\n abort?.()\n return\n }\n this.log(`Send audio chunk (${chunk.length} bytes)`)\n this.socket.send(chunk)\n }\n } else {\n this.log(`Unknown audio type: ${audio}`)\n }\n }\n\n private onClose() {\n if (!this.config) return\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 this.socket = null\n this.config = null\n }\n\n private async onMessage(message: Buffer) {\n if (!Buffer.isBuffer(message)) {\n console.warn(`[CallSocket] 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 if (!this.config) return\n\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 filename = `speech-${Date.now()}.ogg`\n fs.writeFileSync(filename, Buffer.from(await blob.arrayBuffer()))\n this.log(`Saved speech: ${filename}`)\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('[CallSocket]', error)\n // TODO: Implement retry\n }\n }\n\n // Add assistant message and send to client with audio (TTS)\n public async answer(message: string) {\n if (!this.socket || !this.config) return\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 try {\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 // Send audio to client\n await this.sendAudio(audio, abort)\n } catch (error) {\n console.error('[CallSocket]', error)\n // TODO: Implement retry\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, prevMessage?: 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;;;ACAb,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;;;ADIL,IAAM,gBAAgB;AAEtB,IAAM,aAAN,MAAiB;AAAA,EAmBtB,YAAY,QAAmB,QAAoB;AAlBnD,SAAO,SAA2B;AAClC,SAAO,SAA4B;AAEnC,SAAQ,YAAY,KAAK,IAAI;AAC7B,SAAQ,YAAY,KAAK,IAAI;AAG7B;AAAA,SAAQ,cAAc;AAGtB;AAAA,SAAQ,aAAa;AAGrB;AAAA,SAAQ,SAAmB,CAAC;AAM1B,SAAK,SAAS;AACd,SAAK,SAAS;AACd,SAAK,eAAe,CAAC,EAAE,MAAM,UAAU,SAAS,OAAO,aAAa,CAAC;AACrE,SAAK,IAAI,cAAc;AAGvB,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,gBAAgB,KAAK;AACnC,gBAAQ,MAAM;AAAA,MAEhB,CAAC;AAAA,IACL;AAGA,WAAO,GAAG,SAAS,KAAK,QAAQ,KAAK,IAAI,CAAC;AAC1C,WAAO,GAAG,WAAW,KAAK,UAAU,KAAK,IAAI,CAAC;AAAA,EAChD;AAAA;AAAA,EAGO,kBAAkB,cAA4B;AACnD,SAAK,IAAI,oBAAoB;AAC7B,SAAK,eAAe;AAAA,EACtB;AAAA,EAEQ,WAAW,SAA8B;AAC/C,QAAI,CAAC,KAAK,UAAU,CAAC,KAAK,OAAQ;AAClC,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,EAEA,MAAc,UACZ,OACA,OACA;AACA,QAAI,CAAC,KAAK,OAAQ;AAClB,QAAI,KAAK,aAAa;AACpB,cAAQ;AACR;AAAA,IACF;AAEA,QAAI,OAAO,SAAS,KAAK,KAAK,iBAAiB,aAAa;AAE1D,WAAK,IAAI,gBAAgB,MAAM,UAAU,SAAS;AAClD,WAAK,OAAO,KAAK,KAAK;AAAA,IACxB,WAAW,YAAY,OAAO;AAE5B,uBAAiB,SAAS,OAAO;AAC/B,YAAI,KAAK,aAAa;AACpB,kBAAQ;AACR;AAAA,QACF;AACA,aAAK,IAAI,qBAAqB,MAAM,MAAM,SAAS;AACnD,aAAK,OAAO,KAAK,KAAK;AAAA,MACxB;AAAA,IACF,OAAO;AACL,WAAK,IAAI,uBAAuB,KAAK,EAAE;AAAA,IACzC;AAAA,EACF;AAAA,EAEQ,UAAU;AAChB,QAAI,CAAC,KAAK,OAAQ;AAClB,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;AAGD,SAAK,SAAS;AACd,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAc,UAAU,SAAiB;AACvC,QAAI,CAAC,OAAO,SAAS,OAAO,GAAG;AAC7B,cAAQ,KAAK,sCAAsC;AACnD;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;AAC7B,QAAI,CAAC,KAAK,OAAQ;AAGlB,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,WAAW,UAAU,KAAK,IAAI,CAAC;AACrC,QAAG,iBAAc,UAAU,OAAO,KAAK,MAAM,KAAK,YAAY,CAAC,CAAC;AAChE,aAAK,IAAI,iBAAiB,QAAQ,EAAE;AAAA,MACtC;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,gBAAgB,KAAK;AAAA,IAErC;AAAA,EACF;AAAA;AAAA,EAGA,MAAa,OAAO,SAAiB;AACnC,QAAI,CAAC,KAAK,UAAU,CAAC,KAAK,OAAQ;AAClC,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,YAAI;AACF,gBAAM,QAAQ,MAAM,KAAK,OAAO,YAAY,OAAO;AAGnD,gBAAM,QAAQ,MAAM;AAClB,iBAAK,IAAI,iDAAiD;AAC1D,kBAAM,cAAc,KAAK,aAAa,KAAK,aAAa,SAAS,CAAC;AAClE,gBAAI,aAAa,SAAS,aAAa;AACrC,mBAAK,aAAa,IAAI;AACtB,mBAAK,QAAQ,kEAAkD;AAAA,YACjE;AAAA,UACF;AAGA,gBAAM,KAAK,UAAU,OAAO,KAAK;AAAA,QACnC,SAAS,OAAO;AACd,kBAAQ,MAAM,gBAAgB,KAAK;AAAA,QAErC;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,QAAQ,SAAU;AAC5B,UAAM,MAAM,KAAK,IAAI;AACrB,UAAM,QAAQ,MAAM,KAAK;AACzB,SAAK,YAAY;AACjB,YAAQ,IAAI,WAAW,KAAK,OAAO,GAAG,OAAO;AAAA,EAC/C;AACF;;;AEvQO,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 { 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 public socket: WebSocket | null = null\n public config: CallConfig | null = null\n\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(socket: WebSocket, config: CallConfig) {\n this.socket = socket\n this.config = config\n this.conversation = [{ role: 'system', content: config.systemPrompt }]\n this.log(`Call started`)\n\n // Assistant speaks first\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('[CallSocket]', error)\n socket?.close()\n // TODO: Implement retry\n })\n }\n\n // Listen to events\n socket.on('close', this.onClose.bind(this))\n 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 if (!this.socket || !this.config) return\n this.conversation.push(message)\n this.socket.send(`${CallServerCommands.Message} ${JSON.stringify(message)}`)\n this.config.onMessage?.(message)\n }\n\n private async sendAudio(\n audio: ArrayBuffer | NodeJS.ReadableStream,\n abort?: () => void\n ) {\n if (!this.socket) return\n if (this.abortAnswer) {\n abort?.()\n return\n }\n\n if (Buffer.isBuffer(audio) || audio instanceof ArrayBuffer) {\n // Whole audio as a single buffer\n this.log(`Send audio: (${audio.byteLength} bytes)`)\n this.socket.send(audio)\n } else if ('paused' in audio) {\n // Audio as a stream\n for await (const chunk of audio) {\n if (this.abortAnswer) {\n abort?.()\n return\n }\n this.log(`Send audio chunk (${chunk.length} bytes)`)\n this.socket.send(chunk)\n }\n } else {\n this.log(`Unknown audio type: ${audio}`)\n }\n }\n\n private onClose() {\n if (!this.config) return\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 this.socket = null\n this.config = null\n }\n\n private async onMessage(message: Buffer) {\n if (!Buffer.isBuffer(message)) {\n console.warn(`[CallSocket] 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 if (!this.config) return\n\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 filename = `speech-${Date.now()}.ogg`\n fs.writeFileSync(filename, Buffer.from(await blob.arrayBuffer()))\n this.log(`Saved speech: ${filename}`)\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('[CallSocket]', error)\n // TODO: Implement retry\n }\n }\n\n // Add assistant message and send to client with audio (TTS)\n public async answer(message: string | ConversationMessage) {\n if (!this.socket || !this.config) return\n let isEnd = false\n\n let content = typeof message === 'string' ? message : message.content\n const metadata = typeof message === 'string' ? undefined : message.metadata\n\n // Detect end of interview\n if (content.includes(END_INTERVIEW)) {\n content = content.replace(END_INTERVIEW, '').trim()\n isEnd = true\n }\n\n if (content.length) {\n // Send answer to client\n this.log('Assistant message:', message)\n this.addMessage({ role: 'assistant', content, metadata })\n\n // TTS: Generate answer audio\n if (!this.config.disableTTS) {\n try {\n const audio = await this.config.text2Speech(content)\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 // Send audio to client\n await this.sendAudio(audio, abort)\n } catch (error) {\n console.error('[CallSocket]', error)\n // TODO: Implement retry\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 Message = 'message',\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(\n conversation: Conversation\n ): Promise<string | ConversationMessage>\n speech2Text(audioBlob: Blob, prevMessage?: 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<Data = any> {\n role: 'system' | 'user' | 'assistant'\n content: string\n metadata?: Data\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;;;ACAb,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,aAAU;AACV,EAAAA,oBAAA,gCAA6B;AAC7B,EAAAA,oBAAA,kBAAe;AAHL,SAAAA;AAAA,GAAA;;;ADIL,IAAM,gBAAgB;AAEtB,IAAM,aAAN,MAAiB;AAAA,EAmBtB,YAAY,QAAmB,QAAoB;AAlBnD,SAAO,SAA2B;AAClC,SAAO,SAA4B;AAEnC,SAAQ,YAAY,KAAK,IAAI;AAC7B,SAAQ,YAAY,KAAK,IAAI;AAG7B;AAAA,SAAQ,cAAc;AAGtB;AAAA,SAAQ,aAAa;AAGrB;AAAA,SAAQ,SAAmB,CAAC;AAM1B,SAAK,SAAS;AACd,SAAK,SAAS;AACd,SAAK,eAAe,CAAC,EAAE,MAAM,UAAU,SAAS,OAAO,aAAa,CAAC;AACrE,SAAK,IAAI,cAAc;AAGvB,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,gBAAgB,KAAK;AACnC,gBAAQ,MAAM;AAAA,MAEhB,CAAC;AAAA,IACL;AAGA,WAAO,GAAG,SAAS,KAAK,QAAQ,KAAK,IAAI,CAAC;AAC1C,WAAO,GAAG,WAAW,KAAK,UAAU,KAAK,IAAI,CAAC;AAAA,EAChD;AAAA;AAAA,EAGO,kBAAkB,cAA4B;AACnD,SAAK,IAAI,oBAAoB;AAC7B,SAAK,eAAe;AAAA,EACtB;AAAA,EAEQ,WAAW,SAA8B;AAC/C,QAAI,CAAC,KAAK,UAAU,CAAC,KAAK,OAAQ;AAClC,SAAK,aAAa,KAAK,OAAO;AAC9B,SAAK,OAAO,KAAK,0BAA6B,IAAI,KAAK,UAAU,OAAO,CAAC,EAAE;AAC3E,SAAK,OAAO,YAAY,OAAO;AAAA,EACjC;AAAA,EAEA,MAAc,UACZ,OACA,OACA;AACA,QAAI,CAAC,KAAK,OAAQ;AAClB,QAAI,KAAK,aAAa;AACpB,cAAQ;AACR;AAAA,IACF;AAEA,QAAI,OAAO,SAAS,KAAK,KAAK,iBAAiB,aAAa;AAE1D,WAAK,IAAI,gBAAgB,MAAM,UAAU,SAAS;AAClD,WAAK,OAAO,KAAK,KAAK;AAAA,IACxB,WAAW,YAAY,OAAO;AAE5B,uBAAiB,SAAS,OAAO;AAC/B,YAAI,KAAK,aAAa;AACpB,kBAAQ;AACR;AAAA,QACF;AACA,aAAK,IAAI,qBAAqB,MAAM,MAAM,SAAS;AACnD,aAAK,OAAO,KAAK,KAAK;AAAA,MACxB;AAAA,IACF,OAAO;AACL,WAAK,IAAI,uBAAuB,KAAK,EAAE;AAAA,IACzC;AAAA,EACF;AAAA,EAEQ,UAAU;AAChB,QAAI,CAAC,KAAK,OAAQ;AAClB,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;AAGD,SAAK,SAAS;AACd,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,MAAc,UAAU,SAAiB;AACvC,QAAI,CAAC,OAAO,SAAS,OAAO,GAAG;AAC7B,cAAQ,KAAK,sCAAsC;AACnD;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;AAC7B,QAAI,CAAC,KAAK,OAAQ;AAGlB,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,WAAW,UAAU,KAAK,IAAI,CAAC;AACrC,QAAG,iBAAc,UAAU,OAAO,KAAK,MAAM,KAAK,YAAY,CAAC,CAAC;AAChE,aAAK,IAAI,iBAAiB,QAAQ,EAAE;AAAA,MACtC;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,gBAAgB,KAAK;AAAA,IAErC;AAAA,EACF;AAAA;AAAA,EAGA,MAAa,OAAO,SAAuC;AACzD,QAAI,CAAC,KAAK,UAAU,CAAC,KAAK,OAAQ;AAClC,QAAI,QAAQ;AAEZ,QAAI,UAAU,OAAO,YAAY,WAAW,UAAU,QAAQ;AAC9D,UAAM,WAAW,OAAO,YAAY,WAAW,SAAY,QAAQ;AAGnE,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,SAAS,CAAC;AAGxD,UAAI,CAAC,KAAK,OAAO,YAAY;AAC3B,YAAI;AACF,gBAAM,QAAQ,MAAM,KAAK,OAAO,YAAY,OAAO;AAGnD,gBAAM,QAAQ,MAAM;AAClB,iBAAK,IAAI,iDAAiD;AAC1D,kBAAM,cAAc,KAAK,aAAa,KAAK,aAAa,SAAS,CAAC;AAClE,gBAAI,aAAa,SAAS,aAAa;AACrC,mBAAK,aAAa,IAAI;AACtB,mBAAK,QAAQ,kEAAkD;AAAA,YACjE;AAAA,UACF;AAGA,gBAAM,KAAK,UAAU,OAAO,KAAK;AAAA,QACnC,SAAS,OAAO;AACd,kBAAQ,MAAM,gBAAgB,KAAK;AAAA,QAErC;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,QAAQ,SAAU;AAC5B,UAAM,MAAM,KAAK,IAAI;AACrB,UAAM,QAAQ,MAAM,KAAK;AACzB,SAAK,YAAY;AACjB,YAAQ,IAAI,WAAW,KAAK,OAAO,GAAG,OAAO;AAAA,EAC/C;AACF;;;AEpQO,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/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@micdrop/server",
3
- "version": "1.2.1",
3
+ "version": "1.2.2",
4
4
  "description": "A lib for Node.js that helps to use the mic and speaker for voice conversation",
5
5
  "author": "Lonestone",
6
6
  "license": "MIT",