@micdrop/server 2.0.6 → 2.0.8

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/dist/index.js CHANGED
@@ -30,6 +30,12 @@ var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: tru
30
30
  // src/index.ts
31
31
  var index_exports = {};
32
32
  __export(index_exports, {
33
+ AUTO_END_CALL_PROMPT: () => AUTO_END_CALL_PROMPT,
34
+ AUTO_END_CALL_TOOL_NAME: () => AUTO_END_CALL_TOOL_NAME,
35
+ AUTO_IGNORE_USER_NOISE_PROMPT: () => AUTO_IGNORE_USER_NOISE_PROMPT,
36
+ AUTO_IGNORE_USER_NOISE_TOOL_NAME: () => AUTO_IGNORE_USER_NOISE_TOOL_NAME,
37
+ AUTO_SEMANTIC_TURN_PROMPT: () => AUTO_SEMANTIC_TURN_PROMPT,
38
+ AUTO_SEMANTIC_TURN_TOOL_NAME: () => AUTO_SEMANTIC_TURN_TOOL_NAME,
33
39
  Agent: () => Agent,
34
40
  FileSTT: () => FileSTT,
35
41
  Logger: () => Logger,
@@ -53,11 +59,22 @@ module.exports = __toCommonJS(index_exports);
53
59
 
54
60
  // src/agent/Agent.ts
55
61
  var import_eventemitter3 = require("eventemitter3");
62
+
63
+ // src/agent/tools.ts
64
+ var AUTO_END_CALL_TOOL_NAME = "end_call";
65
+ var AUTO_END_CALL_PROMPT = "Call this tool only if user asks to end the call";
66
+ var AUTO_SEMANTIC_TURN_TOOL_NAME = "semantic_turn";
67
+ var AUTO_SEMANTIC_TURN_PROMPT = "Call this tool only if last user message is obviously an incomplete sentence that you need to wait for the end before answering";
68
+ var AUTO_IGNORE_USER_NOISE_TOOL_NAME = "ignore_user_noise";
69
+ var AUTO_IGNORE_USER_NOISE_PROMPT = `Call this tool only if last user message is just an interjection or a sound that expresses emotion, hesitation, or reaction (ex: "Uh", "Ahem", "Hmm", "Ah") but doesn't carry any clear meaning like agreeing, refusing, or commanding`;
70
+
71
+ // src/agent/Agent.ts
56
72
  var Agent = class extends import_eventemitter3.EventEmitter {
57
73
  constructor(options) {
58
74
  super();
59
75
  this.options = options;
60
76
  this.conversation = [{ role: "system", content: options.systemPrompt }];
77
+ this.tools = this.getDefaultTools();
61
78
  }
62
79
  addUserMessage(text, metadata) {
63
80
  this.addMessage("user", text, metadata);
@@ -65,6 +82,15 @@ var Agent = class extends import_eventemitter3.EventEmitter {
65
82
  addAssistantMessage(text, metadata) {
66
83
  this.addMessage("assistant", text, metadata);
67
84
  }
85
+ addTool(tool) {
86
+ this.tools.push(tool);
87
+ }
88
+ removeTool(name) {
89
+ this.tools = this.tools.filter((tool) => tool.name !== name);
90
+ }
91
+ getTool(name) {
92
+ return this.tools.find((tool) => tool.name === name);
93
+ }
68
94
  addMessage(role, text, metadata) {
69
95
  this.log(`Adding ${role} message to conversation: ${text}`);
70
96
  const message = {
@@ -75,28 +101,132 @@ var Agent = class extends import_eventemitter3.EventEmitter {
75
101
  this.conversation.push(message);
76
102
  this.emit("Message", message);
77
103
  }
104
+ addToolMessage(message) {
105
+ this.log("Adding tool message:", message);
106
+ this.conversation.push(message);
107
+ this.emit("Message", message);
108
+ }
78
109
  endCall() {
79
110
  this.log("Ending call");
80
111
  this.emit("EndCall");
81
112
  }
82
113
  cancelLastUserMessage() {
83
114
  this.log("Cancelling last user message");
84
- const lastMessage = this.conversation[this.conversation.length - 1];
85
- if (lastMessage?.role !== "user") return;
86
- this.conversation.pop();
115
+ const lastMessageIndex = this.conversation.findLastIndex(
116
+ (message) => message.role === "user"
117
+ );
118
+ if (lastMessageIndex !== -1) {
119
+ this.conversation.splice(lastMessageIndex, 1);
120
+ }
87
121
  this.emit("CancelLastUserMessage");
88
122
  }
89
- cancelLastAssistantMessage() {
90
- this.log("Cancelling last assistant message");
91
- const lastMessage = this.conversation[this.conversation.length - 1];
92
- if (lastMessage?.role !== "assistant") return;
93
- this.conversation.pop();
94
- this.emit("CancelLastAssistantMessage");
95
- }
96
123
  skipAnswer() {
97
124
  this.log("Skipping answer");
98
125
  this.emit("SkipAnswer");
99
126
  }
127
+ getDefaultTools() {
128
+ const tools = [];
129
+ if (this.options.autoEndCall) {
130
+ tools.push({
131
+ name: AUTO_END_CALL_TOOL_NAME,
132
+ description: typeof this.options.autoEndCall === "string" ? this.options.autoEndCall : AUTO_END_CALL_PROMPT,
133
+ execute: () => this.endCall()
134
+ });
135
+ }
136
+ if (this.options.autoSemanticTurn) {
137
+ tools.push({
138
+ name: AUTO_SEMANTIC_TURN_TOOL_NAME,
139
+ description: typeof this.options.autoSemanticTurn === "string" ? this.options.autoSemanticTurn : AUTO_SEMANTIC_TURN_PROMPT,
140
+ skipAnswer: true,
141
+ execute: () => this.skipAnswer()
142
+ });
143
+ }
144
+ if (this.options.autoIgnoreUserNoise) {
145
+ tools.push({
146
+ name: AUTO_IGNORE_USER_NOISE_TOOL_NAME,
147
+ description: typeof this.options.autoIgnoreUserNoise === "string" ? this.options.autoIgnoreUserNoise : AUTO_IGNORE_USER_NOISE_PROMPT,
148
+ skipAnswer: true,
149
+ execute: () => this.cancelLastUserMessage()
150
+ });
151
+ }
152
+ return tools;
153
+ }
154
+ async executeTool(toolCall) {
155
+ try {
156
+ const tool = this.getTool(toolCall.toolName);
157
+ if (!tool) {
158
+ throw new Error(`Tool not found "${toolCall.toolName}"`);
159
+ }
160
+ this.log("Executing tool:", toolCall.toolName, toolCall.parameters);
161
+ this.addToolMessage(toolCall);
162
+ const parameters = JSON.parse(toolCall.parameters);
163
+ const output = tool.execute ? await tool.execute(parameters) : {};
164
+ this.addToolMessage({
165
+ role: "tool_result",
166
+ toolCallId: toolCall.toolCallId,
167
+ toolName: toolCall.toolName,
168
+ output: JSON.stringify(output ?? null)
169
+ });
170
+ if (tool.emitOutput) {
171
+ this.emit("ToolCall", {
172
+ name: toolCall.toolName,
173
+ parameters,
174
+ output
175
+ });
176
+ }
177
+ return {
178
+ output,
179
+ skipAnswer: tool.skipAnswer
180
+ };
181
+ } catch (error) {
182
+ console.error("[OpenaiAgent] Error executing tool:", error);
183
+ return {
184
+ output: {
185
+ error: error.message
186
+ }
187
+ };
188
+ }
189
+ }
190
+ getExtractOptions() {
191
+ const extract = this.options.extract;
192
+ if (!extract) return void 0;
193
+ if ("json" in extract && extract.json) {
194
+ return { ...extract, startTag: "{", endTag: "}" };
195
+ }
196
+ if ("startTag" in extract && "endTag" in extract) {
197
+ return extract;
198
+ }
199
+ return void 0;
200
+ }
201
+ extract(message) {
202
+ const extractOptions = this.getExtractOptions();
203
+ let metadata = void 0;
204
+ if (extractOptions) {
205
+ const startTagIndex = message.indexOf(extractOptions.startTag);
206
+ if (startTagIndex !== -1) {
207
+ let endTagIndex = message.lastIndexOf(extractOptions.endTag);
208
+ if (endTagIndex === -1) endTagIndex = message.length + 1;
209
+ else endTagIndex += extractOptions.endTag.length;
210
+ const extractedText = message.slice(startTagIndex, endTagIndex).trim();
211
+ try {
212
+ const extractedValue = "json" in extractOptions && extractOptions.json ? JSON.parse(extractedText) : extractedText;
213
+ if (extractOptions.callback) {
214
+ extractOptions.callback(extractedValue);
215
+ }
216
+ if (extractOptions.saveInMetadata) {
217
+ metadata = { extracted: extractedValue };
218
+ }
219
+ } catch (error) {
220
+ console.error(
221
+ `[OpenaiAgent] Error parsing extracted value (${extractedText}):`,
222
+ error
223
+ );
224
+ }
225
+ message = message.slice(0, startTagIndex).trimEnd();
226
+ }
227
+ }
228
+ return { message, metadata };
229
+ }
100
230
  log(...message) {
101
231
  this.logger?.log(...message);
102
232
  }
@@ -207,7 +337,6 @@ var MicdropClientCommands = /* @__PURE__ */ ((MicdropClientCommands2) => {
207
337
  })(MicdropClientCommands || {});
208
338
  var MicdropServerCommands = /* @__PURE__ */ ((MicdropServerCommands2) => {
209
339
  MicdropServerCommands2["Message"] = "Message";
210
- MicdropServerCommands2["CancelLastAssistantMessage"] = "CancelLastAssistantMessage";
211
340
  MicdropServerCommands2["CancelLastUserMessage"] = "CancelLastUserMessage";
212
341
  MicdropServerCommands2["SkipAnswer"] = "SkipAnswer";
213
342
  MicdropServerCommands2["EndCall"] = "EndCall";
@@ -280,10 +409,6 @@ var MicdropServer = class {
280
409
  "CancelLastUserMessage",
281
410
  () => this.socket?.send("CancelLastUserMessage" /* CancelLastUserMessage */)
282
411
  );
283
- this.config.agent.on(
284
- "CancelLastAssistantMessage",
285
- () => this.socket?.send("CancelLastAssistantMessage" /* CancelLastAssistantMessage */)
286
- );
287
412
  this.config.agent.on(
288
413
  "SkipAnswer",
289
414
  () => this.socket?.send("SkipAnswer" /* SkipAnswer */)
@@ -325,7 +450,8 @@ var MicdropServer = class {
325
450
  this.currentUserStream?.end();
326
451
  this.currentUserStream = void 0;
327
452
  const conversation = this.config?.agent.conversation;
328
- if (conversation && conversation[conversation.length - 1].role === "user") {
453
+ const lastMessage = conversation?.[conversation.length - 1];
454
+ if (lastMessage?.role === "user" && this.lastMessageSpeeched !== lastMessage) {
329
455
  this.log(
330
456
  "User stopped speaking and a transcript already exists, answering"
331
457
  );
@@ -351,6 +477,12 @@ var MicdropServer = class {
351
477
  async answer() {
352
478
  if (!this.config) return;
353
479
  this.cancel();
480
+ const lastMessage = this.config.agent.conversation[this.config.agent.conversation.length - 1];
481
+ if (this.lastMessageSpeeched === lastMessage) {
482
+ this.log("Already answered, skipping");
483
+ return;
484
+ }
485
+ this.lastMessageSpeeched = lastMessage;
354
486
  try {
355
487
  const stream = this.config.agent.answer();
356
488
  await this.speak(stream);
@@ -540,6 +672,12 @@ async function waitForParams(socket, validate) {
540
672
  }
541
673
  // Annotate the CommonJS export names for ESM import in node:
542
674
  0 && (module.exports = {
675
+ AUTO_END_CALL_PROMPT,
676
+ AUTO_END_CALL_TOOL_NAME,
677
+ AUTO_IGNORE_USER_NOISE_PROMPT,
678
+ AUTO_IGNORE_USER_NOISE_TOOL_NAME,
679
+ AUTO_SEMANTIC_TURN_PROMPT,
680
+ AUTO_SEMANTIC_TURN_TOOL_NAME,
543
681
  Agent,
544
682
  FileSTT,
545
683
  Logger,
package/dist/index.js.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/index.ts","../src/agent/Agent.ts","../src/agent/MockAgent.ts","../src/audio-convert.ts","../src/errors.ts","../src/Logger.ts","../src/MicdropServer.ts","../src/types.ts","../src/stt/STT.ts","../src/stt/FileSTT.ts","../src/stt/MockSTT.ts","../src/tts/MockTTS.ts","../src/tts/TTS.ts","../src/waitForParams.ts"],"sourcesContent":["export * from './agent'\nexport * from './audio-convert'\nexport * from './errors'\nexport * from './Logger'\nexport * from './MicdropServer'\nexport * from './stt'\nexport * from './tts'\nexport * from './types'\nexport * from './waitForParams'\n","import { EventEmitter } from 'eventemitter3'\nimport { Readable } from 'stream'\nimport { Logger } from '../Logger'\nimport {\n MicdropAnswerMetadata,\n MicdropConversation,\n MicdropConversationMessage,\n MicdropToolCall,\n} from '../types'\n\nexport interface AgentOptions {\n systemPrompt: string\n}\n\nexport interface AgentEvents {\n Message: [MicdropConversationMessage]\n CancelLastUserMessage: []\n CancelLastAssistantMessage: []\n SkipAnswer: []\n EndCall: []\n ToolCall: [MicdropToolCall]\n}\n\nexport abstract class Agent<\n Options extends AgentOptions = AgentOptions,\n> extends EventEmitter<AgentEvents> {\n public logger?: Logger\n\n // Conversation history\n public conversation: MicdropConversation\n\n constructor(protected options: Options) {\n super()\n this.conversation = [{ role: 'system', content: options.systemPrompt }]\n }\n\n abstract answer(): Readable\n abstract cancel(): void\n\n public addUserMessage(text: string, metadata?: MicdropAnswerMetadata) {\n this.addMessage('user', text, metadata)\n }\n\n public addAssistantMessage(text: string, metadata?: MicdropAnswerMetadata) {\n this.addMessage('assistant', text, metadata)\n }\n\n protected addMessage(\n role: 'user' | 'assistant' | 'system',\n text: string,\n metadata?: MicdropAnswerMetadata\n ) {\n this.log(`Adding ${role} message to conversation: ${text}`)\n const message: MicdropConversationMessage = {\n role,\n content: text,\n metadata,\n }\n this.conversation.push(message)\n this.emit('Message', message)\n }\n\n protected endCall() {\n this.log('Ending call')\n this.emit('EndCall')\n }\n\n protected cancelLastUserMessage() {\n this.log('Cancelling last user message')\n const lastMessage = this.conversation[this.conversation.length - 1]\n if (lastMessage?.role !== 'user') return\n this.conversation.pop()\n this.emit('CancelLastUserMessage')\n }\n\n protected cancelLastAssistantMessage() {\n this.log('Cancelling last assistant message')\n const lastMessage = this.conversation[this.conversation.length - 1]\n if (lastMessage?.role !== 'assistant') return\n this.conversation.pop()\n this.emit('CancelLastAssistantMessage')\n }\n\n protected skipAnswer() {\n this.log('Skipping answer')\n this.emit('SkipAnswer')\n }\n\n protected log(...message: any[]) {\n this.logger?.log(...message)\n }\n\n destroy() {\n this.log('Destroyed')\n this.removeAllListeners()\n this.cancel()\n }\n}\n","import { PassThrough } from 'stream'\nimport { Agent } from './Agent'\n\nexport class MockAgent extends Agent {\n private i = 0\n\n constructor() {\n super({ systemPrompt: '' })\n }\n\n answer() {\n const stream = new PassThrough()\n\n // Answer message\n const message = `Assistant Message ${this.i++}`\n this.addAssistantMessage(message)\n stream.write(message)\n stream.end()\n return stream\n }\n\n cancel() {}\n}\n","import ffmpegInstaller from '@ffmpeg-installer/ffmpeg'\nimport ffmpeg from 'fluent-ffmpeg'\nimport { PassThrough, Readable } from 'stream'\n\n// Setup ffmpeg\nffmpeg.setFfmpegPath(ffmpegInstaller.path)\n\n// Convert stream to WAV/PCM\nexport function convertToPCM(\n audioStream: Readable,\n sampleRate = 16000,\n bitDepth = 16\n) {\n const pcmStream = new PassThrough()\n ffmpeg(audioStream)\n .audioChannels(1)\n .audioFrequency(sampleRate)\n .audioCodec(`pcm_s${bitDepth}le`)\n .format(`s${bitDepth}le`)\n .on('error', (error) => {\n console.error('Error converting audio stream:', error.message)\n })\n .pipe(pcmStream)\n return pcmStream\n}\n\n// Convert PCM stream to WebM/Opus\nexport function convertToOpus(audioStream: Readable, sampleRate = 16000) {\n const webmStream = new PassThrough()\n ffmpegToOpus(ffmpeg(audioStream), sampleRate).pipe(webmStream)\n return webmStream\n}\n\n// Convert PCM stream to WebM/Opus\nexport function convertPCMToOpus(audioStream: Readable, sampleRate = 16000) {\n const webmStream = new PassThrough()\n ffmpegToOpus(ffmpeg(audioStream), sampleRate)\n .inputFormat('s16le')\n .inputOptions(['-f s16le', '-ar 16000', '-ac 1'])\n .pipe(webmStream)\n return webmStream\n}\n\nfunction ffmpegToOpus(ffmpegCommand: ffmpeg.FfmpegCommand, sampleRate = 16000) {\n return ffmpegCommand\n .audioChannels(1)\n .audioFrequency(sampleRate)\n .audioCodec('libopus')\n .format('webm')\n .outputOptions([\n '-application audio',\n `-ac 1`,\n `-ar ${sampleRate}`,\n `-b:a 64k`,\n `-f webm`,\n `-map_metadata -1`,\n ])\n .on('error', (error) => {\n console.error('Error converting to Opus: ', error.message)\n })\n}\n","import WebSocket from 'ws'\n\nexport enum MicdropErrorCode {\n BadRequest = 4400,\n Unauthorized = 4401,\n NotFound = 4404,\n}\n\nexport class MicdropError extends Error {\n code: number\n\n constructor(code: number, message: string) {\n super(message)\n this.code = code\n }\n}\n\nexport function handleError(socket: WebSocket, error: unknown) {\n if (error instanceof MicdropError) {\n socket.close(error.code, error.message)\n } else {\n console.error(error)\n socket.close(1011)\n }\n socket.terminate()\n}\n","export class Logger {\n constructor(private readonly name: string) {}\n\n log(...message: any[]) {\n const time = process.uptime().toFixed(3)\n console.log(`[${this.name} ${time}]`, ...message)\n }\n}\n","import { Duplex, PassThrough, Readable } from 'stream'\nimport { WebSocket } from 'ws'\nimport { Logger } from './Logger'\nimport {\n MicdropClientCommands,\n MicdropConfig,\n MicdropServerCommands,\n} from './types'\n\nexport class MicdropServer {\n public socket: WebSocket | null = null\n public config: MicdropConfig | null = null\n public logger?: Logger\n\n private startTime = Date.now()\n\n // When user is speaking, we're streaming chunks for STT\n private currentUserStream?: Duplex\n\n constructor(socket: WebSocket, config: MicdropConfig) {\n this.socket = socket\n this.config = config\n this.log(`Call started`)\n\n // Setup STT\n this.config.stt.on('Transcript', this.onTranscript)\n\n // Setup agent\n this.config.agent.on('Message', (message) =>\n this.socket?.send(\n `${MicdropServerCommands.Message} ${JSON.stringify(message)}`\n )\n )\n this.config.agent.on('CancelLastUserMessage', () =>\n this.socket?.send(MicdropServerCommands.CancelLastUserMessage)\n )\n this.config.agent.on('CancelLastAssistantMessage', () =>\n this.socket?.send(MicdropServerCommands.CancelLastAssistantMessage)\n )\n this.config.agent.on('SkipAnswer', () =>\n this.socket?.send(MicdropServerCommands.SkipAnswer)\n )\n this.config.agent.on('EndCall', () =>\n this.socket?.send(MicdropServerCommands.EndCall)\n )\n this.config.agent.on('ToolCall', (toolCall) =>\n this.socket?.send(\n `${MicdropServerCommands.ToolCall} ${JSON.stringify(toolCall)}`\n )\n )\n\n // Assistant speaks first\n this.sendFirstMessage()\n\n // Listen to events\n socket.on('close', this.onClose)\n socket.on('message', this.onMessage)\n }\n\n private log(...message: any[]) {\n this.logger?.log(...message)\n }\n\n private cancel() {\n this.config?.tts.cancel()\n this.config?.agent.cancel()\n }\n\n private onClose = () => {\n if (!this.config) return\n this.log('Connection closed')\n const duration = Math.round((Date.now() - this.startTime) / 1000)\n\n // Destroy instances\n this.config.agent.destroy()\n this.config.stt.destroy()\n this.config.tts.destroy()\n\n // End call callback\n this.config.onEnd?.({\n conversation: this.config.agent.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 onMessage = async (message: Buffer) => {\n if (message.byteLength === 0) return\n if (!Buffer.isBuffer(message)) {\n this.log('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: ${cmd}`)\n\n if (cmd === MicdropClientCommands.StartSpeaking) {\n // User started speaking\n await this.onStartSpeaking()\n } else if (cmd === MicdropClientCommands.Mute) {\n // User muted the call\n await this.onMute()\n } else if (cmd === MicdropClientCommands.StopSpeaking) {\n // User stopped speaking\n await this.onStopSpeaking()\n }\n }\n\n // Audio chunk\n else if (this.currentUserStream) {\n this.log(`Received chunk (${message.byteLength} bytes)`)\n this.currentUserStream.write(message)\n }\n }\n\n private async onMute() {\n this.currentUserStream?.end()\n this.currentUserStream = undefined\n this.cancel()\n }\n\n private async onStartSpeaking() {\n if (!this.config) return\n this.currentUserStream?.end()\n this.currentUserStream = new PassThrough()\n this.config.stt.transcribe(this.currentUserStream)\n this.cancel()\n }\n\n private async onStopSpeaking() {\n this.currentUserStream?.end()\n this.currentUserStream = undefined\n\n const conversation = this.config?.agent.conversation\n if (conversation && conversation[conversation.length - 1].role === 'user') {\n this.log(\n 'User stopped speaking and a transcript already exists, answering'\n )\n this.answer()\n }\n }\n\n private onTranscript = async (transcript: string) => {\n if (!this.config) return\n this.log(`User transcript: \"${transcript}\"`)\n this.config.agent.addUserMessage(transcript)\n\n // Answer if user stopped speaking\n if (!this.currentUserStream) {\n this.log('User stopped speaking, answering')\n this.answer()\n }\n }\n\n private async sendFirstMessage() {\n if (!this.config) return\n try {\n if (this.config.firstMessage) {\n // Send first message\n this.config.agent.addAssistantMessage(this.config.firstMessage)\n await this.speak(this.config.firstMessage)\n } else if (this.config.generateFirstMessage) {\n // Generate first message\n await this.answer()\n } else {\n // Skip answer if no first message is provided\n // to avoid keeping the client in a processing state\n this.socket?.send(MicdropServerCommands.SkipAnswer)\n }\n } catch (error) {\n console.error('[MicdropServer]', error)\n this.socket?.send(MicdropServerCommands.SkipAnswer)\n }\n }\n\n private async answer() {\n if (!this.config) return\n this.cancel()\n try {\n // LLM: Generate answer\n const stream = this.config.agent.answer()\n\n // TTS: Generate answer audio\n await this.speak(stream)\n } catch (error) {\n console.error('[MicdropServer]', error)\n this.socket?.send(MicdropServerCommands.SkipAnswer)\n }\n }\n\n // Run text-to-speech and send to client\n private async speak(message: string | Readable) {\n if (!this.socket || !this.config) return\n\n // Convert message to stream if needed\n let textStream: Readable\n if (typeof message === 'string') {\n const stream = new PassThrough()\n stream.write(message)\n stream.end()\n textStream = stream\n } else {\n textStream = message\n }\n\n // Run TTS\n const audio = this.config.tts.speak(textStream)\n\n // Send audio to client\n await this.sendAudio(audio)\n }\n\n private async sendAudio(audio: Readable) {\n if (!this.socket) return\n if (!audio.readable) {\n this.log('Non readable audio, skipping', audio)\n return\n }\n\n // Stream audio\n audio.on('data', (chunk) => {\n this.log(`Send audio chunk (${chunk.byteLength} bytes)`)\n this.socket?.send(chunk)\n })\n audio.on('error', (error) => {\n this.log('Error in audio stream', error)\n })\n audio.on('end', () => {\n this.log('Audio stream ended')\n })\n }\n}\n","import type { Agent } from './agent'\nimport type { STT } from './stt'\nimport type { TTS } from './tts'\n\nexport enum MicdropClientCommands {\n StartSpeaking = 'StartSpeaking',\n StopSpeaking = 'StopSpeaking',\n Mute = 'Mute',\n}\n\nexport enum MicdropServerCommands {\n Message = 'Message',\n CancelLastAssistantMessage = 'CancelLastAssistantMessage',\n CancelLastUserMessage = 'CancelLastUserMessage',\n SkipAnswer = 'SkipAnswer',\n EndCall = 'EndCall',\n ToolCall = 'ToolCall',\n}\n\nexport interface MicdropConfig {\n firstMessage?: string\n generateFirstMessage?: boolean\n agent: Agent\n stt: STT\n tts: TTS\n onEnd?(call: MicdropCallSummary): void\n}\n\nexport interface MicdropCallSummary {\n conversation: MicdropConversation\n duration: number\n}\n\nexport type MicdropConversation = MicdropConversationMessage[]\n\nexport type MicdropAnswerMetadata = {\n [key: string]: any\n}\n\nexport interface MicdropConversationMessage<\n Data extends MicdropAnswerMetadata = MicdropAnswerMetadata,\n> {\n role: 'system' | 'user' | 'assistant'\n content: string\n metadata?: Data\n}\n\nexport interface MicdropToolCall {\n name: string\n parameters: any\n output: any\n}\n\nexport type DeepPartial<T> = T extends object\n ? {\n [P in keyof T]?: DeepPartial<T[P]>\n }\n : T\n","import { EventEmitter } from 'eventemitter3'\nimport { Readable } from 'stream'\nimport { Logger } from '../Logger'\n\n// Audio mime type to extension map\nconst MIME_TYPE_TO_EXTENSION = {\n 'audio/wav': 'wav',\n 'audio/ogg': 'ogg',\n 'audio/mpeg': 'mp3',\n 'audio/webm': 'webm',\n 'audio/mp4': 'mp4',\n 'audio/flac': 'flac',\n} as const\n\nexport interface STTEvents {\n Transcript: [string]\n}\n\nexport abstract class STT extends EventEmitter<STTEvents> {\n protected mimeType?: keyof typeof MIME_TYPE_TO_EXTENSION\n public logger?: Logger\n\n // Set stream of audio to transcribe\n transcribe(audioStream: Readable) {\n // Detect mime type at first chunk\n audioStream.once('data', (chunk) => {\n this.mimeType = this.detectMimeType(chunk)\n })\n }\n\n protected log(...message: any[]) {\n this.logger?.log(...message)\n }\n\n destroy() {\n this.log('Destroyed')\n this.removeAllListeners()\n }\n\n protected get extension(): string {\n return (this.mimeType && MIME_TYPE_TO_EXTENSION[this.mimeType]) || 'bin'\n }\n\n private detectMimeType(\n chunk: ArrayBuffer\n ): keyof typeof MIME_TYPE_TO_EXTENSION {\n if (!chunk || chunk.byteLength === 0) {\n throw new Error('Unable to detect mime type (empty chunk)')\n }\n\n const arr = new Uint8Array(chunk)\n\n // WEBM: 1A 45 DF A3\n if (\n arr[0] === 0x1a &&\n arr[1] === 0x45 &&\n arr[2] === 0xdf &&\n arr[3] === 0xa3\n ) {\n return 'audio/webm'\n }\n // OGG: 4F 67 67 53\n if (\n arr[0] === 0x4f &&\n arr[1] === 0x67 &&\n arr[2] === 0x67 &&\n arr[3] === 0x53\n ) {\n return 'audio/ogg'\n }\n // WAV: 52 49 46 46 ... 57 41 56 45\n if (\n arr[0] === 0x52 &&\n arr[1] === 0x49 &&\n arr[2] === 0x46 &&\n arr[3] === 0x46 &&\n arr[8] === 0x57 &&\n arr[9] === 0x41 &&\n arr[10] === 0x56 &&\n arr[11] === 0x45\n ) {\n return 'audio/wav'\n }\n // MP3: 49 44 33\n if (arr[0] === 0x49 && arr[1] === 0x44 && arr[2] === 0x33) {\n return 'audio/mpeg'\n }\n // MP4/M4A: 00 00 00 .. 66 74 79 70\n if (\n arr[4] === 0x66 &&\n arr[5] === 0x74 &&\n arr[6] === 0x79 &&\n arr[7] === 0x70\n ) {\n return 'audio/mp4'\n }\n // FLAC: 66 4c 61 43\n if (\n arr[0] === 0x66 &&\n arr[1] === 0x4c &&\n arr[2] === 0x61 &&\n arr[3] === 0x43\n ) {\n return 'audio/flac'\n }\n this.log('Unable to detect mime type, using default', chunk)\n return 'audio/wav'\n }\n}\n","import { Readable } from 'stream'\nimport { STT } from './STT'\n\n/**\n * Abstract class for STT, converting stream to file before transcribing\n */\n\nexport abstract class FileSTT extends STT {\n abstract transcribeFile(file: File): Promise<string>\n\n transcribe(audioStream: Readable) {\n super.transcribe(audioStream)\n\n // Convert stream to file\n this.log('Converting stream to file...')\n\n const chunks: Buffer[] = []\n audioStream.on('data', (chunk) => {\n chunks.push(chunk)\n })\n\n audioStream.on('end', async () => {\n if (chunks.length === 0) return\n const arrayBuffer = Buffer.concat(chunks)\n const file = new File([arrayBuffer], `audio.${this.extension}`, {\n type: this.mimeType,\n })\n\n // Transcribe file with implementation\n this.log('Transcribing file...')\n const transcript = await this.transcribeFile(file)\n this.emit('Transcript', transcript)\n })\n }\n}\n","import { FileSTT } from './FileSTT'\n\nexport class MockSTT extends FileSTT {\n private i = 0\n\n async transcribeFile(file: File) {\n return `User Message ${this.i++}`\n }\n}\n","import * as fs from 'fs'\nimport { PassThrough, Readable } from 'stream'\nimport { TTS } from './TTS'\n\nexport class MockTTS extends TTS {\n constructor(private audioFilePaths: string[]) {\n super()\n }\n\n speak(textStream: Readable) {\n const audioStream = new PassThrough()\n textStream.once('data', async () => {\n for (const filePath of this.audioFilePaths) {\n await new Promise((resolve) => setTimeout(resolve, 200))\n const audioBuffer = fs.readFileSync(filePath)\n this.log(`Loaded chunk (${audioBuffer.length} bytes)`)\n audioStream.write(audioBuffer)\n }\n audioStream.end()\n })\n return audioStream\n }\n\n cancel() {}\n}\n","import { Readable } from 'stream'\nimport { Logger } from '../Logger'\n\nexport abstract class TTS {\n public logger?: Logger\n\n abstract speak(textStream: Readable): Readable\n abstract cancel(): void\n\n protected log(...message: any[]) {\n this.logger?.log(...message)\n }\n\n destroy() {\n this.log('Destroyed')\n this.cancel()\n }\n}\n","import { WebSocket } from 'ws'\nimport { MicdropError, MicdropErrorCode } 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 MicdropError(MicdropErrorCode.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 MicdropError(MicdropErrorCode.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;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;;;ACAA,2BAA6B;AAuBtB,IAAe,QAAf,cAEG,kCAA0B;AAAA,EAMlC,YAAsB,SAAkB;AACtC,UAAM;AADc;AAEpB,SAAK,eAAe,CAAC,EAAE,MAAM,UAAU,SAAS,QAAQ,aAAa,CAAC;AAAA,EACxE;AAAA,EAKO,eAAe,MAAc,UAAkC;AACpE,SAAK,WAAW,QAAQ,MAAM,QAAQ;AAAA,EACxC;AAAA,EAEO,oBAAoB,MAAc,UAAkC;AACzE,SAAK,WAAW,aAAa,MAAM,QAAQ;AAAA,EAC7C;AAAA,EAEU,WACR,MACA,MACA,UACA;AACA,SAAK,IAAI,UAAU,IAAI,6BAA6B,IAAI,EAAE;AAC1D,UAAM,UAAsC;AAAA,MAC1C;AAAA,MACA,SAAS;AAAA,MACT;AAAA,IACF;AACA,SAAK,aAAa,KAAK,OAAO;AAC9B,SAAK,KAAK,WAAW,OAAO;AAAA,EAC9B;AAAA,EAEU,UAAU;AAClB,SAAK,IAAI,aAAa;AACtB,SAAK,KAAK,SAAS;AAAA,EACrB;AAAA,EAEU,wBAAwB;AAChC,SAAK,IAAI,8BAA8B;AACvC,UAAM,cAAc,KAAK,aAAa,KAAK,aAAa,SAAS,CAAC;AAClE,QAAI,aAAa,SAAS,OAAQ;AAClC,SAAK,aAAa,IAAI;AACtB,SAAK,KAAK,uBAAuB;AAAA,EACnC;AAAA,EAEU,6BAA6B;AACrC,SAAK,IAAI,mCAAmC;AAC5C,UAAM,cAAc,KAAK,aAAa,KAAK,aAAa,SAAS,CAAC;AAClE,QAAI,aAAa,SAAS,YAAa;AACvC,SAAK,aAAa,IAAI;AACtB,SAAK,KAAK,4BAA4B;AAAA,EACxC;AAAA,EAEU,aAAa;AACrB,SAAK,IAAI,iBAAiB;AAC1B,SAAK,KAAK,YAAY;AAAA,EACxB;AAAA,EAEU,OAAO,SAAgB;AAC/B,SAAK,QAAQ,IAAI,GAAG,OAAO;AAAA,EAC7B;AAAA,EAEA,UAAU;AACR,SAAK,IAAI,WAAW;AACpB,SAAK,mBAAmB;AACxB,SAAK,OAAO;AAAA,EACd;AACF;;;ACjGA,oBAA4B;AAGrB,IAAM,YAAN,cAAwB,MAAM;AAAA,EAGnC,cAAc;AACZ,UAAM,EAAE,cAAc,GAAG,CAAC;AAH5B,SAAQ,IAAI;AAAA,EAIZ;AAAA,EAEA,SAAS;AACP,UAAM,SAAS,IAAI,0BAAY;AAG/B,UAAM,UAAU,qBAAqB,KAAK,GAAG;AAC7C,SAAK,oBAAoB,OAAO;AAChC,WAAO,MAAM,OAAO;AACpB,WAAO,IAAI;AACX,WAAO;AAAA,EACT;AAAA,EAEA,SAAS;AAAA,EAAC;AACZ;;;ACtBA,oBAA4B;AAC5B,2BAAmB;AACnB,IAAAA,iBAAsC;AAGtC,qBAAAC,QAAO,cAAc,cAAAC,QAAgB,IAAI;AAGlC,SAAS,aACd,aACA,aAAa,MACb,WAAW,IACX;AACA,QAAM,YAAY,IAAI,2BAAY;AAClC,2BAAAD,SAAO,WAAW,EACf,cAAc,CAAC,EACf,eAAe,UAAU,EACzB,WAAW,QAAQ,QAAQ,IAAI,EAC/B,OAAO,IAAI,QAAQ,IAAI,EACvB,GAAG,SAAS,CAAC,UAAU;AACtB,YAAQ,MAAM,kCAAkC,MAAM,OAAO;AAAA,EAC/D,CAAC,EACA,KAAK,SAAS;AACjB,SAAO;AACT;AAGO,SAAS,cAAc,aAAuB,aAAa,MAAO;AACvE,QAAM,aAAa,IAAI,2BAAY;AACnC,mBAAa,qBAAAA,SAAO,WAAW,GAAG,UAAU,EAAE,KAAK,UAAU;AAC7D,SAAO;AACT;AAGO,SAAS,iBAAiB,aAAuB,aAAa,MAAO;AAC1E,QAAM,aAAa,IAAI,2BAAY;AACnC,mBAAa,qBAAAA,SAAO,WAAW,GAAG,UAAU,EACzC,YAAY,OAAO,EACnB,aAAa,CAAC,YAAY,aAAa,OAAO,CAAC,EAC/C,KAAK,UAAU;AAClB,SAAO;AACT;AAEA,SAAS,aAAa,eAAqC,aAAa,MAAO;AAC7E,SAAO,cACJ,cAAc,CAAC,EACf,eAAe,UAAU,EACzB,WAAW,SAAS,EACpB,OAAO,MAAM,EACb,cAAc;AAAA,IACb;AAAA,IACA;AAAA,IACA,OAAO,UAAU;AAAA,IACjB;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC,EACA,GAAG,SAAS,CAAC,UAAU;AACtB,YAAQ,MAAM,8BAA8B,MAAM,OAAO;AAAA,EAC3D,CAAC;AACL;;;AC1DO,IAAK,mBAAL,kBAAKE,sBAAL;AACL,EAAAA,oCAAA,gBAAa,QAAb;AACA,EAAAA,oCAAA,kBAAe,QAAf;AACA,EAAAA,oCAAA,cAAW,QAAX;AAHU,SAAAA;AAAA,GAAA;AAML,IAAM,eAAN,cAA2B,MAAM;AAAA,EAGtC,YAAY,MAAc,SAAiB;AACzC,UAAM,OAAO;AACb,SAAK,OAAO;AAAA,EACd;AACF;AAEO,SAAS,YAAY,QAAmB,OAAgB;AAC7D,MAAI,iBAAiB,cAAc;AACjC,WAAO,MAAM,MAAM,MAAM,MAAM,OAAO;AAAA,EACxC,OAAO;AACL,YAAQ,MAAM,KAAK;AACnB,WAAO,MAAM,IAAI;AAAA,EACnB;AACA,SAAO,UAAU;AACnB;;;ACzBO,IAAM,SAAN,MAAa;AAAA,EAClB,YAA6B,MAAc;AAAd;AAAA,EAAe;AAAA,EAE5C,OAAO,SAAgB;AACrB,UAAM,OAAO,QAAQ,OAAO,EAAE,QAAQ,CAAC;AACvC,YAAQ,IAAI,IAAI,KAAK,IAAI,IAAI,IAAI,KAAK,GAAG,OAAO;AAAA,EAClD;AACF;;;ACPA,IAAAC,iBAA8C;;;ACIvC,IAAK,wBAAL,kBAAKC,2BAAL;AACL,EAAAA,uBAAA,mBAAgB;AAChB,EAAAA,uBAAA,kBAAe;AACf,EAAAA,uBAAA,UAAO;AAHG,SAAAA;AAAA,GAAA;AAML,IAAK,wBAAL,kBAAKC,2BAAL;AACL,EAAAA,uBAAA,aAAU;AACV,EAAAA,uBAAA,gCAA6B;AAC7B,EAAAA,uBAAA,2BAAwB;AACxB,EAAAA,uBAAA,gBAAa;AACb,EAAAA,uBAAA,aAAU;AACV,EAAAA,uBAAA,cAAW;AAND,SAAAA;AAAA,GAAA;;;ADDL,IAAM,gBAAN,MAAoB;AAAA,EAUzB,YAAY,QAAmB,QAAuB;AATtD,SAAO,SAA2B;AAClC,SAAO,SAA+B;AAGtC,SAAQ,YAAY,KAAK,IAAI;AAsD7B,SAAQ,UAAU,MAAM;AACtB,UAAI,CAAC,KAAK,OAAQ;AAClB,WAAK,IAAI,mBAAmB;AAC5B,YAAM,WAAW,KAAK,OAAO,KAAK,IAAI,IAAI,KAAK,aAAa,GAAI;AAGhE,WAAK,OAAO,MAAM,QAAQ;AAC1B,WAAK,OAAO,IAAI,QAAQ;AACxB,WAAK,OAAO,IAAI,QAAQ;AAGxB,WAAK,OAAO,QAAQ;AAAA,QAClB,cAAc,KAAK,OAAO,MAAM,aAAa,MAAM,CAAC;AAAA;AAAA,QACpD;AAAA,MACF,CAAC;AAGD,WAAK,SAAS;AACd,WAAK,SAAS;AAAA,IAChB;AAEA,SAAQ,YAAY,OAAO,YAAoB;AAC7C,UAAI,QAAQ,eAAe,EAAG;AAC9B,UAAI,CAAC,OAAO,SAAS,OAAO,GAAG;AAC7B,aAAK,IAAI,yBAAyB;AAClC;AAAA,MACF;AAGA,UAAI,QAAQ,aAAa,IAAI;AAC3B,cAAM,MAAM,QAAQ,SAAS;AAC7B,aAAK,IAAI,YAAY,GAAG,EAAE;AAE1B,YAAI,6CAA6C;AAE/C,gBAAM,KAAK,gBAAgB;AAAA,QAC7B,WAAW,2BAAoC;AAE7C,gBAAM,KAAK,OAAO;AAAA,QACpB,WAAW,2CAA4C;AAErD,gBAAM,KAAK,eAAe;AAAA,QAC5B;AAAA,MACF,WAGS,KAAK,mBAAmB;AAC/B,aAAK,IAAI,mBAAmB,QAAQ,UAAU,SAAS;AACvD,aAAK,kBAAkB,MAAM,OAAO;AAAA,MACtC;AAAA,IACF;AA6BA,SAAQ,eAAe,OAAO,eAAuB;AACnD,UAAI,CAAC,KAAK,OAAQ;AAClB,WAAK,IAAI,qBAAqB,UAAU,GAAG;AAC3C,WAAK,OAAO,MAAM,eAAe,UAAU;AAG3C,UAAI,CAAC,KAAK,mBAAmB;AAC3B,aAAK,IAAI,kCAAkC;AAC3C,aAAK,OAAO;AAAA,MACd;AAAA,IACF;AAzIE,SAAK,SAAS;AACd,SAAK,SAAS;AACd,SAAK,IAAI,cAAc;AAGvB,SAAK,OAAO,IAAI,GAAG,cAAc,KAAK,YAAY;AAGlD,SAAK,OAAO,MAAM;AAAA,MAAG;AAAA,MAAW,CAAC,YAC/B,KAAK,QAAQ;AAAA,QACX,0BAAgC,IAAI,KAAK,UAAU,OAAO,CAAC;AAAA,MAC7D;AAAA,IACF;AACA,SAAK,OAAO,MAAM;AAAA,MAAG;AAAA,MAAyB,MAC5C,KAAK,QAAQ,wDAAgD;AAAA,IAC/D;AACA,SAAK,OAAO,MAAM;AAAA,MAAG;AAAA,MAA8B,MACjD,KAAK,QAAQ,kEAAqD;AAAA,IACpE;AACA,SAAK,OAAO,MAAM;AAAA,MAAG;AAAA,MAAc,MACjC,KAAK,QAAQ,kCAAqC;AAAA,IACpD;AACA,SAAK,OAAO,MAAM;AAAA,MAAG;AAAA,MAAW,MAC9B,KAAK,QAAQ,4BAAkC;AAAA,IACjD;AACA,SAAK,OAAO,MAAM;AAAA,MAAG;AAAA,MAAY,CAAC,aAChC,KAAK,QAAQ;AAAA,QACX,4BAAiC,IAAI,KAAK,UAAU,QAAQ,CAAC;AAAA,MAC/D;AAAA,IACF;AAGA,SAAK,iBAAiB;AAGtB,WAAO,GAAG,SAAS,KAAK,OAAO;AAC/B,WAAO,GAAG,WAAW,KAAK,SAAS;AAAA,EACrC;AAAA,EAEQ,OAAO,SAAgB;AAC7B,SAAK,QAAQ,IAAI,GAAG,OAAO;AAAA,EAC7B;AAAA,EAEQ,SAAS;AACf,SAAK,QAAQ,IAAI,OAAO;AACxB,SAAK,QAAQ,MAAM,OAAO;AAAA,EAC5B;AAAA,EAsDA,MAAc,SAAS;AACrB,SAAK,mBAAmB,IAAI;AAC5B,SAAK,oBAAoB;AACzB,SAAK,OAAO;AAAA,EACd;AAAA,EAEA,MAAc,kBAAkB;AAC9B,QAAI,CAAC,KAAK,OAAQ;AAClB,SAAK,mBAAmB,IAAI;AAC5B,SAAK,oBAAoB,IAAI,2BAAY;AACzC,SAAK,OAAO,IAAI,WAAW,KAAK,iBAAiB;AACjD,SAAK,OAAO;AAAA,EACd;AAAA,EAEA,MAAc,iBAAiB;AAC7B,SAAK,mBAAmB,IAAI;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,eAAe,KAAK,QAAQ,MAAM;AACxC,QAAI,gBAAgB,aAAa,aAAa,SAAS,CAAC,EAAE,SAAS,QAAQ;AACzE,WAAK;AAAA,QACH;AAAA,MACF;AACA,WAAK,OAAO;AAAA,IACd;AAAA,EACF;AAAA,EAcA,MAAc,mBAAmB;AAC/B,QAAI,CAAC,KAAK,OAAQ;AAClB,QAAI;AACF,UAAI,KAAK,OAAO,cAAc;AAE5B,aAAK,OAAO,MAAM,oBAAoB,KAAK,OAAO,YAAY;AAC9D,cAAM,KAAK,MAAM,KAAK,OAAO,YAAY;AAAA,MAC3C,WAAW,KAAK,OAAO,sBAAsB;AAE3C,cAAM,KAAK,OAAO;AAAA,MACpB,OAAO;AAGL,aAAK,QAAQ,kCAAqC;AAAA,MACpD;AAAA,IACF,SAAS,OAAO;AACd,cAAQ,MAAM,mBAAmB,KAAK;AACtC,WAAK,QAAQ,kCAAqC;AAAA,IACpD;AAAA,EACF;AAAA,EAEA,MAAc,SAAS;AACrB,QAAI,CAAC,KAAK,OAAQ;AAClB,SAAK,OAAO;AACZ,QAAI;AAEF,YAAM,SAAS,KAAK,OAAO,MAAM,OAAO;AAGxC,YAAM,KAAK,MAAM,MAAM;AAAA,IACzB,SAAS,OAAO;AACd,cAAQ,MAAM,mBAAmB,KAAK;AACtC,WAAK,QAAQ,kCAAqC;AAAA,IACpD;AAAA,EACF;AAAA;AAAA,EAGA,MAAc,MAAM,SAA4B;AAC9C,QAAI,CAAC,KAAK,UAAU,CAAC,KAAK,OAAQ;AAGlC,QAAI;AACJ,QAAI,OAAO,YAAY,UAAU;AAC/B,YAAM,SAAS,IAAI,2BAAY;AAC/B,aAAO,MAAM,OAAO;AACpB,aAAO,IAAI;AACX,mBAAa;AAAA,IACf,OAAO;AACL,mBAAa;AAAA,IACf;AAGA,UAAM,QAAQ,KAAK,OAAO,IAAI,MAAM,UAAU;AAG9C,UAAM,KAAK,UAAU,KAAK;AAAA,EAC5B;AAAA,EAEA,MAAc,UAAU,OAAiB;AACvC,QAAI,CAAC,KAAK,OAAQ;AAClB,QAAI,CAAC,MAAM,UAAU;AACnB,WAAK,IAAI,gCAAgC,KAAK;AAC9C;AAAA,IACF;AAGA,UAAM,GAAG,QAAQ,CAAC,UAAU;AAC1B,WAAK,IAAI,qBAAqB,MAAM,UAAU,SAAS;AACvD,WAAK,QAAQ,KAAK,KAAK;AAAA,IACzB,CAAC;AACD,UAAM,GAAG,SAAS,CAAC,UAAU;AAC3B,WAAK,IAAI,yBAAyB,KAAK;AAAA,IACzC,CAAC;AACD,UAAM,GAAG,OAAO,MAAM;AACpB,WAAK,IAAI,oBAAoB;AAAA,IAC/B,CAAC;AAAA,EACH;AACF;;;AE5OA,IAAAC,wBAA6B;AAK7B,IAAM,yBAAyB;AAAA,EAC7B,aAAa;AAAA,EACb,aAAa;AAAA,EACb,cAAc;AAAA,EACd,cAAc;AAAA,EACd,aAAa;AAAA,EACb,cAAc;AAChB;AAMO,IAAe,MAAf,cAA2B,mCAAwB;AAAA;AAAA,EAKxD,WAAW,aAAuB;AAEhC,gBAAY,KAAK,QAAQ,CAAC,UAAU;AAClC,WAAK,WAAW,KAAK,eAAe,KAAK;AAAA,IAC3C,CAAC;AAAA,EACH;AAAA,EAEU,OAAO,SAAgB;AAC/B,SAAK,QAAQ,IAAI,GAAG,OAAO;AAAA,EAC7B;AAAA,EAEA,UAAU;AACR,SAAK,IAAI,WAAW;AACpB,SAAK,mBAAmB;AAAA,EAC1B;AAAA,EAEA,IAAc,YAAoB;AAChC,WAAQ,KAAK,YAAY,uBAAuB,KAAK,QAAQ,KAAM;AAAA,EACrE;AAAA,EAEQ,eACN,OACqC;AACrC,QAAI,CAAC,SAAS,MAAM,eAAe,GAAG;AACpC,YAAM,IAAI,MAAM,0CAA0C;AAAA,IAC5D;AAEA,UAAM,MAAM,IAAI,WAAW,KAAK;AAGhC,QACE,IAAI,CAAC,MAAM,MACX,IAAI,CAAC,MAAM,MACX,IAAI,CAAC,MAAM,OACX,IAAI,CAAC,MAAM,KACX;AACA,aAAO;AAAA,IACT;AAEA,QACE,IAAI,CAAC,MAAM,MACX,IAAI,CAAC,MAAM,OACX,IAAI,CAAC,MAAM,OACX,IAAI,CAAC,MAAM,IACX;AACA,aAAO;AAAA,IACT;AAEA,QACE,IAAI,CAAC,MAAM,MACX,IAAI,CAAC,MAAM,MACX,IAAI,CAAC,MAAM,MACX,IAAI,CAAC,MAAM,MACX,IAAI,CAAC,MAAM,MACX,IAAI,CAAC,MAAM,MACX,IAAI,EAAE,MAAM,MACZ,IAAI,EAAE,MAAM,IACZ;AACA,aAAO;AAAA,IACT;AAEA,QAAI,IAAI,CAAC,MAAM,MAAQ,IAAI,CAAC,MAAM,MAAQ,IAAI,CAAC,MAAM,IAAM;AACzD,aAAO;AAAA,IACT;AAEA,QACE,IAAI,CAAC,MAAM,OACX,IAAI,CAAC,MAAM,OACX,IAAI,CAAC,MAAM,OACX,IAAI,CAAC,MAAM,KACX;AACA,aAAO;AAAA,IACT;AAEA,QACE,IAAI,CAAC,MAAM,OACX,IAAI,CAAC,MAAM,MACX,IAAI,CAAC,MAAM,MACX,IAAI,CAAC,MAAM,IACX;AACA,aAAO;AAAA,IACT;AACA,SAAK,IAAI,6CAA6C,KAAK;AAC3D,WAAO;AAAA,EACT;AACF;;;ACrGO,IAAe,UAAf,cAA+B,IAAI;AAAA,EAGxC,WAAW,aAAuB;AAChC,UAAM,WAAW,WAAW;AAG5B,SAAK,IAAI,8BAA8B;AAEvC,UAAM,SAAmB,CAAC;AAC1B,gBAAY,GAAG,QAAQ,CAAC,UAAU;AAChC,aAAO,KAAK,KAAK;AAAA,IACnB,CAAC;AAED,gBAAY,GAAG,OAAO,YAAY;AAChC,UAAI,OAAO,WAAW,EAAG;AACzB,YAAM,cAAc,OAAO,OAAO,MAAM;AACxC,YAAM,OAAO,IAAI,KAAK,CAAC,WAAW,GAAG,SAAS,KAAK,SAAS,IAAI;AAAA,QAC9D,MAAM,KAAK;AAAA,MACb,CAAC;AAGD,WAAK,IAAI,sBAAsB;AAC/B,YAAM,aAAa,MAAM,KAAK,eAAe,IAAI;AACjD,WAAK,KAAK,cAAc,UAAU;AAAA,IACpC,CAAC;AAAA,EACH;AACF;;;AChCO,IAAM,UAAN,cAAsB,QAAQ;AAAA,EAA9B;AAAA;AACL,SAAQ,IAAI;AAAA;AAAA,EAEZ,MAAM,eAAe,MAAY;AAC/B,WAAO,gBAAgB,KAAK,GAAG;AAAA,EACjC;AACF;;;ACRA,SAAoB;AACpB,IAAAC,iBAAsC;;;ACE/B,IAAe,MAAf,MAAmB;AAAA,EAMd,OAAO,SAAgB;AAC/B,SAAK,QAAQ,IAAI,GAAG,OAAO;AAAA,EAC7B;AAAA,EAEA,UAAU;AACR,SAAK,IAAI,WAAW;AACpB,SAAK,OAAO;AAAA,EACd;AACF;;;ADbO,IAAM,UAAN,cAAsB,IAAI;AAAA,EAC/B,YAAoB,gBAA0B;AAC5C,UAAM;AADY;AAAA,EAEpB;AAAA,EAEA,MAAM,YAAsB;AAC1B,UAAM,cAAc,IAAI,2BAAY;AACpC,eAAW,KAAK,QAAQ,YAAY;AAClC,iBAAW,YAAY,KAAK,gBAAgB;AAC1C,cAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,GAAG,CAAC;AACvD,cAAM,cAAiB,gBAAa,QAAQ;AAC5C,aAAK,IAAI,iBAAiB,YAAY,MAAM,SAAS;AACrD,oBAAY,MAAM,WAAW;AAAA,MAC/B;AACA,kBAAY,IAAI;AAAA,IAClB,CAAC;AACD,WAAO;AAAA,EACT;AAAA,EAEA,SAAS;AAAA,EAAC;AACZ;;;AErBA,eAAsB,cACpB,QACA,UACqB;AACrB,SAAO,IAAI,QAAoB,CAAC,SAAS,WAAW;AAElD,UAAM,UAAU,WAAW,MAAM;AAC/B,aAAO,IAAI,oCAA0C,gBAAgB,CAAC;AAAA,IACxE,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,oCAA0C,gBAAgB,CAAC;AAAA,MACxE;AAAA,IACF;AAGA,WAAO,GAAG,WAAW,QAAQ;AAAA,EAC/B,CAAC;AACH;","names":["import_stream","ffmpeg","ffmpegInstaller","MicdropErrorCode","import_stream","MicdropClientCommands","MicdropServerCommands","import_eventemitter3","import_stream"]}
1
+ {"version":3,"sources":["../src/index.ts","../src/agent/Agent.ts","../src/agent/tools.ts","../src/agent/MockAgent.ts","../src/audio-convert.ts","../src/errors.ts","../src/Logger.ts","../src/MicdropServer.ts","../src/types.ts","../src/stt/STT.ts","../src/stt/FileSTT.ts","../src/stt/MockSTT.ts","../src/tts/MockTTS.ts","../src/tts/TTS.ts","../src/waitForParams.ts"],"sourcesContent":["export * from './agent'\nexport * from './audio-convert'\nexport * from './errors'\nexport * from './Logger'\nexport * from './MicdropServer'\nexport * from './stt'\nexport * from './tts'\nexport * from './types'\nexport * from './waitForParams'\n","import { EventEmitter } from 'eventemitter3'\nimport { Readable } from 'stream'\nimport type { z } from 'zod'\nimport { Logger } from '../Logger'\nimport {\n MicdropAnswerMetadata,\n MicdropConversation,\n MicdropConversationItem,\n MicdropConversationMessage,\n MicdropConversationToolCall,\n MicdropConversationToolResult,\n MicdropToolCall,\n} from '../types'\nimport {\n AUTO_END_CALL_PROMPT,\n AUTO_END_CALL_TOOL_NAME,\n AUTO_IGNORE_USER_NOISE_PROMPT,\n AUTO_IGNORE_USER_NOISE_TOOL_NAME,\n AUTO_SEMANTIC_TURN_PROMPT,\n AUTO_SEMANTIC_TURN_TOOL_NAME,\n Tool,\n} from './tools'\n\nexport interface AgentOptions {\n systemPrompt: string\n\n // Enable auto ending of the call when user asks to end the call\n // You can provide a custom prompt to use instead of the default one by passing a string\n autoEndCall?: boolean | string\n\n // Enable detection of an incomplete sentence, and skip the answer (assistant waits)\n // You can provide a custom prompt to use instead of the default one by passing a string\n autoSemanticTurn?: boolean | string\n\n // Ignore of the last user message when it's meaningless\n // You can provide a custom prompt to use instead of the default one by passing a string\n autoIgnoreUserNoise?: boolean | string\n\n // Extract a value from the answer\n // Value must be at the end of the answer, in JSON or between tags\n extract?: ExtractJsonOptions | ExtractTagOptions\n}\n\nexport interface AgentEvents {\n Message: [MicdropConversationItem]\n CancelLastUserMessage: []\n SkipAnswer: []\n EndCall: []\n ToolCall: [MicdropToolCall]\n}\n\nexport interface ExtractOptions {\n callback?: (value: string) => void\n saveInMetadata?: boolean\n}\n\nexport interface ExtractJsonOptions extends ExtractOptions {\n json: true\n callback?: (value: any) => void\n}\n\nexport interface ExtractTagOptions extends ExtractOptions {\n startTag: string\n endTag: string\n}\n\nexport abstract class Agent<\n Options extends AgentOptions = AgentOptions,\n> extends EventEmitter<AgentEvents> {\n public logger?: Logger\n public conversation: MicdropConversation\n\n protected tools: Tool[]\n\n constructor(protected options: Options) {\n super()\n this.conversation = [{ role: 'system', content: options.systemPrompt }]\n this.tools = this.getDefaultTools()\n }\n\n abstract answer(): Readable\n abstract cancel(): void\n\n addUserMessage(text: string, metadata?: MicdropAnswerMetadata) {\n this.addMessage('user', text, metadata)\n }\n\n addAssistantMessage(text: string, metadata?: MicdropAnswerMetadata) {\n this.addMessage('assistant', text, metadata)\n }\n\n addTool<Schema extends z.ZodObject>(tool: Tool<Schema>) {\n this.tools.push(tool)\n }\n\n removeTool(name: string) {\n this.tools = this.tools.filter((tool) => tool.name !== name)\n }\n\n getTool(name: string): Tool | undefined {\n return this.tools.find((tool) => tool.name === name)\n }\n\n protected addMessage(\n role: 'user' | 'assistant' | 'system',\n text: string,\n metadata?: MicdropAnswerMetadata\n ) {\n this.log(`Adding ${role} message to conversation: ${text}`)\n const message: MicdropConversationMessage = {\n role,\n content: text,\n metadata,\n }\n this.conversation.push(message)\n this.emit('Message', message)\n }\n\n protected addToolMessage(\n message: MicdropConversationToolCall | MicdropConversationToolResult\n ) {\n this.log('Adding tool message:', message)\n this.conversation.push(message)\n this.emit('Message', message)\n }\n\n protected endCall() {\n this.log('Ending call')\n this.emit('EndCall')\n }\n\n protected cancelLastUserMessage() {\n this.log('Cancelling last user message')\n const lastMessageIndex = this.conversation.findLastIndex(\n (message) => message.role === 'user'\n )\n if (lastMessageIndex !== -1) {\n this.conversation.splice(lastMessageIndex, 1)\n }\n this.emit('CancelLastUserMessage')\n }\n\n protected skipAnswer() {\n this.log('Skipping answer')\n this.emit('SkipAnswer')\n }\n\n protected getDefaultTools() {\n const tools: Tool[] = []\n if (this.options.autoEndCall) {\n tools.push({\n name: AUTO_END_CALL_TOOL_NAME,\n description:\n typeof this.options.autoEndCall === 'string'\n ? this.options.autoEndCall\n : AUTO_END_CALL_PROMPT,\n execute: () => this.endCall(),\n })\n }\n if (this.options.autoSemanticTurn) {\n tools.push({\n name: AUTO_SEMANTIC_TURN_TOOL_NAME,\n description:\n typeof this.options.autoSemanticTurn === 'string'\n ? this.options.autoSemanticTurn\n : AUTO_SEMANTIC_TURN_PROMPT,\n skipAnswer: true,\n execute: () => this.skipAnswer(),\n })\n }\n if (this.options.autoIgnoreUserNoise) {\n tools.push({\n name: AUTO_IGNORE_USER_NOISE_TOOL_NAME,\n description:\n typeof this.options.autoIgnoreUserNoise === 'string'\n ? this.options.autoIgnoreUserNoise\n : AUTO_IGNORE_USER_NOISE_PROMPT,\n skipAnswer: true,\n execute: () => this.cancelLastUserMessage(),\n })\n }\n return tools\n }\n\n protected async executeTool(toolCall: MicdropConversationToolCall) {\n try {\n const tool = this.getTool(toolCall.toolName)\n if (!tool) {\n throw new Error(`Tool not found \"${toolCall.toolName}\"`)\n }\n\n this.log('Executing tool:', toolCall.toolName, toolCall.parameters)\n\n // Save tool call in conversation\n this.addToolMessage(toolCall)\n\n const parameters = JSON.parse(toolCall.parameters)\n const output = tool.execute ? await tool.execute(parameters) : {}\n\n // Save tool result in conversation\n this.addToolMessage({\n role: 'tool_result',\n toolCallId: toolCall.toolCallId,\n toolName: toolCall.toolName,\n output: JSON.stringify(output ?? null),\n })\n\n // Emit output\n if (tool.emitOutput) {\n this.emit('ToolCall', {\n name: toolCall.toolName,\n parameters,\n output,\n })\n }\n\n return {\n output,\n skipAnswer: tool.skipAnswer,\n }\n } catch (error: any) {\n console.error('[OpenaiAgent] Error executing tool:', error)\n return {\n output: {\n error: error.message,\n },\n }\n }\n }\n\n protected getExtractOptions(): ExtractTagOptions | undefined {\n const extract = this.options.extract\n if (!extract) return undefined\n if ('json' in extract && extract.json) {\n return { ...extract, startTag: '{', endTag: '}' }\n }\n if ('startTag' in extract && 'endTag' in extract) {\n return extract\n }\n return undefined\n }\n\n protected extract(message: string) {\n const extractOptions = this.getExtractOptions()\n let metadata: MicdropAnswerMetadata | undefined = undefined\n\n // Extract value?\n if (extractOptions) {\n const startTagIndex = message.indexOf(extractOptions.startTag)\n if (startTagIndex !== -1) {\n // Find end tag\n let endTagIndex = message.lastIndexOf(extractOptions.endTag)\n if (endTagIndex === -1) endTagIndex = message.length + 1\n else endTagIndex += extractOptions.endTag.length\n const extractedText = message.slice(startTagIndex, endTagIndex).trim()\n\n // Parse extracted value\n try {\n const extractedValue =\n 'json' in extractOptions && extractOptions.json\n ? JSON.parse(extractedText)\n : extractedText\n\n // Call callback\n if (extractOptions.callback) {\n extractOptions.callback(extractedValue)\n }\n\n // Save in metadata\n if (extractOptions.saveInMetadata) {\n metadata = { extracted: extractedValue }\n }\n } catch (error) {\n console.error(\n `[OpenaiAgent] Error parsing extracted value (${extractedText}):`,\n error\n )\n }\n\n // Remove extracted value from message\n message = message.slice(0, startTagIndex).trimEnd()\n }\n }\n return { message, metadata }\n }\n\n protected log(...message: any[]) {\n this.logger?.log(...message)\n }\n\n destroy() {\n this.log('Destroyed')\n this.removeAllListeners()\n this.cancel()\n }\n}\n","import type { z } from 'zod'\n\nexport interface Tool<Schema extends z.ZodObject = z.ZodObject> {\n name: string\n description: string\n inputSchema?: Schema\n execute?: (input: z.infer<Schema>) => any | Promise<any>\n skipAnswer?: boolean\n emitOutput?: boolean\n}\n\nexport const AUTO_END_CALL_TOOL_NAME = 'end_call'\nexport const AUTO_END_CALL_PROMPT =\n 'Call this tool only if user asks to end the call'\n\nexport const AUTO_SEMANTIC_TURN_TOOL_NAME = 'semantic_turn'\nexport const AUTO_SEMANTIC_TURN_PROMPT =\n 'Call this tool only if last user message is obviously an incomplete sentence that you need to wait for the end before answering'\n\nexport const AUTO_IGNORE_USER_NOISE_TOOL_NAME = 'ignore_user_noise'\nexport const AUTO_IGNORE_USER_NOISE_PROMPT =\n 'Call this tool only if last user message is just an interjection or a sound that expresses emotion, hesitation, or reaction (ex: \"Uh\", \"Ahem\", \"Hmm\", \"Ah\") but doesn\\'t carry any clear meaning like agreeing, refusing, or commanding'\n","import { PassThrough } from 'stream'\nimport { Agent } from './Agent'\n\nexport class MockAgent extends Agent {\n private i = 0\n\n constructor() {\n super({ systemPrompt: '' })\n }\n\n answer() {\n const stream = new PassThrough()\n\n // Answer message\n const message = `Assistant Message ${this.i++}`\n this.addAssistantMessage(message)\n stream.write(message)\n stream.end()\n return stream\n }\n\n cancel() {}\n}\n","import ffmpegInstaller from '@ffmpeg-installer/ffmpeg'\nimport ffmpeg from 'fluent-ffmpeg'\nimport { PassThrough, Readable } from 'stream'\n\n// Setup ffmpeg\nffmpeg.setFfmpegPath(ffmpegInstaller.path)\n\n// Convert stream to WAV/PCM\nexport function convertToPCM(\n audioStream: Readable,\n sampleRate = 16000,\n bitDepth = 16\n) {\n const pcmStream = new PassThrough()\n ffmpeg(audioStream)\n .audioChannels(1)\n .audioFrequency(sampleRate)\n .audioCodec(`pcm_s${bitDepth}le`)\n .format(`s${bitDepth}le`)\n .on('error', (error) => {\n console.error('Error converting audio stream:', error.message)\n })\n .pipe(pcmStream)\n return pcmStream\n}\n\n// Convert PCM stream to WebM/Opus\nexport function convertToOpus(audioStream: Readable, sampleRate = 16000) {\n const webmStream = new PassThrough()\n ffmpegToOpus(ffmpeg(audioStream), sampleRate).pipe(webmStream)\n return webmStream\n}\n\n// Convert PCM stream to WebM/Opus\nexport function convertPCMToOpus(audioStream: Readable, sampleRate = 16000) {\n const webmStream = new PassThrough()\n ffmpegToOpus(ffmpeg(audioStream), sampleRate)\n .inputFormat('s16le')\n .inputOptions(['-f s16le', '-ar 16000', '-ac 1'])\n .pipe(webmStream)\n return webmStream\n}\n\nfunction ffmpegToOpus(ffmpegCommand: ffmpeg.FfmpegCommand, sampleRate = 16000) {\n return ffmpegCommand\n .audioChannels(1)\n .audioFrequency(sampleRate)\n .audioCodec('libopus')\n .format('webm')\n .outputOptions([\n '-application audio',\n `-ac 1`,\n `-ar ${sampleRate}`,\n `-b:a 64k`,\n `-f webm`,\n `-map_metadata -1`,\n ])\n .on('error', (error) => {\n console.error('Error converting to Opus: ', error.message)\n })\n}\n","import WebSocket from 'ws'\n\nexport enum MicdropErrorCode {\n BadRequest = 4400,\n Unauthorized = 4401,\n NotFound = 4404,\n}\n\nexport class MicdropError extends Error {\n code: number\n\n constructor(code: number, message: string) {\n super(message)\n this.code = code\n }\n}\n\nexport function handleError(socket: WebSocket, error: unknown) {\n if (error instanceof MicdropError) {\n socket.close(error.code, error.message)\n } else {\n console.error(error)\n socket.close(1011)\n }\n socket.terminate()\n}\n","export class Logger {\n constructor(private readonly name: string) {}\n\n log(...message: any[]) {\n const time = process.uptime().toFixed(3)\n console.log(`[${this.name} ${time}]`, ...message)\n }\n}\n","import { Duplex, PassThrough, Readable } from 'stream'\nimport { WebSocket } from 'ws'\nimport type { Agent } from './agent'\nimport { Logger } from './Logger'\nimport type { STT } from './stt'\nimport type { TTS } from './tts'\nimport {\n MicdropCallSummary,\n MicdropClientCommands,\n MicdropConversationItem,\n MicdropServerCommands,\n} from './types'\n\nexport interface MicdropConfig {\n firstMessage?: string\n generateFirstMessage?: boolean\n agent: Agent\n stt: STT\n tts: TTS\n onEnd?(call: MicdropCallSummary): void\n}\n\nexport class MicdropServer {\n public socket: WebSocket | null = null\n public config: MicdropConfig | null = null\n public logger?: Logger\n\n private startTime = Date.now()\n private lastMessageSpeeched?: MicdropConversationItem\n\n // When user is speaking, we're streaming chunks for STT\n private currentUserStream?: Duplex\n\n constructor(socket: WebSocket, config: MicdropConfig) {\n this.socket = socket\n this.config = config\n this.log(`Call started`)\n\n // Setup STT\n this.config.stt.on('Transcript', this.onTranscript)\n\n // Setup agent\n this.config.agent.on('Message', (message) =>\n this.socket?.send(\n `${MicdropServerCommands.Message} ${JSON.stringify(message)}`\n )\n )\n this.config.agent.on('CancelLastUserMessage', () =>\n this.socket?.send(MicdropServerCommands.CancelLastUserMessage)\n )\n this.config.agent.on('SkipAnswer', () =>\n this.socket?.send(MicdropServerCommands.SkipAnswer)\n )\n this.config.agent.on('EndCall', () =>\n this.socket?.send(MicdropServerCommands.EndCall)\n )\n this.config.agent.on('ToolCall', (toolCall) =>\n this.socket?.send(\n `${MicdropServerCommands.ToolCall} ${JSON.stringify(toolCall)}`\n )\n )\n\n // Assistant speaks first\n this.sendFirstMessage()\n\n // Listen to events\n socket.on('close', this.onClose)\n socket.on('message', this.onMessage)\n }\n\n private log(...message: any[]) {\n this.logger?.log(...message)\n }\n\n private cancel() {\n this.config?.tts.cancel()\n this.config?.agent.cancel()\n }\n\n private onClose = () => {\n if (!this.config) return\n this.log('Connection closed')\n const duration = Math.round((Date.now() - this.startTime) / 1000)\n\n // Destroy instances\n this.config.agent.destroy()\n this.config.stt.destroy()\n this.config.tts.destroy()\n\n // End call callback\n this.config.onEnd?.({\n conversation: this.config.agent.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 onMessage = async (message: Buffer) => {\n if (message.byteLength === 0) return\n if (!Buffer.isBuffer(message)) {\n this.log('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: ${cmd}`)\n\n if (cmd === MicdropClientCommands.StartSpeaking) {\n // User started speaking\n await this.onStartSpeaking()\n } else if (cmd === MicdropClientCommands.Mute) {\n // User muted the call\n await this.onMute()\n } else if (cmd === MicdropClientCommands.StopSpeaking) {\n // User stopped speaking\n await this.onStopSpeaking()\n }\n }\n\n // Audio chunk\n else if (this.currentUserStream) {\n this.log(`Received chunk (${message.byteLength} bytes)`)\n this.currentUserStream.write(message)\n }\n }\n\n private async onMute() {\n this.currentUserStream?.end()\n this.currentUserStream = undefined\n this.cancel()\n }\n\n private async onStartSpeaking() {\n if (!this.config) return\n this.currentUserStream?.end()\n this.currentUserStream = new PassThrough()\n this.config.stt.transcribe(this.currentUserStream)\n this.cancel()\n }\n\n private async onStopSpeaking() {\n this.currentUserStream?.end()\n this.currentUserStream = undefined\n\n const conversation = this.config?.agent.conversation\n const lastMessage = conversation?.[conversation.length - 1]\n if (\n lastMessage?.role === 'user' &&\n this.lastMessageSpeeched !== lastMessage\n ) {\n this.log(\n 'User stopped speaking and a transcript already exists, answering'\n )\n this.answer()\n }\n }\n\n private onTranscript = async (transcript: string) => {\n if (!this.config) return\n this.log(`User transcript: \"${transcript}\"`)\n this.config.agent.addUserMessage(transcript)\n\n // Answer if user stopped speaking\n if (!this.currentUserStream) {\n this.log('User stopped speaking, answering')\n this.answer()\n }\n }\n\n private async sendFirstMessage() {\n if (!this.config) return\n try {\n if (this.config.firstMessage) {\n // Send first message\n this.config.agent.addAssistantMessage(this.config.firstMessage)\n await this.speak(this.config.firstMessage)\n } else if (this.config.generateFirstMessage) {\n // Generate first message\n await this.answer()\n } else {\n // Skip answer if no first message is provided\n // to avoid keeping the client in a processing state\n this.socket?.send(MicdropServerCommands.SkipAnswer)\n }\n } catch (error) {\n console.error('[MicdropServer]', error)\n this.socket?.send(MicdropServerCommands.SkipAnswer)\n }\n }\n\n private async answer() {\n if (!this.config) return\n this.cancel()\n\n // Prevent answering twice\n const lastMessage =\n this.config.agent.conversation[this.config.agent.conversation.length - 1]\n if (this.lastMessageSpeeched === lastMessage) {\n this.log('Already answered, skipping')\n return\n }\n this.lastMessageSpeeched = lastMessage\n\n try {\n // LLM: Generate answer\n const stream = this.config.agent.answer()\n\n // TTS: Generate answer audio\n await this.speak(stream)\n } catch (error) {\n console.error('[MicdropServer]', error)\n this.socket?.send(MicdropServerCommands.SkipAnswer)\n }\n }\n\n // Run text-to-speech and send to client\n private async speak(message: string | Readable) {\n if (!this.socket || !this.config) return\n\n // Convert message to stream if needed\n let textStream: Readable\n if (typeof message === 'string') {\n const stream = new PassThrough()\n stream.write(message)\n stream.end()\n textStream = stream\n } else {\n textStream = message\n }\n\n // Run TTS\n const audio = this.config.tts.speak(textStream)\n\n // Send audio to client\n await this.sendAudio(audio)\n }\n\n private async sendAudio(audio: Readable) {\n if (!this.socket) return\n if (!audio.readable) {\n this.log('Non readable audio, skipping', audio)\n return\n }\n\n // Stream audio\n audio.on('data', (chunk) => {\n this.log(`Send audio chunk (${chunk.byteLength} bytes)`)\n this.socket?.send(chunk)\n })\n audio.on('error', (error) => {\n this.log('Error in audio stream', error)\n })\n audio.on('end', () => {\n this.log('Audio stream ended')\n })\n }\n}\n","export enum MicdropClientCommands {\n StartSpeaking = 'StartSpeaking',\n StopSpeaking = 'StopSpeaking',\n Mute = 'Mute',\n}\n\nexport enum MicdropServerCommands {\n Message = 'Message',\n CancelLastUserMessage = 'CancelLastUserMessage',\n SkipAnswer = 'SkipAnswer',\n EndCall = 'EndCall',\n ToolCall = 'ToolCall',\n}\n\nexport interface MicdropCallSummary {\n conversation: MicdropConversation\n duration: number\n}\n\nexport type MicdropConversationItem =\n | MicdropConversationMessage\n | MicdropConversationToolCall\n | MicdropConversationToolResult\n\nexport type MicdropConversation = Array<MicdropConversationItem>\n\nexport type MicdropAnswerMetadata = {\n [key: string]: any\n}\n\nexport interface MicdropConversationMessage<\n Data extends MicdropAnswerMetadata = MicdropAnswerMetadata,\n> {\n role: 'system' | 'user' | 'assistant'\n content: string\n metadata?: Data\n}\n\nexport interface MicdropConversationToolCall {\n role: 'tool_call'\n toolCallId: string\n toolName: string\n parameters: string\n}\n\nexport interface MicdropConversationToolResult {\n role: 'tool_result'\n toolCallId: string\n toolName: string\n output: string\n}\n\nexport interface MicdropToolCall {\n name: string\n parameters: any\n output: any\n}\n\nexport type DeepPartial<T> = T extends object\n ? {\n [P in keyof T]?: DeepPartial<T[P]>\n }\n : T\n","import { EventEmitter } from 'eventemitter3'\nimport { Readable } from 'stream'\nimport { Logger } from '../Logger'\n\n// Audio mime type to extension map\nconst MIME_TYPE_TO_EXTENSION = {\n 'audio/wav': 'wav',\n 'audio/ogg': 'ogg',\n 'audio/mpeg': 'mp3',\n 'audio/webm': 'webm',\n 'audio/mp4': 'mp4',\n 'audio/flac': 'flac',\n} as const\n\nexport interface STTEvents {\n Transcript: [string]\n}\n\nexport abstract class STT extends EventEmitter<STTEvents> {\n protected mimeType?: keyof typeof MIME_TYPE_TO_EXTENSION\n public logger?: Logger\n\n // Set stream of audio to transcribe\n transcribe(audioStream: Readable) {\n // Detect mime type at first chunk\n audioStream.once('data', (chunk) => {\n this.mimeType = this.detectMimeType(chunk)\n })\n }\n\n protected log(...message: any[]) {\n this.logger?.log(...message)\n }\n\n destroy() {\n this.log('Destroyed')\n this.removeAllListeners()\n }\n\n protected get extension(): string {\n return (this.mimeType && MIME_TYPE_TO_EXTENSION[this.mimeType]) || 'bin'\n }\n\n private detectMimeType(\n chunk: ArrayBuffer\n ): keyof typeof MIME_TYPE_TO_EXTENSION {\n if (!chunk || chunk.byteLength === 0) {\n throw new Error('Unable to detect mime type (empty chunk)')\n }\n\n const arr = new Uint8Array(chunk)\n\n // WEBM: 1A 45 DF A3\n if (\n arr[0] === 0x1a &&\n arr[1] === 0x45 &&\n arr[2] === 0xdf &&\n arr[3] === 0xa3\n ) {\n return 'audio/webm'\n }\n // OGG: 4F 67 67 53\n if (\n arr[0] === 0x4f &&\n arr[1] === 0x67 &&\n arr[2] === 0x67 &&\n arr[3] === 0x53\n ) {\n return 'audio/ogg'\n }\n // WAV: 52 49 46 46 ... 57 41 56 45\n if (\n arr[0] === 0x52 &&\n arr[1] === 0x49 &&\n arr[2] === 0x46 &&\n arr[3] === 0x46 &&\n arr[8] === 0x57 &&\n arr[9] === 0x41 &&\n arr[10] === 0x56 &&\n arr[11] === 0x45\n ) {\n return 'audio/wav'\n }\n // MP3: 49 44 33\n if (arr[0] === 0x49 && arr[1] === 0x44 && arr[2] === 0x33) {\n return 'audio/mpeg'\n }\n // MP4/M4A: 00 00 00 .. 66 74 79 70\n if (\n arr[4] === 0x66 &&\n arr[5] === 0x74 &&\n arr[6] === 0x79 &&\n arr[7] === 0x70\n ) {\n return 'audio/mp4'\n }\n // FLAC: 66 4c 61 43\n if (\n arr[0] === 0x66 &&\n arr[1] === 0x4c &&\n arr[2] === 0x61 &&\n arr[3] === 0x43\n ) {\n return 'audio/flac'\n }\n this.log('Unable to detect mime type, using default', chunk)\n return 'audio/wav'\n }\n}\n","import { Readable } from 'stream'\nimport { STT } from './STT'\n\n/**\n * Abstract class for STT, converting stream to file before transcribing\n */\n\nexport abstract class FileSTT extends STT {\n abstract transcribeFile(file: File): Promise<string>\n\n transcribe(audioStream: Readable) {\n super.transcribe(audioStream)\n\n // Convert stream to file\n this.log('Converting stream to file...')\n\n const chunks: Buffer[] = []\n audioStream.on('data', (chunk) => {\n chunks.push(chunk)\n })\n\n audioStream.on('end', async () => {\n if (chunks.length === 0) return\n const arrayBuffer = Buffer.concat(chunks)\n const file = new File([arrayBuffer], `audio.${this.extension}`, {\n type: this.mimeType,\n })\n\n // Transcribe file with implementation\n this.log('Transcribing file...')\n const transcript = await this.transcribeFile(file)\n this.emit('Transcript', transcript)\n })\n }\n}\n","import { FileSTT } from './FileSTT'\n\nexport class MockSTT extends FileSTT {\n private i = 0\n\n async transcribeFile(file: File) {\n return `User Message ${this.i++}`\n }\n}\n","import * as fs from 'fs'\nimport { PassThrough, Readable } from 'stream'\nimport { TTS } from './TTS'\n\nexport class MockTTS extends TTS {\n constructor(private audioFilePaths: string[]) {\n super()\n }\n\n speak(textStream: Readable) {\n const audioStream = new PassThrough()\n textStream.once('data', async () => {\n for (const filePath of this.audioFilePaths) {\n await new Promise((resolve) => setTimeout(resolve, 200))\n const audioBuffer = fs.readFileSync(filePath)\n this.log(`Loaded chunk (${audioBuffer.length} bytes)`)\n audioStream.write(audioBuffer)\n }\n audioStream.end()\n })\n return audioStream\n }\n\n cancel() {}\n}\n","import { Readable } from 'stream'\nimport { Logger } from '../Logger'\n\nexport abstract class TTS {\n public logger?: Logger\n\n abstract speak(textStream: Readable): Readable\n abstract cancel(): void\n\n protected log(...message: any[]) {\n this.logger?.log(...message)\n }\n\n destroy() {\n this.log('Destroyed')\n this.cancel()\n }\n}\n","import { WebSocket } from 'ws'\nimport { MicdropError, MicdropErrorCode } 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 MicdropError(MicdropErrorCode.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 MicdropError(MicdropErrorCode.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;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;;;ACAA,2BAA6B;;;ACWtB,IAAM,0BAA0B;AAChC,IAAM,uBACX;AAEK,IAAM,+BAA+B;AACrC,IAAM,4BACX;AAEK,IAAM,mCAAmC;AACzC,IAAM,gCACX;;;AD6CK,IAAe,QAAf,cAEG,kCAA0B;AAAA,EAMlC,YAAsB,SAAkB;AACtC,UAAM;AADc;AAEpB,SAAK,eAAe,CAAC,EAAE,MAAM,UAAU,SAAS,QAAQ,aAAa,CAAC;AACtE,SAAK,QAAQ,KAAK,gBAAgB;AAAA,EACpC;AAAA,EAKA,eAAe,MAAc,UAAkC;AAC7D,SAAK,WAAW,QAAQ,MAAM,QAAQ;AAAA,EACxC;AAAA,EAEA,oBAAoB,MAAc,UAAkC;AAClE,SAAK,WAAW,aAAa,MAAM,QAAQ;AAAA,EAC7C;AAAA,EAEA,QAAoC,MAAoB;AACtD,SAAK,MAAM,KAAK,IAAI;AAAA,EACtB;AAAA,EAEA,WAAW,MAAc;AACvB,SAAK,QAAQ,KAAK,MAAM,OAAO,CAAC,SAAS,KAAK,SAAS,IAAI;AAAA,EAC7D;AAAA,EAEA,QAAQ,MAAgC;AACtC,WAAO,KAAK,MAAM,KAAK,CAAC,SAAS,KAAK,SAAS,IAAI;AAAA,EACrD;AAAA,EAEU,WACR,MACA,MACA,UACA;AACA,SAAK,IAAI,UAAU,IAAI,6BAA6B,IAAI,EAAE;AAC1D,UAAM,UAAsC;AAAA,MAC1C;AAAA,MACA,SAAS;AAAA,MACT;AAAA,IACF;AACA,SAAK,aAAa,KAAK,OAAO;AAC9B,SAAK,KAAK,WAAW,OAAO;AAAA,EAC9B;AAAA,EAEU,eACR,SACA;AACA,SAAK,IAAI,wBAAwB,OAAO;AACxC,SAAK,aAAa,KAAK,OAAO;AAC9B,SAAK,KAAK,WAAW,OAAO;AAAA,EAC9B;AAAA,EAEU,UAAU;AAClB,SAAK,IAAI,aAAa;AACtB,SAAK,KAAK,SAAS;AAAA,EACrB;AAAA,EAEU,wBAAwB;AAChC,SAAK,IAAI,8BAA8B;AACvC,UAAM,mBAAmB,KAAK,aAAa;AAAA,MACzC,CAAC,YAAY,QAAQ,SAAS;AAAA,IAChC;AACA,QAAI,qBAAqB,IAAI;AAC3B,WAAK,aAAa,OAAO,kBAAkB,CAAC;AAAA,IAC9C;AACA,SAAK,KAAK,uBAAuB;AAAA,EACnC;AAAA,EAEU,aAAa;AACrB,SAAK,IAAI,iBAAiB;AAC1B,SAAK,KAAK,YAAY;AAAA,EACxB;AAAA,EAEU,kBAAkB;AAC1B,UAAM,QAAgB,CAAC;AACvB,QAAI,KAAK,QAAQ,aAAa;AAC5B,YAAM,KAAK;AAAA,QACT,MAAM;AAAA,QACN,aACE,OAAO,KAAK,QAAQ,gBAAgB,WAChC,KAAK,QAAQ,cACb;AAAA,QACN,SAAS,MAAM,KAAK,QAAQ;AAAA,MAC9B,CAAC;AAAA,IACH;AACA,QAAI,KAAK,QAAQ,kBAAkB;AACjC,YAAM,KAAK;AAAA,QACT,MAAM;AAAA,QACN,aACE,OAAO,KAAK,QAAQ,qBAAqB,WACrC,KAAK,QAAQ,mBACb;AAAA,QACN,YAAY;AAAA,QACZ,SAAS,MAAM,KAAK,WAAW;AAAA,MACjC,CAAC;AAAA,IACH;AACA,QAAI,KAAK,QAAQ,qBAAqB;AACpC,YAAM,KAAK;AAAA,QACT,MAAM;AAAA,QACN,aACE,OAAO,KAAK,QAAQ,wBAAwB,WACxC,KAAK,QAAQ,sBACb;AAAA,QACN,YAAY;AAAA,QACZ,SAAS,MAAM,KAAK,sBAAsB;AAAA,MAC5C,CAAC;AAAA,IACH;AACA,WAAO;AAAA,EACT;AAAA,EAEA,MAAgB,YAAY,UAAuC;AACjE,QAAI;AACF,YAAM,OAAO,KAAK,QAAQ,SAAS,QAAQ;AAC3C,UAAI,CAAC,MAAM;AACT,cAAM,IAAI,MAAM,mBAAmB,SAAS,QAAQ,GAAG;AAAA,MACzD;AAEA,WAAK,IAAI,mBAAmB,SAAS,UAAU,SAAS,UAAU;AAGlE,WAAK,eAAe,QAAQ;AAE5B,YAAM,aAAa,KAAK,MAAM,SAAS,UAAU;AACjD,YAAM,SAAS,KAAK,UAAU,MAAM,KAAK,QAAQ,UAAU,IAAI,CAAC;AAGhE,WAAK,eAAe;AAAA,QAClB,MAAM;AAAA,QACN,YAAY,SAAS;AAAA,QACrB,UAAU,SAAS;AAAA,QACnB,QAAQ,KAAK,UAAU,UAAU,IAAI;AAAA,MACvC,CAAC;AAGD,UAAI,KAAK,YAAY;AACnB,aAAK,KAAK,YAAY;AAAA,UACpB,MAAM,SAAS;AAAA,UACf;AAAA,UACA;AAAA,QACF,CAAC;AAAA,MACH;AAEA,aAAO;AAAA,QACL;AAAA,QACA,YAAY,KAAK;AAAA,MACnB;AAAA,IACF,SAAS,OAAY;AACnB,cAAQ,MAAM,uCAAuC,KAAK;AAC1D,aAAO;AAAA,QACL,QAAQ;AAAA,UACN,OAAO,MAAM;AAAA,QACf;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEU,oBAAmD;AAC3D,UAAM,UAAU,KAAK,QAAQ;AAC7B,QAAI,CAAC,QAAS,QAAO;AACrB,QAAI,UAAU,WAAW,QAAQ,MAAM;AACrC,aAAO,EAAE,GAAG,SAAS,UAAU,KAAK,QAAQ,IAAI;AAAA,IAClD;AACA,QAAI,cAAc,WAAW,YAAY,SAAS;AAChD,aAAO;AAAA,IACT;AACA,WAAO;AAAA,EACT;AAAA,EAEU,QAAQ,SAAiB;AACjC,UAAM,iBAAiB,KAAK,kBAAkB;AAC9C,QAAI,WAA8C;AAGlD,QAAI,gBAAgB;AAClB,YAAM,gBAAgB,QAAQ,QAAQ,eAAe,QAAQ;AAC7D,UAAI,kBAAkB,IAAI;AAExB,YAAI,cAAc,QAAQ,YAAY,eAAe,MAAM;AAC3D,YAAI,gBAAgB,GAAI,eAAc,QAAQ,SAAS;AAAA,YAClD,gBAAe,eAAe,OAAO;AAC1C,cAAM,gBAAgB,QAAQ,MAAM,eAAe,WAAW,EAAE,KAAK;AAGrE,YAAI;AACF,gBAAM,iBACJ,UAAU,kBAAkB,eAAe,OACvC,KAAK,MAAM,aAAa,IACxB;AAGN,cAAI,eAAe,UAAU;AAC3B,2BAAe,SAAS,cAAc;AAAA,UACxC;AAGA,cAAI,eAAe,gBAAgB;AACjC,uBAAW,EAAE,WAAW,eAAe;AAAA,UACzC;AAAA,QACF,SAAS,OAAO;AACd,kBAAQ;AAAA,YACN,gDAAgD,aAAa;AAAA,YAC7D;AAAA,UACF;AAAA,QACF;AAGA,kBAAU,QAAQ,MAAM,GAAG,aAAa,EAAE,QAAQ;AAAA,MACpD;AAAA,IACF;AACA,WAAO,EAAE,SAAS,SAAS;AAAA,EAC7B;AAAA,EAEU,OAAO,SAAgB;AAC/B,SAAK,QAAQ,IAAI,GAAG,OAAO;AAAA,EAC7B;AAAA,EAEA,UAAU;AACR,SAAK,IAAI,WAAW;AACpB,SAAK,mBAAmB;AACxB,SAAK,OAAO;AAAA,EACd;AACF;;;AEvSA,oBAA4B;AAGrB,IAAM,YAAN,cAAwB,MAAM;AAAA,EAGnC,cAAc;AACZ,UAAM,EAAE,cAAc,GAAG,CAAC;AAH5B,SAAQ,IAAI;AAAA,EAIZ;AAAA,EAEA,SAAS;AACP,UAAM,SAAS,IAAI,0BAAY;AAG/B,UAAM,UAAU,qBAAqB,KAAK,GAAG;AAC7C,SAAK,oBAAoB,OAAO;AAChC,WAAO,MAAM,OAAO;AACpB,WAAO,IAAI;AACX,WAAO;AAAA,EACT;AAAA,EAEA,SAAS;AAAA,EAAC;AACZ;;;ACtBA,oBAA4B;AAC5B,2BAAmB;AACnB,IAAAA,iBAAsC;AAGtC,qBAAAC,QAAO,cAAc,cAAAC,QAAgB,IAAI;AAGlC,SAAS,aACd,aACA,aAAa,MACb,WAAW,IACX;AACA,QAAM,YAAY,IAAI,2BAAY;AAClC,2BAAAD,SAAO,WAAW,EACf,cAAc,CAAC,EACf,eAAe,UAAU,EACzB,WAAW,QAAQ,QAAQ,IAAI,EAC/B,OAAO,IAAI,QAAQ,IAAI,EACvB,GAAG,SAAS,CAAC,UAAU;AACtB,YAAQ,MAAM,kCAAkC,MAAM,OAAO;AAAA,EAC/D,CAAC,EACA,KAAK,SAAS;AACjB,SAAO;AACT;AAGO,SAAS,cAAc,aAAuB,aAAa,MAAO;AACvE,QAAM,aAAa,IAAI,2BAAY;AACnC,mBAAa,qBAAAA,SAAO,WAAW,GAAG,UAAU,EAAE,KAAK,UAAU;AAC7D,SAAO;AACT;AAGO,SAAS,iBAAiB,aAAuB,aAAa,MAAO;AAC1E,QAAM,aAAa,IAAI,2BAAY;AACnC,mBAAa,qBAAAA,SAAO,WAAW,GAAG,UAAU,EACzC,YAAY,OAAO,EACnB,aAAa,CAAC,YAAY,aAAa,OAAO,CAAC,EAC/C,KAAK,UAAU;AAClB,SAAO;AACT;AAEA,SAAS,aAAa,eAAqC,aAAa,MAAO;AAC7E,SAAO,cACJ,cAAc,CAAC,EACf,eAAe,UAAU,EACzB,WAAW,SAAS,EACpB,OAAO,MAAM,EACb,cAAc;AAAA,IACb;AAAA,IACA;AAAA,IACA,OAAO,UAAU;AAAA,IACjB;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC,EACA,GAAG,SAAS,CAAC,UAAU;AACtB,YAAQ,MAAM,8BAA8B,MAAM,OAAO;AAAA,EAC3D,CAAC;AACL;;;AC1DO,IAAK,mBAAL,kBAAKE,sBAAL;AACL,EAAAA,oCAAA,gBAAa,QAAb;AACA,EAAAA,oCAAA,kBAAe,QAAf;AACA,EAAAA,oCAAA,cAAW,QAAX;AAHU,SAAAA;AAAA,GAAA;AAML,IAAM,eAAN,cAA2B,MAAM;AAAA,EAGtC,YAAY,MAAc,SAAiB;AACzC,UAAM,OAAO;AACb,SAAK,OAAO;AAAA,EACd;AACF;AAEO,SAAS,YAAY,QAAmB,OAAgB;AAC7D,MAAI,iBAAiB,cAAc;AACjC,WAAO,MAAM,MAAM,MAAM,MAAM,OAAO;AAAA,EACxC,OAAO;AACL,YAAQ,MAAM,KAAK;AACnB,WAAO,MAAM,IAAI;AAAA,EACnB;AACA,SAAO,UAAU;AACnB;;;ACzBO,IAAM,SAAN,MAAa;AAAA,EAClB,YAA6B,MAAc;AAAd;AAAA,EAAe;AAAA,EAE5C,OAAO,SAAgB;AACrB,UAAM,OAAO,QAAQ,OAAO,EAAE,QAAQ,CAAC;AACvC,YAAQ,IAAI,IAAI,KAAK,IAAI,IAAI,IAAI,KAAK,GAAG,OAAO;AAAA,EAClD;AACF;;;ACPA,IAAAC,iBAA8C;;;ACAvC,IAAK,wBAAL,kBAAKC,2BAAL;AACL,EAAAA,uBAAA,mBAAgB;AAChB,EAAAA,uBAAA,kBAAe;AACf,EAAAA,uBAAA,UAAO;AAHG,SAAAA;AAAA,GAAA;AAML,IAAK,wBAAL,kBAAKC,2BAAL;AACL,EAAAA,uBAAA,aAAU;AACV,EAAAA,uBAAA,2BAAwB;AACxB,EAAAA,uBAAA,gBAAa;AACb,EAAAA,uBAAA,aAAU;AACV,EAAAA,uBAAA,cAAW;AALD,SAAAA;AAAA,GAAA;;;ADgBL,IAAM,gBAAN,MAAoB;AAAA,EAWzB,YAAY,QAAmB,QAAuB;AAVtD,SAAO,SAA2B;AAClC,SAAO,SAA+B;AAGtC,SAAQ,YAAY,KAAK,IAAI;AAoD7B,SAAQ,UAAU,MAAM;AACtB,UAAI,CAAC,KAAK,OAAQ;AAClB,WAAK,IAAI,mBAAmB;AAC5B,YAAM,WAAW,KAAK,OAAO,KAAK,IAAI,IAAI,KAAK,aAAa,GAAI;AAGhE,WAAK,OAAO,MAAM,QAAQ;AAC1B,WAAK,OAAO,IAAI,QAAQ;AACxB,WAAK,OAAO,IAAI,QAAQ;AAGxB,WAAK,OAAO,QAAQ;AAAA,QAClB,cAAc,KAAK,OAAO,MAAM,aAAa,MAAM,CAAC;AAAA;AAAA,QACpD;AAAA,MACF,CAAC;AAGD,WAAK,SAAS;AACd,WAAK,SAAS;AAAA,IAChB;AAEA,SAAQ,YAAY,OAAO,YAAoB;AAC7C,UAAI,QAAQ,eAAe,EAAG;AAC9B,UAAI,CAAC,OAAO,SAAS,OAAO,GAAG;AAC7B,aAAK,IAAI,yBAAyB;AAClC;AAAA,MACF;AAGA,UAAI,QAAQ,aAAa,IAAI;AAC3B,cAAM,MAAM,QAAQ,SAAS;AAC7B,aAAK,IAAI,YAAY,GAAG,EAAE;AAE1B,YAAI,6CAA6C;AAE/C,gBAAM,KAAK,gBAAgB;AAAA,QAC7B,WAAW,2BAAoC;AAE7C,gBAAM,KAAK,OAAO;AAAA,QACpB,WAAW,2CAA4C;AAErD,gBAAM,KAAK,eAAe;AAAA,QAC5B;AAAA,MACF,WAGS,KAAK,mBAAmB;AAC/B,aAAK,IAAI,mBAAmB,QAAQ,UAAU,SAAS;AACvD,aAAK,kBAAkB,MAAM,OAAO;AAAA,MACtC;AAAA,IACF;AAiCA,SAAQ,eAAe,OAAO,eAAuB;AACnD,UAAI,CAAC,KAAK,OAAQ;AAClB,WAAK,IAAI,qBAAqB,UAAU,GAAG;AAC3C,WAAK,OAAO,MAAM,eAAe,UAAU;AAG3C,UAAI,CAAC,KAAK,mBAAmB;AAC3B,aAAK,IAAI,kCAAkC;AAC3C,aAAK,OAAO;AAAA,MACd;AAAA,IACF;AA1IE,SAAK,SAAS;AACd,SAAK,SAAS;AACd,SAAK,IAAI,cAAc;AAGvB,SAAK,OAAO,IAAI,GAAG,cAAc,KAAK,YAAY;AAGlD,SAAK,OAAO,MAAM;AAAA,MAAG;AAAA,MAAW,CAAC,YAC/B,KAAK,QAAQ;AAAA,QACX,0BAAgC,IAAI,KAAK,UAAU,OAAO,CAAC;AAAA,MAC7D;AAAA,IACF;AACA,SAAK,OAAO,MAAM;AAAA,MAAG;AAAA,MAAyB,MAC5C,KAAK,QAAQ,wDAAgD;AAAA,IAC/D;AACA,SAAK,OAAO,MAAM;AAAA,MAAG;AAAA,MAAc,MACjC,KAAK,QAAQ,kCAAqC;AAAA,IACpD;AACA,SAAK,OAAO,MAAM;AAAA,MAAG;AAAA,MAAW,MAC9B,KAAK,QAAQ,4BAAkC;AAAA,IACjD;AACA,SAAK,OAAO,MAAM;AAAA,MAAG;AAAA,MAAY,CAAC,aAChC,KAAK,QAAQ;AAAA,QACX,4BAAiC,IAAI,KAAK,UAAU,QAAQ,CAAC;AAAA,MAC/D;AAAA,IACF;AAGA,SAAK,iBAAiB;AAGtB,WAAO,GAAG,SAAS,KAAK,OAAO;AAC/B,WAAO,GAAG,WAAW,KAAK,SAAS;AAAA,EACrC;AAAA,EAEQ,OAAO,SAAgB;AAC7B,SAAK,QAAQ,IAAI,GAAG,OAAO;AAAA,EAC7B;AAAA,EAEQ,SAAS;AACf,SAAK,QAAQ,IAAI,OAAO;AACxB,SAAK,QAAQ,MAAM,OAAO;AAAA,EAC5B;AAAA,EAsDA,MAAc,SAAS;AACrB,SAAK,mBAAmB,IAAI;AAC5B,SAAK,oBAAoB;AACzB,SAAK,OAAO;AAAA,EACd;AAAA,EAEA,MAAc,kBAAkB;AAC9B,QAAI,CAAC,KAAK,OAAQ;AAClB,SAAK,mBAAmB,IAAI;AAC5B,SAAK,oBAAoB,IAAI,2BAAY;AACzC,SAAK,OAAO,IAAI,WAAW,KAAK,iBAAiB;AACjD,SAAK,OAAO;AAAA,EACd;AAAA,EAEA,MAAc,iBAAiB;AAC7B,SAAK,mBAAmB,IAAI;AAC5B,SAAK,oBAAoB;AAEzB,UAAM,eAAe,KAAK,QAAQ,MAAM;AACxC,UAAM,cAAc,eAAe,aAAa,SAAS,CAAC;AAC1D,QACE,aAAa,SAAS,UACtB,KAAK,wBAAwB,aAC7B;AACA,WAAK;AAAA,QACH;AAAA,MACF;AACA,WAAK,OAAO;AAAA,IACd;AAAA,EACF;AAAA,EAcA,MAAc,mBAAmB;AAC/B,QAAI,CAAC,KAAK,OAAQ;AAClB,QAAI;AACF,UAAI,KAAK,OAAO,cAAc;AAE5B,aAAK,OAAO,MAAM,oBAAoB,KAAK,OAAO,YAAY;AAC9D,cAAM,KAAK,MAAM,KAAK,OAAO,YAAY;AAAA,MAC3C,WAAW,KAAK,OAAO,sBAAsB;AAE3C,cAAM,KAAK,OAAO;AAAA,MACpB,OAAO;AAGL,aAAK,QAAQ,kCAAqC;AAAA,MACpD;AAAA,IACF,SAAS,OAAO;AACd,cAAQ,MAAM,mBAAmB,KAAK;AACtC,WAAK,QAAQ,kCAAqC;AAAA,IACpD;AAAA,EACF;AAAA,EAEA,MAAc,SAAS;AACrB,QAAI,CAAC,KAAK,OAAQ;AAClB,SAAK,OAAO;AAGZ,UAAM,cACJ,KAAK,OAAO,MAAM,aAAa,KAAK,OAAO,MAAM,aAAa,SAAS,CAAC;AAC1E,QAAI,KAAK,wBAAwB,aAAa;AAC5C,WAAK,IAAI,4BAA4B;AACrC;AAAA,IACF;AACA,SAAK,sBAAsB;AAE3B,QAAI;AAEF,YAAM,SAAS,KAAK,OAAO,MAAM,OAAO;AAGxC,YAAM,KAAK,MAAM,MAAM;AAAA,IACzB,SAAS,OAAO;AACd,cAAQ,MAAM,mBAAmB,KAAK;AACtC,WAAK,QAAQ,kCAAqC;AAAA,IACpD;AAAA,EACF;AAAA;AAAA,EAGA,MAAc,MAAM,SAA4B;AAC9C,QAAI,CAAC,KAAK,UAAU,CAAC,KAAK,OAAQ;AAGlC,QAAI;AACJ,QAAI,OAAO,YAAY,UAAU;AAC/B,YAAM,SAAS,IAAI,2BAAY;AAC/B,aAAO,MAAM,OAAO;AACpB,aAAO,IAAI;AACX,mBAAa;AAAA,IACf,OAAO;AACL,mBAAa;AAAA,IACf;AAGA,UAAM,QAAQ,KAAK,OAAO,IAAI,MAAM,UAAU;AAG9C,UAAM,KAAK,UAAU,KAAK;AAAA,EAC5B;AAAA,EAEA,MAAc,UAAU,OAAiB;AACvC,QAAI,CAAC,KAAK,OAAQ;AAClB,QAAI,CAAC,MAAM,UAAU;AACnB,WAAK,IAAI,gCAAgC,KAAK;AAC9C;AAAA,IACF;AAGA,UAAM,GAAG,QAAQ,CAAC,UAAU;AAC1B,WAAK,IAAI,qBAAqB,MAAM,UAAU,SAAS;AACvD,WAAK,QAAQ,KAAK,KAAK;AAAA,IACzB,CAAC;AACD,UAAM,GAAG,SAAS,CAAC,UAAU;AAC3B,WAAK,IAAI,yBAAyB,KAAK;AAAA,IACzC,CAAC;AACD,UAAM,GAAG,OAAO,MAAM;AACpB,WAAK,IAAI,oBAAoB;AAAA,IAC/B,CAAC;AAAA,EACH;AACF;;;AErQA,IAAAC,wBAA6B;AAK7B,IAAM,yBAAyB;AAAA,EAC7B,aAAa;AAAA,EACb,aAAa;AAAA,EACb,cAAc;AAAA,EACd,cAAc;AAAA,EACd,aAAa;AAAA,EACb,cAAc;AAChB;AAMO,IAAe,MAAf,cAA2B,mCAAwB;AAAA;AAAA,EAKxD,WAAW,aAAuB;AAEhC,gBAAY,KAAK,QAAQ,CAAC,UAAU;AAClC,WAAK,WAAW,KAAK,eAAe,KAAK;AAAA,IAC3C,CAAC;AAAA,EACH;AAAA,EAEU,OAAO,SAAgB;AAC/B,SAAK,QAAQ,IAAI,GAAG,OAAO;AAAA,EAC7B;AAAA,EAEA,UAAU;AACR,SAAK,IAAI,WAAW;AACpB,SAAK,mBAAmB;AAAA,EAC1B;AAAA,EAEA,IAAc,YAAoB;AAChC,WAAQ,KAAK,YAAY,uBAAuB,KAAK,QAAQ,KAAM;AAAA,EACrE;AAAA,EAEQ,eACN,OACqC;AACrC,QAAI,CAAC,SAAS,MAAM,eAAe,GAAG;AACpC,YAAM,IAAI,MAAM,0CAA0C;AAAA,IAC5D;AAEA,UAAM,MAAM,IAAI,WAAW,KAAK;AAGhC,QACE,IAAI,CAAC,MAAM,MACX,IAAI,CAAC,MAAM,MACX,IAAI,CAAC,MAAM,OACX,IAAI,CAAC,MAAM,KACX;AACA,aAAO;AAAA,IACT;AAEA,QACE,IAAI,CAAC,MAAM,MACX,IAAI,CAAC,MAAM,OACX,IAAI,CAAC,MAAM,OACX,IAAI,CAAC,MAAM,IACX;AACA,aAAO;AAAA,IACT;AAEA,QACE,IAAI,CAAC,MAAM,MACX,IAAI,CAAC,MAAM,MACX,IAAI,CAAC,MAAM,MACX,IAAI,CAAC,MAAM,MACX,IAAI,CAAC,MAAM,MACX,IAAI,CAAC,MAAM,MACX,IAAI,EAAE,MAAM,MACZ,IAAI,EAAE,MAAM,IACZ;AACA,aAAO;AAAA,IACT;AAEA,QAAI,IAAI,CAAC,MAAM,MAAQ,IAAI,CAAC,MAAM,MAAQ,IAAI,CAAC,MAAM,IAAM;AACzD,aAAO;AAAA,IACT;AAEA,QACE,IAAI,CAAC,MAAM,OACX,IAAI,CAAC,MAAM,OACX,IAAI,CAAC,MAAM,OACX,IAAI,CAAC,MAAM,KACX;AACA,aAAO;AAAA,IACT;AAEA,QACE,IAAI,CAAC,MAAM,OACX,IAAI,CAAC,MAAM,MACX,IAAI,CAAC,MAAM,MACX,IAAI,CAAC,MAAM,IACX;AACA,aAAO;AAAA,IACT;AACA,SAAK,IAAI,6CAA6C,KAAK;AAC3D,WAAO;AAAA,EACT;AACF;;;ACrGO,IAAe,UAAf,cAA+B,IAAI;AAAA,EAGxC,WAAW,aAAuB;AAChC,UAAM,WAAW,WAAW;AAG5B,SAAK,IAAI,8BAA8B;AAEvC,UAAM,SAAmB,CAAC;AAC1B,gBAAY,GAAG,QAAQ,CAAC,UAAU;AAChC,aAAO,KAAK,KAAK;AAAA,IACnB,CAAC;AAED,gBAAY,GAAG,OAAO,YAAY;AAChC,UAAI,OAAO,WAAW,EAAG;AACzB,YAAM,cAAc,OAAO,OAAO,MAAM;AACxC,YAAM,OAAO,IAAI,KAAK,CAAC,WAAW,GAAG,SAAS,KAAK,SAAS,IAAI;AAAA,QAC9D,MAAM,KAAK;AAAA,MACb,CAAC;AAGD,WAAK,IAAI,sBAAsB;AAC/B,YAAM,aAAa,MAAM,KAAK,eAAe,IAAI;AACjD,WAAK,KAAK,cAAc,UAAU;AAAA,IACpC,CAAC;AAAA,EACH;AACF;;;AChCO,IAAM,UAAN,cAAsB,QAAQ;AAAA,EAA9B;AAAA;AACL,SAAQ,IAAI;AAAA;AAAA,EAEZ,MAAM,eAAe,MAAY;AAC/B,WAAO,gBAAgB,KAAK,GAAG;AAAA,EACjC;AACF;;;ACRA,SAAoB;AACpB,IAAAC,iBAAsC;;;ACE/B,IAAe,MAAf,MAAmB;AAAA,EAMd,OAAO,SAAgB;AAC/B,SAAK,QAAQ,IAAI,GAAG,OAAO;AAAA,EAC7B;AAAA,EAEA,UAAU;AACR,SAAK,IAAI,WAAW;AACpB,SAAK,OAAO;AAAA,EACd;AACF;;;ADbO,IAAM,UAAN,cAAsB,IAAI;AAAA,EAC/B,YAAoB,gBAA0B;AAC5C,UAAM;AADY;AAAA,EAEpB;AAAA,EAEA,MAAM,YAAsB;AAC1B,UAAM,cAAc,IAAI,2BAAY;AACpC,eAAW,KAAK,QAAQ,YAAY;AAClC,iBAAW,YAAY,KAAK,gBAAgB;AAC1C,cAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,GAAG,CAAC;AACvD,cAAM,cAAiB,gBAAa,QAAQ;AAC5C,aAAK,IAAI,iBAAiB,YAAY,MAAM,SAAS;AACrD,oBAAY,MAAM,WAAW;AAAA,MAC/B;AACA,kBAAY,IAAI;AAAA,IAClB,CAAC;AACD,WAAO;AAAA,EACT;AAAA,EAEA,SAAS;AAAA,EAAC;AACZ;;;AErBA,eAAsB,cACpB,QACA,UACqB;AACrB,SAAO,IAAI,QAAoB,CAAC,SAAS,WAAW;AAElD,UAAM,UAAU,WAAW,MAAM;AAC/B,aAAO,IAAI,oCAA0C,gBAAgB,CAAC;AAAA,IACxE,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,oCAA0C,gBAAgB,CAAC;AAAA,MACxE;AAAA,IACF;AAGA,WAAO,GAAG,WAAW,QAAQ;AAAA,EAC/B,CAAC;AACH;","names":["import_stream","ffmpeg","ffmpegInstaller","MicdropErrorCode","import_stream","MicdropClientCommands","MicdropServerCommands","import_eventemitter3","import_stream"]}
package/dist/index.mjs CHANGED
@@ -1,10 +1,21 @@
1
1
  // src/agent/Agent.ts
2
2
  import { EventEmitter } from "eventemitter3";
3
+
4
+ // src/agent/tools.ts
5
+ var AUTO_END_CALL_TOOL_NAME = "end_call";
6
+ var AUTO_END_CALL_PROMPT = "Call this tool only if user asks to end the call";
7
+ var AUTO_SEMANTIC_TURN_TOOL_NAME = "semantic_turn";
8
+ var AUTO_SEMANTIC_TURN_PROMPT = "Call this tool only if last user message is obviously an incomplete sentence that you need to wait for the end before answering";
9
+ var AUTO_IGNORE_USER_NOISE_TOOL_NAME = "ignore_user_noise";
10
+ var AUTO_IGNORE_USER_NOISE_PROMPT = `Call this tool only if last user message is just an interjection or a sound that expresses emotion, hesitation, or reaction (ex: "Uh", "Ahem", "Hmm", "Ah") but doesn't carry any clear meaning like agreeing, refusing, or commanding`;
11
+
12
+ // src/agent/Agent.ts
3
13
  var Agent = class extends EventEmitter {
4
14
  constructor(options) {
5
15
  super();
6
16
  this.options = options;
7
17
  this.conversation = [{ role: "system", content: options.systemPrompt }];
18
+ this.tools = this.getDefaultTools();
8
19
  }
9
20
  addUserMessage(text, metadata) {
10
21
  this.addMessage("user", text, metadata);
@@ -12,6 +23,15 @@ var Agent = class extends EventEmitter {
12
23
  addAssistantMessage(text, metadata) {
13
24
  this.addMessage("assistant", text, metadata);
14
25
  }
26
+ addTool(tool) {
27
+ this.tools.push(tool);
28
+ }
29
+ removeTool(name) {
30
+ this.tools = this.tools.filter((tool) => tool.name !== name);
31
+ }
32
+ getTool(name) {
33
+ return this.tools.find((tool) => tool.name === name);
34
+ }
15
35
  addMessage(role, text, metadata) {
16
36
  this.log(`Adding ${role} message to conversation: ${text}`);
17
37
  const message = {
@@ -22,28 +42,132 @@ var Agent = class extends EventEmitter {
22
42
  this.conversation.push(message);
23
43
  this.emit("Message", message);
24
44
  }
45
+ addToolMessage(message) {
46
+ this.log("Adding tool message:", message);
47
+ this.conversation.push(message);
48
+ this.emit("Message", message);
49
+ }
25
50
  endCall() {
26
51
  this.log("Ending call");
27
52
  this.emit("EndCall");
28
53
  }
29
54
  cancelLastUserMessage() {
30
55
  this.log("Cancelling last user message");
31
- const lastMessage = this.conversation[this.conversation.length - 1];
32
- if (lastMessage?.role !== "user") return;
33
- this.conversation.pop();
56
+ const lastMessageIndex = this.conversation.findLastIndex(
57
+ (message) => message.role === "user"
58
+ );
59
+ if (lastMessageIndex !== -1) {
60
+ this.conversation.splice(lastMessageIndex, 1);
61
+ }
34
62
  this.emit("CancelLastUserMessage");
35
63
  }
36
- cancelLastAssistantMessage() {
37
- this.log("Cancelling last assistant message");
38
- const lastMessage = this.conversation[this.conversation.length - 1];
39
- if (lastMessage?.role !== "assistant") return;
40
- this.conversation.pop();
41
- this.emit("CancelLastAssistantMessage");
42
- }
43
64
  skipAnswer() {
44
65
  this.log("Skipping answer");
45
66
  this.emit("SkipAnswer");
46
67
  }
68
+ getDefaultTools() {
69
+ const tools = [];
70
+ if (this.options.autoEndCall) {
71
+ tools.push({
72
+ name: AUTO_END_CALL_TOOL_NAME,
73
+ description: typeof this.options.autoEndCall === "string" ? this.options.autoEndCall : AUTO_END_CALL_PROMPT,
74
+ execute: () => this.endCall()
75
+ });
76
+ }
77
+ if (this.options.autoSemanticTurn) {
78
+ tools.push({
79
+ name: AUTO_SEMANTIC_TURN_TOOL_NAME,
80
+ description: typeof this.options.autoSemanticTurn === "string" ? this.options.autoSemanticTurn : AUTO_SEMANTIC_TURN_PROMPT,
81
+ skipAnswer: true,
82
+ execute: () => this.skipAnswer()
83
+ });
84
+ }
85
+ if (this.options.autoIgnoreUserNoise) {
86
+ tools.push({
87
+ name: AUTO_IGNORE_USER_NOISE_TOOL_NAME,
88
+ description: typeof this.options.autoIgnoreUserNoise === "string" ? this.options.autoIgnoreUserNoise : AUTO_IGNORE_USER_NOISE_PROMPT,
89
+ skipAnswer: true,
90
+ execute: () => this.cancelLastUserMessage()
91
+ });
92
+ }
93
+ return tools;
94
+ }
95
+ async executeTool(toolCall) {
96
+ try {
97
+ const tool = this.getTool(toolCall.toolName);
98
+ if (!tool) {
99
+ throw new Error(`Tool not found "${toolCall.toolName}"`);
100
+ }
101
+ this.log("Executing tool:", toolCall.toolName, toolCall.parameters);
102
+ this.addToolMessage(toolCall);
103
+ const parameters = JSON.parse(toolCall.parameters);
104
+ const output = tool.execute ? await tool.execute(parameters) : {};
105
+ this.addToolMessage({
106
+ role: "tool_result",
107
+ toolCallId: toolCall.toolCallId,
108
+ toolName: toolCall.toolName,
109
+ output: JSON.stringify(output ?? null)
110
+ });
111
+ if (tool.emitOutput) {
112
+ this.emit("ToolCall", {
113
+ name: toolCall.toolName,
114
+ parameters,
115
+ output
116
+ });
117
+ }
118
+ return {
119
+ output,
120
+ skipAnswer: tool.skipAnswer
121
+ };
122
+ } catch (error) {
123
+ console.error("[OpenaiAgent] Error executing tool:", error);
124
+ return {
125
+ output: {
126
+ error: error.message
127
+ }
128
+ };
129
+ }
130
+ }
131
+ getExtractOptions() {
132
+ const extract = this.options.extract;
133
+ if (!extract) return void 0;
134
+ if ("json" in extract && extract.json) {
135
+ return { ...extract, startTag: "{", endTag: "}" };
136
+ }
137
+ if ("startTag" in extract && "endTag" in extract) {
138
+ return extract;
139
+ }
140
+ return void 0;
141
+ }
142
+ extract(message) {
143
+ const extractOptions = this.getExtractOptions();
144
+ let metadata = void 0;
145
+ if (extractOptions) {
146
+ const startTagIndex = message.indexOf(extractOptions.startTag);
147
+ if (startTagIndex !== -1) {
148
+ let endTagIndex = message.lastIndexOf(extractOptions.endTag);
149
+ if (endTagIndex === -1) endTagIndex = message.length + 1;
150
+ else endTagIndex += extractOptions.endTag.length;
151
+ const extractedText = message.slice(startTagIndex, endTagIndex).trim();
152
+ try {
153
+ const extractedValue = "json" in extractOptions && extractOptions.json ? JSON.parse(extractedText) : extractedText;
154
+ if (extractOptions.callback) {
155
+ extractOptions.callback(extractedValue);
156
+ }
157
+ if (extractOptions.saveInMetadata) {
158
+ metadata = { extracted: extractedValue };
159
+ }
160
+ } catch (error) {
161
+ console.error(
162
+ `[OpenaiAgent] Error parsing extracted value (${extractedText}):`,
163
+ error
164
+ );
165
+ }
166
+ message = message.slice(0, startTagIndex).trimEnd();
167
+ }
168
+ }
169
+ return { message, metadata };
170
+ }
47
171
  log(...message) {
48
172
  this.logger?.log(...message);
49
173
  }
@@ -154,7 +278,6 @@ var MicdropClientCommands = /* @__PURE__ */ ((MicdropClientCommands2) => {
154
278
  })(MicdropClientCommands || {});
155
279
  var MicdropServerCommands = /* @__PURE__ */ ((MicdropServerCommands2) => {
156
280
  MicdropServerCommands2["Message"] = "Message";
157
- MicdropServerCommands2["CancelLastAssistantMessage"] = "CancelLastAssistantMessage";
158
281
  MicdropServerCommands2["CancelLastUserMessage"] = "CancelLastUserMessage";
159
282
  MicdropServerCommands2["SkipAnswer"] = "SkipAnswer";
160
283
  MicdropServerCommands2["EndCall"] = "EndCall";
@@ -227,10 +350,6 @@ var MicdropServer = class {
227
350
  "CancelLastUserMessage",
228
351
  () => this.socket?.send("CancelLastUserMessage" /* CancelLastUserMessage */)
229
352
  );
230
- this.config.agent.on(
231
- "CancelLastAssistantMessage",
232
- () => this.socket?.send("CancelLastAssistantMessage" /* CancelLastAssistantMessage */)
233
- );
234
353
  this.config.agent.on(
235
354
  "SkipAnswer",
236
355
  () => this.socket?.send("SkipAnswer" /* SkipAnswer */)
@@ -272,7 +391,8 @@ var MicdropServer = class {
272
391
  this.currentUserStream?.end();
273
392
  this.currentUserStream = void 0;
274
393
  const conversation = this.config?.agent.conversation;
275
- if (conversation && conversation[conversation.length - 1].role === "user") {
394
+ const lastMessage = conversation?.[conversation.length - 1];
395
+ if (lastMessage?.role === "user" && this.lastMessageSpeeched !== lastMessage) {
276
396
  this.log(
277
397
  "User stopped speaking and a transcript already exists, answering"
278
398
  );
@@ -298,6 +418,12 @@ var MicdropServer = class {
298
418
  async answer() {
299
419
  if (!this.config) return;
300
420
  this.cancel();
421
+ const lastMessage = this.config.agent.conversation[this.config.agent.conversation.length - 1];
422
+ if (this.lastMessageSpeeched === lastMessage) {
423
+ this.log("Already answered, skipping");
424
+ return;
425
+ }
426
+ this.lastMessageSpeeched = lastMessage;
301
427
  try {
302
428
  const stream = this.config.agent.answer();
303
429
  await this.speak(stream);
@@ -486,6 +612,12 @@ async function waitForParams(socket, validate) {
486
612
  });
487
613
  }
488
614
  export {
615
+ AUTO_END_CALL_PROMPT,
616
+ AUTO_END_CALL_TOOL_NAME,
617
+ AUTO_IGNORE_USER_NOISE_PROMPT,
618
+ AUTO_IGNORE_USER_NOISE_TOOL_NAME,
619
+ AUTO_SEMANTIC_TURN_PROMPT,
620
+ AUTO_SEMANTIC_TURN_TOOL_NAME,
489
621
  Agent,
490
622
  FileSTT,
491
623
  Logger,