Merge pull request #316 from harmony-one/voice-to-text

Added VoiceToTextBot

src/bot.ts

@@ -52,6 +52,7 @@ import { ES } from './es'
 import { hydrateFiles } from '@grammyjs/files'
 import { VoiceTranslateBot } from './modules/voice-translate'
 import { TextToSpeechBot } from './modules/text-to-speech'
+import { VoiceToTextBot } from './modules/voice-to-text'
 
 Events.EventEmitter.defaultMaxListeners = 30
 
@@ -221,6 +222,7 @@ const documentBot = new DocumentHandler()
 const telegramPayments = new TelegramPayments(payments)
 const voiceTranslateBot = new VoiceTranslateBot(payments)
 const textToSpeechBot = new TextToSpeechBot(payments)
+const voiceToTextBot = new VoiceToTextBot(payments)
 
 bot.on('message:new_chat_members:me', async (ctx) => {
   try {
@@ -335,6 +337,7 @@ const PayableBots: Record<string, PayableBotConfig> = {
   documentBot: { bot: documentBot },
   translateBot: { bot: translateBot },
   textToSpeech: { bot: textToSpeechBot },
+  voiceToText: { bot: voiceToTextBot },
   openAiBot: {
     enabled: (ctx: OnMessageContext) => ctx.session.openAi.imageGen.isEnabled,
     bot: openAiBot

src/modules/voice-to-text/index.ts

@@ -0,0 +1,90 @@
+import fs from 'fs'
+import pino from 'pino'
+import { InputFile } from 'grammy'
+import type { Logger } from 'pino'
+import type { BotPayments } from '../payment'
+import { speechToText } from '../open-ai/api/openAi'
+import type { OnMessageContext, PayableBot } from '../types'
+import { bot } from '../../bot'
+import { SessionState } from '../types'
+import { download } from '../../utils/files'
+
+export class VoiceToTextBot implements PayableBot {
+  public readonly module = 'VoiceToText'
+  private readonly payments: BotPayments
+
+  private readonly logger: Logger
+
+  constructor (payments: BotPayments) {
+    this.payments = payments
+    this.logger = pino({
+      name: 'VoiceToText',
+      transport: {
+        target: 'pino-pretty',
+        options: { colorize: true }
+      }
+    })
+  }
+
+  public isSupportedEvent (ctx: OnMessageContext): boolean {
+    return ctx.hasCommand('text')
+  }
+
+  public getEstimatedPrice (ctx: OnMessageContext): number {
+    const { voice, audio } = ctx.update.message
+    const seconds = (voice?.duration ?? audio?.duration) ?? 0
+    return seconds * 0.005
+  }
+
+  public async onEvent (ctx: OnMessageContext): Promise<void> {
+    ctx.session.analytics.module = this.module
+    const { voice, audio } = ctx.message.reply_to_message ?? { voice: undefined, audio: undefined }
+
+    if (!voice && !audio) {
+      await ctx.reply('The message must include audio content')
+      return
+    }
+
+    const progressMessage = await ctx.reply('Generating...')
+
+    if (!ctx.chat?.id) {
+      throw Error('chat id is undefined')
+    }
+
+    const fileId = voice?.file_id ?? audio?.file_id
+
+    if (!fileId) {
+      await ctx.reply('The message must include audio content')
+      return
+    }
+
+    const file = await bot.api.getFile(fileId)
+
+    const path = await download(file)
+
+    let ext = 'ogg'
+
+    if (file.file_path) {
+      ext = file.file_path.split('.').pop() ?? ext
+    }
+
+    const filename = path + '.' + ext
+    fs.renameSync(path, filename)
+
+    const resultText = await speechToText(fs.createReadStream(filename))
+    fs.rmSync(filename)
+
+    await ctx.api.deleteMessage(ctx.chat.id, progressMessage.message_id)
+    if (resultText.length > 512) {
+      const translationFile = new InputFile(new TextEncoder().encode(resultText), `From @${ctx.message.from.username}.txt`)
+      await bot.api.sendDocument(ctx.chat.id, translationFile, {
+        reply_to_message_id: ctx.message.message_id,
+        caption: resultText.slice(0, 1024)
+      })
+    } else {
+      await ctx.reply(resultText, { message_thread_id: ctx.message?.message_thread_id })
+    }
+
+    ctx.session.analytics.sessionState = SessionState.Success
+  }
+}

src/utils/files.ts

@@ -0,0 +1,53 @@
+import * as pathLib from 'path'
+import * as fs from 'fs'
+import * as os from 'os'
+import * as https from 'https'
+import type { File } from '@grammyjs/types/manage.js'
+import config from '../config'
+
+export const createTempFile = async (): Promise<string> =>
+  pathLib.join(
+    await fs.promises.mkdtemp(
+      (await fs.promises.realpath(os.tmpdir())) + pathLib.sep
+    ),
+    'filedata'
+  )
+
+export const copyFile = fs.promises.copyFile
+
+export async function downloadFile (url: string, dest: string): Promise<void> {
+  const file = fs.createWriteStream(dest)
+  await new Promise<void>((resolve, reject) => {
+    https.get(url, (res) => {
+      res.pipe(file)
+      file.on('finish', () => {
+        file.close()
+        resolve()
+      })
+    }).on(
+      'error',
+      (err0) => { fs.unlink(dest, (err1) => { reject(err1 ?? err0) }) }
+    )
+  })
+}
+
+const buildFileUrl = (root: string, token: string, path: string): string => `${root}/file/bot${token}/${path}`
+const buildLink = (path: string): string => {
+  return buildFileUrl('https://api.telegram.org', config.telegramBotAuthToken, path)
+}
+function getUrl (file: File): string {
+  const path = file.file_path
+  if (path === undefined) {
+    const id = file.file_id
+    throw new Error(`File path is not available for file '${id}'`)
+  }
+  return pathLib.isAbsolute(path) ? path : buildLink(path)
+}
+
+export async function download (file: File, path?: string): Promise<string> {
+  const url = getUrl(file)
+  if (path === undefined) path = await createTempFile()
+  if (pathLib.isAbsolute(url)) await copyFile(url, path)
+  else await downloadFile(url, path)
+  return path
+}