Transcription (#32)

* refactor: some improves

* build; add nodemon and remove y18n

* chre: remove locales

* feat: adding whisper transcription

* refactor: moving code into controller

* refactor: moving code

* refactor: extracting types

* feature: add middleware to let pass only authorized channels

* refactor: some improves

* refactor: min improves

* chore: dummy change

* a try

* revert

* trying something

* trying++

* trying+2

* ??

* trying bla

* bla2

.env.example

@@ -9,7 +9,9 @@ LOG_CONSOLE_LEVEL=debug
 EXPRESS_PORT=3000
 
 # espiDev
+OWNER_CHANNEL=123456
 MAIN_CHANNEL=454545
+AI_CHANNELS='123456,454545'
 BOT_TOKEN=12345:ABCDEF
 BOT_DOMAIN=https://espi.localhost
 OPENAI_API_KEY=qw-1234567

.eslintrc.json

@@ -9,7 +9,8 @@
   ],
   "env": {
     "node": true,
-    "es2021": true
+    "es2016": true,
+    "jest": true
   },
   "parserOptions": {
     "project": "./tsconfig.eslint.json",
@@ -23,8 +24,11 @@
     },
     "import/resolver": {
       "node": {
-        "extensions": [".js", ".d.ts", ".ts"],
-        "moduleDirectory": ["src", "node_modules"]
+        "extensions": [".d.ts", ".ts"]
+      },
+      "typescript": {
+        "alwaysTryTypes": true,
+        "project": "./tsconfig.eslint.json"
       }
     }
   },
@@ -58,6 +62,6 @@
         }
       }
     ],
-    "import/no-unresolved": ["error", { "commonjs": true }]
+    "import/no-unresolved": "error"
   }
 }

.scripts/start-dev.sh

@@ -13,4 +13,4 @@ echo "API_DOMAIN: $API_DOMAIN"
 echo "BOT_DOMAIN: $BOT_DOMAIN"
 
 # Start the bot
-ts-node src/index.ts
+nodemon --watch 'src/**/*.ts' --exec 'ts-node' ./src/index.ts

Dockerfile

@@ -4,7 +4,6 @@ WORKDIR /app
 
 COPY package.json yarn.lock /app/
 COPY tsconfig.json /app/
-COPY locales/ /app/locales/
 COPY src/ /app/src/
 
 RUN yarn install --frozen-lockfile

package.json

@@ -37,8 +37,7 @@
     "morgan": "^1.10.0",
     "openai": "4.24.1",
     "telegraf": "^4.15.3",
-    "winston": "^3.11.0",
-    "y18n": "^5.0.8"
+    "winston": "^3.11.0"
   },
   "devDependencies": {
     "@babel/core": "^7.23.7",
@@ -61,6 +60,7 @@
     "husky": "^8.0.3",
     "jest": "^29.7.0",
     "lint-staged": "15.2.0",
+    "nodemon": "3.0.2",
     "prettier": "3.1.1",
     "release-it": "^17.0.1",
     "ts-jest": "^29.1.1",

src/bot.ts

@@ -1,15 +1,11 @@
 import { Telegraf } from 'telegraf';
+import { message } from 'telegraf/filters';
 
-import AIController from './controllers/AIController';
+import GPTController from './controllers/GPT/GPTController';
 import StickersController from './controllers/StickersController';
-import Holidays from './features/holidays';
+import Holidays from './core/holidays';
 import logger from './lib/logger';
-
-// eslint-disable-next-line @typescript-eslint/no-var-requires
-const y18n = require('y18n');
-
-// eslint-disable-next-line @typescript-eslint/naming-convention
-const { __ } = y18n({ locale: 'es' });
+import GPTMiddleware from './middlewares/GPTMiddleware';
 
 if (!process.env.BOT_TOKEN) {
   logger.error('You have to define BOT_TOKEN env var');
@@ -19,28 +15,37 @@ if (!process.env.BOT_TOKEN) {
 
 const bot = new Telegraf(process.env.BOT_TOKEN);
 
-// Handler for /start command.
+// Handler for /start command
 bot.start(ctx => ctx.reply('Welcome!'));
 
-// Handler for /help command.
+// Handler for /help command
 bot.help(ctx => ctx.reply('Send me a sticker'));
 
-// Registers middleware for provided update type.
-// bot.on('sticker', ctx => ctx.reply('👍'));
-
-// Registers middleware for handling text messages.
-bot.hears('ping', ctx => ctx.reply('ACK'));
-bot.hears('hi', ctx => ctx.reply(__`hi`));
-
-// Version
+// Basic and Info
+bot.hears(/^ping$/i, ctx => ctx.reply('ACK'));
+bot.hears(/(hi|hola)/, ctx => ctx.reply('👋'));
 bot.hears(/^espi +version/i, ctx => ctx.reply(`Soy ${ctx.botInfo?.username}@${process.env.npm_package_version}`));
-
-// Espi commands
+bot.hears(/^espi +id/i, ctx => {
+  const name = ctx.chat.type === 'private' ? `${ctx.chat.first_name} (@${ctx.chat.username})` : ctx.chat.title;
+  ctx.reply(`
+- Nombre: ${name}
+- ID: ${ctx.chat.id} (${ctx.chat.type})
+`);
+});
+
+// Espi Featuring
 bot.hears(/^espi +feriados/i, Holidays.holidaysAR);
 bot.hears(/^espi +(férié|ferie)/i, Holidays.holidaysCA);
 bot.hears(/^espi +(finde +largo|fl)/i, Holidays.nextLongWeekendAR);
 bot.hears(/^espi +(findes +largos|ffll)/i, Holidays.nextThreeLongWeekendsAR);
-bot.hears(AIController.shouldRespond, AIController.handleQuestion);
+bot.hears(/^espi +(?<question>.+)/i, GPTMiddleware.authorizedChannel, GPTController.handleQuestion);
+
+// Audio
+bot.on(
+  message('voice'),
+  (ctx, next) => GPTMiddleware.authorizedChannel(ctx, next),
+  ctx => GPTController.transcriptAudio(ctx)
+);
 
 // Reply With Stickers
 bot.hears(/\bfacuuu\b/i, StickersController.replyWithMaybeFacu);

src/config.ts

@@ -10,6 +10,8 @@ export const apisCA = {
 
 export const mainChannel = toNumber(process.env.MAIN_CHANNEL || '-1');
 
+export const ownerChannel = toNumber(process.env.OWNER_CHANNEL || '-1');
+
 export const aiChannels = (process.env.AI_CHANNELS ?? `${mainChannel}`).split(',').map(toNumber) || [];
 
 export const espiId = toNumber(process.env.BOT_TOKEN?.split(':')[0] || '-1');

src/controllers/BaseController.ts

@@ -0,0 +1,7 @@
+import { BaseContext } from './types';
+
+export default class BaseController {
+  static showTypingAction(ctx: BaseContext) {
+    ctx.telegram.sendChatAction(ctx.chat.id, 'typing');
+  }
+}

src/controllers/GPT/GPTController.ts

@@ -0,0 +1,53 @@
+import { ChatCompletionMessageParam } from 'openai/resources';
+import { Message } from 'telegraf/typings/core/types/typegram';
+
+import { TextMatchedContext, TranscriptContext } from './types';
+import { espiId } from '../../config';
+import GPT from '../../core/GPT';
+import api from '../../lib/api';
+import BaseController from '../BaseController';
+
+export default class GPTController extends BaseController {
+  private static maxAudioSize = 10 * 1024 * 1024; // 10MB
+
+  /**
+   * Handle incoming question asking to ChatGPT API
+   */
+  static async handleQuestion(ctx: TextMatchedContext) {
+    this.showTypingAction(ctx);
+    const question = (ctx.match?.groups?.question || '').trim();
+    if (question.length < 7) {
+      ctx.reply('Muy cortito amigo');
+      return;
+    }
+
+    const reply = ctx.message.reply_to_message as Message.TextMessage;
+    const contextText = reply?.text || '';
+    const contextRole = reply?.from?.id === espiId ? 'system' : 'user';
+    const context: ChatCompletionMessageParam | undefined = contextText
+      ? { role: contextRole, content: contextText }
+      : undefined;
+    const answer = await GPT.ask(question, context);
+    ctx.reply(answer);
+  }
+
+  /**
+   * Transcript audio using ChatGPT API
+   */
+  static async transcriptAudio(ctx: TranscriptContext) {
+    this.showTypingAction(ctx);
+    const voice = ctx.message?.voice;
+    const link = await ctx.telegram.getFileLink(voice.file_id);
+    const { data: buffer } = await api.get<Uint8Array>(link.href, { responseType: 'arraybuffer' });
+
+    // If audio is too big, don't transcribe it and return
+    if (buffer.length > this.maxAudioSize) {
+      ctx.reply('El audio es muy grande, no puedo transcribirlo');
+      return;
+    }
+
+    // Transcript audio and reply with it
+    const transcription = await GPT.transcript(buffer, ctx.message?.message_id);
+    ctx.reply(transcription);
+  }
+}

src/controllers/GPT/types.ts

@@ -0,0 +1,9 @@
+import { Context, NarrowedContext } from 'telegraf';
+import { Message, Update } from 'telegraf/typings/core/types/typegram';
+
+export type TextMatchedContext = NarrowedContext<
+  Context & { match: RegExpExecArray },
+  Update.MessageUpdate<Message.TextMessage>
+>;
+
+export type TranscriptContext = NarrowedContext<Context, Update.MessageUpdate<Message.VoiceMessage>>;

src/controllers/StickersController.ts

@@ -1,29 +1,28 @@
 import { Context } from 'telegraf';
 import { Update } from 'telegraf/typings/core/types/typegram';
 
-import stickers from '../features/stickers';
+import BaseController from './BaseController';
+import stickers from '../core/stickers';
 
-const StickersController = {
-  replyWithSticker: async (ctx: Context<Update>, stickerPromise: Promise<string>) => {
+export default class StickersController extends BaseController {
+  static async replyWithSticker(ctx: Context<Update>, stickerPromise: Promise<string>) {
     const sticker = await stickerPromise;
     if (!sticker) {
       await ctx.reply('Acá debería venir un sticker pero no lo encontré :(');
     }
 
     await ctx.replyWithSticker(sticker);
-  },
+  }
 
-  replyWithMaybeFacu: async (ctx: Context<Update>) => {
+  static async replyWithMaybeFacu(ctx: Context<Update>) {
     await StickersController.replyWithSticker(ctx, stickers.maybeFacu());
-  },
+  }
 
-  replyWithPaintedDog: async (ctx: Context<Update>) => {
+  static async replyWithPaintedDog(ctx: Context<Update>) {
     await StickersController.replyWithSticker(ctx, stickers.paintedDog());
-  },
+  }
 
-  replyWithPatternMatchingDan: async (ctx: Context<Update>) => {
+  static async replyWithPatternMatchingDan(ctx: Context<Update>) {
     await StickersController.replyWithSticker(ctx, stickers.patternMatchingDan());
-  },
-};
-
-export default StickersController;
+  }
+}

src/controllers/types.ts

@@ -0,0 +1,4 @@
+import { Context, NarrowedContext } from 'telegraf';
+import { Update } from 'telegraf/typings/core/types/typegram';
+
+export type BaseContext = NarrowedContext<Context, Update.MessageUpdate>;

src/core/GPT.ts

@@ -0,0 +1,65 @@
+import fs from 'fs';
+
+import OpenAI from 'openai';
+import { ChatCompletionMessageParam } from 'openai/resources';
+
+import { openAI } from '../config';
+import logger from '../lib/logger';
+
+/**
+ * Wrapper for OpenAI API
+ * @see https://platform.openai.com/docs/api-reference
+ */
+const openai = new OpenAI({ apiKey: openAI.apiKey });
+
+/**
+ * Wrapper class for ChatGPT API
+ */
+export default class GPT {
+  /**
+   * Ask a question to ChatGPT API
+   * @param question Question to ask
+   * @param context Context to provide to the API
+   * @returns Answer from the API
+   */
+  static async ask(question: string, context?: ChatCompletionMessageParam) {
+    const messages: ChatCompletionMessageParam[] = [
+      { role: 'system', content: 'Sus un bot de Telegram llamado Espi. Estás en el grupo "12 Cactus".' },
+    ];
+    if (context) messages.push(context);
+    messages.push({ role: 'user', content: question });
+
+    const response = await openai.chat.completions.create({
+      ...openAI.defaultOptions,
+      messages,
+    });
+
+    const answer = response.choices[0].message?.content || '';
+    return answer || 'Ups... no se que decirte';
+  }
+
+  /**
+   * Transcribe an audio file
+   * @param buffer Audio file buffer
+   * @param messageId Message ID to use as filename
+   * @returns Transcription
+   */
+  static async transcript(buffer: Uint8Array, messageId = 42) {
+    const filename = `./voice-${messageId}.mp3`;
+    try {
+      await fs.promises.writeFile(filename, buffer);
+      const file = fs.createReadStream(filename);
+      const transcript = await openai.audio.transcriptions.create({
+        file,
+        model: 'whisper-1',
+        language: 'es',
+      });
+      return transcript.text;
+    } catch (error: any) {
+      logger.error(error, { filename });
+      return 'Hubo un problema desconocido al transcribir el audio...';
+    } finally {
+      await fs.promises.rm(filename, { force: true });
+    }
+  }
+}