ai-terminal-remote-bot 1.0.2 → 1.0.4

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ai-terminal-remote-bot",
-  "version": "1.0.2",
+  "version": "1.0.4",
   "description": "Secure remote terminal bot with Claude Code integration",
   "type": "module",
   "bin": {
@@ -24,8 +24,6 @@
     "node": ">=18.0.0"
   },
   "dependencies": {
-    "@huggingface/transformers": "^3.4.1",
-    "telegraf": "^4.16.3",
-    "wavefile": "^11.0.0"
+    "telegraf": "^4.16.3"
   }
 }

package/src/i18n/locales/en.json

@@ -52,6 +52,8 @@
     "imageSaved": "🖼️ Image saved: `{{dest}}`",
     "claudeImageWithCaption": "Look at this image I saved at {{dest}} and: {{caption}}",
     "claudeImageAnalyze": "Look at the image I saved at {{dest}} and describe it. If it's a code or error screenshot, analyze it.",
+    "claudeAudioTranscribe": "Transcribe the audio file at {{dest}}. Return only the transcription text.",
+    "claudeAudioWithCaption": "Transcribe the audio file at {{dest}} and then respond to: {{caption}}",
     "audioSaved": "🎤 Audio saved: `{{dest}}`\n\nTranscribing...",
     "transcription": "🎤 *Transcription:*\n\n{{text}}",
     "transcriptionFailed": "⚠️ Could not transcribe: {{error}}\nAudio saved as file.",
@@ -109,6 +111,7 @@
     "truncated": "... (truncated)"
   },
   "transcriber": {
-    "ffmpegMissing": "ffmpeg is not installed. Install it with: sudo apt install ffmpeg"
+    "ffmpegMissing": "ffmpeg is not installed. Install it with: sudo apt install ffmpeg",
+    "unsupportedFormat": "Unsupported audio format: .{{ext}} (supported: ogg, opus, wav)"
   }
 }

package/src/i18n/locales/es.json

@@ -52,6 +52,8 @@
     "imageSaved": "🖼️ Imagen guardada: `{{dest}}`",
     "claudeImageWithCaption": "Mirá esta imagen que guardé en {{dest}} y: {{caption}}",
     "claudeImageAnalyze": "Mirá la imagen que guardé en {{dest}} y describila. Si es un screenshot de código o error, analizalo.",
+    "claudeAudioTranscribe": "Transcribí el archivo de audio en {{dest}}. Devolvé solo el texto de la transcripción.",
+    "claudeAudioWithCaption": "Transcribí el archivo de audio en {{dest}} y luego respondé a: {{caption}}",
     "audioSaved": "🎤 Audio guardado: `{{dest}}`\n\nTranscribiendo...",
     "transcription": "🎤 *Transcripción:*\n\n{{text}}",
     "transcriptionFailed": "⚠️ No se pudo transcribir: {{error}}\nAudio guardado como archivo.",
@@ -109,6 +111,7 @@
     "truncated": "... (truncado)"
   },
   "transcriber": {
-    "ffmpegMissing": "ffmpeg no está instalado. Instalalo con: sudo apt install ffmpeg"
+    "ffmpegMissing": "ffmpeg no está instalado. Instalalo con: sudo apt install ffmpeg",
+    "unsupportedFormat": "Formato de audio no soportado: .{{ext}} (soportados: ogg, opus, wav)"
   }
 }

package/src/server.mjs

@@ -10,7 +10,7 @@ import { TelegramPlatform } from './platforms/telegram.mjs';
 import { TerminalModule } from './modules/terminal.mjs';
 import { ClaudeModule } from './modules/claude.mjs';
 import { MediaModule } from './modules/media.mjs';
-import { TranscriberModule } from './modules/transcriber.mjs';
+
 import { validateCommand, validatePath, validatePid, createRateLimiter } from './security.mjs';
 import { chunkText } from './utils.mjs';
 import { t } from './i18n/index.mjs';
@@ -45,7 +45,7 @@ export async function startServer(config) {
   });
   const claude = new ClaudeModule({ timeout: config.claudeTimeout });
   const media = new MediaModule({ allowedPaths: config.allowedPaths });
-  const transcriber = new TranscriberModule();
+
   const rateLimiter = createRateLimiter();
 
   // === Session Management ===
@@ -444,16 +444,10 @@ export async function startServer(config) {
 
       await mctx.reply(t('server.audioSaved', { dest }));
 
-      try {
-        const transcript = await transcriber.transcribe(dest);
-        await platform.sendMessage(mctx.chatId, t('server.transcription', { text: transcript }));
-
-        if (session.mode === 'claude') {
-          await handleClaude(mctx, transcript, session);
-        }
-      } catch (txErr) {
-        await mctx.reply(t('server.transcriptionFailed', { error: txErr.message }));
-      }
+      const prompt = mctx.caption
+        ? t('server.claudeAudioWithCaption', { dest, caption: mctx.caption })
+        : t('server.claudeAudioTranscribe', { dest });
+      await handleClaude(mctx, prompt, session);
     } catch (e) {
       mctx.reply(`❌ ${e.message}`);
     }

package/src/modules/transcriber.mjs

@@ -1,110 +0,0 @@
-/**
- * Audio transcription module using @huggingface/transformers (Whisper ONNX).
- * Lazy-loads the pipeline on first use. Model cached in ~/.cache/huggingface/.
- */
-
-import { execFile } from 'child_process';
-import { promisify } from 'util';
-import fs from 'fs';
-import path from 'path';
-import os from 'os';
-import { t, getLang } from '../i18n/index.mjs';
-
-const execFileAsync = promisify(execFile);
-
-export class TranscriberModule {
-  constructor() {
-    this._pipeline = null;
-  }
-
-  /**
-   * Lazy-load the whisper pipeline (downloads ~40MB model on first call).
-   */
-  async _getPipeline() {
-    if (this._pipeline) return this._pipeline;
-
-    const { pipeline } = await import('@huggingface/transformers');
-    this._pipeline = await pipeline('automatic-speech-recognition', 'Xenova/whisper-tiny', {
-      dtype: 'fp32',
-    });
-    return this._pipeline;
-  }
-
-  /**
-   * Convert audio file to 16kHz mono WAV using ffmpeg.
-   * @param {string} inputPath - Path to input audio file
-   * @returns {Promise<string>} Path to converted WAV file
-   */
-  async _convertToWav(inputPath) {
-    const wavPath = path.join(os.tmpdir(), `transcribe_${Date.now()}.wav`);
-
-    await execFileAsync('ffmpeg', [
-      '-i', inputPath,
-      '-ar', '16000',
-      '-ac', '1',
-      '-f', 'wav',
-      '-y',
-      wavPath,
-    ], { timeout: 30000 });
-
-    return wavPath;
-  }
-
-  /**
-   * Parse WAV file into Float32Array of audio samples.
-   * @param {string} wavPath - Path to 16kHz mono WAV file
-   * @returns {Promise<Float32Array>}
-   */
-  async _parseWav(wavPath) {
-    const { default: WaveFile } = await import('wavefile');
-    const wavBuffer = fs.readFileSync(wavPath);
-    const wav = new WaveFile(wavBuffer);
-    wav.toBitDepth('32f');
-    wav.toSampleRate(16000);
-
-    const samples = wav.getSamples();
-    // getSamples returns Float64Array for 32f, convert to Float32
-    if (samples instanceof Float64Array || samples instanceof Float32Array) {
-      return new Float32Array(samples);
-    }
-    // Interleaved multi-channel: take first channel
-    if (Array.isArray(samples)) {
-      return new Float32Array(samples[0]);
-    }
-    return new Float32Array(samples);
-  }
-
-  /**
-   * Transcribe an audio file to text.
-   * @param {string} filePath - Path to audio file (OGG, MP3, WAV, etc.)
-   * @returns {Promise<string>} Transcribed text
-   */
-  async transcribe(filePath) {
-    // Check ffmpeg availability
-    try {
-      await execFileAsync('which', ['ffmpeg']);
-    } catch {
-      throw new Error(t('transcriber.ffmpegMissing'));
-    }
-
-    // Convert to 16kHz WAV
-    const wavPath = await this._convertToWav(filePath);
-
-    try {
-      // Parse WAV to Float32Array
-      const audioData = await this._parseWav(wavPath);
-
-      // Run whisper pipeline
-      const pipe = await this._getPipeline();
-      const result = await pipe(audioData, {
-        language: getLang(),
-        task: 'transcribe',
-      });
-
-      return result.text.trim();
-    } finally {
-      // Clean up temp WAV
-      try { fs.unlinkSync(wavPath); } catch {}
-    }
-  }
-}