Fix YouTube downloads - Add Deno runtime for video cipher decryption

[Ixian]

Fix YouTube Downloads - Issue #224

Problem

YouTube downloads were failing with "exit status 1" error. Root cause: YouTube now requires yt-dlp to use a JavaScript runtime for video cipher decryption (see yt-dlp/yt-dlp#14404).

Solution

1. Install Deno Runtime

Added Deno installation to both Dockerfiles (standard and CUDA) since YouTube requires JavaScript execution for video decryption.

Files: Dockerfile lines 71-77, Dockerfile.cuda lines 74-80

# Install Deno (JavaScript runtime required for yt-dlp YouTube downloads)
# YouTube now requires JS execution for video cipher decryption
# See: https://github.com/yt-dlp/yt-dlp/issues/14404
RUN curl -fsSL https://deno.land/install.sh | sh \
  && cp /root/.deno/bin/deno /usr/local/bin/deno \
  && chmod 755 /usr/local/bin/deno \
  && deno --version

2. Upgrade to yt-dlp[default]

Changed from yt-dlp to yt-dlp[default] to include all optional dependencies (brotli, pycryptodomex, websockets, mutagen).

File: internal/transcription/adapters/whisperx_adapter.go line 349

3. Add Error Diagnostics

Added stderr capture to YouTube download command to surface actual yt-dlp errors instead of generic "exit status 1" messages.

File: internal/api/handlers.go lines 2700-2753

var stderr bytes.Buffer
ytDlpCmd.Stderr = &stderr

if err := ytDlpCmd.Run(); err != nil {
    stderrOutput := stderr.String()
    logger.Error("YouTube download failed",
        "error", err.Error(),
        "stderr", stderrOutput)
    c.JSON(http.StatusInternalServerError, gin.H{
        "error":   fmt.Sprintf("Failed to download YouTube audio: %v", err),
        "details": stderrOutput,
    })
    return
}

4. Add Performance Logging

Added timing metrics for YouTube downloads to identify bottlenecks.

File: internal/api/handlers.go - Download timing and file size logging

Testing

Manual verification:

• Deno runtime installed and accessible in container
• yt-dlp[default] with all dependencies installed
• YouTube downloads working successfully
• Error capture properly surfacing stderr output
• Performance logging operational

Test case: 30-minute YouTube video

• Download: ~20 seconds
• Title retrieval: 1-2 seconds
• WhisperX transcription: 12.49x realtime on RTX GPU
• PyAnnote diarization: ~14 minutes
• Total: 16m40s end-to-end

Files Modified

1. Dockerfile - Added Deno installation
2. Dockerfile.cuda - Added Deno installation
3. internal/api/handlers.go - Added stderr capture and YouTube performance logging
4. internal/transcription/adapters/whisperx_adapter.go - Changed to yt-dlp[default]

Related Issues

Closes #224

[rishikanthc]

Thank you so much for your contribution. Once again stellar work just like the gpu PR. Appreciate the time and help :)

[sjsamdrake]

Great that it's fixed. When might this make it into the :latest container image? Thanks!