Add whispercpp support for audio transcription without PyTorch

- Update transcription endpoint to try faster-whisper first, then whispercpp
- Update pre-loading code to support both backends
- Add whispercpp to all requirements files (vulkan, nvidia, default)
- Remove broken llama.cpp fallback (llama.cpp cannot transcribe Whisper)

requirements-nvidia.txt

@@ -13,6 +13,7 @@ psutil>=5.9.0
 
 # Optional: Audio transcription dependencies
 faster-whisper>=0.10.0  # For NVIDIA/CUDA whisper transcription
+whispercpp>=1.0.0  # Alternative whisper library (works without PyTorch)
 
 # Optional: for better performance with NVIDIA GPUs
 bitsandbytes>=0.41.0

requirements-vulkan.txt

@@ -13,4 +13,6 @@ psutil>=5.9.0
 # HuggingFace Hub for downloading GGUF models
 huggingface-hub>=0.19.0
 
-# No PyTorch needed for Vulkan backend - llama-cpp handles everything
+# Optional: Audio transcription without PyTorch (whispercpp)
+# Note: faster-whisper requires PyTorch, but whispercpp works without it
+whispercpp>=1.0.0  # For GGUF-based Whisper transcription without PyTorch

requirements.txt

@@ -38,7 +38,7 @@ procname>=0.3.0
 
 # Optional: Audio transcription dependencies
 faster-whisper>=0.10.0  # For NVIDIA/CUDA whisper transcription
-# whispercpp>=1.0.0  # Alternative whisper library (requires system dependencies)
+whispercpp>=1.0.0  # Alternative whisper library (works without PyTorch)
 
 # Optional: for better performance
 # bitsandbytes>=0.41.0  # for 4-bit/8-bit quantization