The search service can find package by either name (apache), provides(webserver), absolute file names (/usr/bin/apache), binaries (gprof) or shared libraries (libXm.so.2) in standard path. It does not support multiple arguments yet...

The System and Arch are optional added filters, for example System could be "redhat", "redhat-7.2", "mandrake" or "gnome", Arch could be "i386" or "src", etc. depending on your system.

RPM resource whisper-cpp-server

HTTP server for whisper.cpp. WAV (or, with --convert and ffmpeg, other audio) is posted to the model. Config: /etc/sysconfig/whisper-server Unit: whisper.service (default 127.0.0.1:8081, not 8080, so it can run next to llama-server) systemctl enable --now whisper.service Models from https://huggingface.co/ggerganov/whisper.cpp (ggml-*.bin). If MODEL is unset, the unit picks the best readable file from MODEL_ORDER in /etc/sysconfig/whisper-server (large-v3 down to tiny), searching /srv/ai then /usr/share/whisper-cpp/models. Packaged whisper-cpp-model-* RPMs land in the second directory. Override with MODEL=. The unit hides /home; extra model trees need a drop-in ReadOnlyPaths=-/other/models. /usr is already readable. GPU offload uses whatever ggml backends are installed. Disable with --no-gpu. Device index: --device N (or WHISPER_ARG_DEVICE). To transcribe: curl http://127.0.0.1:8081/inference \ -H "Content-Type: multipart/form-data" \ -F file="@audio.wav" \ -F response_format="json" To hot-swap the model (path must be readable, e.g. under /srv/ai): curl http://127.0.0.1:8081/load \ -H "Content-Type: multipart/form-data" \ -F model="/srv/ai/ggml-small.en.bin" There is no built-in API key. Do not bind HOST to a public address without a reverse proxy. --convert shells out to ffmpeg.

Found 1 sites for whisper-cpp-server

• https://github.com/ggml-org/whisper.cpp

Found 2 RPM for whisper-cpp-server

Generated by rpm2html 1.6

Fabrice Bellet