| 42 | } |
| 43 | |
| 44 | void print_help() { |
| 45 | std::cout |
| 46 | << "audiocpp_server --config <server.json> [--host <ip>] [--port <port>] [--backend <backend>]\n" |
| 47 | << " [--device <id>] [--threads <n>] [--busy-timeout-ms <ms>]\n" |
| 48 | << " [--model-spec-override <json-or-directory>]\n" |
| 49 | << " [--log] [--log-file <path>]\n" |
| 50 | << " --backend cpu|cuda|vulkan|metal default cuda\n" |
| 51 | << " --busy-timeout-ms <ms> fail a request with 503 when the model has been\n" |
| 52 | << " busy this long; default 300000, 0 disables\n" |
| 53 | << "\n" |
| 54 | << "Endpoints:\n" |
| 55 | << " GET /health\n" |
| 56 | << " GET /v1/models\n" |
| 57 | << " GET /v1/audio/voices?model=<id>\n" |
| 58 | << " POST /v1/audio/speech\n" |
| 59 | << " POST /v1/audio/transcriptions\n" |
| 60 | << " OpenAI-style streaming: speech stream_format=sse|audio, transcription stream=true\n" |
| 61 | << " POST /v1/tasks/run\n"; |