wavekat

WaveKat CLI

wk is the command-line client for the WaveKat platform. Sign in once with your browser, then inspect projects, list annotations, snapshot exports, and adapt them into HuggingFace-loadable datasets — all from the terminal.

What you can do today

CommandWhat it shows
wk login / wk logoutSign in via your browser, or sign out
wk projects listProjects you can see, with role, counts, and review progress
wk annotations list <project-id>Paginated annotations with inline ASR text
wk exports create <project-id>Snapshot the current label set into a frozen export
wk exports download <export-id>Fetch manifest + every clip
wk exports adapt smart-turn …Convert a downloaded export into HF datasets Parquet shards
wk models list <project-id>Models trained in a project (lineage, metrics, status)
wk models push …Register a trained model + upload its artifacts

Every list command supports --page / --page-size and prints a ready-to-paste Next: line when more pages exist. Add --json to any command for machine-readable output.

This is the highlight reel — see Reference for the full surface (files, model downloads, self-update, etc.), or just run wk --help, wk <group> --help, or wk <group> <command> --help against the binary. Clap-generated help is the authoritative source.

Supported on macOS (Apple Silicon + Intel) and Linux (x86_64 + aarch64).

Where to go next

  • Getting Started — install, sign in, list your first project.
  • Usage — common workflows including the end-to-end dataset pipeline.
  • Reference — every command, every flag, every endpoint.