CLI for the Baseten Inference Platform.
Download the latest release for your platform, extract the archive, and place the baseten executable on your PATH.
Or install v0.1.0 with a one-liner:
curl -sL https://github.com/basetenlabs/baseten-cli/releases/download/v0.1.0/baseten_0.1.0_linux_amd64.tar.gz \
| sudo tar xz -C /usr/local/bin baseten
curl -sL https://github.com/basetenlabs/baseten-cli/releases/download/v0.1.0/baseten_0.1.0_darwin_arm64.tar.gz \
| sudo tar xz -C /usr/local/bin baseten
PowerShell:
Invoke-WebRequest `
https://github.com/basetenlabs/baseten-cli/releases/download/v0.1.0/baseten_0.1.0_windows_amd64.zip `
-OutFile baseten.zip; Expand-Archive -Force baseten.zip .
Then move baseten.exe to a directory on your PATH.
Authenticate via baseten auth login, or set BASETEN_API_KEY in the environment.
Run baseten --help (or baseten <command> --help) for the full command tree.
From inside a model directory containing a config.yaml:
baseten model push
The directory defaults to the current working directory and is configurable via --dir. Useful flags:
--tailstreams build and runtime logs to stderr after the push completes.--waitblocks until the deployment reaches an active status and exits non-zero on terminal failure.
baseten model predict --model-id <model-id> --data '{"prompt":"hello"}'
--model-name is also accepted. Pass --file <path> (or --file - for stdin) to send a request body from a file.
baseten model deployment logs --model-id <model-id> --deployment-id <deployment-id> --tail
Omit --tail and pass --since 1h (or --start/--end) to fetch a historical window.
Run baseten --help for more, and see docs.baseten.co for general Baseten platform documentation.
To build from source, clone this repository and run:
go build ./cmd/baseten
See CONTRIBUTING.md for contribution guidelines.