メインコンテンツへスキップ

Grida FX in the Grida CLI

grida fx runs Grida FX: write a workflow file, and FX plans it, prices it before anything is spent, caches every paid call by its request, and records the run.

grida fx --help
grida fx plan workflow.yaml
grida fx run workflow.yaml --live --max-usd 2

Each grida release carries the Grida FX version it pins, so updating grida updates FX. grida fx --version prints that version. FX owns these commands, their options and their help: grida fx <command> --help describes each one. The FX guide covers workflow files, nodes, costs, caching and runs.

Platforms​

FX runs on macOS and on Linux with glibc 2.28 or later. On Windows, use WSL 2; on Alpine and other musl systems, use a glibc-based image. Elsewhere, grida fx exits with status 1 and explains this. The other grida commands are unaffected.

Provider keys​

FX reads provider keys from the environment, or from a .env file in the FX project. grida fx doctor reports which keys each workflow needs and which it found, never their values. Reading the keys that grida ai providers stores is planned.

FX on its own​

The JavaScript SDK is @grida/fx (npm install @grida/fx), which also installs a standalone grida-fx command. The Python SDK is pip install grida, run as python -m grida.fx. All three run the same engine, and what they print and record is the same apart from the command name they show.