---

name: setup-and-cli description: "Install and inspect FedML, then operate the FedML CLI for login, devices, runs, clusters, storage, network diagnostics, and environment checks." disable-model-invocation: true metadata: disco-role: operating parent_skill: "fedml" license: Apache 2.0

FedML Setup and CLI

Use this sub-skill for package installation, import repair, CLI discovery, account login/logout, device binding, run/cluster inspection, storage operations, backend network diagnostics, and environment snapshots.

Do not use this for

• Building or launching job packages: use ../launch-and-packaging/SKILL.md.
• Training loop implementation: use ../distributed-training/SKILL.md.
• Federated-learning algorithm/workflow setup: use ../federated-learning/SKILL.md.
• Model cards and serving endpoints: use ../model-serving/SKILL.md.
• Workflow DAG composition: use ../workflow-orchestration/SKILL.md.

Setup path

1. Read ../../references/installation.md.

2. Install with pip install fedml or pip install -e python from a source checkout.

3. Run the offline smoke check from the root skill directory:

   python scripts/check_install.py
   

4. If the import fails, read ../../references/troubleshooting.md#import-and-dependency-failures.

CLI discovery path

Use only help/version commands until the user approves remote side effects:

fedml --help
fedml version
fedml login --help
fedml run --help
fedml cluster --help
fedml storage --help
fedml network --help
fedml env --help

Important repo-specific command facts:

• Connectivity diagnostics are under fedml network, not fedml diagnosis.
• The current CLI root does not expose fedml jobs; use fedml launch and fedml run instead.

Account and platform operations

Before running any command that changes remote state:

1. Confirm the target backend version (release, test, dev, or local).
2. Confirm the API key/source of credentials.
3. Confirm whether remote state changes are allowed.

Typical commands:

fedml login <api-key>
fedml logout
fedml device bind --help
fedml cluster list
fedml run list
fedml run status --help
fedml run logs --help
fedml storage list --help

Destructive commands such as cluster stop, cluster kill, run stop, storage delete, and logout require explicit approval.

Python API equivalents

Read ../../references/api-reference.md#public-fedmlapi-helpers when converting CLI work into Python. The usual setup/inspection APIs are:

• fedml.api.fedml_login(api_key)
• fedml.api.run_list(...), run_status(...), run_logs(...), run_stop(...)
• fedml.api.cluster_list(...), cluster_status(...), cluster_start/stop/kill/autostop(...)
• fedml.api.upload(...), download(...), list_storage_objects(...), get_storage_metadata(...)

Exit criteria

A setup/CLI task is complete when the correct target environment is active, import fedml works, offline help/version checks pass, and any remote operation has a clear credential/side-effect decision recorded.