invalid value ' ' for '--provider ': expected one of …

Error message

invalid value '{provider}' for '--provider <PROVIDER>': expected one of {expected}; configured custom providers are accepted only by exec and fleet

What it means

`top_level_provider_override` fails when `--provider` receives a value that is neither a builtin `ProviderKind` nor a legacy antigravity selector, on a command that does not accept raw provider ids. Raw/custom provider ids are restricted to the `exec` and `fleet` subcommands, so everywhere else only builtin provider names are legal.

Solutions

Use a builtin provider name listed in the error's `expected one of {expected}` hint.
Move the custom provider usage to the `exec` or `fleet` subcommand, which are the only commands accepting raw provider ids.
Fix the spelling of the provider name in the command line or script.

Example fix

// before
codewhale --provider my-custom "do work"
// after
codewhale exec --provider my-custom "do work"

Defensive patterns

Strategy: validation

Validate before calling

const BUILTIN = new Set(['openai','anthropic']); // see ProviderKind::names_hint()
const RAW_OK = ['exec','fleet'];
if (!BUILTIN.has(provider) && !RAW_OK.includes(subcommand)) throw new Error(`custom provider '${provider}' only allowed via exec/fleet`);

Prevention

Validate `--provider` against the hint list before invoking.
Route custom providers exclusively through exec/fleet subcommands.

When it happens

Trigger: `codewhale --provider my-custom-provider ...` on any command where `command_accepts_raw_provider` is false; misspelling a builtin name (e.g. `--provider antropic`); using a custom provider id configured in config files at the top level instead of via exec/fleet.

Common situations: Typo'd provider names; users assuming their configured custom providers work everywhere; old scripts using a custom id at top level after the exec/fleet-only restriction was introduced.

Understand the failure class

Background: "Unknown argument", "Invalid value", and "must be one of": invalid CLI argument errors explained — this error's family across 35 libraries.

Related errors

AI-assisted analysis of Hmbown/CodeWhale@73e0f67d83 (2026-09-22). Data as JSON: /api/errors/028f60de649763ac. Report an issue: GitHub.

Appendix: source

Thrown at crates/cli/src/lib.rs:537

fn top_level_provider_override(
    provider: Option<&str>,
    command: Option<&Commands>,
) -> Result<Option<ProviderKind>> {
    let Some(provider) = provider else {
        return Ok(None);
    };
    if is_antigravity_legacy_selector(provider) {
        bail!(codewhale_config::LEGACY_ANTIGRAVITY_TOMBSTONE_MESSAGE);
    }
    if let Some(provider) = builtin_provider_arg(provider) {
        return Ok(Some(provider));
    }
    if command_accepts_raw_provider(command) {
        return Ok(None);
    }

    let expected = ProviderKind::names_hint();
    bail!(
        "invalid value '{provider}' for '--provider <PROVIDER>': expected one of {expected}; configured custom providers are accepted only by exec and fleet"
    )
}

fn prepare_raw_provider_tui_dispatch(
    cli: &Cli,
    command: Option<&Commands>,
    runtime_overrides: &CliRuntimeOverrides,
) -> Result<Option<(ResolvedRuntimeOptions, Vec<String>)>> {
    let Some(provider) = cli.provider.as_deref() else {
        return Ok(None);
    };
    if builtin_provider_arg(provider).is_some() || !command_accepts_raw_provider(command) {
        return Ok(None);
    }

    let passthrough = match command {
        Some(Commands::Exec(args)) => {

View on GitHub (pinned to 73e0f67d83)