Hmbown/CodeWhale · error · anyhow::Error

Unrecognized --provider {provider_arg:?}. Known providers: {

Error message

Unrecognized --provider {provider_arg:?}. Known providers: {} or a configured [providers.<name>] custom provider

What it means

`--provider` must resolve to either a built-in ApiProvider id (parsed case-insensitively, e.g. deepseek, deepseek-cn, zai, moonshot, openrouter, xai, together, ...) or the name of a custom provider declared in a `[providers.<name>]` table in the loaded config (checked via custom_provider_config). If both lookups miss, the CLI bails and prints the built-in names this build knows.

Source

Thrown at crates/tui/src/lib.rs:861

fn apply_exec_provider_override(config: &mut Config, provider_arg: &str) -> Result<()> {
    let provider_arg = provider_arg.trim();
    if provider_arg.is_empty() {
        return Ok(());
    }
    if config
        .providers
        .as_ref()
        .and_then(|providers| providers.custom_provider_config(provider_arg))
        .is_some()
    {
        config.provider = Some(provider_arg.to_string());
        return Ok(());
    }
    if let Some(provider) = crate::config::ApiProvider::parse(provider_arg) {
        config.provider = Some(provider.as_str().to_string());
        return Ok(());
    }
    bail!(
        "Unrecognized --provider {provider_arg:?}. Known providers: {} \
         or a configured [providers.<name>] custom provider",
        crate::config::ApiProvider::names_hint()
    );
}

fn exec_model_env_override() -> Option<String> {
    let read = || {
        ["CODEWHALE_MODEL", "DEEPSEEK_MODEL"]
            .into_iter()
            .find_map(|key| {
                std::env::var(key)
                    .ok()
                    .map(|model| model.trim().to_string())
                    .filter(|model| !model.is_empty())
            })
    };
    #[cfg(test)]

View on GitHub (pinned to 0c42157ee5)

Solutions

Re-run and read the printed known-provider list; use the exact id with correct dashes/underscores
For a custom endpoint, add a `[providers.<name>]` table (kind + base_url) to the config and pass that name
Verify which config file is in effect (--config flag, profile) and that the table is inside it
Check case-insensitive spelling: deepseek-cn, deepseek_china, deepseekcn, deepseek-china all parse

Example fix

# before
codewhale exec --provider open_ai "hi"          # Unrecognized --provider "open_ai"

# after (built-in)
codewhale exec --provider openai "hi"

# after (custom, in codewhale config)
[providers.lm-studio]
kind = "openai-compatible"
base_url = "http://127.0.0.1:1234/v1"
# then: codewhale exec --provider lm-studio "hi"

Defensive patterns

Strategy: validation

Validate before calling

# Fail fast on unknown provider ids before launching
PROVIDER="$1"
codewhale exec --provider "$PROVIDER" --help >/dev/null 2>&1 || true
# authoritative check: known built-ins or a [providers.<name>] table in config
grep -q "^\[providers\.$PROVIDER\]" "$(codewhale config path 2>/dev/null || echo config.toml)" \
  || codewhale --help | grep -q "\b$PROVIDER\b" \
  || { echo "unknown provider: $PROVIDER"; exit 2; }

Try / catch

capture stderr; if it matches /Unrecognized --provider/, print the known list from the message and fail fast instead of retrying with the same id

Prevention

Source provider ids from the error's own hint list, never from memory
Define custom endpoints as [providers.<name>] tables before referencing them
Assert the provider name in CI before long exec runs

When it happens

Trigger: Passing a typo'd or unsupported id (`--provider gpt4`, `--provider open_ai`, `--provider Open AI`); naming a custom provider whose `[providers.<name>]` table is absent from the config file actually loaded (for example when --config points elsewhere); using an id renamed or removed in a newer release.

Common situations: Scripting `codewhale exec --provider` from memory instead of the hint list; adding a local LM Studio/Ollama-compatible endpoint but forgetting the [providers.<name>] table; config drift between machines; legacy aliases like deepseek-china that only ApiProvider::parse accepts.

Related errors

AI-assisted analysis of Hmbown/CodeWhale@0c42157ee5 (2026-08-20). Data as JSON: /api/errors/836bd139d49714f9. Report an issue: GitHub.