headroomlabs-ai/headroom · error · ValueError

Unsupported api_style: {api_style}

Error message

Unsupported api_style: {api_style}

What it means

Provider requests are dispatched through _CLIENT_TRANSPORTS, which currently maps exactly two api_style keys: "anthropic" and "openai". dispatch_client_call() looks up the style and re-raises KeyError as ValueError('Unsupported api_style: ...') when a client was built with any other style string. This indicates a provider/client configuration mismatch rather than a network problem.

Source

Thrown at headroom/providers/registry.py:282

        return f"{provider_config.display_name} via LiteLLM (region={bedrock_region})"
    return f"{provider_config.display_name} via LiteLLM"


def call_client_transport(
    api_style: str,
    client: Any,
    *,
    model: str,
    messages: list[dict[str, Any]],
    stream: bool,
    metrics: Any,
    **kwargs: Any,
) -> Any:
    """Dispatch the SDK request to the provider-specific transport handler."""
    try:
        transport = _CLIENT_TRANSPORTS[api_style]
    except KeyError as exc:
        raise ValueError(f"Unsupported api_style: {api_style}") from exc

    return transport(
        client,
        model=model,
        messages=messages,
        stream=stream,
        metrics=metrics,
        **kwargs,
    )


def _load_anyllm_backend() -> Any:
    global AnyLLMBackendType
    if AnyLLMBackendType is None:
        from headroom.backends.anyllm import AnyLLMBackend

        AnyLLMBackendType = AnyLLMBackend
    return AnyLLMBackendType

View on GitHub (pinned to 322425c43b)

Solutions

1. Set the provider's api_style to "anthropic" or "openai" — the only styles _CLIENT_TRANSPORTS registers in this version.
2. Check for case/whitespace issues: the lookup is an exact dict key match, so "OpenAI" or " anthropic" fail.
3. If you expected a newer style (e.g. for Gemini pass-through), upgrade headroom or route Gemini traffic through its dedicated handler instead of the generic client dispatch.

Example fix

# before
client = build_client(provider_cfg)  # api_style="gemini"
response = dispatch_client_call(client, api_style="gemini", ...)  # ValueError

# after
response = dispatch_client_call(client, api_style="openai", ...)  # gemini via openai-compatible endpoint

SUPPORTED_API_STYLES = frozenset({"anthropic", "openai"})

def api_style_ok(style: str) -> bool:
    return style in SUPPORTED_API_STYLES

from typing import Literal

ApiStyle = Literal["anthropic", "openai"]

def is_api_style(value: str) -> TypeGuard[ApiStyle]:
    return value in ("anthropic", "openai")

try:
    result = dispatch_client_call(client, api_style=style, ...)
except ValueError as exc:
    if "Unsupported api_style" in str(exc):
        raise ConfigError(f"{style!r} not supported here; use anthropic|openai") from exc
    raise

When it happens

Trigger: Creating a provider client with an api_style outside {"anthropic", "openai"} — e.g. "gemini", "litellm", "generic", a typo like "openAi", or a style added by a newer/older headroom version — and then issuing a request that goes through dispatch_client_call().

Common situations: Copy-pasting a provider config from docs for a version that supported a different style set; hand-rolling a provider registration with an invented style name; version skew between the config file and the installed headroom package.

Related errors

• Unknown provider: {provider}
• Unknown provider: {self.llm_config.provider}
• recommendations IO error at {path}: {source}
• bedrock_eventstream_parse_failed
• bedrock_eventstream_crc_mismatch

AI-assisted analysis of headroomlabs-ai/headroom@322425c43b (2026-08-15). Data as JSON: /api/errors/e0d8a2871da3ac60. Report an issue: GitHub.