mvanhorn/last30days-skill · error · RuntimeError

Unsupported reasoning provider: {provider_name}

Error message

Unsupported reasoning provider: {provider_name}

What it means

RuntimeError raised in mock_runtime when LAST30DAYS_REASONING_PROVIDER (after 'auto' maps to 'gemini') is not a key in _MODEL_DEFAULTS. Mock mode resolves model pins without live credentials but still requires a known provider name; unknown names are rejected here.

Source

Thrown at skills/last30days/scripts/lib/providers.py:226

        default_rerank = GEMINI_PRO

    planner_model = config.get("LAST30DAYS_PLANNER_MODEL") or default_planner
    rerank_model = config.get("LAST30DAYS_RERANK_MODEL") or default_rerank

    if provider_name == "gemini":
        _require_gemini_31(planner_model, role="planner")
        _require_gemini_31(rerank_model, role="rerank")

    return planner_model, rerank_model


def mock_runtime(config: dict[str, Any], depth: str) -> schema.ProviderRuntime:
    """Resolve model pins for mock mode without requiring live credentials."""
    provider_name = (config.get("LAST30DAYS_REASONING_PROVIDER") or "gemini").lower()
    if provider_name == "auto":
        provider_name = "gemini"
    if provider_name not in _MODEL_DEFAULTS:
        raise RuntimeError(f"Unsupported reasoning provider: {provider_name}")

    planner_model, rerank_model = _resolve_model_pins(config, depth, provider_name)
    return schema.ProviderRuntime(
        reasoning_provider=provider_name,
        planner_model=planner_model,
        rerank_model=rerank_model,

        x_search_backend=_resolve_x_backend(config),
    )


def resolve_runtime(config: dict[str, Any], depth: str) -> tuple[schema.ProviderRuntime, ReasoningClient | None]:
    """Resolve the reasoning provider and pinned models."""
    provider_name = (config.get("LAST30DAYS_REASONING_PROVIDER") or "auto").lower()
    google_key = config.get("GOOGLE_API_KEY") or config.get("GEMINI_API_KEY") or config.get("GOOGLE_GENAI_API_KEY")
    openai_token = config.get("OPENAI_API_KEY")
    xai_key = config.get("XAI_API_KEY")

View on GitHub (pinned to c7460f6114)

Solutions

1. Set LAST30DAYS_REASONING_PROVIDER to a supported value (gemini, openai, xai, openrouter) or 'auto'
2. Check the exact spelling in ~/.config/last30days/.env and the session environment
3. If you added a custom provider, register its defaults in _MODEL_DEFAULTS as well

Example fix

# before
export LAST30DAYS_REASONING_PROVIDER=gemni  # typo

# after
export LAST30DAYS_REASONING_PROVIDER=gemini

SUPPORTED = {"gemini", "openai", "xai", "openrouter", "auto"}
provider = (os.environ.get("LAST30DAYS_REASONING_PROVIDER") or "gemini").strip().lower()
assert provider in SUPPORTED, f"set LAST30DAYS_REASONING_PROVIDER to one of {sorted(SUPPORTED)}"

def is_supported_provider(name: str | None) -> bool:
    normalized = (name or "gemini").strip().lower()
    return normalized == "auto" or normalized in {"gemini", "openai", "xai", "openrouter"}

try:
    runtime = mock_runtime(config, depth)
except RuntimeError as e:
    if "Unsupported reasoning provider" in str(e):
        config["LAST30DAYS_REASONING_PROVIDER"] = "gemini"
        runtime = mock_runtime(config, depth)

When it happens

Trigger: Running the engine in mock mode with LAST30DAYS_REASONING_PROVIDER set to a typo ('gemni', 'Gemeni' after .lower(), 'azure') or a provider added to resolve_runtime but not to _MODEL_DEFAULTS.

Common situations: Typos in .env; copy-pasting provider names from other tools; adding a new provider to the real resolution path but forgetting the mock path; CI mock runs with a stale env file.

Related errors

• Gemini selected but no Google API key is configured.
• xAI selected but XAI_API_KEY is not configured.
• OpenRouter selected but OPENROUTER_API_KEY is not configured
• No sources are available for this run.
• No X backend is available.

AI-assisted analysis of mvanhorn/last30days-skill@c7460f6114 (2026-08-15). Data as JSON: /api/errors/697152a772dedc78. Report an issue: GitHub.