PaddlePaddle/PaddleOCR · error · ValueError

Unknown provider: {provider}

Error message

Unknown provider: {provider}

What it means

ValueError from the paddleocr_mcp __main__ provider dispatch: after handling the appstore and self_hosted branches (and qianfan above), any other --provider value falls into the else and raises 'Unknown provider'. The printed message uses the raw provider string, and async_main catches ValueError, prints it, and exits with code 2.

Source

Thrown at mcp_server/paddleocr_mcp/__main__.py:219

            poll_timeout=float(args.aistudio_poll_timeout),
        )
    elif provider == InferenceProvider.QIANFAN.value:
        return create_inference(
            model=model,
            provider=provider,
            base_url=args.qianfan_base_url,
            api_key=args.qianfan_api_key,
            http_timeout=args.http_timeout,
        )
    elif provider == InferenceProvider.SELF_HOSTED.value:
        return create_inference(
            model=model,
            provider=provider,
            base_url=args.self_hosted_base_url,
            http_timeout=args.http_timeout,
        )
    else:
        raise ValueError(f"Unknown provider: {provider}")


async def async_main() -> None:
    """Asynchronous main entry point."""
    args = _parse_args()
    _validate_args(args)

    try:
        model = resolve_model(args.model, args.ppocr_source)
    except ValueError as e:
        print(f"Error: {e}", file=sys.stderr)
        sys.exit(2)

    inference = _create_inference_from_args(args, model)

    try:
        await inference.start()

View on GitHub (pinned to 2661c7c0ef)

Solutions

1. Use one of the exact InferenceProvider values (e.g. appstore, qianfan, self_hosted) — check the enum in paddleocr_mcp for the authoritative list.
2. If the provider comes from an environment variable, strip it: os.environ[...].strip().
3. Check `--help` output of the mcp server for the allowed choices.

Example fix

// before
paddleocr-mcp --provider local
Error: Unknown provider: local

// after
paddleocr-mcp --provider self_hosted

from paddleocr_mcp import InferenceProvider  # or import from its selection/config module

def provider_valid(provider: str) -> bool:
    return (provider or "").strip() in {p.value for p in InferenceProvider}

from typing import Any

def is_known_provider(value: Any) -> bool:
    if isinstance(value, InferenceProvider):
        return True
    return isinstance(value, str) and value.strip() in {
        p.value for p in InferenceProvider
    }

try:
    create_provider(args)
except ValueError as e:
    if "Unknown provider" in str(e):
        # exit code 2 already set by async_main; log allowed values for the operator
        log.error("provider must be one of %s", [p.value for p in InferenceProvider])
    raise

When it happens

Trigger: Starting the MCP server with --provider set to a value not in InferenceProvider (e.g. 'local', 'qianfan ' with whitespace, wrong casing like 'SelfHosted'); a typo on the CLI.

Common situations: CLI flags copied from outdated docs where provider names changed; env var supplying the provider with trailing newline/whitespace.

Related errors

• Unsupported model: {normalized!r}. Supported models: {suppor
• Model {normalized!r} is not supported with qianfan source. S
• Invalid data URL: expected a comma after the MIME type.
• Invalid Base64 input: {e}. Ensure the string is complete and
• The input data is inconsistent with expectations.

AI-assisted analysis of PaddlePaddle/PaddleOCR@2661c7c0ef (2026-08-14). Data as JSON: /api/errors/9144eeb1cfddda3c. Report an issue: GitHub.