What causes "Unsupported model: {model!r}" in PaddlePaddle/PaddleOCR?

List valid values: `from langchain_paddleocr... import Model; print([m.value for m in Model])` and pass one exactly.

PaddlePaddle/PaddleOCR · error · ValueError

Unsupported model: {model!r}

Error message

Unsupported model: {model!r}

What it means

ValueError from _resolve_vl_model in the langchain-paddleocr PaddleOCRVLLoader when a string model name cannot be converted to the Model enum. The loader accepts either a Model enum member or one of the enum's exact string values; anything else (wrong casing, typo, unknown model id) is rejected with 'Unsupported model: {model!r}'.

Source

Thrown at langchain-paddleocr/langchain_paddleocr/document_loaders/paddleocr_vl.py:43

from paddleocr import Model, PaddleOCRClient, PaddleOCRVLOptions
from pydantic import SecretStr

logger = logging.getLogger(__name__)

_DEFAULT_BASE_URL = "https://paddleocr.aistudio-app.com"

_PAGES_DELIMITER = "\n\f"


def _resolve_vl_model(model: str | Model) -> Model:
    if isinstance(model, Model):
        resolved = model
    else:
        try:
            resolved = Model(model)
        except ValueError as exc:
            msg = f"Unsupported model: {model!r}"
            raise ValueError(msg) from exc

    return resolved


class PaddleOCRVLLoader(BaseLoader):
    """Load documents using the PaddleOCR-VL document parsing API via SDK."""

    def __init__(
        self,
        file_path: str | Iterable[str],
        *,
        access_token: SecretStr | None = None,
        base_url: str | None = None,
        model: str | Model | None = None,
        use_doc_orientation_classify: bool | None = False,
        use_doc_unwarping: bool | None = False,
        use_layout_detection: bool | None = None,
        use_chart_recognition: bool | None = None,

View on GitHub (pinned to 2661c7c0ef)

Solutions

List valid values: `from langchain_paddleocr... import Model; print([m.value for m in Model])` and pass one exactly.
Upgrade langchain-paddleocr if the model you want was added in a newer release.
Pass the Model enum member directly instead of a string to get static checking.

Example fix

// before
loader = PaddleOCRVLLoader(path, model="paddleocr-vl")

// after
from langchain_paddleocr.document_loaders.paddleocr_vl import Model
loader = PaddleOCRVLLoader(path, model=Model.PADDLEOCR_VL)  # or the exact enum string value

Defensive patterns

Strategy: type-guard

Validate before calling

from langchain_paddleocr.document_loaders import paddleocr_vl

def model_supported(model: str) -> bool:
    values = {m.value for m in paddleocr_vl.Model}
    return model in values

Type guard

from typing import Any

def is_vl_model(value: Any) -> bool:
    """True when value is a Model member or one of its exact string values."""
    if isinstance(value, paddleocr_vl.Model):
        return True
    return isinstance(value, str) and value in {m.value for m in paddleocr_vl.Model}

Try / catch

try:
    loader = PaddleOCRVLLoader(path, model=name)
    docs = loader.load()
except ValueError as e:
    if "Unsupported model" in str(e):
        valid = [m.value for m in Model]
        raise ValueError(f"{name!r} invalid; choose from {valid}") from e
    raise

Prevention

Derive model names from the Model enum at runtime instead of hardcoding strings.
Run a startup assert that configured model names are enum members so failures surface at boot, not mid-load.
Upgrade langchain-paddleocr when adopting newly released VL models.

When it happens

Trigger: Passing model="paddleocr-vl" when the enum expects e.g. "PaddleOCR-VL" or a specific version string; passing an arbitrary API model id not present in the local Model enum; passing None-like or empty strings.

Common situations: New server-side model names not yet in the installed langchain-paddleocr version; casing/whitespace differences between docs and enum values.

Related errors

AI-assisted analysis of PaddlePaddle/PaddleOCR@2661c7c0ef (2026-08-14). Data as JSON: /api/errors/b71a8b8191ff87d5. Report an issue: GitHub.