PaddlePaddle/PaddleOCR · error · ValueError

Invalid data URL: expected a comma after the MIME type.

Error message

Invalid data URL: expected a comma after the MIME type.

What it means

ValueError from extract_base64_payload in paddleocr_mcp/utils.py when an input starts with `data:` but contains no comma. Data URLs must look like `data:<mime>;base64,<payload>`; the function splits on the first comma, so a missing comma means the MIME header was never terminated.

Source

Thrown at mcp_server/paddleocr_mcp/utils.py:41

def is_url(value: str) -> bool:
    if not (value.startswith("http://") or value.startswith("https://")):
        return False

    result = urlparse(value)
    return all([result.scheme, result.netloc]) and result.scheme in ("http", "https")


def is_base64(value: str) -> bool:
    pattern = r"^[A-Za-z0-9+/]+={0,2}$"
    return bool(re.fullmatch(pattern, value))


def extract_base64_payload(input_data: str) -> str:
    if input_data.startswith("data:"):
        if "," not in input_data:
            raise ValueError("Invalid data URL: expected a comma after the MIME type.")
        return input_data.split(",", 1)[1]
    return input_data


def decode_base64_payload(payload: str) -> bytes:
    try:
        return base64.b64decode(payload, validate=True)
    except Exception as e:
        raise ValueError(
            f"Invalid Base64 input: {e}. "
            "Ensure the string is complete and correctly padded."
        ) from e


def infer_file_type_from_bytes(data: bytes) -> Optional[str]:
    import puremagic

    mime = puremagic.from_string(data, mime=True)

View on GitHub (pinned to 2661c7c0ef)

Solutions

1. Send a well-formed data URL: `data:image/png;base64,<payload>` with the comma present.
2. Or send the bare base64 string without the `data:` prefix — extract_base64_payload passes it through unchanged.
3. Build data URLs with a helper (f"data:{mime};base64,{b64}") instead of string concatenation.

Example fix

// before
await tool.ocr("data:image/png;base64")

// after
b64 = base64.b64encode(png_bytes).decode()
await tool.ocr(f"data:image/png;base64,{b64}")

def data_url_well_formed(s: str) -> bool:
    if not s.startswith("data:"):
        return True  # bare base64 passes through
    return "," in s

def is_parseable_data_url(value: object) -> bool:
    return isinstance(value, str) and (not value.startswith("data:") or "," in value)

try:
    payload = extract_base64_payload(user_input)
except ValueError as e:
    if "comma after the MIME type" in str(e):
        raise ValueError(
            "expected 'data:<mime>;base64,<payload>' or a bare base64 string"
        ) from e
    raise

When it happens

Trigger: MCP client sends `data:image/png;base64` (payload omitted), `data:image/png` (no parameters and no payload), or a truncated paste where the comma was cut off.

Common situations: Template string built by concatenation that forgets the ',' separator; client-side truncation of long base64 strings at a fixed buffer size; LLM tool callers emitting malformed data URLs.

Related errors

• Invalid Base64 input: {e}. Ensure the string is complete and
• Unknown provider: {provider}
• Unsupported model: {normalized!r}. Supported models: {suppor
• The input data is inconsistent with expectations.
• Unsupported model: {model!r}

AI-assisted analysis of PaddlePaddle/PaddleOCR@2661c7c0ef (2026-08-14). Data as JSON: /api/errors/0a504668a839645c. Report an issue: GitHub.