BerriAI/litellm · error · ValueError

Invalid data URL format: {data_url[:50]}...

Error message

Invalid data URL format: {data_url[:50]}...

What it means

Raised by the Amazon Nova embedding/multimodal transformation when a supplied media reference is not a data URL: it must start with 'data:' (format data:<mediatype>;base64,<payload>). Only the first 50 characters are echoed in the error.

Source

Thrown at litellm/llms/bedrock/embed/amazon_nova_transformation.py:66

                optional_params["embedding_dimension"] = v
            elif k in self.get_supported_openai_params():
                optional_params[k] = v
        return optional_params

    def _parse_data_url(self, data_url: str) -> tuple:
        """
        Parse a data URL to extract the media type and base64 data.

        Args:
            data_url: Data URL in format: data:image/jpeg;base64,/9j/4AAQ...

        Returns:
            tuple: (media_type, base64_data)
                media_type: e.g., "image/jpeg", "video/mp4", "audio/mpeg"
                base64_data: The base64-encoded data without the prefix
        """
        if not data_url.startswith("data:"):
            raise ValueError(f"Invalid data URL format: {data_url[:50]}...")

        # Split by comma to separate metadata from data
        # Format: data:image/jpeg;base64,<base64_data>
        if "," not in data_url:
            raise ValueError(f"Invalid data URL format (missing comma): {data_url[:50]}...")

        metadata, base64_data = data_url.split(",", 1)

        # Extract media type from metadata
        # Remove 'data:' prefix and ';base64' suffix
        metadata = metadata[5:]  # Remove 'data:'

        if ";" in metadata:
            media_type = metadata.split(";")[0]
        else:
            media_type = metadata

        return media_type, base64_data

View on GitHub (pinned to 6c2dcb801b)

Solutions

1. Convert the media to a data URL: f"data:{mime};base64,{b64data}"
2. For remote files, download the bytes and base64-encode them with the correct MIME type before building the data URL
3. Ensure the prefix uses lowercase 'data:' exactly

Example fix

# before
{"image": "https://example.com/cat.jpg"}

# after
import base64, mimetypes
def to_data_url(path):
    mime = mimetypes.guess_type(path)[0] or "application/octet-stream"
    return f"data:{mime};base64," + base64.b64encode(open(path, 'rb').read()).decode()

def is_data_url(v: str) -> bool:
    return isinstance(v, str) and v.startswith("data:") and "," in v

def to_data_url(raw: str, mime: str = "image/jpeg") -> str:
    if raw.startswith("data:"):
        return raw
    if raw.startswith(("http://", "https://")):
        import base64, urllib.request
        b = urllib.request.urlopen(raw).read()
        return f"data:{mime};base64," + base64.b64encode(b).decode()
    raise ValueError("unsupported media reference")

When it happens

Trigger: Passing an https:// image URL, a local file path, or a raw base64 string without the data: prefix as image/video/audio data to Nova embed calls.

Common situations: Receiving file references from webhooks/frontend uploads that use hosted URLs, or base64 blobs generated upstream whose data-URL prefix was stripped during JSON transport.

Related errors

• Invalid data URL format (missing comma): {data_url[:50]}...
• output_s3_uri is required for async invoke requests
• Unable to determine bedrock embedding provider for model: {m
• Input type '{input_type}' requires async_invoke route. Use m
• Unsupported image format: {image_format}. Supported formats:

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/916419ef39dab534. Report an issue: GitHub.