sgl-project/sglang · error · ValueError

data URI header is too large

Error message

data URI header is too large

What it means

For data: URIs, the header (everything before the ',') must not exceed MINIMAX_H3_BASE64_HEADER_MAX_CHARS. This bounds-check prevents absurdly long headers from being parsed and acts as a lightweight abuse guard against malformed or hostile URIs.

Source

Thrown at python/sglang/multimodal_gen/runtime/pipelines_core/stages/model_specific_stages/minimax_h3/material_io.py:99

            paths = registry.pop(owner, [])
            if isinstance(paths, (list, tuple)):
                for path in paths:
                    shutil.rmtree(str(path), ignore_errors=True)
        if not registry:
            batch.extra.pop(MINIMAX_H3_TEMP_DIRS_EXTRA_KEY, None)
    if owners is None or "material" in selected:
        batch.extra.pop(MINIMAX_H3_MATERIAL_CACHE_EXTRA_KEY, None)
        batch.extra.pop(MINIMAX_H3_MATERIAL_PROBE_EXTRA_KEY, None)


def _base64_uri_payload_start(uri: str) -> tuple[int, str | None]:
    media_type = None
    if uri.startswith("data:"):
        separator = uri.find(",")
        if separator < 0:
            raise ValueError("data URI must contain a comma separator")
        if separator > MINIMAX_H3_BASE64_HEADER_MAX_CHARS:
            raise ValueError("data URI header is too large")
        header = uri[:separator]
        if ";base64" not in header:
            raise ValueError("data URI must use ;base64 encoding")
        media_type = header[5:].split(";", 1)[0].lower() or None
        payload_start = separator + 1
    elif uri.startswith("base64://"):
        payload_start = len("base64://")
        separator = uri.find(",", payload_start)
        if separator >= 0:
            if separator - payload_start > MINIMAX_H3_BASE64_HEADER_MAX_CHARS:
                raise ValueError("base64 URI header is too large")
            header = uri[payload_start:separator]
            media_type = header.split(";", 1)[0].lower() or None
            payload_start = separator + 1
    else:  # pragma: no cover - guarded by the caller
        raise ValueError("not a base64 material URI")
    return payload_start, media_type

View on GitHub (pinned to 0132848349)

Solutions

1. Shorten the data URI header to a plain 'data:image/png;base64' style
2. Check the comma position: uri.find(',') must be present and small; if missing, fix URI generation
3. Sanitize/validate material URIs at request ingestion with the same header cap

Example fix

// before
uri = "data:image/png;base64" + payload  # missing comma -> whole string is header
// after
uri = "data:image/png;base64," + payload

sep = uri.find(',')
if uri.startswith('data:') and (sep < 0 or sep > MINIMAX_H3_BASE64_HEADER_MAX_CHARS):
    raise ValueError('data URI header missing or too large')

def data_uri_header_ok(uri: str, cap: int) -> bool:
    sep = uri.find(',')
    return 0 < sep <= cap

When it happens

Trigger: A data: URI whose media-type/parameter header exceeds the configured character cap, e.g. a header stuffed with junk parameters or a URI where the comma appears very late because the payload was concatenated onto the header.

Common situations: Malformed URIs missing the comma so the whole payload counts as header, prompt-injected material URIs with huge headers, or a comma lost during JSON transport.

Related errors

• data URI must contain a comma separator
• data URI must use ;base64 encoding
• base64 URI header is too large
• tar material URI encoded header is too large
• material URI has an invalid percent escape

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/b54da1635781ffd9. Report an issue: GitHub.