opendatalab/MinerU · error · ValueError

Unsupported middle json backend for client-side output gener

Error message

Unsupported middle json backend for client-side output generation: {backend}

What it means

ValueError from _select_union_make(): client-side output generation dispatches markdown/content-list rendering by the middle-json's backend tag, and the backend must be one of the supported families (pipeline -> pipeline_middle_json_mkcontent, vlm/hybrid -> vlm module, office -> office module). Anything else cannot be rendered locally.

Source

Thrown at mineru/cli/client_side_output.py:31

SUPPORTED_BACKENDS = {*PDF_BACKENDS, "office"}


def _select_union_make(backend: str) -> Callable[[list, str, str], Any]:
    """根据 middle json 后端选择对应的 Markdown/content list 渲染函数。"""
    if backend == "pipeline":
        from mineru.backend.pipeline.pipeline_middle_json_mkcontent import union_make

        return union_make
    if backend in {"vlm", "hybrid"}:
        from mineru.backend.vlm.vlm_middle_json_mkcontent import union_make

        return union_make
    if backend == "office":
        from mineru.backend.office.office_middle_json_mkcontent import union_make

        return union_make

    raise ValueError(
        f"Unsupported middle json backend for client-side output generation: {backend}"
    )


def _write_json(path: Path, payload: Any) -> None:
    """按项目现有格式写入 JSON 文件，保持中文内容可读。"""
    path.write_text(
        json.dumps(payload, ensure_ascii=False, indent=4),
        encoding="utf-8",
    )


def regenerate_client_side_outputs(
    parse_dir: str | Path,
    doc_stem: str,
) -> tuple[Path, ...]:
    """读取服务端 staged/finalized middle json，并在客户端覆盖生成最终输出产物。"""
    parse_dir = Path(parse_dir)

View on GitHub (pinned to 4fe4bde114)

Solutions

1. Regenerate the middle json with the current mineru version so _backend is one of pipeline/vlm/hybrid/office
2. If editing middle.json manually, keep _backend intact and valid
3. Re-download the artifacts from the server instead of reusing stale local staged files
4. Pin matching versions between the server that produced the json and the client that renders it

Example fix

# before: stale middle json with "_backend": "vlm-auto-engine"

# after: regenerate via the server, yielding "_backend": "vlm"

SUPPORTED = {"pipeline", "vlm", "hybrid", "office"}
backend = json.loads(middle_json_path.read_text(encoding="utf-8")).get("_backend")
if backend not in SUPPORTED:
    raise SystemExit(f"Cannot render middle json: backend {backend!r} not supported; regenerate artifacts with this mineru version")

def is_renderable_middle_json(payload: object) -> bool:
    return (
        isinstance(payload, dict)
        and payload.get("_backend") in {"pipeline", "vlm", "hybrid", "office"}
        and isinstance(payload.get("pdf_info"), list)
    )

try:
    regenerate_client_side_outputs(parse_dir, doc_stem)
except ValueError as e:
    if "Unsupported middle json backend" in str(e):
        # artifacts are stale — refetch from server instead of rendering locally
        refetch_results()
    else:
        raise

When it happens

Trigger: Calling regenerate_client_side_outputs() (indirectly via client-side output generation after a parse) where the staged {doc}_middle.json has _backend set to an unexpected value such as an old backend name or a hand-edited/custom tag.

Common situations: Middle-json files produced by an older mineru version whose _backend tags differ; users editing or post-processing middle.json and changing/removing _backend; mixing outputs from forks that add backend families.

Related errors

• Missing middle json file: {middle_json_path}
• middle_json must be a dict.
• middle_json must contain a list field named pdf_info.
• backend={backend} requires server_url
• Please install transformers to use the transformers backend.

AI-assisted analysis of opendatalab/MinerU@4fe4bde114 (2026-08-14). Data as JSON: /api/errors/b57556b0642f5e98. Report an issue: GitHub.