BerriAI/litellm · error · HuggingFaceError

sentence-similarity requires 2+ sentences

Error message

sentence-similarity requires 2+ sentences

What it means

Raised by the HuggingFace embedding handler when the model's pipeline tag is 'sentence-similarity' (or 'similarity') and the input list contains fewer than 2 strings. The sentence-similarity API shape requires one source_sentence plus at least one comparison sentence, so litellm refuses the call with HTTP 400 before sending a request.

Source

Thrown at litellm/llms/huggingface/embedding/handler.py:80

    pipeline_tag: Final[str | None] = model_info_dict.get("pipeline_tag", None)

    return pipeline_tag


class HuggingFaceEmbedding(BaseLLM):
    _client_session: httpx.Client | None = None
    _aclient_session: httpx.AsyncClient | None = None

    def __init__(self) -> None:
        super().__init__()

    def _transform_input_on_pipeline_tag(self, input: list, pipeline_tag: str | None) -> dict:
        if pipeline_tag is None:
            return {"inputs": input}
        if pipeline_tag == "sentence-similarity" or pipeline_tag == "similarity":
            if len(input) < 2:
                raise HuggingFaceError(
                    status_code=400,
                    message="sentence-similarity requires 2+ sentences",
                )
            return {"inputs": {"source_sentence": input[0], "sentences": input[1:]}}
        elif pipeline_tag == "rerank":
            if len(input) < 2:
                raise HuggingFaceError(
                    status_code=400,
                    message="reranker requires 2+ sentences",
                )
            return {"inputs": {"query": input[0], "texts": input[1:]}}
        return {"inputs": input}  # default to feature-extraction pipeline tag

    async def _async_transform_input(
        self,
        model: str,
        task_type: str | None,
        embed_url: str,

View on GitHub (pinned to 6c2dcb801b)

Solutions

1. Pass at least 2 strings in the input list: the first becomes source_sentence, the rest become sentences to compare.
2. If you only want vector embeddings, use a feature-extraction model (e.g. sentence-transformers/all-MiniLM-L6-v2) instead of a similarity model.
3. If you intended cross-encoder scoring of one pair, still pass both members: input=[query, candidate].

Example fix

# before
litellm.embedding(model='huggingface/sentence-transformers/all-MiniLM-L6-v2', input=['hello world'])

# after
litellm.embedding(model='huggingface/sentence-transformers/all-MiniLM-L6-v2', input=['hello world', 'hi there'])

def validate_similarity_input(input_list: list[str]) -> bool:
    # sentence-similarity path needs a source sentence + >=1 comparison
    return isinstance(input_list, list) and len(input_list) >= 2 and all(isinstance(x, str) and x.strip() for x in input_list)

try:
    resp = litellm.embedding(model=model, input=texts)
except litellm.llms.huggingface.common_utils.HuggingFaceError as e:
    if 'requires 2+ sentences' in str(e):
        raise ValueError(f'Need >=2 texts for similarity model {model}') from e
    raise

When it happens

Trigger: Calling litellm.embedding(model='huggingface/BAAI/bge-...', input=['only one sentence']) or input=[] on a model whose HuggingFace pipeline tag is sentence-similarity/similarity, or explicitly passing task='sentence-similarity' with a single-element input list.

Common situations: Developer reuses an embedding call written for feature-extraction models (which accept a single string) against a cross-encoder/similarity model; or splits a document and the chunking step yields one chunk.

Related errors

• Invalid task_type={task_type}. Expected one of={hf_tasks_emb
• reranker requires 2+ sentences
• sentence transformers requires 2+ sentences
• model is required
• {embeddings[error]}

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/fd4fc563c44787ae. Report an issue: GitHub.