stanford-oval/storm · error · ValueError

Please provide an embedding model.

Error message

Please provide an embedding model.

What it means

VectorRM's __init__ requires an embedding model because it builds a langchain HuggingFaceEmbeddings instance (self.model) used to embed queries against the Qdrant vector store. Without an embedding model, vectors cannot be compared to stored embeddings. The check raises ValueError immediately at construction time.

Source

Thrown at knowledge_storm/rm.py:214

        k: int = 3,
    ):
        from langchain_huggingface import HuggingFaceEmbeddings

        """
        Params:
            collection_name: Name of the Qdrant collection.
            embedding_model: Name of the Hugging Face embedding model.
            device: Device to run the embeddings model on, can be "mps", "cuda", "cpu".
            k: Number of top chunks to retrieve.
        """
        super().__init__(k=k)
        self.usage = 0
        # check if the collection is provided
        if not collection_name:
            raise ValueError("Please provide a collection name.")
        # check if the embedding model is provided
        if not embedding_model:
            raise ValueError("Please provide an embedding model.")

        model_kwargs = {"device": device}
        encode_kwargs = {"normalize_embeddings": True}
        self.model = HuggingFaceEmbeddings(
            model_name=embedding_model,
            model_kwargs=model_kwargs,
            encode_kwargs=encode_kwargs,
        )

        self.collection_name = collection_name
        self.client = None
        self.qdrant = None

    def _check_collection(self):
        from langchain_qdrant import Qdrant

        """
        Check if the Qdrant collection exists and create it if it does not.

View on GitHub (pinned to fb951af774)

Solutions

1. Pass embedding_model explicitly, e.g. embedding_model='sentence-transformers/all-MiniLM-L6-v2'
2. Ensure the same embedding model was used to build the stored collection, otherwise retrieval quality degrades
3. Pre-download the model (huggingface) if running in an offline/sandboxed environment

Example fix

// before
rm = VectorRM(collection_name='my_docs', source='offline', vector_store_path='./qdrant')
// after
rm = VectorRM(collection_name='my_docs', source='offline', vector_store_path='./qdrant', embedding_model='sentence-transformers/all-MiniLM-L6-v2')

from knowledge_storm.rm import VectorRM
EMBED_MODEL = 'sentence-transformers/all-MiniLM-L6-v2'
if not EMBED_MODEL:
    raise SystemExit('embedding_model is required')
rm = VectorRM(collection_name='docs', source='offline', vector_store_path='./qdrant', embedding_model=EMBED_MODEL)

def valid_vectorrm_config(collection_name: str, embedding_model: str | None) -> bool:
    return bool(collection_name) and bool(embedding_model)

try:
    rm = VectorRM(collection_name='docs', source='offline', vector_store_path='./qdrant', embedding_model=MODEL)
except ValueError as e:
    if 'embedding model' in str(e):
        raise SystemExit(f'Config error: {e}')
    raise

When it happens

Trigger: Constructing VectorRM(...) without passing embedding_model (e.g. VectorRM(collection_name='docs', source='offline', vector_store_path='./qdrant')). The parameter defaults to None, so any call omitting it fails.

Common situations: Copying a minimal example that only sets collection_name and vector_store_path; upgrading knowledge-storm where older examples omitted embedding_model; assuming the model is stored in the vector store and not needed at query time.

Understand the failure class

Background: Missing required parameter errors: what 'X is required' and 'the required X param is missing' mean, and how to fix them — this error's family across 27 libraries.

Related errors

• Qdrant client is not initialized.
• Collection {self.collection_name} does not exist. Please cre
• Please provide a url for the Qdrant server.
• Please provide an api key.
• Error occurs when connecting to the server: {e}

AI-assisted analysis of stanford-oval/storm@fb951af774 (2026-08-28). Data as JSON: /api/errors/b76082b85c159380. Report an issue: GitHub.