b'{"status":{"error":"Wrong input: Conversion between sparse and regular vectors failed"},"time":0.010832087}'

[HarshSonaiya]

def create_sparse_vector(docs: Document, sparse_text_embedding_model) -> SparseVector:

embeddings = list(sparse_text_embedding_model.embed([docs.page_content]))[0]
logging.debug(f"Sparse embeddings: {embeddings}")

if hasattr(embeddings, 'indices') and hasattr(embeddings, 'values'):
    return SparseVector(
        indices=embeddings.indices.tolist(),
        values=embeddings.values.tolist()
    )
else:
    raise ValueError("The embeddings object does not have 'indices' and 'values' attributes.")

def create_sparse_collection(client: QdrantClient):
"""
Create a collection for sparse vectors in Qdrant if it does not exist.

Args:
    client (QdrantClient): An instance of QdrantClient.
"""
if not client.collection_exists(collection_name=sparse_collection):
    client.create_collection(
        collection_name=sparse_collection,
        vectors_config= {},
        sparse_vectors_config={
            "sparse": models.SparseVectorParams(),
        }
    )

logging.info(f"Created sparse vector collection '{sparse_collection}' in Qdrant.")

def index_documents(file: str):

# Initialize client and embedding models
client = create_qdrant_client()

# Extract and split PDF content
documents = extract_content_from_pdf(file)

# Create Qdrant collection
create_sparse_collection(client)

# Index sparse embeddings into the sparse collection
sparse_points = []
for i, doc in enumerate(tqdm(documents, total=len(documents))):
    sparse_vector = create_sparse_vector(doc, sparse_embedding_model)
    sparse_point = models.PointStruct(
        id=i,
        vector={"sparse":sparse_vector}
    )
    # logging.info(f"Sparse points:{sparse_vector}")
    sparse_points.append(sparse_point)
logging.info(type(sparse_points[0]),len(sparse_points))
client.upsert(
    collection_name=sparse_collection,
    points=sparse_points
)
logging.info(f"Upserted {len(sparse_points)} points with sparse vectors into Qdrant.")

[Anush008]

From the error, it looks like the collection was created with dense vectors earlier.
You can try deleting the collection and re-running

def create_sparse_collection(client: QdrantClient):

so it's created with the appropriate config.