Skip to Content
ConceptsVectors

Vectors

Vectors are the fundamental data units stored in an index. Each vector has a unique ID, an embedding array, and optional metadata and filter fields.

Dense Vector Fields

FieldRequiredDescription
IDYesUnique string identifier for the vector
VectorYesDense embedding array (must match the index dimension)
MetaNoArbitrary metadata object — returned in query results
FilterNoKey-value pairs used to filter results during queries
index = client.get_index(name="my_index")
 
index.upsert([
    {
        "id": "doc1",
        "vector": [0.12, -0.34, 0.89, ...],
        "meta": {"title": "Introduction to ML"},
        "filter": {"category": "tech", "year": 2024}
    },
    {
        "id": "doc2",
        "vector": [0.55, 0.21, -0.10, ...],
        "meta": {"title": "Deep Learning Basics"},
        "filter": {"category": "tech", "year": 2023}
    }
])

Hybrid Vector Fields

For hybrid indexes, include sparse vector components alongside the dense vector:

FieldRequiredDescription
Sparse IndicesYesNon-zero term positions in the sparse vector
Sparse ValuesYesBM25 weights corresponding to each index position
index.upsert([
    {
        "id": "doc1",
        "vector": [0.12, -0.34, 0.89, ...],
        "sparse_indices": [4821, 19043, 73201],
        "sparse_values":  [1.42,  0.87,  1.15],
        "meta": {"title": "Introduction to ML"}
    }
])

sparse_indices and sparse_values must have the same length. Each position in sparse_indices maps to the weight at the same position in sparse_values.

Maximum batch size is 1,000 vectors per upsert call.

Precision

The precision parameter controls how vectors are stored internally. Lower precision reduces memory and speeds up search at the cost of some accuracy.

PrecisionBitsStorageSpeedAccuracy
BINARY1-bitSmallestFastestLower
INT88-bitSmallFastGood
INT1616-bit integerMediumMediumHigher
FLOAT1616-bit floatMediumMediumHigh
FLOAT3232-bit floatLargestSlowerHighest
from endee import Precision
 
# INT8 — recommended for most use cases
client.create_index(name="my_index", dimension=384,
                    space_type="cosine", precision=Precision.INT8)
 
# FLOAT32 — maximum accuracy
client.create_index(name="precise_index", dimension=384,
                    space_type="cosine", precision=Precision.FLOAT32)
 
# BINARY — minimum memory
client.create_index(name="large_index", dimension=384,
                    space_type="cosine", precision=Precision.BINARY)

Recommendations:

  • INT8 — best balance of speed, memory, and accuracy for most use cases (default)
  • FLOAT32 — use when maximum recall accuracy is critical and memory is not a concern
  • BINARY — use for very large indexes where memory is the primary constraint

Precision is set at index creation time and cannot be changed without recreating the index.

IndexesSearch