The NeumVector is the object used to organize data extracted for a given data source. It is analogous to similar constructs (i.e. Document) used in frameworks like Langchain and LlamaIndex. The goal of this interface is to abstract three properties:
  • id (str): This is a unique identifier for a given vector. The id is constructed throughout the pre-processing of the data and used as the vector id within the vector database. It is used at synchronization to ensure vectors are not being re-computed and duplicated.
  • vector (List[float]): Vector embedding generated from the content in the NeumDocument.
  • metadata (dict): This value contains the attached metadata for a given document. This can include values extracted from the data source or loader, as well as any calculated values. Metadata includes the content from the NeumDocument.

​
Usage

NeumVector
from neumai.Shared.NeumVector import NeumVector
neum_vector = NeumVector(
    id = 'abc', 
    vector = [.....],
    metadata = {'text':'Hello', 'createdDate':'2023-01-01'}
)