Introduction
Vector fields and vector field definitions allow for storing vectors of floating point numbers. Such vectors are for example generated by an AI Connector and used for Similarity Search (POC). An example of a vector with 4 dimensions is [0.5, -0.5, -0.5, -0.5].
Floating point
The vector is stored in the metadata as a hexadecimal string using the following conversion:
Each floating point number is more precisely a 32-bit precision IEEE 754 floating point.
Each floating point is converted to an 8-character hexadecimal string, e.g.
0.5→3F000000All 8-character hexadecimal strings are concatenated into a single string
A vector with dimensions
Nwill be stored as a hexadecimal string of length8 * N
Example
The vector [0.5, -0.5, -0.5, -0.5] in hexadecimal form is 3F000000BF000000BF0000003F000000