Redis Vector Lookup
Redis Vector Lookup processor allows you to find the similar items from vector data set for the given query vector. It uses vector embedding data. Distance between two vectors represents similarity between two vectors. For calculating distance following three methods are supported: Euclidean Distance, Cosine Distance and Dot Product.
Redis Vector Lookup Configuration
Select the processor and join it with the pipeline on the canvas. Click on it to configure. Below are the available fields:
Connection Name
Select connection name out of the list of saved connections. Create one from Connections page as explained in the topic - Redis Connection.
Index
Select the index of Redis to read the vector data from.
Index Info
Details of the selected index is shown.
Index Vector Field
Select the index vector field with which the lookup has to be performed with the input vector data.
Input Column
Select the input vector column with which the lookup has to be performed with the index vector data.
Top Matches
Specify the number of Top Matches to retrieve.
Example: Enter ‘5’ to get the top five matches.
Output Configuration
Id
Enter a column name where the ID attribute obtained from the matched results should be captured.
Score
Enter a column name where the Score attribute obtained from the matched results should be captured.
Values
Enter a column name where the Value attribute obtained from the matched results should be captured.
ADD OBJECT
Select the option to add more object(s). You will be required to provide attribute, output column and datatype details.
Attribute
Select the name of the index attribute to be a part of the output result.
Output Column
Select or provide the output column in which the result will be stored of the corresponding attribute.
Datatype
Select the datatype of the output column in which the output results will be stored.
Add Configuration: Click the option to add more key-value based configuration(s) to the component.
Environment Params: Click the + ADD PARAM option to add more environment configuration(s).
If you have any feedback on Gathr documentation, please email us!