FG
💻 Software🤖 AI & LLMs

Suport Vector Sum Aggregation

Fresh5 days ago
Mar 14, 20260 views
Confidence Score49%
49%

Problem

First, love this extension and I'm very happy RDS started supporting it. Apologies if this exists and I missed it but it would be great to add support for an aggregate sum function. There are some use cases where it is preferable to sum embedding vectors. Additionally, it would be very convenient to be able to use this extension for general purpose vector operations (as opposed to custom array_agg functions/aggregates).

Unverified for your environment

Select your OS to check compatibility.

1 Fix

Canonical Fix
Unverified Fix
New Fix – Awaiting Verification

Implement Aggregate Sum Function for Vector Operations

Medium Risk

The current implementation of the pgvector extension does not support an aggregate sum function for embedding vectors, which limits its usability for certain applications that require summing vector embeddings. This is likely due to the design focus on individual vector operations rather than aggregate functions.

Awaiting Verification

Be the first to verify this fix

  1. 1

    Define Aggregate Sum Function

    Create a new aggregate function that can sum embedding vectors. This involves defining a state transition function that takes two vectors and returns their sum.

    sql
    CREATE AGGREGATE vector_sum (  
    SFUNC = vector_add,  
    STYPE = vector,  
    INITCOND = '0'  
);
  2. 2

    Implement Vector Addition Function

    Implement the vector_add function that will handle the addition of two vectors. This function should ensure that the vectors are of the same dimension and return their sum.

    sql
    CREATE FUNCTION vector_add(v1 vector, v2 vector) RETURNS vector AS $$  
BEGIN  
    RETURN (SELECT ARRAY(SELECT v1[i] + v2[i] FROM generate_series(1, array_length(v1, 1)) AS i));  
END;  
$$ LANGUAGE plpgsql;
  3. 3

    Test the Aggregate Function

    Run tests to ensure that the new vector_sum function works correctly. Create sample data and verify that the sum of vectors is calculated accurately.

    sql
    SELECT vector_sum(ARRAY[1, 2, 3]::vector), vector_sum(ARRAY[4, 5, 6]::vector);
  4. 4

    Document Usage

    Update the documentation for the pgvector extension to include examples of how to use the new aggregate sum function for embedding vectors.

Validation

To confirm the fix worked, execute the vector_sum function with multiple embedding vectors and verify that the output matches the expected summed vector. Additionally, check the documentation for clarity and completeness.

Sign in to verify this fix

Environment

Submitted by

AC

Alex Chen

2450 rep

Tags

pgvectorembeddingsvector-search