Combine filter types

Build complex queries with must, should, and must-not conditions.

Combine must, should, and must-not filter types to create complex queries. This approach handles scenarios requiring strict requirements, flexible alternatives, and explicit exclusions.

This page uses a products collection with 128-dimensional vectors and payload fields: category (string), price (float), rating (float), and condition (string).

Python

from actian_vectorai import VectorAIClient, FilterBuilder, Field
import random

# Connect to VectorAI DB server
with VectorAIClient("localhost:6574") as client:
    # Generate query vector
    query_vector = [random.gauss(0, 1) for _ in range(128)]
    
    # Must be electronics
    # Should be either cheap OR highly rated
    # Must NOT be refurbished
    filter = FilterBuilder()\
        .must(Field("category").eq("electronics"))\
        .should(Field("price").lt(50.0))\
        .should(Field("rating").gt(4.0))\
        .must_not(Field("condition").eq("refurbished"))\
        .build()
    
    # Search with combined filters
    results = client.points.search(
        "products",  # Collection name
        vector=query_vector,  # Query vector
        limit=10,  # Maximum results
        filter=filter  # Apply filter
    )
    
    # Display results
    for result in results:
        print(f"Category: {result.payload['category']}")
        print(f"Price: ${result.payload['price']}")
        print(f"Rating: {result.payload.get('rating', 'N/A')}")
        print("-" * 50)

import { VectorAIClient, FilterBuilder, Field } from '@actian/vectorai-client';

async function main() {
    // Connect to VectorAI DB server
    const client = new VectorAIClient('localhost:6574');

    // Generate query vector
    const queryVector = Array.from({ length: 128 }, () => Math.random() * 2 - 1);

    // Must be electronics
    // Should be either cheap OR highly rated
    // Must NOT be refurbished
    const filter = new FilterBuilder()
        .must(new Field('category').eq('electronics'))
        .should(new Field('price').lt(50.0))
        .should(new Field('rating').gt(4.0))
        .mustNot(new Field('condition').eq('refurbished'))
        .build();

    // Search with combined filters
    const results = await client.points.search('products', queryVector, {
        filter: filter,
        limit: 10
    });

    // Display results
    for (const result of results) {
        console.log(`Category: ${result.payload.category}`);
        console.log(`Price: $${result.payload.price}`);
        console.log(`Rating: ${result.payload.rating ?? 'N/A'}`);
        console.log('-'.repeat(50));
    }
}

main().catch(console.error);

Each result includes these fields:

• id: The unique identifier of the matching point.
• score: Similarity score based on vector distance.
• payload: Full metadata dictionary for the matching point.