Skip to main content
Apply filters to payload fields during search to combine semantic similarity with business rules. VectorAI DB returns only results whose payloads match the filter conditions. If no points match your filters, you receive an empty response.
Before you begin, make sure you have an existing collection with payload-bearing points and a query vector that matches the collection’s dimension.

​
Search with payload filters

The following example finds semantically similar items that match specific criteria. 
import random
from actian_vectorai import VectorAIClient, FilterBuilder, Field

DIMENSION = 128
COLLECTION = "products"

# Connect to VectorAI DB server
with VectorAIClient("localhost:50051") as client:
    # Search for electronics under $600 with similarity
    query_vector = [random.gauss(0, 1) for _ in range(DIMENSION)]
    
    # Build filter for electronics under $600
    filter = FilterBuilder()\
        .must(Field("category").eq("electronics"))\
        .must(Field("price").lt(600.0))\
        .build()
    
    # Search with filter
    results = client.points.search(
        COLLECTION,  # Collection name
        vector=query_vector,  # Query vector
        limit=5,  # Maximum results
        filter=filter  # Apply filter
    )
    
    # Display results
    print("Search results (electronics under $600):")
    for result in results:
        print(f"  - {result.payload} (score: {result.score:.4f})")
For more complex filter operations, see the filtering documentation which covers must, should, and must-not filter types.
Each result includes these fields:
  • id: The unique identifier of the matching point.
  • score: Similarity score for points that passed the filter.
  • payload: Full metadata dictionary showing filtered product attributes.