Skip to main content
VectorAI DB provides high-performance gRPC endpoints for applications that require low-latency operations and efficient binary communication. gRPC uses Protocol Buffers for serialization and HTTP/2 for transport, making it ideal for production workloads with demanding performance requirements. From this page you can learn about the gRPC connection details, understand when to choose gRPC over REST, and run a sample workflow using the Python SDK.

​
Connection details

  • Default port: 50051.
  • Protocol: gRPC over HTTP/2.
  • Serialization: Protocol Buffers (protobuf).

​
Benefits of gRPC

gRPC offers several advantages over REST for vector database operations.
  • High performance: binary serialization is faster and more compact than JSON.
  • Low latency: HTTP/2 multiplexing reduces connection overhead.
  • Type safety: strong typing with Protocol Buffers.
  • Bidirectional streaming: support for streaming requests and responses.
  • Code generation: automatic client generation from .proto files.

​
When to use gRPC

Choose gRPC endpoints when you need:
  • Minimal latency for vector operations.
  • High-throughput batch processing.
  • Microservices architecture with service-to-service communication.
  • Streaming data operations.
  • Strong typing and contract-first API design.
For standard web applications and quick prototyping, the REST API may be simpler to integrate.

​
Client libraries

The Python SDK supports gRPC endpoints for communication with VectorAI DB. For more information about using the Python SDK with gRPC endpoints, see Python SDK. The following example connects to a VectorAI DB gRPC endpoint, creates a collection with 128-dimensional cosine vectors, inserts a point, and runs a similarity search. After running this code you should see the search results printed to the console, confirming that your gRPC connection is working. 
from actian_vectorai import (
    Distance,
    PointStruct,
    VectorAIClient,
    VectorParams,
)

# Connect to gRPC endpoint (default port 50051)
with VectorAIClient("localhost:50051") as client:
    # Health check
    info = client.health_check()
    print(f"Connected to {info['title']} v{info['version']}")

    # Create a collection
    client.collections.create(
        "my_collection",
        vectors_config=VectorParams(size=128, distance=Distance.Cosine),
    )

    # Insert a point
    client.points.upsert(
        "my_collection",
        points=[
            PointStruct(
                id=1,
                vector=[0.1] * 128,
                payload={"category": "example"}
            )
        ]
    )

    # Perform search operations
    results = client.points.search(
        "my_collection",
        vector=[0.1] * 128,
        limit=10
    )
    print(f"Found {len(results)} results")