Skip to main content

Vector Search Architecture

Vector Search is designed for fast, secure semantic search over event data while keeping all data inside your own GCP project.

System Overview​

Vector search architecture overview

Pipeline Summary​

  • Event ingestion: Camera events flow into cloud storage and analytics systems.
  • Embedding generation: Text and image representations are created for search.
  • Vector search: BigQuery performs similarity search on embeddings.
  • Results delivery: The UI returns ranked matches with relevant context.

This architecture emphasizes low latency, predictable cost, and clear security boundaries.

Internal Architecture Details

Detailed implementation, model configuration, and query examples are available to authorized team members and NDA customers.

Internal Vector Search Architecture