Enhancing Apache Cassandra Drivers with Vector Support
summary
Three drivers (Java, Python, Node.js) now include native vector support for Astra Vector and DSE 7.0 preview with planned compatibility for Apache Cassandra 5.0. Vectors are implemented as a parameterized custom type over the binary protocol; current work supports fixed-size element types (floats) while variable-size element serialization remains unresolved. Java exposes vectors via Java: SQLVector class instead of raw arrays, Python returns a Python: seekable native type, and Node maps to Node.js: TypedArray mapping.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 3:12Purpose of vector support
- 4:37What is a vector
- 5:40Vector support origins
- 7:21Vectors as custom types
- 9:08Row-level messaging
- 12:19Parameterized custom types
- 13:24Schema type specs
- 15:31Reading/writing vectors
- 16:34Driver serialization differences
- 18:38Language design choices
- 21:27Java driver mapping
- 22:52Python/Node.js vectors
- 24:41Using existing APIs
- 25:46Cassandra 5 vectors
- 26:49Challenges and future work
- 27:15Q&A
related talks
11mCassandra Vector Search Performance Tuning
18mKeynote: Data = AI: Leading the Future of Gen AI with Apache Cassandra
24mStaying Ahead of the Curve with Spring and Cassandra 4
53mVectors: how to better support a nasty data type in PostgreSQL
19mAccelerate AppDev with Database 23ai
57mEnter The Brave New World of GenAI with Vector Search
35mImpact of Vector Search
16mMariaDB Vector: Why your AI data should be in an RDBMS