Vespa Roadmap

This page lists large upcoming features in Vespa. To make it easy to keep up to date, we're not listing smaller items, nor work that we consider improvements to existing features. These are often tracked by GitHub issues.

Launching soon

  • Tensor support in Streaming Search.
  • Support multilingual embedding models using sentencepiece tokenization.

In progress

  • New data logging feature for simplified Amazon Kinesis export.
  • New sample applications for hybrid/vector search.
  • Maximum inner product search (MIPS) similarity function with norms different from 1.
  • Native Vespa embedders using GPU acceleration when available.

Priority backlog

  • Online learning: Global tensors mutable at a high continuous rate.
  • Global multi-region feeding: Feed once, globally, with automatic resync.