AI that’s good at AI: Sphinx Release v0.5.1

We’re happy to announce a set of new features, which shipped this morning, that make Sphinx smarter, faster, and easier to use. These updates are available in the latest version of Sphinx for all users:

1. Sphinx is now equipped with an AI library that unlocks easy use of language and embedding modes as part of your data pipelines. We abstract away fast-changing brittle APIs so Sphinx can neatly process large-scale unstructured data using multiple smaller models in parallel.

   Coming soon: BYO keys/models, and Sphinx will be able to use them automatically in your data process!

2. Sphinx takes checkpoints of notebook state, allowing you to cleanly undo its last steps.
3. We’ve added the ability to create a new notebook from Sphinx’s UI, so you can get into an analysis even faster.
4. Sphinx now has feedback buttons, which aren’t just for fun – when you approve/disapprove of Sphinx’s actions, it will try to immediately improve based on that feedback.
5. Performance optimizations for Sphinx’s runtime check process – Sphinx is now around 2x more efficient at identifying when code is taking too long or not doing what it needs to do, and stepping in to make a fix.
6. Performance improvements to streaming code generation to make Sphinx up to 30% faster when responding to you, especially when having longer conversations.