Sphinx 0.7 — Keeping on the Rails

Sphinx 0.7 just shipped! Our focus for this release was ensuring that Sphinx can be more strictly controlled in response to your preferences and guidelines, and that when Sphinx does make mistakes it can recover significantly faster. Here’s what’s new:

1. Guardrails: Certain behaviors can now be caught by Sphinx before it runs any code. When a guardrail is activated in the Sphinx Dashboard, the agent will either avoid a behavior altogether or ask for manual confirmation before proceeding.

   Coming soon for our enterprise customers: adding and enforce guardrails across your entire team/organization!

2. Improved Editing and Deletion: Sphinx is now much better at making small, targeted edits to its code. This lets it recover from mistakes faster, and shows you a clearer diff when you review Sphinx’s work. Likewise, Sphinx is now better at getting rid of cells that it no longer needs for the analysis, keeping your notebook clean and error-free.
3. Revamped IDE UI: We redesigned our user interface based on your feedback — Sphinx’s features are now more discoverable than ever, and easier to access as you work.
4. Revamped CLI UI: The Sphinx CLI now has significantly improved logging to let you track and monitor the autonomous data scientist as it works through a task.
5. Streamlit Environment Building: Streamlit Mode now tries to capture and reproduce your notebook environment alongside the dashboards it builds for you. This makes it easier than ever to build rich interactive visualizations, and makes Sphinx’s dashboards portable to run anywhere you need them to.