Change8

arize-phoenix-v11.25.0

📦 arize-phoenix
5 features🐛 8 fixes🔧 2 symbols

Summary

Version 11.25.0 introduces several UI and backend enhancements, including improved span sorting, better experiment filtering, and fixes related to database configuration and serialization.

✨ New Features

  • The `get_spans_dataframe` function now sorts spans by newest/oldest.
  • Added CC support for email in disk usage notifications.
  • Implemented evaluation/annotation quick filters in experiments view.
  • Enabled data hydration in the experiment compare list page.
  • Special case added for `sqlite:///:memory:` in Helm configurations.

🐛 Bug Fixes

  • Added OpenAI timeout configuration to the playground client.
  • Removed redundant `id` field during bulk insertion of annotation objects.
  • Defaulted to SQLite usage when `postgresql.enabled` is set to false.
  • Handled null scores in metrics calculations.
  • Made metric query errors scrollable.
  • Allowed bounded slider fields in the playground to be unset (addressing issue #8940).
  • Restored numpy array handling for `orjson` serialization.
  • Updated the comment for legacy `phoenix.trace.openai` usage.

🔧 Affected Symbols

get_spans_dataframephoenix.trace.openai