arize-phoenix-evals-v2.12.0
📦 arize-phoenixView on GitHub →
✨ 1 features🐛 3 fixes⚡ 1 deprecations🔧 1 symbols
Summary
Version 2.12.0 introduces support for structured data inputs in evals and patches several bugs, including a security-related pin on the LiteLLM dependency.
✨ New Features
- Evals now accept structured data as input for evaluations.
🐛 Bug Fixes
- Built-in classification evaluators now correctly allow setting 'prompt_template' and 'kwargs'.
- LiteLLM dependency pinned to versions less than 1.82.7 to mitigate a supply chain attack.
- Prevented OpenTelemetry (OTel) context leak in the playground that was causing duplicate 'trace_id' in dataset runs.
Affected Symbols
⚡ Deprecations
- The documentation for evals 1.0 is deprecated.