arize-phoenix-v17.3.0
📦 arize-phoenixView on GitHub →
✨ 18 features🐛 5 fixes🔧 3 symbols
Summary
This release introduces significant enhancements to the playground and agent capabilities, including new PXI tools for experiment recording, dataset management, and support for the claude-fable-5 model. Several bug fixes address UI alignment and server claim extraction issues.
✨ New Features
- Add copy trace ID chat action.
- Add playground repetitions tool.
- Add playground run cancellation tool.
- Agents: Evals for trace-debug skill.
- Agents: Subagents support.
- Agents: User invokable skill menu.
- App: Inline, editable time range selector.
- Playground: Add set_appended_messages_path PXI tool.
- Playground: Add support for claude-fable-5.
- Playground: Set template variables path tool.
- Preserve place and auto truncate for large tools.
- PXI: Add playground experiment recording tool.
- PXI: Add playground prompt instance tools.
- PXI: Agent-driven validated evaluator submit.
- PXI: Dataset management tools for the agent.
- PXI: Gate write_prompt_tools behind accept/reject diff.
- PXI: Surface dataset-run experiment context.
- Update PXI base instructions.
🐛 Bug Fixes
- Agents: Pin forced-floating PXI panel to the FAB corner over drawers.
- Agents: Show streaming PXI bash summary.
- Align assistant chat scroll before first paint.
- PXI: Treat /dev/null as a discard sink in bash tool filesystem policy.
- Server: Tolerate JMESPath type errors in OAuth2 claim extraction.