Highlights
Switched conversational prompt context to interleaved notebook format with paired code and outputs
Improved AutoLab run-state and stop-experiment UX with clearer controls and safer actions
Fixed sidebar chat regressions including Clear chat and Fix with AI response handling
Polished thinking UI, scrolling behavior, and large-variable column summarization
Improved
- Stabilized and polished three major areas: conversational notebook UX, AutoLab experiments UX, and version/release flow.
- Changed conversational prompt context to an interleaved
[notebook]format with code and outputs paired per cell. - Improved output capture so manual notebook executions are reflected more reliably in prompt context.
- Auto-run follow-up prompts now keep full content when sent, while older history entries are compacted for readability.
- Updated variable column summarization to show the first 15 and last 15 columns when total columns exceed 30.
- Tuned scroll behavior in conversation views to reduce jumpiness and top-jumps during rerenders.
- Refined thinking UI in sidebar and conversational views with calmer styling, better spacing, consistent pending states, and typography fixes.
- Introduced safer dependency-install policy UX in AutoLab with simpler user-facing options and clearer form labels/styles.
- Improved AutoLab run-state UI, including run monitor layout/height behavior and reduced nested-scroll issues.
- Redesigned stop-experiment controls in AutoLab: top placement, running-only visibility, and a confirmation dialog before stopping.
- Added and adjusted experiment state logging and improved sidebar state/polling behavior while switching contexts.
Fixed
- Fixed a sidebar Clear chat regression where prompts could no longer be sent after clearing.
- Fixed Fix with AI flow so backend replies are correctly surfaced in sidebar chat.