agent + UI: token-lean planner loop, gated tool selector, hero/table/persist UI fixes

gradio: Figma UI redesign, long-run stability, clarify pause, per-run project lifecycle

predict: align inference frame to training schema (fix Lexsi #05-018)

Add P1/P2 agent tools: data-quality, sampling, join-paths, feature/target engineering

Add table finding tools

Add new tools and dab runs

Add dab benchmark integration

multi-turn sessions, lexsi-sdk 0.1.51, family-aware pods, dark code css

v1 marquee tools, clarify pause/resume, gradio rewrite, bench docs

End to end run for pred

Fixes for tabular tools

fixed conftest

added summary and sdk cross-check fixes

added tools and tests

Add sample benchmark queries and tests