websmaLLMs session explorer
$ python3 smaLLMs.py run --all-local --samples 25
$ python3 smaLLMs.py export

Every benchmark run, all the way down to prompt and response level.

The site now reads the exported session bundle directly. Keep using the current terminal aesthetic, but with a real data contract behind it: leaderboard, benchmark coverage, evaluation metrics, raw sample traces, system metadata, and artifact provenance in one scrollable surface.