remove debugging

format prompts and add debug log

make hub download once per app load

support loading leaderboard/votes from persistent source

standardize new scheduled locations

shorten scheduler commit window for testing

refresh and persistent logging

update data pipeline

remove cogito for gpu-hour reasons and add approved llama models

add additional models

fix typo

gemma 3 4b misaligned bug, so removing that as an option for now

update model list

revert to sequential processing

typo fix

prepare for zeroGPU

specified device

vllm backend swap v1

removed backup vote_details (#19)

proper threaded generation interrupt

Handled interruption (#10)

Added shared interruption flag (#12)

ready hyperlinks for leaderboard

Rewrote the processor to make it more robust to edge cases (#5)

Implemented elo (#6)

Component to store vote details to a csv (#8)

sanitized arena_df

first pass at async handling

switched models to using full context

update data loading (#1)

fix models

remove gated models for now

v1 inference code

add UI and data processing

initial app commit