fix: one-liner curl commands for Windows CMD

fix: module-level session store for cross-instance state

fix: stateless demo using /grader endpoint

fix: single worker to preserve session state

interactive demo landing page with curl commands

fix grader: other type penalty, gradient_not_zeroed loss check

fix landing page route to /ui

add landing page, blog link, HF Space UI

Block A B C: partial observability, LLM judge, adversarial scheduler

v3: compound tasks, hardened graders, other type, 8 tasks total

add 6 tasks, fix log format, multi-turn retry, grader improvements

fix: scores strictly between 0 and 1 exclusive

fix: use injected API_KEY and API_BASE_URL

fix: 20/20 all tasks 1.0

fix: use API_BASE_URL and API_KEY env vars for LLM proxy

ML Debug Environment - OpenEnv Hackathon submission