fix: query budget 500, sync all hardcoded refs across codebase

fix: UI polish — 9 scenario count, styled empty state, richer Try It tab with How It Works + curl examples

feat: playground all-fields-visible grid, loop detection, 200 query budget

feat: playground overhaul — tool-specific fields, scenario picker, live status dashboard

fix: brutal audit — reset tool_call_counts, date dedup, unused vars, playground overhaul with scenario picker + status dashboard

fix: render placeholders in reset observation message

fix: python 3.11 f-string compat, inference OPENENV_BASE_URL fix, README action/observation spaces

feat: investigation-grade overhaul + procedural generation

feat: /grader endpoint, validate-submission.sh, session cleanup, final polish

fix: DEBUG to stderr, scores (0.001,0.999), OPENENV_BASE_URL support, clean imports

fix: 3 showstoppers (dockerignore, END format, efficiency gaming) + license + unused imports

fix: all critical audit issues — [END] format, finding matching, methodology scoring, LLM timeout, error handling

fix: typed Action model, OPENAI_API_KEY support, proper spec compliance

fix: auto-grade on budget exhaustion + max steps, no more 0.01 dead scores

fix: replace oversized mermaid with clean text flow diagrams

fix: readable state graphs (TD layout, short labels, progress-only)

feat: test suite (9 tests), dynamic leaderboard, mermaid graphs

feat: mermaid state-graph diagrams in scenarios tab

fix: mount gradio at / (maverick pattern)

fix: mount gradio at /web (hf iframe path), disable openenv default ui

feat: custom gradio dashboard (6 tabs, charcoal+gold, leaderboard, playground, architecture)

fix: enable web interface, mount custom ui at /dashboard

feat: premium ui (charcoal+gold authority palette), multi-model eval, /web mount fix

fix: robust gradio mount with html fallback for hf spaces

feat: eu ai act compliance auditor — mcp-based openenv environment