Redesign env as incident operations workflow

Add LLM-backed grading and planning

Harden env isolation and proxy validation

Fix reset endpoint and Python 3.14 compatibility

Add grounded optimization modes and wire dedicated optimizer module

Improve citation guidance, preserve prompt structure, harden baseline, and validate inference

Fix grounding truthfulness, baseline fallback honesty, docs, and validator coverage

Add PyTorch context tuning with demo-seeded retrieval and UI grounding stats

Add gold citations and multi-pack enterprise corpus switching

Expose grounding and citation signals in frontend and baseline

Add citation-aware grading and swappable corpus path

Finalize single-template UI and enterprise workflow framing

Force fresh template reads and add visible build marker

Extract UI template and reframe benchmark as enterprise workflow

Fix reset default and harden baseline runner semantics

Use full-height left-right prompt workspace layout

Restyle UI with cleaner Apple-like design language

Fix stale optimization summary after prompt edits

Fix optimize summary updates and enforce shorter output

Make optimize-prompt actually shorten the prompt

Simplify UI to prompt-in optimized-prompt-out flow

Show optimized prompt output and real compression effect

Rebuild UI and harden frontend-backend action flow

Make UI actions auto-start episodes from custom prompt

Fix custom prompt UI and token optimizer flow

Add interactive root UI and Streamlit app

Initial OpenEnv submission