fix: revert inference format and make /reset payload optional

This is the End

Submission polish: compliance hardening, baseline matrix, dashboard UX, tests, and docs

Improve shift_surge grading and phraseology-based protocol reward

Expand action space: stage, reassign, mutual aid, severity change

Unify OpenEnv and benchmark episode scoring

Unify benchmark scoring with OpenEnv episode reward

Fix triage reward lookup for enum-qualified metadata

Finalize OpenEnv baseline: OpenAI client, PORT binding, and docs

Add OpenEnv runtime endpoints

Make openenv validate pass

Apply changes.md compliance fixes

Fix OpenEnv metadata, docker data, tasks endpoint, and demo

Dispatch environment: rewards, dashboard, docs, and passing tests

Rename project slug; remove architecture.md

Initial import: 911 dispatch RL environment