Keep inference runner exit code stable

Redesign env as incident operations workflow

Harden env isolation and proxy validation

Align inference with submission env contract

Require proxy env vars for submission inference

Use API_KEY for LiteLLM proxy calls

Finalize submission checks and long inference output format

Align inference output with strict hackathon contract

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

Expose grounding and citation signals in frontend and baseline

Add citation-aware grading and swappable corpus path

Fix reset default and harden baseline runner semantics

Clean repo artifacts and harden submission assets

Initial OpenEnv submission