{"data":{"kind":"file","path":"README.md","version_id":"wy6xjgup493kjz0hxudn0fi0","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":995,"modified_at":"2026-05-06T18:30:28.658000","content_hash":"567ee7a52a5de09dd1ebc88ae0d524ad979eedfff0a0c5a9f1c977608610bc16"},"entries":[],"content":"# crystal-relaxation-rlm\n\nWIP: this repository is an in-progress migration of the original `crystal_relaxation` environment to an RLM-native implementation.\n\n## Status\n\n- Migration is active and not feature-complete.\n- APIs, scoring details, and training configs may change.\n- Outputs in this repo should be treated as experimental.\n\n## Goal\n\nPort the legacy crystal relaxation pipeline into a composable RLM environment with:\n\n- sandbox-backed rubric scoring,\n- checker-based CIF validation,\n- hosted training/eval compatibility, and\n- improved observability for skill and rubric execution.\n\n## Quickstart\n\n```bash\nprime eval run crystal-relaxation-rlm\n```\n\n## Sandbox image\n\nThe default sandbox image is:\n\n```text\ncmhkq5thw0000wlnw8zsnj8ze/crystal-relaxation-rlm:latest\n```\n\nBuild and push it with:\n\n```bash\nprime images push crystal-relaxation-rlm:latest --dockerfile docker/Dockerfile --context .\n```\n\n## Notes\n\n- Environment ID: `crystal-relaxation-rlm`\n- This is not yet a stable release.\n","encoding":"utf-8","truncated":false,"total_bytes":995},"status":null}