Comprehensive score clamping: fix ALL code paths in graders, app.py baseline/grader endpoints, inference.py, environment.py + add grader fields to openenv.yaml + fix reward range

Clamp ALL grader scores to strictly (0.01, 0.99) - fix every return path in all 3 graders + environment exception handler

Remove external .db files, generate all data synthetically - fixes 'file is not a database' on evaluator

Cleanup server/app.py API fallback keys and commit pending Dockerfile updates

Fix ModuleNotFoundError in Native Fallback crashing the LLM Proxy and restore strict JSON mode

fix: orchestrate server+baseline startup to ensure stdout is captured correctly and priority use of HF_TOKEN for proxy authentication

fix: /baseline endpoint now uses LLM via os.environ[API_BASE_URL]+API_KEY proxy - was using hardcoded heuristic with zero LLM calls

Fix final deployment validation bugs (uv.lock, openenv version, app.py main)

Allow /reset to accept empty payload for validation

Update grader documentation

Update generator docs

Remove pycache payload and apply gitignore

Initial commit with LFS for databases