artifacts/annotations-training.manifest.json

{
  "requested": 9065,
  "valid": 2781,
  "coverage_fraction": 0.3068,
  "run_status": "interrupted_before_completion",
  "note": "No completion manifest was written by the annotation process (process did not reach its final write step); this file was reconstructed from the cache after the fact. Remaining 6284/9065 train examples are not yet annotated. --require-complete-annotations will fail until the run resumes and finishes.",
  "accounted_cost_usd_from_cache": 89.8479,
  "budget_ceiling_usd": 185,
  "model": "gpt-5.6-terra",
  "resume_command": "bash scripts/annotate_local.sh full  # reuses artifacts/annotation-cache, does not re-spend on already-cached examples"
}