# Cloud-OpsBench v1 — Anthropic slice (claude-4-sonnet).
#
# Provider slice of cloudopsbench_v1.yml. Use this when running just the
# Anthropic model (no OPENAI_API_KEY / DEEPSEEK_API_KEY required). The
# full grid (all 4 paper models in one run) is still
# cloudopsbench_v1.yml — keep that for the publication-grade comparison
# once all provider credits are available.
#
# Sibling slices:
#   - cloudopsbench_v1_openai.yml      (gpt-4o + gpt-5)
#   - cloudopsbench_v1_deepseek.yml    (deepseek-v3.2)
#
# Required env at run time: ANTHROPIC_API_KEY.
#
# Run with --dev first to verify the chain, then drop --dev for production:
#   uv run python -m tests.benchmarks._framework.cli run \
#       tests/benchmarks/cloudopsbench/configs/cloudopsbench_v1_anthropic.yml --dev

benchmark: cloudopsbench

modes:
  - opensre+llm

llms:
  - claude-4-sonnet

model_versions:
  claude-4-sonnet: claude-sonnet-4-5-20250929

runs_per_case: 3
workers: 4
cost_budget_usd: 500.0
seed: 42

output_dir: .bench-results/cloudopsbench_v1_anthropic/

pre_registration_path: tests/benchmarks/cloudopsbench/configs/preregistrations/cloudopsbench_v1.yml

filters:
  limit: 30
  seen_shape: [true]
  systems: []
  fault_categories: []

report_formats:
  - json
  - markdown
  - html