frontierscience_judge:
  resources_servers:
    frontierscience_judge:
      entrypoint: app.py
      domain: other
      verified: false
      description: FrontierScience answer grading via single-pass LLM judge
      value: Evaluate FrontierScience Olympiad short answers or Research rubric-scored answers

      judge_model_server:
        type: responses_api_models
        name: judge_model
      judge_responses_create_params:
        input: []
        max_output_tokens: 2048
      judge_mode: olympiad
      rubric_pass_score_threshold: 7.0
      rubric_max_score: 10.0
      judge_prompt_path: resources_servers/frontierscience_judge/prompts/judge.yaml
      use_chat_completions_for_judge: true

frontierscience_judge_simple_agent:
  responses_api_agents:
    simple_agent:
      entrypoint: app.py
      resources_server:
        type: resources_servers
        name: frontierscience_judge
      model_server:
        type: responses_api_models
        name: policy_model
      datasets:
      - name: example
        type: example
        jsonl_fpath: resources_servers/frontierscience_judge/data/example.jsonl
        license: Apache 2.0

judge_base_url: https://integrate.api.nvidia.com/v1
judge_api_key: ${oc.env:NVIDIA_API_KEY,null}
judge_model_name: openai/gpt-oss-20b

judge_model:
  responses_api_models:
    openai_model:
      entrypoint: app.py
      openai_base_url: ${judge_base_url}
      openai_api_key: ${judge_api_key}
      openai_model: ${judge_model_name}