env.py

import os
OWNER = "cyberco"
VERSION = "2025_v1"
REPO_ID = f"{OWNER}/CAIA-Benchmark-Leaderboard"

TOKEN = os.getenv("HF_TOKEN")


SUBMISSION_DATASET_PUBLIC = f"{OWNER}/public_submissions"  # 添加缺失的变量

INTERNAL_DATASET = f"{OWNER}/caia_internal"
EVALUATE_RESULT_DATASET = f"{OWNER}/public_results"
SUBMISSION_DATASET = f"{OWNER}/submissions_internal"
CONTACT_DATASET = f"{OWNER}/contact_info"


BENCHMARK_INTERNAL_EVALUATE_DATASET_FILE = f"{VERSION}/{os.getenv('BENCHMARK_INTERNAL_EVALUATE_DATASET', 'example_evaluate_data.json')}"
EVALUATE_RESULT_DATASET_FILE = f"{VERSION}/{os.getenv('EVALUATE_RESULT_DATASET', 'example_result.json')}"
CONTACT_DATASET_FILE = f"{os.getenv('CONTACT_DATASET_FILE', 'example_contact_info.json')}"


llm_config = {
    "parse_llm_config": {
        "model_name": "gpt-4.1-mini-2025-04-14",
        "api_key": os.getenv("OPENAI_API_KEY", None),
        "model_params": {
            "temperature": 0
        }
    },
    "evaluate_llm_configs": [
        {
            "model_name": "o3-2025-04-16",
            "api_key": os.getenv("OPENAI_API_KEY", None),
            "model_params": {
                "reasoning_effort": "medium"
            }
        },
        {
            "model_name": "gpt-4.1",
            "api_key": os.getenv("OPENAI_API_KEY", None),
            "model_params": {
                "temperature": 0.2
            }
        },
        {
            "model_name": "deepseek-r1-250120",
            "api_key": os.getenv("DEEPSEEK_API_KEY", None),
            "base_url": os.getenv("DEEPSEEK_BASE_URL", None),
            "model_params": {
                "temperature": 0.2
            }
        }
    ]
}