config.py

"""
配置文件
"""
import os
from pathlib import Path
from typing import List

# 项目根目录
BASE_DIR = Path(__file__).parent

# API配置
API_CONFIG = {
    "title": "HR对话质量评估API",
    "description": "基于ML模型的HR对话质量评估系统",
    "version": "v1.0",
    "host": os.getenv("API_HOST", "0.0.0.0"),
    "port": int(os.getenv("API_PORT", 8000)),
    "debug": os.getenv("DEBUG", "false").lower() == "true",
    "reload": os.getenv("DEBUG", "false").lower() == "true",
}

# 模型配置 - 使用 HuggingFace Hub 上的微调模型（已设为公开）
HF_SBERT_MODEL = "KarenYYH/sbert-hr-v2"
HF_SENTIMENT_MODEL = "KarenYYH/sentiment-hr"

local_sbert_path = BASE_DIR / "models" / "sbert-hr-v2"
# 优先使用 HuggingFace Hub 上的微调模型
default_sbert = HF_SBERT_MODEL

# 情绪分析微调模型路径
local_sentiment_path = BASE_DIR / "models" / "sentiment-hr" / "final_model"
default_sentiment = HF_SENTIMENT_MODEL

MODEL_CONFIG = {
    "cache_dir": os.getenv("MODEL_CACHE_DIR", str(BASE_DIR / "models" / "checkpoints")),
    "sentence_transformer_model": os.getenv(
        "SENTENCE_TRANSFORMER_MODEL",
        default_sbert
    ),
    # 情绪分析模型（微调版本）
    "sentiment_model_path": os.getenv(
        "SENTIMENT_MODEL_PATH",
        default_sentiment
    ),
    # 本地对话模型已移除，推荐使用 LLM API
    "dialogue_model_path": None,
    "device": "mps" if os.getenv("USE_MPS", "true").lower() == "true" and os.uname().sysname == "Darwin" else ("cuda" if os.getenv("USE_CUDA", "true").lower() == "true" else "cpu"),
}

# 数据路径
DATA_CONFIG = {
    "knowledge_base_path": os.getenv(
        "KNOWLEDGE_BASE_PATH",
        str(BASE_DIR / "data" / "knowledge_base.json")
    ),
    "sensitive_words_path": os.getenv(
        "SENSITIVE_WORDS_PATH",
        str(BASE_DIR / "data" / "sensitive_words.json")
    ),
    "whitelist_path": os.getenv(
        "WHITELIST_PATH",
        str(BASE_DIR / "data" / "whitelist.json")
    ),
}

# 评估阈值
EVALUATION_CONFIG = {
    "correctness": {
        "threshold": float(os.getenv("CORRECTNESS_THRESHOLD", 0.85)),
        "partial_threshold": float(os.getenv("PARTIAL_THRESHOLD", 0.6)),
    },
    "sentiment": {
        "confidence_threshold": float(os.getenv("SENTIMENT_CONFIDENCE_THRESHOLD", 0.6)),
    },
    "compliance": {
        "severity_levels": {
            "high": ["discrimination", "illegal_employment"],
            "medium": ["privacy_violation", "threat"],
            "low": ["other"]
        }
    }
}

# CORS配置
CORS_CONFIG = {
    "origins": os.getenv("CORS_ORIGINS", "*").split(","),
    "allow_credentials": os.getenv("CORS_ALLOW_CREDENTIALS", "true").lower() == "true",
    "allow_methods": ["*"],
    "allow_headers": ["*"],
}

# 日志配置
LOG_CONFIG = {
    "level": os.getenv("LOG_LEVEL", "INFO"),
    "format": "{time:YYYY-MM-DD HH:mm:ss} | {level} | {name}:{function}:{line} | {message}",
    "rotation": "1 day",
    "retention": "7 days",
    "compression": "zip",
}

# API认证（可选）
AUTH_CONFIG = {
    "enabled": os.getenv("API_KEY") is not None,
    "api_key": os.getenv("API_KEY", ""),
}

# 缓存配置
CACHE_CONFIG = {
    "enabled": True,
    "ttl": 3600,  # 1小时
    "backend": "redis" if os.getenv("USE_REDIS", "false").lower() == "true" else "memory",
    "redis": {
        "host": os.getenv("REDIS_HOST", "localhost"),
        "port": int(os.getenv("REDIS_PORT", 6379)),
        "password": os.getenv("REDIS_PASSWORD", None),
        "db": int(os.getenv("REDIS_DB", 0)),
        "socket_timeout": 5,
        "max_connections": 100,  # 连接池优化
    }
}

# 意图识别模型配置
INTENT_MODEL_CONFIG = {
    "model_name": os.getenv("INTENT_MODEL_NAME", "bert-base-chinese"),
    "model_path": os.getenv("INTENT_MODEL_PATH", str(BASE_DIR / "models" / "intent_bert")),
    "confidence_threshold": 0.85,
}

# 对话生成模型 API 配置
LLM_API_CONFIG = {
    # 启用 API 模式（true=使用API, false=使用本地模型）
    "enabled": os.getenv("LLM_API_ENABLED", "false").lower() == "true",

    # API 提供商: openai, anthropic, deepseek, moonshot, zhipu, dashscope, ollama, custom
    "provider": os.getenv("LLM_API_PROVIDER", "openai"),

    # API 密钥
    "api_key": os.getenv("LLM_API_KEY", ""),

    # API 端点（用于自定义 API）
    "base_url": os.getenv("LLM_API_BASE_URL", ""),

    # 模型名称
    "model": os.getenv("LLM_API_MODEL", "gpt-4o-mini"),

    # API 参数
    "temperature": float(os.getenv("LLM_API_TEMPERATURE", "0.7")),
    "max_tokens": int(os.getenv("LLM_API_MAX_TOKENS", "512")),
    "top_p": float(os.getenv("LLM_API_TOP_P", "0.9")),

    # 超时设置（秒）
    "timeout": int(os.getenv("LLM_API_TIMEOUT", "30")),
}