agent/core/level2_config.py

"""
Level 2 Configuration - Single model configuration via environment variables
"""

import os
from dataclasses import dataclass
from typing import Optional


@dataclass
class LLMConfig:
    """Single LLM model configuration from environment"""
    model: str
    api_key: str
    base_url: Optional[str] = None
    temperature: float = 0.7
    max_tokens: int = 4096
    
    @classmethod
    def from_env(cls) -> 'LLMConfig':
        """Load configuration from environment variables"""
        return cls(
            model=os.getenv("LLM_MODEL", "gpt-4-turbo-preview"),
            api_key=os.getenv("LLM_API_KEY", ""),
            base_url=os.getenv("LLM_BASE_URL"),
            temperature=float(os.getenv("LLM_TEMPERATURE", "0.7")),
            max_tokens=int(os.getenv("LLM_MAX_TOKENS", "4096"))
        )


@dataclass
class Level2Config:
    """Level 2 advanced configuration"""
    # Reasoning
    enable_multi_pass_reasoning: bool = True
    max_reasoning_depth: int = 5
    complexity_threshold_high: int = 70
    complexity_threshold_low: int = 30
    
    # Caching
    enable_semantic_cache: bool = True
    cache_similarity_threshold: float = 0.92
    cache_ttl_seconds: int = 604800  # 7 days
    
    # Optimization
    enable_parallel_execution: bool = True
    enable_auto_retry: bool = True
    max_tool_retries: int = 3
    tool_timeout_base: float = 30.0
    
    # Observability
    enable_metrics: bool = True
    enable_anomaly_detection: bool = True
    anomaly_threshold_multiplier: float = 2.5
    
    # Memory
    enable_contextual_memory: bool = True
    max_context_tokens: int = 2000
    
    @classmethod
    def from_env(cls) -> 'Level2Config':
        """Load Level 2 configuration from environment"""
        return cls(
            enable_multi_pass_reasoning=os.getenv("ENABLE_MULTI_PASS", "true").lower() == "true",
            max_reasoning_depth=int(os.getenv("MAX_REASONING_DEPTH", "5")),
            complexity_threshold_high=int(os.getenv("COMPLEXITY_HIGH", "70")),
            complexity_threshold_low=int(os.getenv("COMPLEXITY_LOW", "30")),
            enable_semantic_cache=os.getenv("ENABLE_SEMANTIC_CACHE", "true").lower() == "true",
            cache_similarity_threshold=float(os.getenv("CACHE_SIMILARITY", "0.92")),
            cache_ttl_seconds=int(os.getenv("CACHE_TTL", "604800")),
            enable_parallel_execution=os.getenv("ENABLE_PARALLEL", "true").lower() == "true",
            enable_auto_retry=os.getenv("ENABLE_AUTO_RETRY", "true").lower() == "true",
            max_tool_retries=int(os.getenv("MAX_RETRIES", "3")),
            tool_timeout_base=float(os.getenv("TOOL_TIMEOUT", "30.0")),
            enable_metrics=os.getenv("ENABLE_METRICS", "true").lower() == "true",
            enable_anomaly_detection=os.getenv("ENABLE_ANOMALY", "true").lower() == "true",
            anomaly_threshold_multiplier=float(os.getenv("ANOMALY_THRESHOLD", "2.5")),
            enable_contextual_memory=os.getenv("ENABLE_MEMORY", "true").lower() == "true",
            max_context_tokens=int(os.getenv("MAX_CONTEXT_TOKENS", "2000"))
        )


# Global configurations
llm_config = LLMConfig.from_env()
level2_config = Level2Config.from_env()