config.py

"""Configuration settings for LLM Inference Dashboard."""

from dataclasses import dataclass, field
from typing import Optional
import os


@dataclass
class Config:
    """Dashboard configuration with sensible defaults."""

    # vLLM Connection
    vllm_host: str = "localhost"
    vllm_port: int = 8000
    model_path: Optional[str] = None

    # Dashboard
    refresh_interval: float = 1.0
    history_length: int = 300  # 5 minutes at 1s intervals

    # Database
    db_path: str = "data/metrics.db"

    # Alert Thresholds
    alert_kv_cache_threshold: float = 90.0
    alert_gpu_memory_threshold: float = 95.0
    alert_ttft_multiplier: float = 2.0
    alert_throughput_drop_pct: float = 50.0

    # Webhooks
    slack_webhook: Optional[str] = None
    pagerduty_routing_key: Optional[str] = None
    generic_webhooks: list = field(default_factory=list)

    # Load Testing Defaults
    loadtest_concurrent_users: int = 10
    loadtest_rps: float = 5.0
    loadtest_duration: int = 60

    @property
    def metrics_endpoint(self) -> str:
        return f"http://{self.vllm_host}:{self.vllm_port}/metrics"

    @property
    def openai_endpoint(self) -> str:
        return f"http://{self.vllm_host}:{self.vllm_port}/v1"

    @property
    def health_endpoint(self) -> str:
        return f"http://{self.vllm_host}:{self.vllm_port}/health"

    @classmethod
    def from_env(cls) -> "Config":
        """Create config from environment variables."""
        return cls(
            vllm_host=os.getenv("VLLM_HOST", "localhost"),
            vllm_port=int(os.getenv("VLLM_PORT", "8000")),
            model_path=os.getenv("MODEL_PATH"),
            refresh_interval=float(os.getenv("REFRESH_INTERVAL", "1.0")),
            db_path=os.getenv("DB_PATH", "data/metrics.db"),
            slack_webhook=os.getenv("SLACK_WEBHOOK"),
            pagerduty_routing_key=os.getenv("PAGERDUTY_KEY"),
        )


# Global config instance
config = Config.from_env()