process_kpi/kpi_health_check/benchmarks.py

import pandas as pd
import numpy as np

def calculate_sla_metrics(
    df: pd.DataFrame, 
    kpi: str, 
    rules_df: pd.DataFrame | None = None
) -> dict:
    """
    Calculates simple metrics for the given KPI trace:
    - SLA value (if exists)
    - Median (recent window)
    
    Returns a dict with: 'sla': float|None, 'median': float|None
    """
    res = {"sla": None, "median": None}
    
    if df is None or df.empty or kpi not in df.columns:
        return res

    # 1. Get SLA from rules
    if rules_df is not None and not rules_df.empty:
        # Assuming rules_df has 'KPI' and 'sla' columns
        # We also need to match RAT? usually passed or handled outside.
        # Here we do a simplistic lookup.
        try:
            row = rules_df[rules_df["KPI"] == kpi]
            if not row.empty:
                val = row.iloc[0].get("sla")
                res["sla"] = float(val) if pd.notna(val) else None
        except Exception:
            pass

    # 2. Calculate Median (entire passed df, usually it's the recent window)
    try:
        vals = pd.to_numeric(df[kpi], errors="coerce").dropna()
        if not vals.empty:
            res["median"] = float(vals.median())
    except Exception:
        pass
        
    return res