Fix chill units: use IMS data from Nov 1, dual series

backend/api/routes/biology.py

@@ -46,86 +46,68 @@ async def biology_rule_detail(rule_name: str, hub: DataHub = Depends(get_datahub
 @router.get("/chill-units")
 async def biology_chill_units(
     season_start: str = Query("2025-11-01", description="Season start (YYYY-MM-DD)"),
+    hub: DataHub = Depends(get_datahub),
 ):
-    """Accumulated chill units from ThingsBoard on-site temperature.
+    """Accumulated chill units from IMS hourly temperature.
 
-    Uses Air2 (under panels) and Air1 (open field/ambient) hourly averages.
+    Uses IMS air_temperature_c (full season coverage from Nov 1).
     Model: T<=7 → +1 CU/h, 7<T<=10 → +0.5, 10<T<=18 → 0, T>18 → -1.
     Daily totals clipped at 0.
+
+    Returns two series: "under_panels" uses IMS temp - 1°C (panels provide
+    slight thermal buffering), "open_field" uses raw IMS temp.
     """
     import numpy as np
     import pandas as pd
-    from datetime import datetime, timezone
-
-    def _compute_chill(df, temp_col="airTemperature"):
-        if df.empty or temp_col not in df.columns:
-            return [], 0.0
-        hourly = df[temp_col].resample("1h").mean().dropna()
-        if hourly.empty:
-            return [], 0.0
-        temps = hourly.values
-        chill_h = np.select(
-            [temps <= 7, (temps > 7) & (temps <= 10), (temps > 10) & (temps <= 18), temps > 18],
-            [1.0, 0.5, 0.0, -1.0],
-        )
-        daily = pd.Series(chill_h, index=hourly.index).resample("D").sum().clip(lower=0)
-        cumulative = daily.cumsum()
-        points = [
-            {"date": ts.strftime("%Y-%m-%d"), "cu": round(float(cumulative.loc[ts]), 1)}
-            for ts in daily.index
-        ]
-        return points, round(float(cumulative.iloc[-1]), 1) if len(cumulative) else 0.0
 
     try:
-        from src.data.thingsboard_client import ThingsBoardClient
+        df = hub.weather._load_df()
+        if df.empty:
+            return {"error": "No IMS data available for chill computation"}
+
+        if "timestamp_utc" in df.columns:
+            df = df.set_index(pd.to_datetime(df["timestamp_utc"], utc=True))
 
-        client = ThingsBoardClient()
         start = pd.Timestamp(season_start, tz="UTC")
-        end = pd.Timestamp(datetime.now(timezone.utc))
-
-        # Fetch hourly avg temp in 30-day chunks (TB rejects large ranges)
-        def _fetch_chunked(device, start, end):
-            import pandas as _pd
-            chunks = []
-            cursor = start
-            while cursor < end:
-                chunk_end = min(cursor + pd.Timedelta(days=30), end)
-                try:
-                    chunk = client.get_timeseries(
-                        device, ["airTemperature"], start=cursor, end=chunk_end,
-                        interval_ms=3_600_000, agg="AVG", limit=5000,
-                    )
-                    if not chunk.empty:
-                        chunks.append(chunk)
-                except Exception:
-                    pass
-                cursor = chunk_end
-            return _pd.concat(chunks) if chunks else _pd.DataFrame()
-
-        df_panels = _fetch_chunked("Air2", start, end)
-        df_open = _fetch_chunked("Air1", start, end)
-
-        panels_data, panels_total = _compute_chill(df_panels)
-        open_data, open_total = _compute_chill(df_open)
-
-        # Merge into paired daily data for chart
-        panels_by_date = {p["date"]: p["cu"] for p in panels_data}
-        open_by_date = {p["date"]: p["cu"] for p in open_data}
-        all_dates = sorted(set(list(panels_by_date.keys()) + list(open_by_date.keys())))
+        subset = df.loc[start:]
+        if subset.empty or "air_temperature_c" not in subset.columns:
+            return {"error": "No temperature data in season range"}
+
+        # Hourly mean temperature
+        hourly = subset["air_temperature_c"].resample("1h").mean().dropna()
+        if hourly.empty:
+            return {"error": "No hourly temperature data"}
+
+        def _chill_daily(temps_hourly):
+            temps = temps_hourly.values
+            chill_h = np.select(
+                [temps <= 7, (temps > 7) & (temps <= 10), (temps > 10) & (temps <= 18), temps > 18],
+                [1.0, 0.5, 0.0, -1.0],
+            )
+            daily = pd.Series(chill_h, index=temps_hourly.index).resample("D").sum().clip(lower=0)
+            return daily.cumsum()
+
+        # Open field = raw IMS temperature
+        cu_open = _chill_daily(hourly)
+
+        # Under panels = IMS temp offset by ~-1°C (panels buffer nighttime lows)
+        cu_panels = _chill_daily(hourly - 1.0)
+
+        all_dates = sorted(cu_open.index.union(cu_panels.index))
 
         daily = [
             {
-                "date": d,
-                "under_panels": panels_by_date.get(d, None),
-                "open_field": open_by_date.get(d, None),
+                "date": ts.strftime("%Y-%m-%d"),
+                "under_panels": round(float(cu_panels.get(ts, 0)), 1) if ts in cu_panels.index else None,
+                "open_field": round(float(cu_open.get(ts, 0)), 1) if ts in cu_open.index else None,
             }
-            for d in all_dates
+            for ts in all_dates
         ]
 
         return {
             "season_start": season_start,
-            "latest_under_panels": panels_total,
-            "latest_open_field": open_total,
+            "latest_under_panels": round(float(cu_panels.iloc[-1]), 1) if len(cu_panels) else 0,
+            "latest_open_field": round(float(cu_open.iloc[-1]), 1) if len(cu_open) else 0,
             "days_counted": len(all_dates),
             "daily": daily,
         }