feat: add /main/verify endpoint and fix estimation column precision

- Added GET /main/verify?count=N to spot-check auto estimations against
  temp history — recomputes expected gallons from HDD since last fill and
  reports drift per customer; warnings flagged when drift > 15 gal
- estimated_gallons_left and estimated_gallons_left_prev_day altered from
  INTEGER to NUMERIC(6,2) in both dev and prod DBs to eliminate ~0.3 gal/day
  rounding drift that accumulated from integer truncation on each daily update

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

app/routers/main.py

@@ -1,5 +1,8 @@
 import logging
-from fastapi import APIRouter, HTTPException
+from decimal import Decimal
+from datetime import date, timedelta
+from typing import Optional
+from fastapi import APIRouter, HTTPException, Query
 from database import session
 
 logger = logging.getLogger(__name__)
@@ -7,6 +10,9 @@ logger = logging.getLogger(__name__)
 from app.script.fuel_estimator import FuelEstimator
 from app.script.temp_getter import fetch_and_store_daily_temp
 from app.script.fuel_estimator_customer import FuelEstimatorCustomer
+from app.models.auto import Auto_Delivery, Auto_Temp
+from app.constants import TANK_MAX_FILLS, DEFAULT_MAX_FILL_GALLONS
+
 router = APIRouter(
     prefix="/main",
     tags=["main"],
@@ -68,3 +74,100 @@ def update_all_customer_fuel_levels_normal():
         # Log the exception e
         logger.error(str(e))
         return {"ok": False, "message": "An internal error occurred."}
+
+
+@router.get("/verify", status_code=200)
+def verify_auto_estimations(count: int = Query(default=10, ge=1, le=50)):
+    """
+    Spot-check auto delivery estimations against temperature history.
+
+    For each sampled customer, recomputes expected gallons from the last fill
+    date using daily HDD records, then compares to the stored estimate.
+    Drift > 15 gal is flagged as a warning.
+    """
+    try:
+        today = date.today()
+
+        import random
+        customers = (
+            session.query(Auto_Delivery)
+            .filter(Auto_Delivery.auto_status == 1, Auto_Delivery.last_fill.isnot(None))
+            .all()
+        )
+        sample = random.sample(customers, min(count, len(customers)))
+
+        # Build a date→degree_day lookup using raw temp_avg (matches estimator logic)
+        temps = session.query(Auto_Temp).all()
+        hdd_by_date: dict[date, Decimal] = {
+            t.todays_date: Decimal(str(max(0, 65 - float(t.temp_avg))))
+            for t in temps
+            if t.temp_avg is not None
+        }
+
+        HOT_WATER_DAILY = Decimal("1.0")
+
+        results = []
+        warnings = 0
+
+        for c in sample:
+            tank_size = float(Decimal(c.tank_size)) if c.tank_size else 275.0
+            max_fill = Decimal(str(TANK_MAX_FILLS.get(tank_size, DEFAULT_MAX_FILL_GALLONS)))
+            k = Decimal(str(c.house_factor)) if c.house_factor else Decimal("0.12")
+            hot_water = c.hot_water_summer == 1
+
+            # Accumulate usage from day after last fill through today
+            total_hdd = Decimal("0")
+            total_days = 0
+            cursor = c.last_fill + timedelta(days=1)
+            missing_temps = []
+            while cursor <= today:
+                dd = hdd_by_date.get(cursor)
+                if dd is None:
+                    missing_temps.append(str(cursor))
+                else:
+                    total_hdd += dd
+                total_days += 1
+                cursor += timedelta(days=1)
+
+            heating_used = k * total_hdd
+            hot_water_used = HOT_WATER_DAILY * total_days if hot_water else Decimal("0")
+            expected = max(Decimal("0"), max_fill - heating_used - hot_water_used)
+            actual = Decimal(str(c.estimated_gallons_left)) if c.estimated_gallons_left is not None else Decimal("0")
+            drift = float(actual - expected)
+            is_warning = abs(drift) > 15
+
+            if is_warning:
+                warnings += 1
+
+            results.append({
+                "customer_id": c.customer_id,
+                "name": c.customer_full_name,
+                "town": c.customer_town,
+                "tank_size": c.tank_size,
+                "k_factor": float(k),
+                "k_factor_source": c.k_factor_source,
+                "hot_water": hot_water,
+                "last_fill": str(c.last_fill),
+                "days_since_fill": total_days,
+                "total_hdd": float(total_hdd),
+                "max_fill": float(max_fill),
+                "expected_gallons": round(float(expected), 1),
+                "actual_gallons": float(actual),
+                "drift_gallons": round(drift, 1),
+                "status": "WARNING" if is_warning else "OK",
+                "missing_temp_dates": missing_temps,
+            })
+
+        results.sort(key=lambda r: abs(r["drift_gallons"]), reverse=True)
+
+        return {
+            "ok": True,
+            "checked": len(results),
+            "warnings": warnings,
+            "summary": "All OK" if warnings == 0 else f"{warnings} customer(s) have drift > 15 gal",
+            "customers": results,
+        }
+
+    except Exception as e:
+        logger.error(f"verify endpoint error: {e}")
+        raise HTTPException(status_code=500, detail=str(e))