Body

backend/app/api/health.py

@@ -144,8 +144,9 @@ async def intraday_health(
     )
     metric = result.scalar_one_or_none()
     return {
-        "hr_values":    metric.intraday_hr if metric else None,
-        "body_battery": metric.body_battery if metric else None,
+        "hr_values":          metric.intraday_hr if metric else None,
+        "body_battery":       metric.body_battery if metric else None,
+        "body_battery_hires": metric.body_battery_hires if metric else None,
     }
 
 

backend/app/models/user.py

@@ -244,8 +244,9 @@ class HealthMetric(Base):
     active_calories = Column(Float, nullable=True)
     total_calories = Column(Float, nullable=True)
     spo2_avg = Column(Float, nullable=True)
-    intraday_hr = Column(JSON, nullable=True)   # [[epoch_ms, bpm], ...] — not in API list response
-    body_battery = Column(JSON, nullable=True)  # {charged,drained,start_level,end_level,values:[[ts_ms,level,type,stress]...]}
+    intraday_hr = Column(JSON, nullable=True)           # [[epoch_ms, bpm], ...] — not in API list response
+    body_battery = Column(JSON, nullable=True)          # {charged,drained,start_level,end_level,values:[[ts_ms,level,type,stress]...]}
+    body_battery_hires = Column(JSON, nullable=True)    # [[ts_ms, level], ...] interpolated from bb + HR; higher resolution than raw values
 
     __table_args__ = (
         UniqueConstraint("user_id", "date", name="uq_health_user_date"),

backend/app/services/garmin_connect_sync.py

@@ -258,12 +258,19 @@ def sync_wellness(garmin, user_id: int, since: Optional[datetime], db,
                 row["body_battery"] = _json.dumps(bb)
 
         # Intraday heart rate — store non-null [epoch_ms, bpm] pairs
+        intraday = None
         if hr_raw:
             raw_vals = hr_raw.get("heartRateValues") or []
             intraday = [[int(ts), int(v)] for ts, v in raw_vals if v is not None]
             if intraday:
                 row["intraday_hr"] = intraday
 
+        # High-resolution body battery derived from BB checkpoints + intraday HR
+        if bb and intraday:
+            hires = _compute_body_battery_hires(bb.get("values") or [], intraday)
+            if hires:
+                row["body_battery_hires"] = _json.dumps(hires)
+
         if not row:
             continue
 
@@ -351,6 +358,72 @@ def _parse_body_battery(bb_response, day_str: str):
     }
 
 
+def _compute_body_battery_hires(bb_values, intraday_hr):
+    """
+    Produce a higher-resolution body battery series by interpolating between
+    sparse BB checkpoints using intraday HR as a proxy for effort.
+
+    During drain segments (BB falling) the drain is distributed proportionally
+    to how much each HR reading exceeds the day's median — peaks spend battery
+    faster than valleys.  During recovery segments (BB rising) recovery is
+    spread uniformly over time.
+
+    Returns [[ts_ms, level], ...] at the granularity of intraday HR, or None
+    if inputs are insufficient.
+    """
+    if not bb_values or not intraday_hr or len(bb_values) < 2:
+        return None
+
+    bb  = sorted(bb_values,  key=lambda x: x[0])
+    hr  = sorted(intraday_hr, key=lambda x: x[0])
+
+    hr_vals = [bpm for _, bpm in hr if bpm is not None and bpm > 0]
+    if not hr_vals:
+        return None
+
+    hr_median = sorted(hr_vals)[len(hr_vals) // 2]
+
+    result = []
+    for i in range(len(bb) - 1):
+        t1, L1 = bb[i][0], bb[i][1]
+        t2, L2 = bb[i + 1][0], bb[i + 1][1]
+        delta = L2 - L1
+
+        seg_hr = [(ts, bpm) for ts, bpm in hr if t1 <= ts <= t2 and bpm is not None]
+        result.append([t1, round(float(L1), 1)])
+
+        if not seg_hr or abs(delta) < 1:
+            continue
+
+        if delta < 0:
+            # Drain: weight each reading by HR above median
+            efforts = [max(0.0, bpm - hr_median) for _, bpm in seg_hr]
+            total   = sum(efforts) or 1.0
+            cumul   = 0.0
+            for j, (ts, bpm) in enumerate(seg_hr):
+                cumul += efforts[j] * delta / total
+                level  = max(0.0, min(100.0, L1 + cumul))
+                result.append([ts, round(level, 1)])
+        else:
+            # Recovery: linear over time
+            span = max(1, t2 - t1)
+            for ts, _ in seg_hr:
+                frac  = (ts - t1) / span
+                level = max(0.0, min(100.0, L1 + delta * frac))
+                result.append([ts, round(level, 1)])
+
+    result.append([bb[-1][0], round(float(bb[-1][1]), 1)])
+
+    # Deduplicate and sort
+    seen, out = set(), []
+    for item in sorted(result, key=lambda x: x[0]):
+        if item[0] not in seen:
+            seen.add(item[0])
+            out.append(item)
+
+    return out if len(out) > 4 else None
+
+
 def _safe(fn, *args):
     try:
         return fn(*args)