Keep bot operational when forecast model is unavailable

crypto_spot_bot/time_series.py

@@ -213,6 +213,7 @@ class TimeSeriesForecaster:
             probability=self.settings.time_series_min_probability_up,
             confidence=self.settings.time_series_min_confidence,
         )
+        symbol_eligible = _calibration_symbol_eligible(calibration, symbol)
         entry = _torch_recurrent_entry(symbol, artifact)
         model = _torch_recurrent_model_name(symbol, artifact)
         clip = _clamp(_float_entry(entry or {}, "clip", 8.0), 1.0, 50.0)
@@ -274,7 +275,8 @@ class TimeSeriesForecaster:
             )
             conservative_return_percent = min(expected_return_percent, q50_percent)
             block_entry = bool(
-                (expected_return_percent <= -min_edge and probability_up <= 0.45)
+                not symbol_eligible
+                or (expected_return_percent <= -min_edge and probability_up <= 0.45)
                 or (q50_percent <= -min_edge and probability_up <= 0.48)
             )
             reason = _reason(
@@ -284,6 +286,8 @@ class TimeSeriesForecaster:
                 skill=skill,
                 block_entry=block_entry,
             )
+            if not symbol_eligible:
+                reason = "symbol excluded by train-only calibration"
             return TimeSeriesForecast(
                 enabled=True,
                 usable=True,
@@ -344,7 +348,9 @@ class TimeSeriesForecaster:
             min_edge=min_edge,
             max_adjustment=self.settings.time_series_max_adjustment,
         )
-        block_entry = bool(expected_return_percent <= -min_edge and probability_up <= 0.45)
+        block_entry = bool(
+            not symbol_eligible or (expected_return_percent <= -min_edge and probability_up <= 0.45)
+        )
         reason = _reason(
             model=model,
             expected_return_percent=expected_return_percent,
@@ -352,6 +358,8 @@ class TimeSeriesForecaster:
             skill=skill,
             block_entry=block_entry,
         )
+        if not symbol_eligible:
+            reason = "symbol excluded by train-only calibration"
         return TimeSeriesForecast(
             enabled=True,
             usable=True,
@@ -496,6 +504,16 @@ def _calibrated_thresholds(
     }
 
 
+def _calibration_symbol_eligible(calibration: dict[str, Any], symbol: str | None) -> bool:
+    if not isinstance(calibration, dict) or "eligible_symbols" not in calibration:
+        return True
+    eligible = calibration.get("eligible_symbols")
+    if not isinstance(eligible, list) or not symbol:
+        return False
+    allowed = {str(value).strip().upper() for value in eligible if str(value).strip()}
+    return symbol.strip().upper() in allowed
+
+
 def _model_freshness(artifact: dict[str, Any], max_age_hours: float) -> tuple[str, float | None, bool]:
     raw = str(artifact.get("created_at", "")).strip() if isinstance(artifact, dict) else ""
     if not raw:
@@ -998,7 +1016,12 @@ def _torch_recurrent_entry(symbol: str | None, artifact: dict[str, Any]) -> dict
         entry = default if isinstance(default, dict) else None
     if not isinstance(entry, dict):
         return None
-    if not isinstance(entry.get("state_dict"), dict):
+    members = entry.get("ensemble_members")
+    has_member_state = isinstance(members, list) and any(
+        isinstance(member, dict) and isinstance(member.get("state_dict"), dict)
+        for member in members
+    )
+    if not isinstance(entry.get("state_dict"), dict) and not has_member_state:
         return None
     return entry