Improve PPO diagnostics and recharge behavior

agent_ppo/conf/monitor_builder.py

@@ -125,6 +125,10 @@ def build_monitor():
             metrics_name="recharge_escape_count",
             expr="avg(recharge_escape_count{})",
         )
+        .add_metric(
+            metrics_name="recharge_steps",
+            expr="avg(recharge_steps{})",
+        )
         .end_panel()
         .add_panel(
             name="NPC危险接近",
@@ -172,6 +176,42 @@ def build_monitor():
             expr="avg(remaining_charge{})",
         )
         .end_panel()
+        .add_panel(
+            name="动作掩码健康",
+            name_en="mask_health",
+            type="line",
+        )
+        .add_metric(
+            metrics_name="mask_final_avg",
+            expr="avg(mask_final_avg{})",
+        )
+        .add_metric(
+            metrics_name="mask_one_action_steps",
+            expr="avg(mask_one_action_steps{})",
+        )
+        .add_metric(
+            metrics_name="mask_two_or_less_action_steps",
+            expr="avg(mask_two_or_less_action_steps{})",
+        )
+        .add_metric(
+            metrics_name="mask_zero_final_steps",
+            expr="avg(mask_zero_final_steps{})",
+        )
+        .end_panel()
+        .add_panel(
+            name="回充动作掩码",
+            name_en="recharge_mask",
+            type="line",
+        )
+        .add_metric(
+            metrics_name="mask_recharge_active",
+            expr="avg(mask_recharge_active{})",
+        )
+        .add_metric(
+            metrics_name="mask_recharge_changed",
+            expr="avg(mask_recharge_changed{})",
+        )
+        .end_panel()
         .end_group()
         .build()
     )