Disable writing data during learning process

assume/common/outputs.py

@@ -373,7 +373,6 @@ async def on_stop(self):
                 with self.db.begin() as db:
                     df.to_sql("kpis", self.db, if_exists="append", index=None)
         except ProgrammingError as e:
-            # self.db.rollback() not working, no rollback function
             logger.error(f"No scenario run Yet {e}")
 
     def learning_queries(self):

assume/common/units_operator.py

@@ -86,7 +86,7 @@ async def add_unit(
         """
         self.units[unit.id] = unit
 
-        if not self.context.data_dict.get("learning_mode"):
+        if self.context.data_dict.get("learning_agent_addr") is None:
             db_aid = self.context.data_dict.get("output_agent_id")
             db_addr = self.context.data_dict.get("output_agent_addr")
             if db_aid and db_addr:
@@ -236,7 +236,7 @@ def write_actual_dispatch(self):
             filter(lambda x: x["end_time"] >= now, self.valid_orders)
         )
 
-        if not self.context.data_dict.get("learning_mode"):
+        if self.context.data_dict.get("learning_agent_addr") is None:
             db_aid = self.context.data_dict.get("output_agent_id")
             db_addr = self.context.data_dict.get("output_agent_addr")
             if db_aid and db_addr:
@@ -397,7 +397,7 @@ def write_learning_to_output(self, start: datetime, marketconfig: MarketConfig):
 
                 output_agent_list.append(output_dict)
 
-        if self.context.data_dict.get("learning_mode"):
+        if self.context.data_dict.get("learning_agent_addr"):
             db_aid = self.context.data_dict.get("output_agent_id")
             db_addr = self.context.data_dict.get("output_agent_addr")
             if db_aid and db_addr:
@@ -420,9 +420,6 @@ def write_to_learning(
         device: str,
         learning_unit_count: int,
     ):
-        learning_role_id = "learning_agent"
-        learning_role_addr = self.context.addr
-
         all_observations = []
         all_rewards = []
         try:
@@ -453,6 +450,9 @@ def write_to_learning(
         all_rewards = np.array(all_rewards)
         rl_agent_data = (np.array(all_observations), all_actions, all_rewards)
 
+        learning_role_id = self.context.data_dict.get("learning_agent_id")
+        learning_role_addr = self.context.data_dict.get("learning_agent_addr")
+
         self.context.schedule_instant_acl_message(
             receiver_id=learning_role_id,
             receiver_addr=learning_role_addr,

assume/markets/base_market.py

@@ -406,19 +406,17 @@ async def store_order_book(self, orderbook: Orderbook):
         :param orderbook: The order book to be stored
         :type orderbook: Orderbook
         """
-        if self.context.data_dict.get("learning_mode"):
-            return
-
-        message = {
-            "context": "write_results",
-            "type": "store_order_book",
-            "sender": self.marketconfig.name,
-            "data": orderbook,
-        }
 
         db_aid = self.context.data_dict.get("output_agent_id")
         db_addr = self.context.data_dict.get("output_agent_addr")
+
         if db_aid and db_addr:
+            message = {
+                "context": "write_results",
+                "type": "store_order_book",
+                "sender": self.marketconfig.name,
+                "data": orderbook,
+            }
             await self.context.send_acl_message(
                 receiver_id=db_aid,
                 receiver_addr=db_addr,
@@ -434,18 +432,16 @@ async def store_market_results(self, market_meta):
         :type market_meta: any
         """
 
-        if self.context.data_dict.get("learning_mode"):
-            return
-
-        message = {
-            "context": "write_results",
-            "type": "store_market_results",
-            "sender": self.marketconfig.name,
-            "data": market_meta,
-        }
         db_aid = self.context.data_dict.get("output_agent_id")
         db_addr = self.context.data_dict.get("output_agent_addr")
+
         if db_aid and db_addr:
+            message = {
+                "context": "write_results",
+                "type": "store_market_results",
+                "sender": self.marketconfig.name,
+                "data": market_meta,
+            }
             await self.context.send_acl_message(
                 receiver_id=db_aid,
                 receiver_addr=db_addr,

assume/world.py

@@ -148,7 +148,10 @@
             # if self.same_process:
             # separate process does not support buffer and learning
             if True:
-                rl_agent = RoleAgent(self.container, suggested_aid="learning_agent")
+                self.learning_agent_addr = (self.addr, "learning_agent")
+                rl_agent = RoleAgent(
+                    self.container, suggested_aid=self.learning_agent_addr[1]
+                )
                 rl_agent.add_role(self.learning_role)
             else:
 
@@ -213,7 +216,12 @@
         unit_operator_agent._role_context.data_dict = {
             "output_agent_addr": self.output_agent_addr[0],
             "output_agent_id": self.output_agent_addr[1],
-            "learning_mode": self.learning_mode,
+            "learning_agent_addr": self.learning_agent_addr[0]
+            if self.learning_mode
+            else None,
+            "learning_agent_id": self.learning_agent_addr[1]
+            if self.learning_mode
+            else None,
         }
 
     async def async_add_unit(
@@ -304,9 +312,12 @@
 
         # after creation of an agent - we set additional context params
         market_operator_agent._role_context.data_dict = {
-            "output_agent_addr": self.output_agent_addr[0],
-            "output_agent_id": self.output_agent_addr[1],
-            "learning_mode": self.learning_mode,
+            "output_agent_addr": None
+            if self.learning_mode
+            else self.output_agent_addr[0],
+            "output_agent_id": None
+            if self.learning_mode
+            else self.output_agent_addr[1],
         }
         self.market_operators[id] = market_operator_agent