update

src/interface.jl

@@ -442,7 +442,7 @@ function conversation(a::T, userinput::Dict) where {T<:agent}
           )  
         )
       bestplan = runMCTS(a, initialState, decisionMaker, progressValueEstimator, reflector,
-                        isterminal, 2, 3, 2, 1.0)
+                        isterminal, 2, 3, 3, 1.0)
       error("---> bestplan")
 
       # actor loop(bestplan)

src/llmfunction.jl

@@ -101,7 +101,7 @@ function virtualWineCustomerReccommendbox(a::T1, input::T2)::String where {T1<:a
     )
   )
   @show outgoingMsg
-  result = GeneralUtils.sendReceiveMqttMsg(outgoingMsg)
+  result = GeneralUtils.sendReceiveMqttMsg(outgoingMsg; timeout=120)
   response = result[:response][:text]
 
   return response
@@ -157,7 +157,7 @@ function virtualWineCustomerChatbox(a::T1, input::T2)::String where {T1<:agent,
     )
   )
   @show outgoingMsg
-  result = GeneralUtils.sendReceiveMqttMsg(outgoingMsg)
+  result = GeneralUtils.sendReceiveMqttMsg(outgoingMsg; timeout=120)
   response = result[:response][:text]
 
   return response
@@ -291,7 +291,7 @@ function jsoncorrection(a::T1, input::T2,
               )
             )
           )
-        result = GeneralUtils.sendReceiveMqttMsg(outgoingMsg)
+        result = GeneralUtils.sendReceiveMqttMsg(outgoingMsg; timeout=120)
         incorrectjson = result[:response][:text]
       end
     else

src/mcts.jl

@@ -348,8 +348,9 @@ function selectChildNode(node::MCTSNode)::MCTSNode
 
   # loop thought node children dictionary to find the highest progress value
   for (k, childNode) in node.children
+    thisNodeProgressValue = childNode.statevalue + childNode.reward
     if childNode.statevalue > highestProgressValue
-      highestProgressValue = childNode.statevalue + childNode.reward
+      highestProgressValue = thisNodeProgressValue
       nodekey = childNode.nodekey
     end
   end
@@ -443,10 +444,10 @@ function runMCTS(
 
     expand(a, node, decisionMaker, progressValueEstimator, isterminal, n=n)
     
-    leaf_node = UCTselect(node, w)
-    simTrajectoryReward = simulate(a, leaf_node, decisionMaker, progressValueEstimator,
-                      isterminal, maxDepth, n=n)
-    backpropagate(leaf_node, simTrajectoryReward)
+    leafNode = UCTselect(node, w)
+    simTrajectoryReward = simulate(a, leafNode, decisionMaker, progressValueEstimator,
+                                  isterminal, maxDepth, n=n)
+    backpropagate(leafNode, simTrajectoryReward)
   end
 
   best_child_state = argmax([child.total_reward / child.visits for child in values(root.children)])