diff --git a/src/interface.jl b/src/interface.jl
index 061bec0..6c09caa 100755
--- a/src/interface.jl
+++ b/src/interface.jl
@@ -351,12 +351,13 @@ function actor_mistral_openorca(a::agentReflex)
   prompt = 
   """
     <|im_start|>system
-    {role}
+    {context}
     {tools}
+    {role}  
     {thinkingFormat}
     <|im_end|>
-    {context}
     {shorttermMemory}
+    Thought $(a.step):
   """
   
   prompt = replace(prompt, "{role}" => a.roles[a.role])
@@ -392,7 +393,7 @@ function actor_mistral_openorca(a::agentReflex)
   # context = replace(context, "{earlierConversation}" => "My earlier talk with the user:\n$(a.earlierConversation)")
   context = replace(context, "{env state}" => "")
   context = replace(context, "{longterm memory}" => "")
-  context = replace(context, "{plan}" => "Plan:\n$(a.memory[:shortterm]["Plan 1:"])")
+  context = replace(context, "{plan}" => "My plan:\n$(a.memory[:shortterm]["Plan 1:"])")
   prompt = replace(prompt, "{context}" => context)
 
   return prompt
@@ -618,11 +619,12 @@ function work(a::agentReflex, usermsg::String)
     a.earlierConversation = conversationSummary(a)
     _ = addNewMessage(a, "user", usermsg)
     a.memory[:shortterm]["user:"] = usermsg
-    @show a.memory[:shortterm]
+    a.memory[:log]["user:"] = usermsg
   elseif a.thinkingmode == :continue_thinking #TODO
     error("continue_thinking $(@__LINE__)")
     _ = addNewMessage(a, "user", usermsg)
-    a.thought *= "Obs $(a.attempt): $usermsg\n"
+    a.memory[:shortterm]["Obs $(a.step):"] = usermsg
+    a.memory[:log]["Obs $(a.step):"] = usermsg
   else
     error("undefined condition thinkingmode = $thinkingmode $(@__LINE__)")
   end
@@ -641,13 +643,16 @@ function work(a::agentReflex, usermsg::String)
       respond = sendReceivePrompt(a, prompt)
 
       # sometimes LLM add not-need word I don't want
-      plan = split(respond, "<|im_end|>")[1]
-      plan = split(plan, "Response:")[1]
-      plan = split(plan, "Execution:")[1]
-      plan = split(plan, "Result:")[1]
+      # plan = split(respond, "<|im_end|>")[1]
+      # plan = split(plan, "Response:")[1]
+      # plan = split(plan, "Execution:")[1]
+      # plan = split(plan, "Result:")[1]
+      # plan = split(plan, "Recommendation:")[1]
+      plan = splitext(plan, ["<|im_end|>", "Response:", "Execution:", "Result:", "Recommendation:"])
 
       plan = replace(plan, "Plan:"=>"")
       a.memory[:shortterm]["Plan $(a.attempt):"] = plan
+      a.memory[:log]["Plan $(a.attempt):"] = plan
       actorstate, msgToUser = actor(a)
 
       if actorstate == "chatbox"
@@ -658,7 +663,8 @@ function work(a::agentReflex, usermsg::String)
 
         respond = formulateRespond(a)
         error("10")
-        a.memory[:shortterm]["Respond:"] = respond
+        a.memory[:shortterm]["Respond $(a.attempt):"] = respond
+        a.memory[:log]["Respond $(a.attempt):"] = respond
 
         # evaluate. if score < 8/10 try again.
         guideline = writeEvaluationGuideline(a, a.memory[:shortterm]["user:"])
@@ -668,11 +674,12 @@ function work(a::agentReflex, usermsg::String)
         if score >= 8 # good enough answer
           @show a.memory[:shortterm]
           a.memory[:shortterm] = OrderedDict{String, Any}()
+          a.memory[:log] = OrderedDict{String, Any}()
           break
         else  # self evaluate and reflect then try again
           analysis = analyze(a, a.memory[:shortterm])
           @show analysis
-
+          error(12)
           lessonwithcontext = selfReflext(a, analysis)
           @show lessonwithcontext
           a.memory[:shortterm] = OrderedDict{String, Any}()
@@ -784,10 +791,10 @@ function actor(a::agentReflex)
         @show toolresult
         a.memory[:shortterm]["Obs $(a.step):"] = toolresult
 
-        #WORKING goNogo
         go = goNogo(a)
-        @show go
-        error(11)
+        if go == "No"   # in case there is a cancel, go straight to evaluation
+          a.step = totalsteps
+        end
       end
     else  #TODO finish all steps
       actorState = "all steps done"
@@ -1099,7 +1106,7 @@ julia> shorttermMemory = OrderedDict{String, Any}(
 julia> decision = goNogo(agent)
 "Yes"
 ```
-""" #WORKING
+"""
 function goNogo(a)
   stimulus = a.memory[:shortterm]["user:"]
   work = ""
diff --git a/src/llmfunction.jl b/src/llmfunction.jl
index aef36eb..d752195 100644
--- a/src/llmfunction.jl
+++ b/src/llmfunction.jl
@@ -47,11 +47,11 @@ function wikisearch(a::agentReflex, phrase::T) where {T<:AbstractString}
     result = "No info available for your search query."
   end
 
-  if result == ""
-    result = "No info available for your search query."
-  else
-    result = makeSummary(a, result)
-  end
+  # if result == ""
+  #   result = "No info available for your search query."
+  # else
+  #   result = makeSummary(a, result)
+  # end
   
   return result
 end
@@ -106,7 +106,6 @@ end
 
 
 
-
 
 
   
diff --git a/src/type.jl b/src/type.jl
index 7f5252e..61c63ed 100644
--- a/src/type.jl
+++ b/src/type.jl
@@ -34,7 +34,6 @@ abstract type agent end
   # messages= [Dict(:role=>"system", :content=> "", :timestamp=> Dates.now()),]
   messages = Vector{Dict{Symbol, Any}}()
   tools::Union{Dict, Nothing} = nothing 
-  thoughtlog::String = ""  # logs unfinished thoughts
   attempt::Int = 0 # attempted number
   step::Int = 0 # step number
   attemptlimit::Int = 5 # thinking round limit
@@ -43,6 +42,7 @@ abstract type agent end
   memory::Dict = Dict(
     :shortterm=> OrderedDict{String, Any}(),
     :longterm=> OrderedDict{String, Any}(),
+    :log=> OrderedDict{String, Any}(),  # span from user stimulus -> multiples attempts -> final respond
   )
 end
 
@@ -100,7 +100,7 @@ function agentReflex(
     :actor=> 
     """
     Use the following format:
-    Thought: your should always think about how to do step {step} of the plan (pay attention to correct numeral calculation and commonsense).
+    Thought: think about how to do step {step} of the plan? (pay attention to correct numeral calculation and commonsense).
     Act: the action to take that match your thought, should be one of [{toolnames}]
     ActInput: the input to the action (pay attention to the tool's input)
     Obs: the result of the action
diff --git a/src/utils.jl b/src/utils.jl
index 5750e44..31a1c47 100644
--- a/src/utils.jl
+++ b/src/utils.jl
@@ -3,7 +3,7 @@ module utils
 export makeSummary, sendReceivePrompt, chunktext, extractStepFromPlan, checkTotalStepInPlan,
         detectCharacters, findDetectedCharacter, extract_number, toolNameBeingCalled,
         chooseThinkingMode, conversationSummary, checkReasonableness, addStepNumber,
-        addShortMem!
+        addShortMem!, splittext
 
 using UUIDs, Dates, DataStructures
 using CommUtils, GeneralUtils
@@ -27,8 +27,7 @@ function makeSummary(a::T1, input::T2) where {T1<:agent, T2<:AbstractString}
   prompt = replace(prompt, "{input}" => input)
   result = sendReceivePrompt(a, prompt)
   result = GeneralUtils.getStringBetweenCharacters(result, "{", "}")
-  println("///// ", result)
-  if result == "Yes"
+  if result == "Yes"  # seperate summary part 
     prompt = 
     """
       <|im_start|>system
@@ -44,6 +43,9 @@ function makeSummary(a::T1, input::T2) where {T1<:agent, T2<:AbstractString}
       summary = result
     end
   end
+  input_summary = input
+  @show input_summary
+  @show summary
 
   return summary
 end
@@ -327,7 +329,7 @@ end
 
 function chooseThinkingMode(a::agentReflex, usermsg::String)
   thinkingmode = nothing
-  if length(a.thoughtlog) != 0
+  if length(a.memory[:log]) != 0
     thinkingmode = :continue_thinking
   else
     prompt =
@@ -531,8 +533,29 @@ end
 
 
 
+""" Split text using all keywords in a list. Start spliting from rightmost of the text.
 
+Args:
+  text = a text you want to split
+  list = a list of keywords you want to split
 
+Return:
+  a leftmost text after split
+
+# Example
+```jldoctest
+julia> text = "Consider the type of food, occasion and temperature at the serving location."
+julia> list = ["at", "and"]
+"Consider the type of food, occasion "
+```
+"""
+function splittext(text, list)
+  newtext = text
+  for i in list
+    newtext = split(newtext, i)[1]
+  end
+  return newtext
+end