diff --git a/src/interface.jl b/src/interface.jl
index 6dbceee..573051e 100755
--- a/src/interface.jl
+++ b/src/interface.jl
@@ -3,7 +3,8 @@ module interface
 
 export agentReact, agentReflex,
         addNewMessage, clearMessage, removeLatestMsg, conversation, writeEvaluationGuideline,
-        grading, analyze, selfReflext, actor_mistral_openorca2, formulateUserRespond
+        grading, analyze, selfReflext, actor_mistral_openorca2, formulateUserRespond,
+        extractinfo, updateEnvState
 
 using JSON3, DataStructures, Dates, UUIDs, HTTP
 using CommUtils, GeneralUtils
@@ -453,25 +454,24 @@ function conversation(a::agentReflex, usermsg::String; attemptlimit::Int=3)
   respond = nothing
 
   a.earlierConversation = conversationSummary(a)
+  _ = addNewMessage(a, "user", usermsg)
+  isusetools = isUseTools(a, usermsg)
 
-  # determine thinking mode
-  a.thinkingmode = chooseThinkingMode(a, usermsg)
-  @show a.thinkingmode
-
-  if a.thinkingmode == :no_thinking
-    _ = addNewMessage(a, "user", usermsg)
-    prompt = chat_mistral_openorca(a, usermsg)
-    println("")
-    @show prompt
-    respond = sendReceivePrompt(a, prompt)
-    respond = split(respond, "<|im_end|>")[1]
-    respond = replace(respond, "\n" => "")
-    _ = addNewMessage(a, "assistant", respond)
-    println("")
-    @show respond
-  else
+  #WORKING
+  if isusetools # use tools before responding
     respond = work(a, usermsg)
   end
+  
+  #WORKING
+  prompt = chat_mistral_openorca(a, usermsg)
+  println("")
+  @show prompt
+  respond = sendReceivePrompt(a, prompt)
+  respond = split(respond, "<|im_end|>")[1]
+  respond = replace(respond, "\n" => "")
+  _ = addNewMessage(a, "assistant", respond)
+  println("")
+  @show respond
 
   return respond
 end
@@ -1110,13 +1110,68 @@ function goNogo(a)
 end
 
 
+#WORKING
+""" Determine whether LLM should go to next step.
+
+Args:
+  a, one of ChatAgent's agent.
+
+Return:
+  "Yes" or "no" decision to go next step.
+
+# Example
+```jldoctest
+julia> using ChatAgent, CommUtils
+julia> agent = ChatAgent.agentReflex("Jene")
+julia> shorttermMemory = OrderedDict{String, Any}(
+  "user:"        => "What's the latest AMD GPU?",
+  "Plan 1:"     => "   To answer this question, I will need to search for the latest AMD GPU using the wikisearch tool.\n",
+  "Act 1:"      => " wikisearch\n",
+  "Actinput 1:" => " amd gpu latest\n",
+  "Obs 1:"      => "No info available for your search query.",
+  "Act 2:"      => "  wikisearch\n",
+  "Actinput 2:" => "  amd graphics card latest\n",
+  "Obs 2:"      => "No info available for your search query.")
+
+julia> decision = goNogo(agent)
+"Yes"
+```
+"""
+function extractinfo(a, msg)
+  
+  prompt =
+  """
+    <|im_start|>system
+    User message:
+    $msg
+    
+    Your job is to extract important info from user's message into keys and values using this format: key=value ,.
+    p.s.1 you can extract many key-value pairs.
+    <|im_end|>
+
+    """
+
+  respond = sendReceivePrompt(a, prompt, temperature=0.0)
+end
 
 
+function updateEnvState(a, currentinfo, newinfo)
+  prompt =
+  """
+    <|im_start|>system
+    Current state:
+    $currentinfo
 
+    New info:
+    $newinfo
+    
+    Your job is to update or add information from new info into the current state which use key-value format.
+    <|im_end|>
+    Updated Current State:\n
+  """
 
-
-
-
+  respond = sendReceivePrompt(a, prompt, temperature=0.0)
+end
 
 
 
diff --git a/src/type.jl b/src/type.jl
index 47bd076..f3d3be7 100644
--- a/src/type.jl
+++ b/src/type.jl
@@ -38,7 +38,7 @@ abstract type agent end
   attemptlimit::Int = 5 # thinking round limit
   attempt::Int = 0 # attempted number
   step::Int = 0 # step number
-  thinkingmode::Symbol = :no_thinking
+  env::Dict = Dict()
   thinkingFormat::Union{Dict, Nothing} = nothing
   memory::Dict = Dict(
     :shortterm=> OrderedDict{String, Any}(),
diff --git a/src/utils.jl b/src/utils.jl
index 21a926a..29686fb 100644
--- a/src/utils.jl
+++ b/src/utils.jl
@@ -2,7 +2,7 @@ module utils
 
 export makeSummary, sendReceivePrompt, chunktext, extractStepFromPlan, checkTotalStepInPlan,
         detectCharacters, findDetectedCharacter, extract_number, toolNameBeingCalled,
-        chooseThinkingMode, conversationSummary, checkReasonableness, replaceHeaders,
+        isUseTools, conversationSummary, checkReasonableness, replaceHeaders,
         addShortMem!, splittext, dictToString, removeHeaders, keepOnlyKeys, experience
 
 using UUIDs, Dates, DataStructures
@@ -89,12 +89,14 @@ end
                                         )
   ```
 """
-function sendReceivePrompt(a::T, prompt::String; max_tokens=256, timeout::Int=120) where {T<:agent}
+function sendReceivePrompt(a::T, prompt::String; max_tokens=256, timeout::Int=120,  
+                            temperature::AbstractFloat=0.2) where {T<:agent}
   a.msgMeta[:msgId] = "$(uuid4())"  # new msg id for each msg
   msg = Dict(
     :msgMeta=> a.msgMeta,
     :txt=> prompt,
-    :max_tokens=>max_tokens
+    :max_tokens=> max_tokens,
+    :temperature=> temperature,
   )
   payloadChannel = Channel(1)
 
@@ -340,52 +342,6 @@ function toolNameBeingCalled(text::T, tools::Dict) where {T<:AbstractString}
   return toolNameBeingCalled
 end
 
-
-
-function chooseThinkingMode(a::agentReflex, usermsg::String)
-  thinkingmode = nothing
-  if length(a.memory[:log]) != 0
-    thinkingmode = :continue_thinking
-  else
-    prompt =
-    """
-    <|im_start|>system
-    {systemMsg}
-    You always use tools if there is a chance to impove your respond.
-    You have access to the following tools:
-    {tools}
-    Your job is to determine whether you will use tools or actions to respond.
-
-    Choose one of the following choices:
-      Choice 1: If the user want to get wine say, "{yes}".
-      Choice 2: If you don't need to use tools or actions to respond to the stimulus say, "{no}".
-    <|im_end|>
-
-    <|im_start|>user
-    {input}
-    <|im_end|>
-    <|im_start|>assistant
-
-    """
-    toollines = ""
-    for (toolname, v) in a.tools
-      if toolname ∉ ["chatbox", "nothing"]
-        toolline = "$toolname: $(v[:description]) $(v[:input]) $(v[:output])\n"
-        toollines *= toolline
-      end
-    end
-    prompt = replace(prompt, "{systemMsg}" => a.roles[a.role])
-    prompt = replace(prompt, "{tools}" => toollines)
-    prompt = replace(prompt, "{input}" => usermsg)
-    result = sendReceivePrompt(a, prompt)
-    willusetools = GeneralUtils.getStringBetweenCharacters(result, "{", "}")
-    thinkingmode = willusetools == "yes" ? :new_thinking : :no_thinking
-  end
-
-  return thinkingmode 
-end
-
-
 # function chooseThinkingMode(a::agentReflex, usermsg::String)
 #   thinkingmode = nothing
 #   if length(a.memory[:log]) != 0
@@ -429,32 +385,26 @@ end
 #   return thinkingmode 
 # end
 
-function chooseThinkingMode(a::agentReflex, usermsg::String)
-  thinkingmode = nothing
-  if length(a.memory[:log]) != 0
-    thinkingmode = :continue_thinking
-  else
-    prompt =
+function isUseTools(a::agentReflex, usermsg::String)
+  prompt =
     """
     <|im_start|>system
     {systemMsg}
-    You always use tools if there is a chance to impove your respond.
     You have access to the following tools:
     {tools}
 
     User message:
     {input}
 
-    Your job is to determine the following topics:
-      Topic 1: Is the user message show that they would like to get some wine? {Yes/No}
-      Topic 2: Is the user message show that they would like to get some cloth? {Yes/No}
+    Your job is to answer the following questions:
+      Question 1: From the user's message, Do you need to any tools before responding? Answer: {Yes/No/Not sure}. What will the you do?
     <|im_end|>
     <|im_start|>assistant
-    
+    Answer: 
     """
     toollines = ""
     for (toolname, v) in a.tools
-      if toolname ∉ ["chatbox", "nothing"]
+      if toolname ∉ [:chatbox]
         toolline = "$toolname: $(v[:description]) $(v[:input]) $(v[:output])\n"
         toollines *= toolline
       end
@@ -462,19 +412,17 @@ function chooseThinkingMode(a::agentReflex, usermsg::String)
     prompt = replace(prompt, "{systemMsg}" => a.roles[a.role])
     prompt = replace(prompt, "{tools}" => toollines)
     prompt = replace(prompt, "{input}" => usermsg)
-    result = sendReceivePrompt(a, prompt)
-    headers = detectCharacters(lessonwithcontext, ["Topic 1:", "Topic 2:"])
-    chunkedtext = chunktext(result, headers)
-    @show chunkedtext
-    error(11)
-    willusetools = GeneralUtils.getStringBetweenCharacters(result, "{", "}")
-    thinkingmode = willusetools == "Yes" ? :new_thinking : :no_thinking
-  end
+    
+    result = sendReceivePrompt(a, prompt, temperature=0.2)
 
-  @show result
-  error(11)
+    # headers = detectCharacters(result, ["Question 1:", "Question 2:"])
+    # chunkedtext = chunktext(result, headers)
 
-  return thinkingmode 
+    if occursin("Yes", result)
+      return true
+    else
+      return false
+    end
 end