diff --git a/src/interface.jl b/src/interface.jl
index 6f2494c..322597d 100755
--- a/src/interface.jl
+++ b/src/interface.jl
@@ -2,7 +2,8 @@ module interface
     
 
 export agentReact, addNewMessage, clearMessage, removeLatestMsg, generatePrompt_tokenPrefix,
-        generatePrompt_tokenSuffix, conversation, work, detectCharacters, chunktext
+        generatePrompt_tokenSuffix, conversation, work, detectCharacters, chunktext,
+        findDetectedCharacter
 
 using JSON3, DataStructures, Dates, UUIDs
 using CommUtils, GeneralUtils
@@ -130,6 +131,12 @@ function agentReact(
       :input => "Input should be a search query with as much details as possible.",
       :output => "" ,
     ),
+    :nothing=>Dict(
+      :name => "nothing",
+      :description => "useful for when you don't need to use tools or actions",
+      :input => "No input is needed",
+      :output => "" ,
+    ),
   ),
   msgMeta::Dict=Dict(
     :msgPurpose=> "updateStatus",
@@ -359,7 +366,7 @@ function generatePrompt_react_mistral_openorca(a::T, usermsg::String,
   return prompt
 end
 
-#WORKING
+
 function conversation(a::T, usermsg::String) where {T<:agent}
   userintend = identifyUserIntention(a, usermsg)
   @show userintend
@@ -367,7 +374,7 @@ function conversation(a::T, usermsg::String) where {T<:agent}
 
   # AI thinking mode
   if userintend == "chat" 
-    a.context = conversationSummary(a)  #BUG should be long conversation before use summary because it leaves out details
+    a.context = conversationSummary(a)  #TODO should be long conversation before use summary because it leaves out details
     _ = addNewMessage(a, "user", usermsg)
     prompt = generatePrompt_mistral_openorca(a, usermsg)
     @show prompt
@@ -421,7 +428,7 @@ function conversation(a::T, usermsg::String) where {T<:agent}
   return respond
 end
 
-#WORKING
+
 function work(a::T, prompt::String) where {T<:agent}
   respond = nothing
   while true
@@ -430,10 +437,14 @@ function work(a::T, prompt::String) where {T<:agent}
 
     respond = sendReceivePrompt(a, prompt)
     @show respond
+    try
+      respond = split(respond, "Obs:")[1]  
+    catch
+    end
+    
     headers = detectCharacters(respond, 
       ["QTS:", "Plan:", "Thought:", "Act:", "ActInput:", "Obs:", ".....", "ANS:"])
     @show headers
-    respond = split(respond, "Obs:")[1]
     chunkedtext = chunktext(respond, headers)
     @show chunkedtext
 
@@ -443,9 +454,8 @@ function work(a::T, prompt::String) where {T<:agent}
       break
     else
       # check for tool being called
-      ActInd = findDetectedCharacter(headers, "Act:")
-      println(">>> ", chunkedtext[ActInd][:body])
-      # toolname = toolNameBeingCalled(chunkedtext[ActInd][:body])
+      ActInd = findDetectedCharacter(headers, "Act:")[1]
+      toolname = toolNameBeingCalled(chunkedtext[ActInd][:body], )  #WORKING
 
       # function call