update

src/interface.jl

@@ -292,11 +292,14 @@ function planner_mistral_openorca(a::agentReflex)
     Objective: the objective you intend to do
     Plan: first you should always think about the objective, the info you need and the info you have thoroughly then extract and devise a step by step plan (pay attention to correct numeral calculation and commonsense).
     p.s.1 each step of the plan should be a single action.
+    p.s.2 make sure you gather all the required information.
     <|im_end|>
     $conversation
     <|im_start|>assistant
     Objective: 
     """
+    # p.s.2 the last step should be about responding.
+
     result = sendReceivePrompt(a, assistant_plan_prompt, max_tokens=1024, temperature=0.2)
     @show raw_plan = result
     x = split(result, "<|im_end|>")[1]
@@ -396,7 +399,6 @@ end
 #   return prompt
 # end
 
-
 function actor_mistral_openorca(a::agentReflex)
   """
     general prompt format:
@@ -430,20 +432,16 @@ function actor_mistral_openorca(a::agentReflex)
   prompt = 
   """
     <|im_start|>system
-    {role} 
+    $(a.roles[a.role])
     {tools}
-    {thinkingFormat}
+    $(a.thinkingFormat[:actor])
     {context}
     <|im_end|>
     {shorttermMemory}
     Thought $(a.step): 
   """
   
-  prompt = replace(prompt, "{role}" => a.roles[a.role])
-  prompt = replace(prompt, "{thinkingFormat}" => a.thinkingFormat[:actor])
-  prompt = replace(prompt, "{step}" => a.step)
-  
-  s = dictToString(a.memory[:shortterm], ["user:", "Plan 1:"])
+  s = dictToString(a.memory[:shortterm], skiplist=["user:", "Plan 1:"])
   prompt = replace(prompt, "{shorttermMemory}" => s)
 
   toolnames = ""
@@ -455,9 +453,13 @@ function actor_mistral_openorca(a::agentReflex)
   end
   prompt = replace(prompt, "{toolnames}" => toolnames)
   prompt = replace(prompt, "{tools}" => "You have access to the following tools:\n$toollines")
+
+  conversation = messagesToString_nomark(a.messages, addressAIas="I")
   
   context = 
   """
+  Your talk with the user:
+  $conversation
   {env state}
   {longterm memory}
   {plan}
@@ -471,6 +473,80 @@ function actor_mistral_openorca(a::agentReflex)
   return prompt
 end
 
+# function actor_mistral_openorca(a::agentReflex)
+#   """
+#     general prompt format:
+
+#     "
+#     <|im_start|>system
+#     {role}
+#     {tools}
+#     {thinkingFormat}
+#     <|im_end|>
+#     {context}
+#     <|im_start|>user
+#     {usermsg}
+#     <|im_end|>
+#     <|im_start|>assistant
+
+#     "
+
+#     Note:
+#       {context} = 
+#         "
+#         {earlierConversation}
+#         {env state}
+#         {shortterm memory}
+#         {longterm memory}
+#         "
+#   """
+
+#   mark = "$(a.step)"
+
+#   prompt = 
+#   """
+#     <|im_start|>system
+#     {role} 
+#     {tools}
+#     {thinkingFormat}
+#     {context}
+#     <|im_end|>
+#     {shorttermMemory}
+#     Thought $(a.step): 
+#   """
+  
+#   prompt = replace(prompt, "{role}" => a.roles[a.role])
+#   prompt = replace(prompt, "{thinkingFormat}" => a.thinkingFormat[:actor])
+#   prompt = replace(prompt, "{step}" => a.step)
+  
+#   s = dictToString(a.memory[:shortterm], skiplist=["user:", "Plan 1:"])
+#   prompt = replace(prompt, "{shorttermMemory}" => s)
+
+#   toolnames = ""
+#   toollines = ""
+#   for (toolname, v) in a.tools
+#     toolline = "$toolname: $(v[:description]) $(v[:input]) $(v[:output])\n"
+#     toollines *= toolline
+#     toolnames *= "$toolname, "
+#   end
+#   prompt = replace(prompt, "{toolnames}" => toolnames)
+#   prompt = replace(prompt, "{tools}" => "You have access to the following tools:\n$toollines")
+  
+#   context = 
+#   """
+#   {env state}
+#   {longterm memory}
+#   {plan}
+#   """
+#   # context = replace(context, "{earlierConversation}" => "My earlier talk with the user:\n$(a.earlierConversation)")
+#   context = replace(context, "{env state}" => "")
+#   context = replace(context, "{longterm memory}" => "")
+#   context = replace(context, "{plan}" => "My plan:\n$(a.memory[:shortterm]["Plan $(a.attempt):"])")
+#   prompt = replace(prompt, "{context}" => context)
+
+#   return prompt
+# end
+
 
 
 """
@@ -526,7 +602,7 @@ function conversation(a::agentReflex, usermsg::String; attemptlimit::Int=3)
   end
 
   # if LLM using chatbox, use returning msg form chatbox as conversation response
-  if workstate == "chatbox" ?
+  if workstate == "chatbox"
     #TODO paraphrase msg so that it is human friendlier word.
   else
     response = chat_mistral_openorca(a)
@@ -916,8 +992,6 @@ function actor(a::agentReflex)
         a.memory[:shortterm] = removeHeaders(a.memory[:shortterm], a.step, ["Plan"])
         a.memory[:log] = removeHeaders(a.memory[:log], a.step, ["Plan"])
         println("repeating step $(a.step)")
-
-      #WORKING
       else
         error("undefined condition decision = $decision $(@__LINE__)")
       end
@@ -1238,7 +1312,7 @@ julia> report = formulateUserresponse(agent, shorttermMemory)
 ```
 """
 function formulateUserresponse(a)
-  stimulus = a.memory[:shortterm]["user:"]
+  stimulus = a.memory[:shortterm]["user:"]  #BUG no "user" key
 
   work = dictToString(a.memory[:shortterm], ["user:"])
 
@@ -1294,34 +1368,10 @@ julia> shorttermMemory = OrderedDict{String, Any}(
 julia> decision = goNogo(agent)
 "Yes"
 ```
-"""
+""" #WORKING add choice skip, if LLM already completed the step
 function goNogo(a)
-  stimulus = a.memory[:shortterm]["user:"]
-  work = dictToString(a.memory[:shortterm], ["user:"])
-
-  # prompt =
-  # """
-  #   <|im_start|>system
-  #   Symbol meaning:
-  #   Stimulus: the input user gives to you and you must response
-  #   Plan: a plan
-  #   Thought: your thought
-  #   Act: the action you took
-  #   Actinput: the input to the action
-  #   Obs: the result of the action
-
-  #   Stimulus:
-  #   $stimulus
-
-  #   Your work:
-  #   $work
-    
-  #   From your work, you job is to decide what to do next by choosing one of the following choices:
-  #   If you are ready to do the next step of the plan say, "{Yes}". And what is the rationale behind the decision?
-  #   If you need to repeat the latest step say, "{No}". And what is the rationale behind the decision?
-  #   If you are ready to formulate a final response to user original stimulus say, {formulateUserresponse}. And what is the rationale behind the decision?
-  #   <|im_end|>
-  # """
+  # stimulus = a.memory[:shortterm]["user:"]
+  work = dictToString(a.memory[:shortterm])
 
   prompt =
   """
@@ -1334,24 +1384,18 @@ function goNogo(a)
     Actinput: the input to the action
     Obs: the result of the action
 
-    Stimulus:
-    $stimulus
-
     Your work:
     $work
     
     Your job is to check whether step $(a.step) of your work is completed according to the plan and choose only one of the following choices.
-    choice 1: If you are ready to do the next step of the plan say, "{Yes}". And what is the rationale behind the decision?
-    choice 2: If you need to repeat the latest step say, "{No}". And what is the rationale behind the decision?
+    choice 1: If you get what you intend to do and you are ready to do the next step of the plan say, "{Yes}". And what is the rationale behind the decision to do the next step?
+    choice 2: If you didn't get what you intend to do and you need to repeat the latest step say, "{No}". And what is the rationale behind the decision to repeat the latest step?
     <|im_end|>
     <|im_start|>assistant
 
   """
 
   response = sendReceivePrompt(a, prompt)
-  
-
-
   decision = nothing
   reason = nothing
   if occursin("Yes", response)