separate and improve conversation agent

backend/agent.py

@@ -75,92 +75,6 @@ class DebugAgent(Agent):
             thread_id=chat_view.current_thread_id,
         )
 
-class ConversationAgent(Agent):
-    identifier = "conversation"
-    schema_fields = []
-    tools = []
-
-    def __init__(self, app, data):
-        Agent.__init__(self, app, data)
-
-        # Constant system prompt for developing purposes 
-        self.system_prompt = "You are a friendly AI having a casual conversation with the user."
-
-    async def prompt(self, content):
-        # TODO: Advanced features
-        messages = [
-            LlamaMessage("system", self.system_prompt),
-            LlamaMessage("user", content)
-        ]
-        response = await replicate_llama_completion(self.app, messages)
-        return response
-
-    async def received_message(self, message):
-
-        # TODO: before retrieving the chat context, acquire a lock for it
-
-        saved_context = ThreadContext(self.app.get_agent_resource(self.id, 'last-thread-context', message.chat_id))
-        chat_view = self.app.get_chat_view(message.chat_id)
-        print("ChatView:\n" + chat_view.debug_string())
-
-        if message.thread_id != saved_context.current_thread_id:
-            # the incoming message is not from the same thread that the agent was last aware of
-            # store details about the old thread in the agent's long term memory
-            old_thread = chat_view.get_thread(saved_context.current_thread_id)
-            await self.process_thread(old_thread)
-
-            # TODO: This is a good place to store the new context, its fine that the agent hasnt responded
-
-        current_thread = chat_view.get_thread(message.thread_id)
-
-        # REMOVE THIS. Only for testing, while the function only prints...
-        await self.process_thread(current_thread)
-
-        chat_summary = self.serialize_thread_history(current_thread)
-        messages = [
-            LlamaMessage("system", f"{self.system_prompt}\nYou are given the context of the conversation, respond with your next message in the conversation."),
-            LlamaMessage("user", f"Conversation History:\n{chat_summary}")
-        ]
-
-        response = await replicate_llama_completion(self.app, messages)
-        self.app.send_chat_message(
-            chat_id=message.chat_id,
-            sender=self.id,
-            content=response,
-            thread_id=current_thread.id,
-        )
-
-        # TODO: drop the chat context lock
-
-    # Encode a thread's messages as a string to pass to the agent
-    def serialize_thread_history(self, thread):
-        def sender_name(message):
-            sender = message.sender
-            is_digit = isinstance(sender, (int, str)) and str(sender).isdigit()
-            is_self = is_digit and str(sender) == str(self.id)
-            return self.name if is_self else sender
-        
-        return "\n".join(f"[{sender_name(message)}]: {message.content}" for message in thread.messages)
-    
-    # Compute thread summnary and concrete memories, then save the progress done
-    async def process_thread(self, thread):
-        # TODO: extract memories from the thread messages, then store them in the agent's vector memory db
-        # TODO: somewhere here it must be known if the thread was seen before: the same memory should not be stored multiple times
-        print(f"PROCESSING THREAD: {thread.id}")
-
-        instruction = "Your task is to identify facts from a conversation that its participants would remember in the long term. Each line in your response is a different memory, if there's nothing relevant to remember, simply give an empty response."
-        intro = f"Identify the facts that {self.name} will remember long term after the conversation. Following the line break is the personality prompt for {self.name}."
-
-        chat_history = self.serialize_thread_history(thread)
-        messages = [
-            LlamaMessage("system", f"{instruction}\n{intro}\n\n{self.system_prompt}"),
-            LlamaMessage("user", f"Conversation History:\n{chat_history}")
-        ]
-        response = await replicate_llama_completion(self.app, messages)
-        memories = [line for line in response.split('\n') if line.strip()]
-        
-        print("New agent memories:", memories)
-
 class ToolUsingAgent:
     identifier = "tool-using"
     schema_fields = [
@@ -353,14 +267,6 @@ class CoderV1Agent(Llama8bAgent):
     async def gral_update_file(self, message_history: list, filepath: str, file_content: str):
         return await gral.gral_update_file(self.app, self.config['system_prompt'], message_history, filepath, file_content)
 
-# Used by ConversationAgent to keep track of the last chat thread it participated in
-class ThreadContext:
-    def __init__(self, data):
-        if data:
-            self.current_thread_id = data.get('current_thread_id')
-        else:
-            self.current_thread_id = None
-
 class ParsedResponse:
     def __init__(self):
         self.response = None