add(agents): loggers + system prompt in base agent + filtering AI Messages in human resume

pmcp/agents/executor.py

@@ -5,8 +5,8 @@ from langchain_core.tools import BaseTool
 from langchain_openai import ChatOpenAI
 from langchain_core.messages import HumanMessage
 
-from pmcp.models.plan import PlanStep
 from pmcp.models.state import PlanningState
+from loguru import logger
 
 
 class ExecutorAgent:
@@ -18,16 +18,21 @@ class ExecutorAgent:
         )
 
     def call_executor_agent(self, state: PlanningState):
-        current_step = state.plan_step
-        plan_step: PlanStep = state.plan.steps[current_step]
-        return {
-            "plan_step": current_step + 1,
-            "messages": [
+        plan_step_index = state.plan_step
+        current_step = None
+        messages = []
+        if len(state.plan.steps) > plan_step_index:
+            current_step = state.plan.steps[plan_step_index]
+            messages = [
                 HumanMessage(
-                    content=f"The {plan_step.agent} agent should perform the following action:\n{plan_step.description}"
+                    content=f"The {current_step.agent} agent should perform the following action:\n{current_step.description}"
                 )
-            ],
-            "current_step": plan_step,
+            ]
+        logger.info(f"The Executor is executing step: {current_step}")
+        return {
+            "plan_step": plan_step_index + 1,
+            "messages": messages,
+            "current_step": current_step,
         }
 
     async def acall_executor_agent(self, state: PlanningState):
@@ -36,10 +41,12 @@ class ExecutorAgent:
         messages = []
         if len(state.plan.steps) > plan_step_index:
             current_step = state.plan.steps[plan_step_index]
-            messages = [HumanMessage(
+            messages = [
+                HumanMessage(
                     content=f"The {current_step.agent} agent should perform the following action:\n{current_step.description}"
-                )]
-                
+                )
+            ]
+        logger.info(f"The Executor is executing step: {current_step}")
         return {
             "plan_step": plan_step_index + 1,
             "messages": messages,

pmcp/agents/github_agent.py

@@ -2,10 +2,10 @@ from typing import List, Optional
 
 from pmcp.agents.agent_base import AgentBlueprint
 from langchain_core.tools import BaseTool
-from langchain_core.messages import SystemMessage
 from langchain_openai import ChatOpenAI
 
 from pmcp.models.state import PlanningState
+from loguru import logger
 
 SYSTEM_PROMPT = """
 You are an assistant that can manage Trello boards and projects.
@@ -24,13 +24,11 @@ class GithubAgent:
         )
 
     def call_github_agent(self, state: PlanningState):
-        response = self.agent.call_agent(
-            [SystemMessage(content=self.agent.system_prompt)] + state.messages
-        )
+        logger.info("Calling Github agent...")
+        response = self.agent.call_agent(state.messages)
         return {"messages": [response]}
 
     async def acall_github_agent(self, state: PlanningState):
-        response = await self.agent.acall_agent(
-            [SystemMessage(content=self.agent.system_prompt)] + state.messages
-        )
+        logger.info("Calling Github agent...")
+        response = await self.agent.acall_agent(state.messages)
         return {"messages": [response]}

pmcp/agents/planner.py

@@ -2,11 +2,11 @@ from typing import List, Optional
 
 from pmcp.agents.agent_base import AgentBlueprint
 from langchain_core.tools import BaseTool
-from langchain_core.messages import SystemMessage
 from langchain_openai import ChatOpenAI
 
 from pmcp.models.plan import Plan
 from pmcp.models.state import PlanningState
+from loguru import logger
 
 SYSTEM_PROMPT = """
 You are a Planner Agent responsible for breaking down high-level project goals into clear, actionable steps. You do not execute tasks yourself — instead, you delegate them to two specialized agents:
@@ -38,15 +38,19 @@ class PlannerAgent:
         )
 
     def call_planner_agent(self, state: PlanningState):
+        logger.info("Calling Planner agent...")
         response = self.agent.call_agent_structured(
-            messages=[SystemMessage(content=self.agent.system_prompt)] + state.messages,
+            messages=state.messages,
             clazz=Plan,
         )
+        logger.info(f"Building plan: {response}")
         return {"plan": response, "plan_step": 0, "current_step": None}
 
     async def acall_planner_agent(self, state: PlanningState):
+        logger.info("Calling Planner agent...")
         response = await self.agent.acall_agent_structured(
-            messages=[SystemMessage(content=self.agent.system_prompt)] + state.messages,
+            messages=state.messages,
             clazz=Plan,
         )
+        logger.info(f"Building plan: {response}")
         return {"plan": response, "plan_step": 0, "current_step": None}

pmcp/agents/trello_agent.py

@@ -2,10 +2,11 @@ from typing import List
 
 from pmcp.agents.agent_base import AgentBlueprint
 from langchain_core.tools import BaseTool
-from langchain_core.messages import SystemMessage
 from langchain_openai import ChatOpenAI
 
 from pmcp.models.state import PlanningState
+from loguru import logger
+
 
 SYSTEM_PROMPT = """
 You are an assistant that can manage Trello boards and projects.
@@ -24,13 +25,11 @@ class TrelloAgent:
         )
 
     def call_trello_agent(self, state: PlanningState):
-        response = self.agent.call_agent(
-            [SystemMessage(content=self.agent.system_prompt)] + state.messages
-        )
+        logger.info("Calling Trello Agent...")
+        response = self.agent.call_agent(state.messages)
         return {"messages": [response]}
 
     async def acall_trello_agent(self, state: PlanningState):
-        response = await self.agent.acall_agent(
-            [SystemMessage(content=self.agent.system_prompt)] + state.messages
-        )
+        logger.info("Calling Trello Agent...")
+        response = await self.agent.acall_agent(state.messages)
         return {"messages": [response]}

pmcp/nodes/human_interrupt_node.py

@@ -2,9 +2,10 @@ from typing import List, Optional
 
 from pmcp.agents.agent_base import AgentBlueprint
 from langchain_core.tools import BaseTool
-from langchain_core.messages import SystemMessage, AIMessage
+from langchain_core.messages import AIMessage
 from langchain_openai import ChatOpenAI
 from langgraph.types import Command, interrupt
+from langchain_core.messages.utils import filter_messages
 
 from pmcp.models.state import PlanningState
 
@@ -27,28 +28,31 @@ class HumanInterruptNode:
         )
 
     def call_human_interrupt_agent(self, state: PlanningState):
-        last_message = state.messages[-1]
-    
-        #TODO: chiedere a Gerlax lo strumento
+        last_message = filter_messages(state.messages, include_types=[AIMessage])[-1]
+
+        # TODO: chiedere a Gerlax lo strumento
         try:
             tool_call = last_message.tool_calls[-1]
         except Exception:
-            last_message = state.messages[-2]
+            last_message = filter_messages(state.messages, include_types=[AIMessage])[
+                -2
+            ]
             tool_call = last_message.tool_calls[-1]
 
-
-
         if tool_call.get("name", "").startswith("get_"):
-            return Command(goto="tool")            
-        
+            return Command(goto="tool")
+
         response = self.agent.call_agent(
-            messages=[SystemMessage(content=self.agent.system_prompt), AIMessage(content= f"Tool Calling details: {str(tool_call)}")] + state.messages,
+            messages=[
+                AIMessage(content=f"Tool Calling details: {str(tool_call)}"),
+            ]
+            + state.messages,
         )
         human_review = interrupt(response.content)
 
         confirm_action = human_review.confirm_action
         changes_description = human_review.changes_description
-        
+
         if confirm_action:
             return Command(goto="tool")
 

pmcp/nodes/human_resume_node.py

@@ -2,11 +2,12 @@ from typing import List, Optional
 
 from pmcp.agents.agent_base import AgentBlueprint
 from langchain_core.tools import BaseTool
-from langchain_core.messages import SystemMessage, HumanMessage
+from langchain_core.messages import HumanMessage
 from langchain_openai import ChatOpenAI
 from langgraph.types import Command
 
 from pmcp.models.resume_trigger import ResumeTrigger
+from loguru import logger
 
 
 SYSTEM_PROMPT = """
@@ -29,8 +30,11 @@ class HumanResumeNode:
         )
 
     def call_human_interrupt_agent(self, user_message: str):
+        logger.info("Human resumer agent...")
         response = self.agent.call_agent_structured(
-            [SystemMessage(content=self.agent.system_prompt), HumanMessage(content= user_message)],
+            [
+                HumanMessage(content=user_message),
+            ],
             clazz=ResumeTrigger,
         )