made chnages in graph and tools

app/agents/agents.py

@@ -2,6 +2,10 @@ from langchain_groq import ChatGroq
 from app.schemas.pydanticschema import ResumeExtract,JobDescriptionExtract,SkillGapAnalysis
 from app.core.config import settings
 from app.tools.tools import roadmap_planner_agent_tools
+from app.prompts.roadmap_planner_agent_prompt import roadmap_planner_agent_prompt
+from typing import Any
+from langchain.agents import create_agent
+from langchain.agents.middleware import ToolCallLimitMiddleware
 import os
 
 if "GROQ_API_KEY" not in os.environ:
@@ -49,11 +53,15 @@ gap_analysis_agent=gap_analysis_agent.with_structured_output(
     strict=True
 )
 
-
-
-roadmap_planner_agent=ChatGroq(
-    model="qwen/qwen3-32b",
-    temperature=0.2,
+roadmap_planner_agent = create_agent(
+    model="qwen/qwen3-32b", 
+    tools=roadmap_planner_agent_tools,
+    system_prompt=roadmap_planner_agent_prompt,
+    middleware=[
+        ToolCallLimitMiddleware[Any, None](
+            tool_name="search_courses",
+            run_limit=4,
+            thread_limit=10,
+        )
+    ],
 )
-
-roadmap_planner_agent=roadmap_planner_agent.bind_tools(roadmap_planner_agent_tools)

app/graph.py

@@ -1,42 +1,50 @@
 from app.state.state import OnboardingState
 from app.nodes.graphnodes import *
 from langgraph.prebuilt import ToolNode ,tools_condition
+from app.agents.agents import roadmap_planner_agent
 from langgraph.graph import StateGraph,END,START
 
 builder = StateGraph(OnboardingState)
 
+# Define Nodes
 # Define Nodes
 builder.add_node("input_node", input_node)
 builder.add_node("resume_data_extraction", extractResumeDataNode)
 builder.add_node("jd_data_extraction", extractJDDataNode)
 builder.add_node("skill_gap_analysis", skill_gap_node)
-builder.add_node("roadmap_planning_agent", roadmap_planning_node)
-builder.add_node("tools", tool_node) # Named 'tools' for tools_condition compatibility
+
+# The ReAct Agent Node
+builder.add_node("roadmap_planning_agent", roadmap_planner_agent)
+
+# The Tool Execution Node (Required for the loop)
+builder.add_node("tools", ToolNode(roadmap_planner_agent_tools))
+
 builder.add_node("finalize_state", finalize_state_node)
 
-# Define Entry Point and initial Extraction Parallelism
-builder.set_entry_point("input_node")
+# 5. Define Edges and Workflow
+builder.add_edge(START, "input_node")
 builder.add_edge("input_node", "resume_data_extraction")
 builder.add_edge("input_node", "jd_data_extraction")
 
-# Join Extractions into Gap Analysis
+# Join Parallel Extractions
 builder.add_edge("resume_data_extraction", "skill_gap_analysis")
 builder.add_edge("jd_data_extraction", "skill_gap_analysis")
 
-# Transition from Analysis to Planning Agent
+# Start the Planning Phase
 builder.add_edge("skill_gap_analysis", "roadmap_planning_agent")
 
-# Agentic ReAct Loop (Planning Agent <-> Tools)
+# Agentic ReAct Loop
 builder.add_conditional_edges(
     "roadmap_planning_agent",
-    tools_condition,
+    tools_condition, # Built-in: routes to "tools" if the model calls a tool
     {
-        "tools": "tools",            
-        "__end__": "finalize_state"  # tools_condition returns '__end__' when no tools are called
+        "tools": "tools",
+        END: "finalize_state" # Routes to finalize if the model gives a final answer
     }
 )
 
+# Loop back to agent after tool execution
 builder.add_edge("tools", "roadmap_planning_agent")
-builder.add_edge("finalize_state", END)
 
+# 6. Compile
 graph = builder.compile()

app/nodes/graphnodes.py

@@ -124,24 +124,6 @@ def skill_gap_node(state: OnboardingState):
     return {"skill_gap_analysis_data": result["parsed"]}
 
 
-    
-
-def roadmap_planning_node(state: OnboardingState):
-
-    """
-    The agent's 'thinking' node. It looks at the Skill Gaps and 
-    decides which tool to call next.
-    """
-    skill_gap_data = state["skill_gap_analysis_data"]
-
-    skill_gap_data= skill_gap_data.model_dump()
-
-    system_prompt = SystemMessage(content=roadmap_planner_agent_prompt)
-    input_msg = HumanMessage(content=f"<skill_gap_data> {skill_gap_data} </skill_gap_data>")
-    
-    response = roadmap_planner_agent.invoke([system_prompt, input_msg] + state["messages"])
-    
-    return {"messages": [response]}
 
 
 def finalize_state_node(state: OnboardingState):

app/schemas/pydanticschema.py

@@ -317,10 +317,12 @@ class SkillGapAnalysis(BaseModel):
     )
 
 
+
+# Schema for learning_roadmap tool
 class RoadmapStep(BaseModel):
     course_id: str
     title: str
-    reasoning: str = Field(..., description="Why this specific course was chosen for this user")
+    reasoning: str = Field(..., description="Why this specific course was chosen for this user tell in short 10-15 words strictly")
     is_foundation: bool
     sequence_order: int = Field(..., description="The order in which the course should be taken")
 
@@ -328,4 +330,12 @@ class LearningRoadmap(BaseModel):
     candidate_name: str
     target_role: str
     roadmap: List[RoadmapStep]
-    onboarding_summary: str
+    onboarding_summary: str
+
+
+
+
+# Schema for search_courses tool
+
+class SearchCourse(BaseModel):
+    query:str=Field(..., description="The skill to find with  semantic terms (e.g., 'FastAPI', 'PostgreSQL', 'Docker','Enterprise VMS Strategy','Utilization Management')")

app/tools/tools.py

@@ -1,7 +1,7 @@
 from langchain_core.tools import tool
 from typing import Optional
 from app.utils.vectordatabase import retriever
-from app.schemas.pydanticschema import LearningRoadmap
+from app.schemas.pydanticschema import LearningRoadmap,SearchCourse
 import json
 from typing import Dict, List,Any
 from pathlib import Path
@@ -68,7 +68,7 @@ def submit_final_roadmap(candidate_name, target_role, roadmap, onboarding_summar
     return result
 
 
-@tool
+@tool(args_schema=SearchCourse)
 def submit_mermaid_visualization(mermaid_code: str):
     """
     STRICTLY call this tool to save the Mermaid.js visualization of the roadmap.