Update app.py

app.py

@@ -39,27 +39,56 @@ def ticket_creation_agent(state: State) -> Command[Literal["priority_classificat
 # Priority classification agent
 def priority_classification_agent(state: State) -> Command[Literal["escalation_classification_agent"]]:
     """Classify priority based on the issue description."""
-    prompt = f"Classify the following issue as urgent, critical, or normal: {state['issue_description']}"
-    priority = llm.invoke(prompt).strip()  # Use .invoke() instead of __call__
-    
-    # Update the priority and proceed to escalation classification
-    return Command(update={"priority": priority}, goto="escalation_classification_agent")
+    prompt = (
+        f"You are a support assistant. Classify the following issue as "
+        f"'urgent', 'critical', or 'normal' based on its severity:\n\n"
+        f"Issue: {state['issue_description']}"
+    )
+    priority = llm.invoke(prompt).strip().lower()
+
+    # Ensure valid priority classification
+    if priority not in ["urgent", "critical", "normal"]:
+        priority = "normal"  # Default to 'normal' if the classification fails
 
-# Escalation classification agent
-def escalation_classification_agent(state: State) -> Command[Literal["generate_response_agent"]]:
-    """Classify whether escalation is needed based on priority."""
-    escalation_needed = state["priority"].lower() in ["urgent", "critical"]
-    return Command(update={"escalation_needed": escalation_needed}, goto="generate_response_agent")
+    return Command(update={"priority": priority}, goto="escalation_classification_agent")
 
 # Generate response agent
 def generate_response_agent(state: State) -> Dict[str, str]:
     """Generate response based on ticket priority and escalation need."""
-    escalation = "Escalate the issue to a senior team member immediately." if state["escalation_needed"] else "No escalation needed."
-    prompt = f"Generate a response for the following issue: {state['issue_description']}. The priority is {state['priority']}."
-    response = llm.invoke(prompt).strip()  # Use .invoke() and ensure the response is clean
+    escalation = (
+        "Escalate the issue to a senior team member immediately."
+        if state["escalation_needed"]
+        else "No escalation needed."
+    )
+
+    # Updated prompt to guide the model explicitly
+    prompt = (
+        f"You are a customer service assistant. Generate a concise and actionable response "
+        f"for the following issue:\n\n"
+        f"Issue: {state['issue_description']}\n"
+        f"Priority: {state['priority']}.\n\n"
+        f"Your response should directly address the issue and provide next steps."
+    )
+    response = llm.invoke(prompt).strip()
 
     return {"response": response, "escalation": escalation}
 
+# Gradio Interface function to process the ticket
+def process_ticket(issue_description: str):
+    """Process the issue ticket through the multi-agent flow."""
+    state = {"issue_description": issue_description}
+    try:
+        print(f"Initial Issue Description: {issue_description}")  # Debug log
+        result = graph.invoke(state)
+        print(f"Graph Result: {result}")  # Debug log
+        response = result.get("response", "No response generated")
+        escalation = result.get("escalation", "No escalation specified")
+        return response, escalation
+    except Exception as e:
+        print(f"Error occurred: {e}")  # Debug log
+        return f"Error occurred: {e}", "Unable to determine escalation"
+
+
 # Add nodes to the graph
 builder.add_edge(START, "ticket_creation_agent")
 builder.add_node("ticket_creation_agent", ticket_creation_agent)