Conversational Optimization Tutorial

from sematryx.client.sdk import SematryxClient

client = SematryxClient(api_key="YOUR_API_KEY")

# Start conversation with natural language description
result = client.start_conversational_optimization(
    description="I want to optimize my marketing budget across Google, Facebook, and LinkedIn to maximize ROI"
)

conversation_id = result["conversation_id"]
print(f"✅ Conversation started: {conversation_id}")

import time

# Check conversation status
status = client.get_conversation_status(conversation_id)

if status["status"] == "waiting_for_input":
    print(f"🤖 Agent: {status.get('question')}")
    print(f"💡 Examples: {', '.join(status.get('examples', []))}")
elif status["status"] == "ready_to_optimize":
    print("✅ All parameters collected! Ready to optimize.")
elif status["status"] == "processing":
    print("⏳ Agent is still processing...")
    time.sleep(2)  # Wait and check again

# Respond to agent's question
response = "$50000"  # User's answer
result = client.continue_conversation(conversation_id, response)
print(f"✅ Response accepted: {result.get('accepted', False)}")

from sematryx.client.sdk import SematryxClient
import time

client = SematryxClient(api_key="YOUR_API_KEY")

# Step 1: Start conversation
result = client.start_conversational_optimization(
    description="I want to optimize my marketing budget for maximum ROI"
)
conversation_id = result["conversation_id"]

# Step 2: Interactive loop
while True:
    status = client.get_conversation_status(conversation_id)
    
    if status["status"] == "waiting_for_input":
        # Agent has a question
        print(f"\n🤖 Agent: {status.get('question')}")
        
        # Get user response (in production, from user input)
        user_response = input("You: ")
        
        # Continue conversation
        client.continue_conversation(conversation_id, user_response)
        
    elif status["status"] == "ready_to_optimize":
        print("\n✅ All parameters collected! Ready to optimize.")
        break
        
    elif status["status"] == "processing":
        print("⏳ Processing...")
        time.sleep(2)
    else:
        break

# Step 3: Complete and optimize
result = client.complete_conversational_optimization(
    conversation_id,
    max_evaluations=2000,
    mode="balanced"
)

optimization_id = result['optimization_id']
print(f"\n🚀 Optimization started!")
print(f"📊 Optimization ID: {optimization_id}")

# Step 4: Poll for results
print("\n⏳ Waiting for optimization to complete...")
while True:
    opt_status = client.get_optimization_status(optimization_id)
    
    if opt_status["status"] == "completed":
        print("\n🎉 Optimization complete!")
        
        # Get detailed results
        results = client.get_optimization_results(optimization_id)
        
        # Display results in a user-friendly format
        print(f"\n📊 Optimal Budget Allocation:")
        solution = results.get("solution", {})
        for channel, allocation in solution.items():
            dollar_sign = "$"
            formatted_amount = f"{dollar_sign}{allocation:,.0f}"
            print(f"   • {channel}: {formatted_amount}")
        
        print(f"\n💰 Performance Metrics:")
        print(f"   • Total ROI: {results.get('roi', 0):.1%}")
        revenue = results.get('revenue', 0)
        dollar_sign = "$"
        formatted_revenue = f"{dollar_sign}{revenue:,.0f}"
        print(f"   • Expected Revenue: {formatted_revenue}")
        print(f"   • Total Conversions: {results.get('conversions', 0):,.0f}")
        
        # Agent can also provide insights
        if "insights" in results:
            print(f"\n💡 Key Insights:")
            for insight in results["insights"]:
                print(f"   • {insight}")
        
        break
    elif opt_status["status"] == "failed":
        print(f"\n❌ Optimization failed: {opt_status.get('error')}")
        break
    else:
        time.sleep(2)  # Poll every 2 seconds