Agent Model and Display Name Fixes

Agent Model and Display Name Fixes

Version Implemented: 0.229.001

Summary

Fixed UX issues with message metadata and display when using Semantic Kernel agents.

Issues Fixed

1. Model Selection in Message Metadata ✅

Problem: Messages showed the fallback model (gpt-5-chat) even when using agents with different models.

Example: World News Agent uses gpt-4o but messages showed model_deployment_name: "gpt-5-chat"

Fix:

• Modified route_backend_chats.py to use agent's actual model when available
• Added logic to extract deployment_name from selected_agent object
• Falls back to final_model_used when no agent is selected

Code Changes:

# Determine the actual model used and agent information
actual_model_used = final_model_used
if selected_agent:
    # When using an agent, use the agent's actual model deployment
    if hasattr(selected_agent, 'deployment_name') and selected_agent.deployment_name:
        actual_model_used = selected_agent.deployment_name

2. Message Sender Display ✅

Problem: Messages showed "AI (gpt-4o)" even when using named agents like "World News Agent".

Expected:

• With agents: Show agent display name (e.g., "World News Agent")
• Without agents: Show "AI (model-name)" format

Fix:

• Added agent_display_name and agent_name to backend response
• Modified frontend appendMessage function to use agent display name when available
• Updated all appendMessage calls to include new parameters

Code Changes:

// Frontend logic in chat-messages.js
if (agentDisplayName) {
  senderLabel = agentDisplayName;
} else if (modelName) {
  senderLabel = `AI <span style="color: #6c757d; font-size: 0.8em;">(${modelName})</span>`;
} else {
  senderLabel = "AI";
}

3. UI Model Selection Metadata ✅ NEW FIX

Problem: The "Selected Model" in the UI metadata panel showed fallback model instead of agent's model.

Example: UI showed "Selected Model: gpt-5-chat" even when World News Agent (gpt-4o) was being used.

Fix:

• Added logic to update user message metadata after agent model is determined
• Updates the model_selection.selected_model field to reflect actual model used
• Ensures UI consistency between agent selection and model display

Code Changes:

# Update the user message metadata with the actual model used
try:
    user_message_doc = cosmos_messages_container.read_item(
        item=user_message_id, 
        partition_key=conversation_id
    )
    
    # Update the model selection in metadata to show actual model used
    if 'metadata' in user_message_doc and 'model_selection' in user_message_doc['metadata']:
        user_message_doc['metadata']['model_selection']['selected_model'] = actual_model_used
        cosmos_messages_container.upsert_item(user_message_doc)
        
except Exception as e:
    print(f"Warning: Could not update user message metadata: {e}")

Files Modified

Backend

• route_backend_chats.py: Added agent model extraction and metadata update logic
• config.py: Updated version to 0.226.078

Frontend

• static/js/chat/chat-messages.js: Updated appendMessage function and all call sites

Testing

• functional_tests/test_agent_model_display_fixes.py: Validates message display logic
• functional_tests/test_model_selection_metadata_fix.py: Validates UI metadata consistency

Result

✅ Agent messages: Show agent display name and correct model
✅ Non-agent messages: Show "AI (model)" format as before
✅ Message metadata: Contains accurate model and agent information
✅ UI metadata panel: Shows consistent agent model information
✅ Backward compatibility: Existing functionality preserved

Example Behavior

Before Fixes

Message from World News Agent:
- Sender: "AI (gpt-5-chat)"  ❌ Wrong model, generic sender
- Message Metadata: model_deployment_name: "gpt-5-chat"  ❌ Fallback model
- UI Metadata Panel: "Selected Model: gpt-5-chat"  ❌ Inconsistent with agent

After Fixes

Message from World News Agent:
- Sender: "World News Agent"  ✅ Agent display name
- Message Metadata: model_deployment_name: "gpt-4o"  ✅ Agent's actual model
- UI Metadata Panel: "Selected Model: gpt-4o"  ✅ Consistent with agent

Impact

Users now see consistent, accurate information about which agent and model are being used, providing better transparency and user experience when working with AI agents.