Uncategorized

Overview 

The latest update to our Magic platform introduces significant new performance enhancements, aimed at improving security and expanding the capabilities of our Assets, ensuring a more efficient platform. 

Announcement 

• Integration of Classifier and Extractor Workflows: We are pleased to announce the integration of two powerful workflows—Classifier and Extractor—into a unified category known as Classic AI. Users can now create a Classic AI asset and select its type, either Classifier or Extractor, during the creation process.
  
  
  • User experience: This update simplifies the user experience, making the process more intuitive and user-friendly.
    
    
  • Streamline Workflow: This update streamlines the procedure and reduces complexity, allowing for a more efficient workflow.

Updates

• Asset Visibility and Privacy: Added the ability for asset creators to designate assets as Public or Private while creating it. This provides full control over who can view and manage these assets, resulting in stronger privacy controls and more secure, effective asset management. It is applicable for the following assets.
  
  • Classifier Asset (Classic AI)
  • Extractor Asset  (Classic AI)
  • Use case Asset
    
    
• Filter Enhancement: The update introduces significant enhancements to the filtering system, providing users with a more intuitive, efficient, and secure experience when managing and utilizing assets.
  
  
  • Improved Filter Selection: Enhanced the filter selection process to offer a more user-friendly and intuitive asset filtering experience.
    
    
  • Saved Filters: Introduced the ability to save filter settings, allowing users to quickly apply their preferred filters and display results based on saved preferences.
    
    
  • Filter Name Validation: Implemented a validation feature for filter names, ensuring all validation rules are followed to prevent errors or security threats, thereby enhancing the robustness and security of the filtering process.
    
    
  • Improved Filter Experience: Asset tags and user list options have been enhanced for a smoother and more efficient user experience.
    
    
  • Scope of Enhancement: These filter enhancements have been applied across four key areas, with slight variations based on the specific context:
    
    
    • Asset Studio: Filters help find and manage assets efficiently within the Asset Studio. For more information, see Utilizing Filter.
      
    • Asset Monitor: Filters assist in tracking and monitoring asset status within the Asset Monitor. For more information, see Utilizing Filter.
      
      
      

Updates

• Complete Trace View with Decision Logic: Users can now access a comprehensive, chronological view of all actions and interactions in the traces feature, including the LLM’s thought and action reasoning behind each decision. This enriched trace view allows users to validate each step, understand decision-making in context and seamlessly navigate between timeline and full trace views for efficient monitoring and insights. For more information, see Traces in Conversational Agent.
  
  
• Feedback on LLM Output: Users can now provide feedback on LLM responses during the debug journey and post-publishing of an agent by liking or disliking the output, with the ability to add comments. This feedback helps drive continuous improvement in the quality and relevance of LLM responses. For more information, see Provide Feedback.
  
  
• GPT-4o 16K Model for Conversational and Automation Agents: With the addition of GPT-4 16K support, users of both conversational and automation agents can now leverage a larger context window to handle more complex data inputs, ensuring continuity and accuracy in extended interactions. For more information, see Model Capability Matrix.
  
  
• UI Enhancements for Automation Assets: The automation asset interface has been restructured for clarity, with separate sections for input and output configuration. This streamlined design improves the user journey by providing clear, organized transaction steps.
  
  

Traces in conversational agents refer to the records of interactions and processes that occur during a conversation. They serve as a detailed log of the user-agent interaction, capturing various elements that help in improving the user experience.

Users must have the Gen AI User policy to create, and manage the conversational agents and view the traces in the agents. 

Note: Traces are applicable only for the ReAct Template.

1. In the conversational response from the agent, select  Traces.
   
   
   
   
2. In the Trace window that appears, you can view the conversational traces in the Timeline view.
   
   
   
   

Timeline 

Timeline trace view is a chronological overview of the Agent’s interactions and processes that occur during a conversation. It allows users to see a detailed timeline of the tasks/agent.

1. Click    to expand the agent’s conversational timelines and select the respective agent to view its components on a task.
   
   
   
   
2. In the left side pane, you can view the components of the traces on a particular task. 

 
Components of the Traces on a task in Timeline view

1. Inputs: User input is the text that users provide when communicating with the Agent during the interaction. These inputs are crucial as they guide the agent in understanding the user’s intent and generating appropriate responses. Input can be from users or from agents. If it is a query at the start, then it is from the user. After that, the inputs of the agents can come from other agents as well.
   
   Example: “Check with Sarah and James about their capabilities.”
   
   
2. Decision Logics: This is the set of rules and processes that an agent follows to determine the appropriate responses to user inputs. It is also known as an overview of an LLM’s thought process to arrive at a decision or action point.
   
3. Action: An action is a process that facilitates the fulfillment of user requests by coordinating across different AI agents, and tools.
   
4. Output: Refers to the responses or actions generated by an AI agent in reaction to user inputs or the agent’s query/input. 

Complete

Complete trace view allows users to combine views for a complete understanding of the agent’s operational track, and how the agent navigates complex processes with all relevant information in one place for a conversation.Click Complete to view the complete trace information.

Finding  a Knowledge Base refers to the process of  locating and accessing a specific Knowledge Base. Once the Knowledge Base creation is initiated in the Platform, you can easily search and access that Knowledge Base to save your valuable time.

This guide will walk you through the steps on how to find a knowledge base from the platform. 

Accessing the search bar 

1. Head to the Asset Studio module and choose Knowledge hub tab.
   
   
   
   
2. Locate the Search bar, enter the name of the  knowledge base you wish to find, and  press Enter.
   
   
   
   
3. The platform will dynamically display relevant results.

Utilizing filter

1. On the Asset Studio page, choose Knowledge hub tab.
   
   
   
   
2. In the Knowledge hub tab, click Filter.
   
   
   
   
3. In the filter window that appears, choose the required option to filter the knowledge base. 
   
   • In Status, select the required status.
     
     • Creation in progress: Choose this option to filter the knowledge bases that are currently in the process of being created. 
     • Creation failed: Choose this option to filter the knowledge bases that failed during the creation process.
       
     • Created: Choose this option to filter the knowledge bases that havealready been created.
       
   • In Users, select the user whose knowledge bases you want to filter. You can use the search bar to find specific users.
     
   • In Date modified, select the date time options to filter the assets within a specific period.
     
     • 24 Hrs: Choose this option to filter the assets created within 24 hours. 
     • Custom: Choose this option to filter the assets created within a custom date range. You can select both the Start and End dates to filter the assets.  
       
       
       

Save filter 

This option allows you to save specific filter criteria for easy reuse. It helps to quickly apply the same filter settings without setting them up each time. 

1. In the filter window, click and select Save as New Filter.
   
   
   
   
2. In the Save Filter window that appears, enter the Filter name and click Save.
   
   
   
   
3. The custom filter is saved and you can access the filter whenever you need. 

Reuse the Saved Filter

1. In the filter option,  click and select the Saved filter and choose the saved filter to reuse.
   
   
   
   

Finding  a tool refers to the process of  locating and accessing a specific tool. Once the tool creation is initiated in the Platform, you can easily search and access that tool to save your valuable time.

This guide will walk you through the steps on how to find the tools from the platform. 

Accessing the search bar 

1. Head to the Asset Studio module and choose Tools tab, then locate the Search bar.
   
   
   
   
2. In the Search bar, enter the name of the tool you wish to find and then press Enter.
   
   
   
   
3. The platform will dynamically display relevant results.

Utilizing filter

1. On the Asset Studio page, choose Tools tab.
   
   
   
   
2. In the Tools tab, click Filter.
   
   
   
   
3. In the filter window that appears, choose the desired option to filter the tools. 
   
   
   
   
4. In Users, select the user whose tools you want to filter. You can use the search bar to find specific users.
   
5. In Date modified, select the date time options to filter the assets within a specific period.
   
   • 24 Hrs: Choose this option to filter the assets created within 24 hours. 
   • Custom: Choose this option to filter the assets created within a custom date range. You can select both the Start and End dates to filter the assets.  
     
     
     
     

Save filter

This option allows you to save specific filter criteria for easy reuse. It helps to quickly apply the same filter settings without setting them up each time. 

1. In the filter window, click and select Save as New Filter.
   
   
   
   
2. In the Save Filter window that appears, enter the Filter name and click Save.
   
   
   
   
3. The custom filter is saved and you can access the filter whenever you need. 

Reuse the Saved Filter

1. In the filter option,  click and select the Saved filter and choose the saved filter to reuse.
   
   
   
   

The platform  may use feedback that it receives about its agent-generated messages to improve the quality of its Agent. Feedback for a conversational agent is essential information from users about their interactions with it. This feedback helps developers understand how well the agent is performing and where improvements are needed.

By collecting user feedback, developers can optimize the agent’s accuracy, responsiveness, and overall performance. This continuous feedback loop not only boosts the agent’s capabilities but also ensures it aligns better with user needs, making interactions more engaging and effective.

Users must have the Gen AI User policy to provide feedback for the agents. 

1. In the conversational message, select the response you’d like to provide feedback on. 
2. For the text response, click like or  dislike depending on the feedback you wish to provide.
   

Like

1. Click   like if you found the response helpful and wish to provide additional feedback comments.
   
2. In the comments window that appears,  explain what you liked, provide any additional feedback you wish to share, and then click Submit.
   
   
   
   

Dislike

1. Click dislike if you wish something can be improved.
   
2. In the comments window that appears, explain what could be better, such as inaccuracies, unclear explanations, or missing information and then click Submit.