ACE

Introduction

The ACE Platform is a comprehensive AI-powered solution that offers multiple interaction modes, document integration, and personalized responses. This documentation provides a thorough overview of the platform's components, features, and configuration options to help users effectively utilize the system.

Table of Contents

1. ACE Chat
2. Knowledge Management
3. Organization Policies
4. Prompt Management
5. Applications
6. Configuration
7. Feature-Specific Documentation
   • Brand Voice
   • Persona Management
   • Extension Management
   • Action Management
   • Analytics
   • Search

ACE Chat

ACE Chat is an AI-powered chat interface that offers multiple interaction modes, document integration, and personalized responses.

Chat Types

ACE Chat offers three distinct chat modes:

1. Search Knowledge

   • Retrieves information from connected knowledge sources
   • Uses data from Knowledge Management system
   • Provides source citations for responses

2. Search on Web

   • Retrieves information from the internet
   • Uses extensions like Wikipedia, DuckDuckGo, and others
   • Provides links to web sources used in responses

3. Ask AI

   • General-purpose AI chat functionality
   • Uses foundation Large Language Models (LLMs)
   • Can be configured to use different AI models
   • Handles a wide range of queries without external sources

Features

Persona

Located at the top middle of the interface, the Persona feature allows you to:

• Select from multiple personalities
• Get responses tailored to the selected persona
• Add or edit personas through Persona Management

Brand Voice

Accessible through Settings, Brand Voice enables:

• Typewriter-style response generation
• Custom voice types for responses
• Enable/disable via toggle button in Settings

Document Integration

ACE Chat supports two types of document integration:

Attached Documents

• Upload documents up to 10MB each
• Supported formats: .PDF, .DOCX, .HTML, .JSON, .TXT, .PPTX, .MD
• Get responses based directly on document content
• View document sources and relevance scores
• See reasoning steps showing how the AI analyzed the document

Retrieved Documents

• View documents fetched for your query
• See relevance scores for each source
• Hover over sources to view specific citations

Conversation Management

The platform provides several options for managing conversations:

Left Sidebar Options

• Start a New Conversation: Reset the chat to begin fresh
• Search Conversations: Find previous chats
• Conversation History: Access past conversations organized
• Clear All Conversations: Remove all conversation history
• Export Conversations: Download conversations in various formats

Conversation Renaming

• Click the pencil icon next to a conversation
• Enter a new name
• Confirm to save the change

Keyboard Shortcuts

• Alt+L: Rephrase your query automatically
• / (Slash): Access prompts

• (Hash): Access available Ace applications

Settings

Settings are accessible via the settings icon and vary by chat type:

Search Knowledge Settings

• Sources: Select knowledge types (File, GitHub, Image, SharePoint, Confluence, Website)
• Knowledge Sets: Choose grouped knowledge sources
• Time Range: Filter by when knowledge was added (24 hours, 7 days, 30 days)

Search on Web Settings

• Select Extension: Choose which web extensions to use for searches

Ask AI Settings

• Select Model: Choose from available LLM models

Common Settings

• Brand Voice: Toggle to enable/disable brand voice features

Response Options

When you receive a response, several options are available:

• Like/Dislike: Provide feedback on response quality (Additional feedback options appear after selection)
• Listen: Activate text-to-speech to hear the response
• Copy: Copy the entire response with formatting
• Modify Response: Adjust the response format (Simpler, Shorter, Longer)
• Regenerate: Create a new response to your query

Additional Features

Follow-up Questions

• AI generates 3-4 relevant follow-up questions after each response
• Click any suggestion to automatically ask that question

Source Citations

• Hover over cited sources to view detailed information
• See which parts of documents were used for responses
• View relevance scores for each source

Ace Applications

• Access specialized applications using # in the input box
• Select from available apps to switch the conversation context

Knowledge Management

The Knowledge Management module enables users to connect various data sources, manage knowledge objects, create FAQs, and organize information into knowledge sets for efficient retrieval and utilization in chat and search functionalities.

Adding Knowledge

To add new knowledge to the system:

1. From the navigation bar, click on Knowledge Management
2. Select Add Knowledge
3. Choose from the available connectors in the dropdown menu

The system supports the following connectors:

• Web: Scrapes data from websites
• Files: Upload documents (PDF, DOCX, HTML, JSON, TXT)
• SharePoint: Connect to SharePoint resources
• Minio: Connect to Minio storage
• API: Connect via API endpoints
• Confluence: Import Confluence content
• YouTube: Extract data from YouTube videos
• GitHub: Connect to GitHub repositories
• Google Drive: Access Google Drive documents
• Jira: Import Jira tickets and documentation
• Postgres: Connect to PostgreSQL databases

Web Connector Configuration

To add a website as a knowledge source:

1. Select Web from the connectors dropdown
2. Choose a scrape method:
   • Recursive
   • Sitemap
   • Single page
   • XML
3. Enter the website URL
4. Provide a knowledge name
5. Configure recording settings (optional):
   • Enable recording to schedule recurring data retrieval
   • Set frequency (e.g., daily, weekly)
6. Select user groups to grant access permissions
7. Click Save

After saving, you will be redirected to the Existing Website tab showing:

• Website link
• Knowledge name
• Scrape method
• Status (Enable/Success)
• Connected details
• Actions (Delete)

Managing Knowledge

The Manage Knowledge section allows you to review and monitor your knowledge sources.

Review Knowledge To review a knowledge source:

1. Click on the Review button in the Connected details column
2. This redirects to the Manage Knowledge page with the following information:
   • Knowledge name
   • Configuration details (website URL, scrape method)
   • Cron job schedule
   • File path (for uploaded files)
   • Creator information
   • Group permissions
   • Indexing attempts

To check indexing logs:

1. Click the Logs button in the Indexing attempts section
2. A pop-up will display the logs showing progress or errors

Knowledge Objects

To view the content extracted from knowledge sources:

1. Navigate to Knowledge Objects
2. This displays all objects (e.g., extracted URLs from sitemaps)
3. Click the Preview button to view the scraped data

My Knowledge

The My Knowledge section allows individual users to add personal knowledge sources that won't be shared with others.

• This feature becomes available after admin configuration of connectors
• Users can add personal knowledge sources following the same process as Add Knowledge
• Personal knowledge is only accessible to the user who created it

The My Knowledge dashboard displays:

• Knowledge name
• Type (Website, File, GitHub, etc.)
• Status
• Creation date
• Actions (Edit, Delete)

FAQs

To manage Frequently Asked Questions:

1. Navigate to the FAQ section
2. Create categories to organize FAQs
3. Add questions and answers within categories:
   • Click Add New Question
   • Enter the question
   • Provide the response
   • Set group permissions
   • Click Add

To view or manage existing FAQs:

1. Click on the FAQ name to view details
2. The popup shows:
   • Creator information
   • Last modification date
   • Category
   • Actions (Edit, Group Permissions, Delete)

Knowledge Sets

Knowledge Sets allow you to group related knowledge sources together:

1. Click Add Knowledge Set
2. Provide:
   • Knowledge set name
   • Description
   • Group permissions
3. Add existing knowledge to the set:
   • Use Existing Knowledge option to add complete knowledge sources
   • Use Add Information Using Search to add specific knowledge objects

After adding a knowledge set, you can:

• Edit the knowledge set details
• Delete the knowledge set

Result Ranking

The Result Ranking feature helps prioritize search results:

1. Enter a search term (e.g., "katonic")
2. View all relevant knowledge related to the search term
3. Adjust relevance scores for results
4. Hide irrelevant results

Organization Policies

Organization policies allow administrators to define rules that govern AI interactions, implement risk controls, and ensure responsible use of AI capabilities across the organization.

Creating a New Organization Policy

Prerequisites

• Administrator access to the Responsible AI module
• Understanding of your organization's AI usage guidelines

Steps to Create a Policy

1. From the Organization Policy screen, click the Add New button
2. Complete the following fields in the policy creation form:
   • Policy Name: Enter a descriptive name for the policy
   • Policy Description: Provide detailed information about the policy's purpose and function
   • Risk Level: Select the appropriate risk level from the dropdown menu
   • Action: Choose the action to be taken when the policy is triggered (e.g., Warn)
3. Click the Add button to create and save the new policy

Example:

• Policy Name: Sensitive Data Protection
• Policy Description: This policy prevents AI systems from processing or exposing personally identifiable information (PII).
• Risk Level: High
• Action: Block

Managing Organization Policies

The Organization Policy interface provides several options for managing existing policies:

Enabling/Disabling Policies

• Toggle the switch in the Enable column to activate or deactivate a policy
• Disabled policies will not be enforced during AI interactions

Editing Policies

1. Locate the policy you wish to modify
2. Click the Edit icon in the actions column
3. Update the policy details as needed
4. Save your changes

Deleting Policies

1. Locate the policy you wish to remove
2. Click the Delete icon in the actions column
3. Confirm the deletion when prompted

Policy Components Explained

Policy Name A clear, concise identifier for the policy that indicates its purpose.

Policy Description Detailed explanation of what the policy controls and why it exists. This description helps users understand why certain AI interactions might be restricted.

Risk Level Categorizes the potential impact of policy violations:

• Low: Minimal potential for harm or misuse
• Medium: Moderate potential for inappropriate content or usage
• High: Significant risk of harm, misuse, or violation of ethical guidelines

Action Types

• Warn: Displays a warning message to users but allows the interaction to proceed
• Block: Prevents the AI system from completing the requested interaction

Prompt Management

The Prompt Management system is a powerful feature that enables users to create, organize, and utilize pre-defined prompts across the platform. This centralized repository of prompts improves workflow efficiency, maintains consistency in AI interactions, and allows teams to leverage proven prompt templates for various use cases.

Platform Integration

Prompt Management serves as a central repository for pre-defined prompts that integrate with multiple components across the platform:

ACE Chat In ACE Chat, users can access and utilize pre-defined prompts to:

• Quickly trigger specific AI capabilities without rewriting complex instructions
• Maintain consistency in AI interactions across users and teams
• Access specialized or optimized prompts for technical tasks
• Execute standardized workflows with a single prompt selection

Users select pre-defined prompts from the Browse prompts button in the ACE Chat UI, where they can search or filter by category to find relevant prompts.

Smart Copilot Smart Copilot offers similar prompt selection functionality, allowing users to:

• Access domain-specific prompt templates
• Leverage proven prompt patterns for common tasks
• Ensure consistent AI assistance across different sessions
• Save time by reusing optimized prompts instead of rewriting instructions

The interface mirrors ACE Chat's prompt selection mechanism, providing a seamless experience across platform components.

Agent Applications For Agent Applications, Prompt Management provides:

• Pre-built instruction sets for specialized agent behavior
• Standardized prompts for common agent tasks and functions
• The ability to share optimized prompts across different agent implementations
• Centralized management of agent conversation patterns

Accessing Prompt Management

To access the Prompt Management interface:

1. Navigate to the main application dashboard
2. Click on Prompt Management in the navigation menu
3. The Prompt Management dashboard will display, showing all categories and their associated prompts

Working with Categories

Categories help organize prompts into logical groups, making them easier to find and manage.

Adding a New Category

1. From the Prompt Management page, click the Add New Category button
2. In the form that appears, enter:
   • Category Name: A descriptive name for your category (e.g., "Data Analysis", "Content Creation")
   • Category Description: Details about the category's purpose and the types of prompts it contains
3. Click the Add button to create the category
4. The new category will appear in the category list

Editing a Category

1. Locate the category you wish to edit
2. Click the Edit button next to the category name
3. Update the category name and/or description as needed
4. Click Save to apply your changes

Deleting a Category When you delete a category, all prompts within that category will also be deleted.

1. Locate the category you wish to delete
2. Click the Delete Category button associated with that category
3. A confirmation dialog will appear warning about the deletion of all prompts within the category
4. Click Confirm to proceed with deletion or Cancel to abort

Managing Prompts

Adding a New Prompt To add a new prompt to an existing category:

1. Locate the category where you want to add the prompt
2. Click the dropdown arrow to expand the category
3. Click the Add New Prompt button
4. In the prompt creation form, enter:
   • Prompt Name: A descriptive title for your prompt (e.g., "Sentiment Analysis", "Technical Documentation Generator")
   • From Category: Select the appropriate category from the dropdown
   • Prompt Description: Enter the prompt text/instructions in detail
   • Tags (Optional): Add relevant tags to improve searchability
5. Click the Save button to add the new prompt
6. The prompt will now appear under the selected category

Editing a Prompt To modify an existing prompt:

1. Locate the prompt you wish to edit
2. Click the Edit button associated with that prompt
3. In the edit form, you can modify:
   • Prompt Name
   • Prompt Category
   • Prompt Description
   • Tags
4. Click the Save button to apply your changes

Deleting a Prompt To remove a prompt:

1. Find the prompt you wish to delete
2. Click the Delete button associated with that prompt
3. Confirm the deletion when prompted
4. The prompt will be permanently removed from the system

Using Prompts

Selecting Prompts in ACE To use a pre-defined prompt in ACE Chat:

1. Open the ACE Chat interface
2. Click the Browse prompts button located near the chat input field
3. A prompt selection pop-up will appear showing available prompts
4. You can:
   • Search for specific prompts using the search bar
   • Filter prompts by category using the category dropdown
   • Browse through the list of available prompts
5. Once you find the desired prompt, click the Use button
6. The prompt will be automatically inserted into your conversation with ACE
7. ACE will immediately process the pre-defined prompt and respond accordingly

Selecting Prompts in Smart Copilot The process for using prompts in Smart Copilot is similar to ACE:

1. Open the Smart Copilot interface
2. Click the Browse prompts button in the chat interface
3. Use the search functionality or category filters to locate the desired prompt
4. Click Use to select the prompt
5. The Smart Copilot will process the selected prompt and provide assistance based on the predefined instructions

Utilizing Prompts in Agent Applications For Agent Applications:

1. During agent configuration, access the prompt library through the agent setup interface
2. Select appropriate prompts for different agent functions or conversation scenarios
3. Assign prompts to specific agent triggers or conversation contexts
4. Test agent behavior with the selected prompts to ensure proper functionality

Applications

Applications serve as specialized chat interfaces that can leverage knowledge sources, extensions, and actions to perform various business functions.

Creating a New Application

Basic Information

1. Navigate to the Applications section of the platform
2. Click the Create New button
3. In the Create New Application page, provide the following details:
   • Application Name: Enter a descriptive name for your application
   • Application Description: Provide a brief description of the application's purpose
   • Application Category: Select from available categories (HR, Sales, Finance, Accounting, Marketing, Operations, IT)

Sharing Settings

1. Use the Share Type switch button to toggle between private and shared access:
   • Private: Application accessible only to you
   • Share: Application accessible to specified groups
2. If Share is selected, expand the group dropdown menu and select the appropriate group(s)
3. Users in the selected groups will be able to access the application

Knowledge Sources You can configure your application to use one or more of the following sources:

• Knowledge: Add knowledge sources from your knowledge management system
  • Knowledge objects
  • Existing knowledge sets
  • Entire knowledge repositories

Note: If you select Knowledge sources, you will not be able to select Extensions or Actions.

Extensions and Actions As alternatives to Knowledge sources, you can configure:

• Extensions: Add capabilities like Wikipedia or other web extensions configured in extension management
• Actions: Integrate with external systems to perform specific tasks
  • Example: Jira actions to create or delete tickets
  • Users can trigger actions through natural language queries

Note: When selecting Actions, you must provide custom instructions manually.

Instructions

• For Knowledge and Extension-based applications, instructions are pre-filled automatically
• For Action-based applications, you must add instructions manually
• These instructions define how the application will behave and respond

Model Selection

1. Expand the Select LLM Models section
2. Choose the appropriate foundation model according to your requirements

Add-ons

1. Expand the Add-on section to access additional features:
   • Citation: Enable source references in responses
   • Conversation Starter: Configure initial prompts
   • Text-to-Speech: Enable voice interaction

Preview and Testing Before creating your application:

1. Click the Preview Application button to see how the application will appear
2. Test the application by entering sample queries
3. Review the generated responses
4. Use the Clear Conversation option to reset the preview if needed

Managing Applications

Viewing Application Details

1. Click on an application from the list to view its details:
   • Instructions
   • Category
   • Configuration settings

Editing Applications

1. Select an application from the list
2. Click the Edit button
3. Make necessary changes to the application settings
4. Click Save to apply your changes

Filtering Applications Filter your application list using the available options:

• Category: Filter by application type (Accounting, HR, etc.)
• Created By Me: Show only applications you've created
• Shared By Me: Show only applications you've shared

Using Applications

1. Click on the application name in your application list
2. You will be redirected to a chat interface branded with your application name
3. The chat will behave according to the configured settings and instructions
4. Interact with the application using natural language queries

Note: The application's behavior is determined by its configuration. Knowledge-based applications will retrieve information, extension-based applications will access external sources, and action-based applications will perform tasks in integrated systems.

Configuration

This document provides a comprehensive guide to the configuration options available in the system. It outlines how to access and modify application settings, customize prompts, and manage usage limits to tailor the system to your specific requirements.

Accessing Configuration

To access the configuration options:

1. Click on Configuration in the navigation bar
2. Choose from the three available options:
   • Application Settings
   • Prompt Personalization
   • Usage Limit

Application Settings

Model Configuration Configure the foundation models used by the system:

• Default Foundation Model: Select the primary LLM to be used
• Fallback LLM Model: Secondary model used if the primary fails
• Embedding Model: Select the embedding model (can only be changed by resetting)
  • To change the embedding model, click the Reset Embedding Model button
  • Note: This will remove all existing embedding data
• Multi-model Configuration: Enable multiple models for embedding indexing

After making changes, click the Save button to apply your configurations.

Search on Web settings:

• Search on Web: Add default extensions
• Default Foundation Model: Select model for search on web
• Groups for Access: Set access permissions

Ask AI Configuration Configure the Ask AI feature:

• Default Foundation Model: Select model for Ask AI
• Groups: Configure access permissions

Guardrails Set boundaries for the AI assistant:

• Enable/disable guardrails
• Select foundation models used specifically for guardrails

File Upload Limits Expand this section to set file upload limits for knowledge management across all users.

Chat Accuracy Settings Adjust settings that affect chat accuracy:

• Chat History Number: Configure how much history is retained
• Number of Chunks: Set how many chunks are used in chat

Advanced Settings

Conversation Starter

• Enable to add multiple predefined questions that appear in the chat interface
• Users can click these to start a conversation

Text to Speech and Speech to Text

• Enable text-to-speech output for responses
• Enable speech-to-text for asking questions verbally

Attached Documents When enabled:

• Adds an option in chat to upload documents
• Allows chatting with document content
• Requires selecting:
  • LLM foundation model to generate responses
  • Embedding model to store data in vector space

Source Citation

• When enabled, citations appear in the chat with responses
• Shows which parts of the response came from which document

Feedback Mechanism

• Enable to add feedback options (like/dislike) for responses

Browse Source

• When enabled, displays the source files used in generating responses

Reasoning Engine

• Enable to see the step-by-step thinking process behind query responses

Application Labels

Customize labels for:

• Chat reasoning/thinking text
• Search reasoning/thinking text
• Search creating text

Brand Voice

• Enable to allow users to generate responses according to your organization's brand voice
• Helps control the tone and style of AI-generated responses

Persona

• When enabled, adds an option in the top left of chat to select a persona
• Chat will respond according to the selected persona
• Responses are generated based on user-created personas

Web Scraping Limits

Configure limits for web scraping operations by users.

Prompt Personalization

Accessing prompt personalization settings:

1. Click on Prompt Personalization from the Configuration page
2. Configure various prompt types as needed

Available Prompt Types Each prompt type allows you to customize the system's behavior:

• Search Knowledge Prompt: Controls how responses are generated when searching knowledge
• Search Web Prompt: Controls responses for web searches
• RCA Prompt: For root cause analysis functions
• Brand Voice Prompt: Used with "Generate with AI" to create branded responses
• Task Prompt: Used for generating task-specific responses
• Search Summary Prompt: Controls how search summaries are generated
• Conversation Name Prompt: Determines how conversation names are generated based on queries
• Follow-up Questions Prompt: Configures how follow-up questions are generated
• Question Validation Prompt: Controls the validation process for questions
• Answer Validation Prompt: Controls the validation process for answers
• Query Rephrase Prompt: Used when rephrasing queries (triggered by Alt+L)
• Organization Policy Prompt: Controls how organization policies are applied
• Response Length Prompt: Determines how response length is managed

Moderation Panel

Configure content moderation settings:

• Enable/disable filters for threatening content
• Control filters for sexual content, minors-related content
• Manage filters for self-harm content
• Configure destructive content filters

When enabled, these filters will block corresponding content from responses.

Usage Limits

Adding Usage Limits

1. Click on Usage Limit from the Configuration page
2. In the "Add Usage Limit" popup, configure:
   • Select Limit Level: Choose user, group, or application level
   • Application: Select from ACE Companion, ACE, AI Studio, or other options
   • Select Limit Type: Choose budget amount or number of requests
   • Set Frequency: Configure when thresholds reset (daily, weekly, or monthly)
3. Click the Add button to create the limit

Managing Usage Limits For existing usage limits, you can:

• View: See limit usage details including:
  • Amount used
  • Cost type consumed
  • Number of requests made
  • Use the Refresh button to update the limit data
• Edit: Modify settings including:
  • Target application
  • Budget type
  • Number of requests
  • Reset frequency
• Delete: Remove usage limits that are no longer needed

Feature-Specific Documentation

Brand Voice

Overview Brand Voice is a feature that enables:

• Typewriter-style response generation
• Custom voice types for responses
• Enable/disable via toggle button in Settings

This feature allows users to generate responses according to your organization's brand voice, helping control the tone and style of AI-generated responses.

Persona Management

Overview The Persona feature allows you to:

• Select from multiple personalities
• Get responses tailored to the selected persona
• Add or edit personas through Persona Management

When enabled, this adds an option in the top left of chat to select a persona, and the chat will respond according to the selected persona. Responses are generated based on user-created personas.

Extension Management

Overview The Extension Management feature allows users to add and configure various extensions that enhance the platform's capabilities, particularly for the Search on Web functionality.

Action Management

Overview The Action Management feature enables integration with external systems to perform specific tasks through natural language queries in applications.

Analytics

Overview The Analytics feature provides comprehensive data visualization and reporting capabilities for monitoring platform usage, costs, and conversation metrics.

Search

Overview The Search Knowledge feature retrieves information from connected knowledge sources, uses data from the Knowledge Management system, and provides source citations for responses.

Conclusion

The ACE Platform provides a comprehensive suite of tools for AI-powered interactions, knowledge management, and customizable applications. By following this documentation, users can effectively utilize the platform's features to enhance productivity and streamline workflows.

For additional information or specific use cases, refer to the feature-specific documentation sections or contact your system administrator.