This document aims to simplify the process of integrating various agents into your Powell Intranet, enhancing its functionality and user experience.
Navigating the Agent Management Interface
Upon successful login to the AI agent management platform, you will be greeted with a navigation menu on the left side of your screen. This menu is your gateway to exploring and managing the diverse functionalities offered by our solution.
To manage your agents, select Agents > Agents from the left navigation menu.
This will direct you to a comprehensive list of all active agents available within the platform.
Modifying, Removing, or Creating an Agent
To edit or delete an agent, navigate to the agent's homepage. Each entry has a “...” menu to update or remove it.
To create a new agent:
Click the “Create” button at the top of the agent list.
Fill in the agent configuration by navigating through the following four tabs:
Identity: Define name, description, icon, and color.
Knowledge: Select the type (e.g., content creation), the AI model, additional prompt and any linked knowledge base.
Behavior: Assign personas, suggested prompts, and configure user goals.
Plugins: Add external data sources or actions via API integrations.
You can preview and test your agent directly in the agent edit interface to quickly validate its behavior. To start using the preview, you need at least one AI model configured. If you make any changes to the agent configuration, remember to save again to apply them.
Agent Configuration
Each agent can be configured in detail through four structured tabs:
1. Identity Tab
This section defines the agent’s identity and visual representation.
Name: Define the agent’s display name (e.g., "Marketing Assistant").
You can localize the name for different languages using the “Add a localization” option.Description: Internal or user-facing description of the agent's purpose.
Also supports localization for multilingual environments.Icon: Choose an icon to visually represent the agent.
Icon background color: Select a background color in hex format (e.g.,
#000000).
2. Knowledge Tab
This section configures how the agent responds, including AI model, data access, and behavioral context.
Type: Defines the agent's core function:
Content Creation: The agent generates text-based content.Image Creation: The agent generates visual content based on user queries.
Additional Prompt: (Optional) A system-level instruction to guide the agent’s tone, focus, or restrictions.
Example: “Always write in a professional tone suitable for corporate communication.”AI Model: Select the model to power the agent (e.g., GPT-4o mini).
Knowledge Base: Optionally link a structured knowledge source to enhance answers with company-specific content.
Demonstrations: Add example user inputs and expected outputs to help shape the agent’s behavior through concrete examples.
Each demonstration consists of:User Input: A sample question or request.
Answer: The ideal response the agent should produce.
3. Behavior Tab
This tab defines user-facing features and fine-tuned user experience.
Personas: Assign one or multiple personas to the agent to personalize tone and response style.
Personas are created globally and shared across agents.Attachments: Toggle this option to allow users to upload a document as part of their question.
Useful for scenarios like summarizing, extracting, or analyzing content from a file. This option can be activated once the embeding specialized AI model is configured in the AI models page.- Text to Speech (TTS): Enables the agent to read text aloud for voice interactions or audio guides. This option can only be activated if the TTS specialized AI model is configured in the AI models page.
- Speech to Text (STT): Allows the agent to convert voice or audio files into text for processing or analysis. This option can only be activated if the STT specialized AI model is configured in the AI models page.
Suggested Prompts (max 5): Provide predefined example prompts that users can click to initiate a conversation.
Each prompt can be localized by language to match the user’s environment.
Example: “Summarize this document” or “Write an engaging LinkedIn post”.Goals: Define reusable actions (shortcuts) at the bottom of each agent reply.
Each goal contains:Title: Name of the goal (e.g., “Convert to Newsletter”).
Goal titles can be localized if needed.Prompt: Instruction sent to the agent when the user selects the goal.
Prompts can be localized if needed.
4. Plugins Tab
Plugins extend the agent’s capabilities by connecting it to external services, APIs, or internal business systems.
A plugin allows the agent to:
Retrieve dynamic data (e.g., weather, HR data)
Perform external actions (e.g., sending emails, calling APIs)
Augment its responses with real-time or contextual information
You can attach multiple plugins to a single agent. Each one can be enabled or disabled independently.
Creating and Configuring a Plugin
Click Add Plugin to begin. The following options are available:
Plugin Type:
HTTP Request: Configure any custom REST API call.Lucca: Integrate with the Lucca HR system.Web Search: Retrieve information from web search APIs.OpenAPI: Describe and interact with any OpenAPI-compatible service.
Display Name: A user-friendly name identifying the plugin.
Description: Clarify the plugin’s purpose or data source.
Enabled: Toggle the plugin on or off.
Endpoint URL: The base API URL.
HTTP Method: Define the HTTP method (e.g.,
GET,POST).Request URL Template: Use
{{parameter}}placeholders to insert values dynamically.
Example:&q={{city name}}Request Body Template (Optional) : Define the payload for methods like
POST.Authentication: Choose from:
No Authentication
API Key
Bearer Token
Pass User Token
OAuth 2.0 Client Credentials
Pass Header Token
Plugin Prompt: Add an internal instruction to describe how the plugin should be used by the agent.
Example: “Use this plugin to fetch weather forecast if the user question contains a city name.”