Agents in Brightsy AI are configurable AI assistants that can be customized with specific instructions, knowledge bases, and tool access. This guide provides step-by-step instructions on how to use the Brightsy AI Agent Form to create and edit AI agents. The form is designed to be intuitive and user-friendly, allowing you to manage your agents efficiently.
Accessing the Form
- Navigate to the Agent Form:
- Access the form through your Brightsy AI account dashboard.
- If editing an existing agent, select the agent from your list to open the form pre-filled with its details.
Agent Properties
Property | Type | Description | Required | Constraints |
---|---|---|---|---|
id | string | Unique identifier for the agent | Auto-generated | Read-only |
name | string | Name of the agent | Yes | 2-50 characters |
description | string | Description of the agent's purpose | Yes | Max 5000 characters |
status | enum | Current state of the agent | Yes | Values: "disabled", "enabled" |
assistantName | string | Name of the assistant that generates content | Yes | 2-50 characters |
systemInstruction | string | Instructions for the AI on how to behave | Yes | Max 5000 characters |
userInstruction | string | Instructions for what the user should do | Yes | Max 5000 characters |
tokenLimit | number | Maximum number of tokens for responses | Yes | Range: 1-8000 |
temperature | number | Level of randomness/creativity in responses | Yes | Range: 0-1 |
webSearch | boolean | Enables web search capability | No | Default: false |
internetAccess | boolean | Allows internet access outside of application tools | No | Default: false |
fileTools | boolean | Allows agent to access to file management tools | No | Default: false |
cms | boolean | Allows agent to access to the content management system | No | Default: false |
excludeReasoning | boolean | Hides intermediate reasoning in responses | No | Default: false |
reasoningEffort | number | Depth of reasoning applied to queries | No | Range: 0-1, Default: 0.5 |
memoryDepth | enum | Controls conversation history retention | No | Values: "workspace", "agent", "user", Default: "workspace" |
superAgent | boolean | Enables enhanced capabilities and permissions | No | Default: false |
fileAttachments | array | Files attached to the agent for reference | No | Supported file types: PDF, DOC, MD, DOCX, TXT, CSV, TSV, XLS, XLSX |
createdAt | datetime | When the agent was created | Auto-generated | Read-only |
updatedAt | datetime | When the agent was last updated | Auto-generated | Read-only |
Form Fields and Instructions
Agent Information
Agent ID (Read-Only):
- Displays the unique identifier for the agent. This field is only visible when editing an existing agent.
- You can copy the ID to your clipboard using the copy button.
Name:
- Enter a name for your agent.
- The name must be between 2 and 50 characters.
Description:
- Provide a detailed description of the agent's purpose.
- The description can be up to 5000 characters long.
Status:
- Select the current status of the agent: Enabled or Disabled.
- The status determines whether the agent is active or inactive.
Agent Configuration
Agent's Assistant Name:
- Specify the name of the assistant that will generate content for this agent.
- The name must be between 2 and 50 characters.
System Instruction:
- Define instructions for the AI on how to behave.
- This field supports up to 5000 characters.
User Instruction:
- Describe what you want the user to do.
- This field supports up to 5000 characters.
Advanced Settings
Token Limit:
- Set the maximum number of tokens for the agent's responses (range: 1-8000).
- Adjust the slider to your desired token limit.
Temperature:
- Adjust the creativity of the agent's responses (range: 0 to 1).
- A lower value results in more deterministic responses, while a higher value increases creativity.
Web Search:
- Enable or disable the agent's ability to search the web for information.
- When enabled, the agent can retrieve real-time information from the internet.
Internet Access:
- Allow agent to access the internet outside of using application tools.
- This allows the agent to access and process information from websites.
Super Agent:
- Enables enhanced capabilities and permissions.
- When enabled, you can use it to manage all Brighsty entities.
File Access:
- Allow agent to have access to file management tools.
- This allows the agent to save, update, delete, move and list files.
CMS:
- Allow agent to the the content management system.
- This allows the agent to save, update, delete, move and record types.
Exclude Reasoning:
- Toggle whether to hide the agent's reasoning process in responses.
- When enabled, only final answers will be shown without intermediate thinking steps.
Reasoning Effort:
- Adjust the depth of reasoning the agent applies to queries.
- Higher values result in more thorough analysis of complex questions.
Memory Depth:
- Set how much conversation history the agent retains and considers.
- A higher value allows for more context-aware responses across longer interactions.
- Memory access options:
- Team: All team members in the workspace will have access to the memory with any agent (default)
- Agent: All team members will have access to memory when using the specific agent
- User: The user who creates the memory will have access with any agent
File Attachments
- Attach Files:
- Click the "Attach Files" button to upload files that the AI can use.
- Supported file types include PDF, DOC, MD, DOCX, TXT, CSV, TSV, XLS, and XLSX.
- Attached files will be listed with options to remove them if needed.
Saving and Canceling
Cancel:
- Click the "Cancel" button to discard changes and close the form.
Save:
- Click the "Save" button to submit the form.
- If the form is in the process of saving, the button will display "Saving."
Error Handling
- Error Alert:
- If an error occurs while saving, an alert will appear indicating the issue.
- Review the form for any missing or incorrect information and try again.
Additional Features
- Agent Danger Zone:
- When editing an existing agent, additional options may be available for advanced management.