Configuring AI Inputs

Found in: Settings > AI Inputs

• Overview
  • Location & Access
  • Input Modes
• Configuration Tabs
  • Basic Information
  • Prompt
    • Variable Syntax
    • Example Prompt
  • Field Input
    • URL Pattern
    • CSS Selector
    • Execution Settings
  • Access Control
  • Files
    • Filter Interface
    • Filter Options
    • Example Configurations
• Using AI Inputs
  • For End Users
  • Preview Tab
• Export and Import
• Best Practices
  • Writing Effective Prompts
  • Performance Tips
• Related Pages

AI Inputs allow administrators to enhance the platform with AI-powered assistants. These can be attached to text fields or inspection templates to generate content with AI, or used to provide predefined prompt templates in the chat interface.

This feature requires Super Admin access and the AI & Analytics Package subscription.

Overview

Location & Access

Navigate to Settings > AI Inputs to manage AI Inputs.

Permissions: Only Super Admin users can create and configure AI Inputs. End users interact with configured inputs based on access control settings.

Input Modes

AI Inputs operate in two modes:

Field Input Mode — Attach AI assistants to specific text fields or use them in inspection templates. When users interact with a configured field, they can trigger AI-powered suggestions based on the current record, attached files, and custom prompts. When this mode is selected, an additional Field Input tab appears for targeting configuration.

Chat Mode — Enhance the chat interface by providing predefined prompt templates that users can reuse. Instead of typing complex queries each time, users select from configured templates that include the right context and instructions. This helps standardize common queries across the organization.

---

Configuration Tabs

Each AI Input is configured through multiple tabs:

Basic Information

Field	Description
Input Mode	Choose between Chat Mode (predefined templates) or Field Input Mode (targets specific fields). Selecting Field Input Mode reveals an additional Field Input tab.
Title	Display name shown in the AI dropdown menu
Description	Internal notes about when and how to use this input
Weight	Determines processing order when multiple inputs match. Lower weight inputs are processed first.

Weight usage:

• In Chat Mode, a lower weight input could provide general context or directions, while higher weight inputs add specific instructions related to the user’s question
• In Field Input Mode, weight determines the order inputs are processed on the backend when multiple inputs target the same page

---

Prompt

The prompt defines what the AI should do. Use variable substitution to include contextual data from the current record.

Variable Syntax

Syntax	Example	Description
Dot Notation	{{context.model.id}}	Access nested properties
Nested Properties	{{context.model.client.acronym}}	Deep property access
File Content	{{files.document.content}}	Include file content
JSONPath Basic	{{$.context.model.id}}	Same as dot notation
JSONPath Wildcard	{{$..name}}	Find all “name” properties
JSONPath Array	{{$.files[0].data}}	Access first file’s data
JSONPath Filter	{{$.context.model[?(@.status=="active")]}}	Filter by condition

Variables that do not exist are left as-is in the output. Objects and arrays are automatically JSON-formatted.

Example Prompt

You are assisting with a product risk assessment using ISO 10377:2013.

--- PRODUCT DESCRIPTION ---
{{$.context.model.inspectionModel.data.product_description.value}}

--- APPLICABLE LEGISLATION ---
{{$.context.model.products[*].legislation}}

Analyze the product and provide a structured risk assessment.

---

Field Input

This tab appears when Field Input Mode is selected. Configure which pages and form fields trigger this AI input.

URL Pattern

Determines on which pages the input will be available. You can switch between Simple and Advanced modes:

• Simple Mode — Select from a list of predefined page types (Bookings, Products, Suppliers, Inspection Jobs, etc.)
• Advanced Mode — Manually edit the URL pattern for fine-grained control, such as enabling the input only on a specific supplier page

Leave empty to make the input available on all pages.

CSS Selector

Determines which fields on the page will trigger this AI input. Target specific input fields using CSS selectors:

Pattern	Example	Description
Mobile field code	[data-mobile-field-code="product_risk_assessment"]	Target mobile form fields
Input by name	input[name="description"]	Target by name attribute
Input by ID	#result	Target by element ID
CSS class	.custom-field-class	Target by CSS class

Execution Settings

Setting	Description
Show in browser	Display the AI dropdown when users interact with matching fields. Users can see responses in real-time via a chat-like interface and adjust the output before applying.
Run on server	Process asynchronously on the backend without user interaction

Server execution modes:

Mode	Behavior
Never	Backend processing disabled
Always	Process on every form submission
If value is empty	Process only when the target field has no existing value

Browser mode is useful when users need to review and refine AI-generated content before applying it. Backend processing is useful for automated tasks that don’t require user interaction, or when users should be restricted from editing—for example, when a supplier submits a form and backend processing fills in missing details automatically.

---

Access Control

Configure which roles and permission groups can use this AI input. See Access Control Matrix for details on the permission system.

---

Files

Configure which files should be included as context for AI processing. Context files are typically all files associated with the current page—you can see which files are available in the Preview tab when editing from a specific page.

Filter Interface

The file filter uses a combination of AND and OR logic to select specific files:

• Within a filter group: All conditions are combined with AND logic (all must match)
• Between filter groups: Groups are combined with OR logic (any group can match)

Use + AND to add conditions within a group, and Add Filter Group to create additional OR groups.

Filter Options

Criterion	Description
Filename	Match by filename (Equals, Contains, Starts with, Ends with)
File Size	Filter by file size
Folder	Match exact folder name
Subfolder	Match exact subfolder name
Properties (JSON Path)	Filter by file metadata properties
Object Table	Filter by associated object table

Example Configurations

Include only PDF files:

• Filter Group 1: Filename → Ends with → .pdf

Include Word documents OR PDF files:

• Filter Group 1: Filename → Ends with → .docx
• Filter Group 2: Filename → Ends with → .pdf

Include files from a specific folder that are PDFs:

• Filter Group 1:
  • Folder → Equals → specifications
  • AND Filename → Ends with → .pdf

Include inspection reports or any file from the compliance folder:

• Filter Group 1: Filename → Contains → inspection_report
• Filter Group 2: Folder → Equals → compliance

---

Using AI Inputs

For End Users

1. Navigate to a page with configured AI fields
2. Hover or focus on the target input field
3. Click the AI dropdown that appears
4. Select an AI input from the available options
5. Optionally attach additional files for context
6. Review the AI-generated response
7. Click Apply to insert the response into the field

Preview Tab

When editing an AI input from a target page (by clicking the pencil icon in the dropdown), a Preview tab becomes available. This tab lets you test how prompts will be processed with real data from the current page.

Prompt Preview (with context substitution) Shows the prompt template with variables replaced by actual values from the current page context. This helps verify variable paths are correct and data is being pulled as expected.

Current Context Data (JSON) Displays the complete JSON object available for variable substitution. Use this to explore accessible data and find correct paths for prompt variables.

Matched Files Lists files matching the configured file filter, showing which documents will be available for AI processing.

Copy Context & Help / Copy Context Copy context information to your clipboard for testing prompts with external AI tools like ChatGPT or Claude

---

Export and Import

AI Inputs support Export / Import via the Save dropdown menu for backing up, duplicating, or sharing configurations.

---

Best Practices

Writing Effective Prompts

1. Be specific - Use clear, detailed instructions
2. Include context - Reference relevant data using variable substitution
3. Set expectations - Specify the desired output format
4. Test thoroughly - Use the Preview tab to validate with real data

Performance Tips

• Use file filters to include only relevant files
• Set appropriate weights when multiple inputs might match
• Enable server execution only when needed to reduce processing load

---

Related Pages

• AI Chat
• Using AI Inputs
• MCP Server Integration
• Access Control Matrix
• Workflow Processes