Intent

The hardest part of any AI product isn’t the AI: it’s getting the user to ask the first question. Intents are how you put a hundred good first questions, in your customer’s voice, on the chat home screen. And Vibe Coding is how you write all of them in five minutes.

The Intent feature (/admin/intent/instance) manages categorized prompt suggestions that appear in the Chat interface. Intents provide users with pre-built conversation starters organized by topic, making it easier to discover and use the AI capabilities of Viglet Turing.

Each intent represents a category (e.g., “Boost my work productivity”) containing one or more actions: each action is a labeled prompt that is sent directly to the chat when clicked.

Think of intents as the hook on the home screen of every conversation. The customer didn’t type a question yet: they’re staring at the prompt box wondering what to ask. Intents say “here’s what people like you ask”, in their language. Conversion starts here.

---

Intent Listing

The listing page shows all configured intents as a grid of cards. Each card displays:

• Title: the intent category name
• Description: a brief explanation of the category
• Icon: a Lucide icon representing the category

A New Intent button at the top opens the creation form. Clicking an existing card navigates to its detail page.

---

Creating or Editing an Intent

The intent form is divided into three sections.

General Information

Field	Max Length	Description
Title	150 chars	Category name displayed in the chat interface (required)
Description	500 chars	Brief explanation of what prompts this intent provides
Icon	n/a	Lucide icon picker: select an icon to represent this category

Actions

Actions are the individual prompts within an intent category. Each action has:

Field	Description
Label	Short action label shown to the user (e.g., “Create daily schedule”), required
Prompt	The full prompt text sent to the chat when the user clicks this action, required

Actions support drag-and-drop reordering: drag the handle on each action card to change the display order. Use the Add Action button to create new actions, and the trash icon to remove existing ones.

Write actions in the customer’s voice

The label is what the user clicks. The prompt is what the LLM receives. They don’t have to be the same. The label should sound like the user asking themselves a question (“Help me prepare for my next 1-on-1”); the prompt can be a more elaborate brief for the LLM (“Help me prepare for a 1-on-1 meeting with my manager. Suggest topics, questions to ask, and a way to bring up career growth.”). The pairing is what makes intents feel like a friendly assistant rather than a search box.

Status

Field	Description
Enabled	Toggle to show or hide this intent in the chat interface. Disabled intents are not visible to users.

---

Vibe Coding: Author Intents With AI

Writing 4 categories × 6 actions × 2 languages = 48 prompts is exactly the kind of work the AI is best at. The Intent form has a Vibe Coding mode: you describe what you want; the assistant drafts the entire intent.

You: “Make me an intent for new product owners. Six actions covering the first 30 days on the job: backlog grooming, stakeholder mapping, OKRs, the first sprint demo, a retro plan, and a simple metric to track.”

Assistant: “Done. Title: ‘New Product Owner: First 30 Days’. Icon: graduation-cap. Six actions ready: Map your stakeholders → ‘Help me build a stakeholder map for my first month as PO. Roles, influence, what they need from me, and one question I should ask each of them.’ Groom the backlog → ‘Walk me through grooming a backlog I just inherited. What’s the right order to clean it?’ Set OKRs for the first quarter → … Want me to also write a Portuguese version?”

The same pattern as Chat Flow and Custom Tools. Two interaction modes:

• First-turn creation: describe the intent, get the full structure (title, description, icon, all actions).
• Revision turns: “add an action about pricing”, “make the third one shorter”, “translate everything to Portuguese”.

Vibe Coding is especially powerful here because intents are usually authored en masse: you’re populating the home screen, not writing a one-off. Three minutes of prompting beats two hours of typing.

---

Pre-loaded Intents

Turing ships with four default intents:

Intent	Icon	Actions
Boost my work productivity	Briefcase	Create daily schedule, Draft a difficult email, Prepare meeting agenda
Learn something new	Graduation Cap	Explain complex topics, 30-day study plan, Test my knowledge
Create or improve content	Pen Tool	Social Media Ideas, Proofread & Enhance, Video Scripting
Health & Mindset	Heart Pulse	Quick Meditation Guide, Rapid Meal Ideas, At-home Workout

These can be edited, disabled, or deleted. New intents can be added at any time.

---

REST API

All endpoints require authentication via the Key header.

Method	Endpoint	Description
GET	/api/intent	List all intents (sorted by title)
GET	/api/intent/enabled	List enabled intents only (sorted by display order)
GET	/api/intent/{id}	Get a specific intent with its actions
POST	/api/intent	Create a new intent
PUT	/api/intent/{id}	Update an intent
DELETE	/api/intent/{id}	Delete an intent and all its actions

Request / Response Body

{
  "id": "intent-productivity",
  "title": "Boost my work productivity",
  "description": "Prompts to help you be more productive at work",
  "icon": "lucide:briefcase",
  "enabled": 1,
  "sortOrder": 0,
  "actions": [
    {
      "id": "action-1",
      "label": "Create daily schedule",
      "prompt": "Help me create an efficient daily schedule...",
      "sortOrder": 0
    },
    {
      "id": "action-2",
      "label": "Draft a difficult email",
      "prompt": "Help me draft a professional email...",
      "sortOrder": 1
    }
  ]
}

Example: list enabled intents:

curl "http://localhost:2700/api/intent/enabled" \
  -H "Key: <YOUR_API_TOKEN>"

Example: create an intent:

curl -X POST "http://localhost:2700/api/intent" \
  -H "Key: <YOUR_API_TOKEN>" \
  -H "Content-Type: application/json" \
  -d '{
    "title": "Data Analysis",
    "description": "Prompts for analyzing and visualizing data",
    "icon": "lucide:bar-chart-3",
    "enabled": 1,
    "sortOrder": 4,
    "actions": [
      {
        "label": "Summarize a dataset",
        "prompt": "Analyze the following dataset and provide key statistics, trends, and insights.",
        "sortOrder": 0
      }
    ]
  }'

---

Permissions

Intent management uses role-based access control:

Permission	Description
INTENT_VIEW	View intent categories
INTENT_CREATE	Create new intents
INTENT_EDIT	Edit existing intents
INTENT_DELETE	Delete intents

All permissions also require the ROLE_ADMIN role.

---

Related Pages

Page	Description
Chat	The chat interface where intents are displayed to users
AI Agents	Configure AI Agents that respond to chat prompts
Administration Guide	User, group, and role management

---