[Meta] - UX / Design Conversational Interface

Created on 2 July 2025, about 1 month ago

Desciption

Modern conversational AI interfaces are rapidly evolving beyond simple chat windows into sophisticated, multi-modal experiences that blend conversation with direct manipulation, collaborative editing, and intelligent content generation.

This meta issue is for creating a UX concept covering current and future aspects of conversational interface design for Drupal AI.

Once we have a UX, we can proceed to further expand the API between frontend applications (the conversational interface) and Drupal's chat connection.

The work should be executed in the following steps

  1. Create and approve UX
  2. Expand / define chat API
  3. Select most important features
  4. Implement selected functionality

The UX work will be done in this issue, implementation work can be done in child issues.

UX Briefing

Project Goal: To design a flexible and intuitive chat interface for Drupal AI that enables users to interact with Drupal websites using agentic AI functionality. The interface must adapt to various display contexts and effectively present diverse AI responses, including rich media and "artifacts."

Scope: This briefing focuses specifically on the chat interface UX, not a full workspace like Claude or Gemini.

Core Principles & Context

Adaptability: The chat UI must be launchable from different locations (header bar, sidebar, bottom right) and utilize varying amounts of space (slide-up chat, sidebar integration, content section embed, modal window).

Rich Interaction: Support for diverse AI response types beyond plain text is crucial, including visual elements, code snippets, and "artifacts" (documents, complex media).

User-Centric: Prioritize clear visual feedback, contextual guidance, and user control to foster trust and efficiency.

Key Visual Components & Requirements

Based on the provided concepts, the following visual components are essential for the Drupal AI chat interface:

Main Conversational Interface

Chat History Display:

  • Clear distinction between user queries and AI responses (e.g., distinct chat bubbles, avatars).
  • Support for plain text, bold text, and clear typography.

User Input Field:

  • Prominently placed at the bottom.
  • Includes a "Send" button.
  • Attachment Button: A paperclip icon or "Attach File" button for direct document/file uploads.
  • Voice Input Button: For speech-to-text activation (consider large touch target for mobile).
  • Visual Feedback Indicators (for AI Processing States):
  • Typing Indicator: Classic three-dot ellipsis animation or "AI is typing..." for short responses.
  • Spinner/Loader: Circular or animated AI icon for brief processing times ("AI is thinking...", "Searching...").
  • Progress Bar: Linear or circular with percentage/descriptive text ("Generating draft...", "Synthesizing content...") for longer operations.
  • Skeleton Screens: For complex content generation, a visual outline of forthcoming content.
  • Checkmarks: To indicate successful completion.
  • Cancel Option: An "X" icon or "Stop" button adjacent to active indicators to allow users to cancel long-running AI operations.

Follow-Up Questions & Suggestions

Quick Reply Buttons:

  • Styled as pills or compact rectangular buttons.
  • Dynamically generated, contextually relevant follow-up questions or actionable prompts (e.g., "Tell me more," "Summarize this," "Save as draft").
  • "Show More Suggestions" option if applicable.

Contextual Prompts: AI-generated conversation starters or nudges.

Document & Artifact Display

In-line Artifacts: For small, self-contained items like code snippets, short data tables, or single-image previews, rendered directly within the chat bubble for immediate comprehension.

Collapsible Sidebar:

  • For document previews (e.g., technical specs, research papers) when AI references or generates them.
  • Allows scrolling, viewing key highlights/summaries, and related document suggestions.
  • Easily expandable/collapsible to manage screen real estate.
  • Full-Screen Overlay (Modal Window):
  • For detailed document viewing (when clicking a preview or link).
  • Includes standard functionalities: zoom, full-text search, download options, potentially AI-powered annotations/highlights.
  • Maintains conversational context in the background (dimmed).
  • Canvas Mode: A dedicated workspace for editing or deep interaction with complex AI-generated content (e.g., a structured report, a content outline). This should be accessible from the chat but open as a distinct view.

Rich Content Rendering

  • Interactive Image Gallery/Carousel:
  • For multiple images (product images, design variations).
  • Swipe/click navigation, AI-generated captions.
  • Options to expand, save, or generate similar images.
  • Embedded Slideshow Player:
  • For AI-generated presentations or curated content sets.
  • "Next"/"Previous" controls, AI-generated captions/key takeaways.
  • Options to download or export to Drupal content types.
  • Interactive Elements: Clickable tools or quick-response buttons embedded within responses (e.g., "Show more like this," "Generate different style").

Customizable Settings & AI Model Selection

  • Settings Access Point: A clearly visible "Settings" icon (gear icon) or menu option.
  • AI Persona/Tone Selector: Dropdown/radio buttons for tone (Professional, Friendly, Concise, Creative).
  • Domain/Model Selection ("AI Modes"): Toggles/radio buttons/dropdown for pre-configured AI behaviors (e.g., Research Assistant, Code Generator, Content Editor, Marketing Strategist), with brief descriptions.
  • Output Refinement Options: Checkboxes/toggles for preferences like "Generate detailed responses," "Include source citations," "Suggest follow-up questions automatically."
  • Feedback Mechanism: Prominent "Send Feedback" button or a structured section for ratings/issues (e.g., thumbs-up/down).
  • Reset/Clear Conversation Options: Buttons to "Clear Current Conversation" or "Reset AI State."

Wireframe & Prototype Considerations (for Figma)

The UX designer should use this briefing to create low-fidelity visual representations (Wireframes) of the chat interface, demonstrating how the above components work in in different contexts:

  • Launching the chat from different entry points.
  • Sending a query and observing AI processing feedback.
  • Interacting with quick replies and follow-up suggestions.
  • Viewing in-line artifacts.
  • Opening a document in the sidebar and then in a full-screen overlay.
  • Interacting with an image gallery/slideshow.
  • Accessing and adjusting AI settings.

In addition a prototype should be created with the following focus areas:

  • Responsiveness: Show how the layout adapts to different screen sizes (mobile, tablet, desktop).
  • Transitions: Illustrate smooth transitions when the chat opens/closes or changes display modes.
  • Interaction States: Clearly show hover, active, and disabled states for interactive elements.

Additional information / Screenshots / Prototype Videos

As preparation for this issue a UX research was done using Gemini and Claude.

Screenshots Drupal Chat

More Screenshots Drupal Chat

Screenshots Claude Deep Research

More Screenshots Claude Deep Research

Screenshots Gemini Research

More Screenshots Gemini Research

Figma wirefame from XB and demo video

Screenshot and Demo video

Figma Link XB

Todo

  • Create UX Briefing
  • Create UX and approve
  • Select important features
  • Do issue breakdown as child issues for implementation

Accept

  • We have an approved UX
  • We have selected the most important features
  • We have broken the most important features down for implementation
πŸ“Œ Task
Status

Active

Version

1.1

Component

AI Chatbot

Created by

πŸ‡©πŸ‡ͺGermany breidert

Live updates comments and jobs are added and updated live.
Sign in to follow issues

Comments & Activities

  • Issue created by @breidert
  • Hello,

    I'm currently working in the high-fidelity wireframes for desktop and mobile. The issue status is very advanced but not finished.

    You can check the design in the following Figma File: Conversational AI Interface

  • πŸ‡΅πŸ‡ΉPortugal bbruno Poland

    Current status

    The current wireframes had the idea to introduce the chatbot as a button in the side admin navigation, which when opened pushed down the rest of the menu items and allowed for the chatbot to exist on the page without having to take up any extra real estate on the screen (which also did not cover the content on the page).

    However, we have heard from the Gin team that moving forward, the navigation bar is not planned to be around for very long, or rather mostly hidden with emphasis on the top bar for tools and actions - which breaks the idea above as the chatbot needs to always be standing out and not tucked away with the rest of the navigation.

    As for now, as also suggested by the Gin team, it is best if the AI Chatbot button lives somewhere in the top bar on the far left side, similar to the current implementation in Experience Builder, to provide the most similar feeling. This would be somewhere where currently the breadcrumbs are present, or as in the wireframe, the "Take a tour" button, although it will mostly be a space for tools in the future.
    As for the chatbot placement itself, this is still to be tried out and discussed, but once again staying similar to the XB experience, the top candidate is having the chatbot grow from the left of the screen - pushing the entire page to the right. Although there are some concerns if this is not too heavy on browser resources.

  • πŸ‡΅πŸ‡ΉPortugal bbruno Poland

    Current Status

    Following up from last week's meeting where we got a pretty clear idea on how to move forward, we have created a small demo with the following suggestions that align best with Gin's ideas moving forward, and also provides a similar experience to Experience Builder in terms of design for a very nice consistency across both:

    • - Do not place any AI related functionality on the side navigation itself, this will eventually be 'hidden' in Gin, which we do not want since we want to really show off the AI button and functionalities
    • - Place the AI Chatbot button in the top left of the toolbar, this will be the designated spot for Tools in Gin and also follows what XB is also doing
    • - Have the Chatbot itself 'slide' out from the left of the screen, pushing the rest of the page to the right (Not overlayed). This allows us to not have to occupy any existing real-estate within menus or sections of the page, and also not have to overlay any content - This was important as in the future it is planned for the chatbot to provide real times changes occurring on the page which should be visible to the user. There were concerns if this will impact performance since we would be moving content and admin menus, but this so far seems to not be the case.

    There was also the suggestion that the chatbot should remember its opened state, this to ensure consistency in the case that the Chatbot takes you elsewhere, or you are navigating around the site with the Chatbot already opened

    Here is the current look of the demo presented in the AI alignment meeting with the changes above taken into account. I have also attached a video with the opening and closing animations and message design:

    However, there are still some more things left to discuss:

    How should this best be shown on mobile? The most popular idea in the alignment meeting was for the AI Chatbot to take up the entire screen when opened, but instead of sliding from the left, to be draggable from bottom/top. This allows for an easy hiding/showing of the rest of the content on the page.

    How should we best label the Assistant button? This will also impact how users will refer to the AI assistant itself. Currently it is just 'Assistant', maybe that is fine.

  • πŸ‡ΊπŸ‡ΈUnited States Kristen Pol Santa Cruz, CA, USA

    Nice! I like this, and it aligns with the placement of the AI chatbot in XB πŸ‘

  • πŸ‡©πŸ‡ͺGermany breidert

    This looks awesome

Production build 0.71.5 2024