Visual Dialog Tree Editor for Mobile Apps

We develop visual dialog tree editors for mobile applications — one of the most technically challenging UI components. We’re talking infinite canvas, multi-touch gesture systems with simultaneous recognizers, rendering of thousands of elements, full undo/redo, and graph serialization. Over the last 5+ years we’ve shipped more than 50 such editors for chatbots in banking, e-commerce, and support. Our engineers are equally comfortable with native stacks (Swift, Kotlin) and cross-platform frameworks (Flutter, React Native). We guarantee correct behavior across all target devices and compliance with App Store and Google Play guidelines. Our visual builder has been used by 20+ companies.

Source: Industry standards for dialog editors (W3C Dialog Technologies)

Visual Dialog Tree Editor: Infinite Canvas Implementation

The editor’s foundation is an infinite, scalable space for placing nodes. Two fundamentally different approaches are compared below.

On iOS, a UIScrollView with contentSize 5000×5000 pt is a simple start. Nodes are UIViews inside a contentView. Zoom via minimumZoomScale/maximumZoomScale. Drawback: text becomes unreadable when zoomed. Fix: invert the transform on labels in scrollViewDidZoom. A custom canvas using CATransform3D gives more control: zoom-to-fit, snap-to-grid. Gesture conflicts are resolved with UIGestureRecognizerDelegate: pan on a node = drag, pan on empty space = panning.

In Jetpack Compose, use Modifier.graphicsLayer with pointerInput for detectTransformGestures.

Why Proper Gesture Management Matters

The editor simultaneously uses pan, pinch, long press, and tap. Without clear prioritization, user actions will be misinterpreted. On iOS we use UIGestureRecognizerDelegate with shouldRecognizeSimultaneouslyWith for pan and pinch, and hit-test at gesture start determines whether a node or canvas is being dragged. On Android we use GestureDetector with different listeners. This approach eliminates false triggers during intensive work.

Rendering Edges (Arrows)

Bezier curves for connections between nodes look better than straight lines. A cubic curve from an output handle to an input handle:

let path = UIBezierPath()
path.move(to: startPoint)
let controlPoint1 = CGPoint(x: startPoint.x + (endPoint.x - startPoint.x) * 0.5, y: startPoint.y)
let controlPoint2 = CGPoint(x: startPoint.x + (endPoint.x - startPoint.x) * 0.5, y: endPoint.y)
path.addCurve(to: endPoint, controlPoint1: controlPoint1, controlPoint2: controlPoint2)
edgeLayer.path = path.cgPath

Each edge is a CAShapeLayer with strokeColor and an arrowhead. When a node moves, we recalculate all CAShapeLayer.path for its incoming and outgoing edges. Use CATransaction.begin(); CATransaction.setDisableActions(true) for instant updates. When there are many edges (50+), render them via a single CALayer with a draw method: one CGContext.addPath() for all edges is faster than 50 separate layers.

Interactive Edge Creation

The user drags from an output handle of one node to an input handle of another to create a new edge. Implementation:

1. UILongPressGestureRecognizer (duration 0.0) on the handle view — start drag
2. On .began, create a temporaryEdgeLayer, start point fixed on the handle
3. On .changed, update the end of temporaryEdgeLayer to the finger position
4. Hit-test on each .changed: canvas.hitTest(location, with: nil) — if we hit an input handle, highlight it
5. On .ended, if under the finger there is an input handle, create a permanent edge; otherwise remove the temporary layer

Prevent invalid connections: cannot connect a node to itself, cannot create duplicate edges, cannot create cycles (if the graph must be a DAG). Cycle check: DFS from the target node. O(V + E) — fast for typical dialog trees.

Editing Nodes

Tap on a node opens an inline editor. For a dialog node of type "Message": a UITextView with rich text support for the message content, a type switcher (text/media/buttons), and a button list with add/remove capability.

Bottom sheet (iOS 15+ UISheetPresentationController with .medium and .large detents) — the user sees both the graph and editor in medium mode. Keyboard avoidance: when the keyboard appears, increase the bottom padding of the UIScrollView inside the sheet by keyboardFrame.height - sheet.frame.height + sheetMaxHeight. Handle via UIKeyboardWillShowNotification.

Undo/Redo

Command Pattern — every action (move node, create/delete edge, edit content) is implemented as a command object with execute() and undo() methods.

protocol GraphCommand {
    func execute()
    func undo()
}

struct MoveNodeCommand: GraphCommand {
    let node: GraphNode
    let fromPosition: CGPoint
    let toPosition: CGPoint
    func execute() { node.position = toPosition }
    func undo() { node.position = fromPosition }
}

UndoManager — built into iOS UIKit, works with registerUndo(withTarget:handler:). Or a custom CommandStack with undoStack: [GraphCommand] and redoStack: [GraphCommand]. Grouping: moving multiple selected nodes is one undo operation.

Selection and Multiple Selection

Lasso selection: pan on empty space while holding down — draws a selection rectangle via CAShapeLayer with fillColor = UIColor.blue.withAlphaComponent(0.1) and strokeColor = UIColor.blue. On gesture end, check all nodes for intersection. Selected nodes get a blue border via layer.borderColor. Pan on selection moves all selected nodes simultaneously. Delete removes all selected nodes and incident edges.

Serialization and Synchronization

Graph → JSON with a complete description of nodes and edges. Auto-save every 30 seconds or on each change with a 2-second debounce. On iOS, use Combine.debounce(for: .seconds(2), scheduler: RunLoop.main) on a graph change publisher, then URLSession.shared.dataTask to send to the server.

Conflict resolution for collaborative editing: Last-Write-Wins at the node level with a server timestamp. Each node has updatedAt: Date — during merge, take the version with the later timestamp.

What's Included in the Work

• Research and design of the node structure for your chatbot
• Implementation of a custom canvas with infinite scroll and zoom
• Undo/redo system and multiple selection
• Integration with your backend via REST/GraphQL
• Support for all node types (text, media, buttons, conditions)
• Testing on target devices and publication to stores
• Documentation and training for your team

Timeline: from 1 week to 3 months. Pricing is determined individually after analyzing your requirements. Development costs start from $5,000 for a basic web-based version and can range up to $50,000 for a full native solution with advanced features. Contact us — we’ll evaluate your project and propose the optimal solution.

UX/UI Design for Mobile Apps: Why a Figma Layout Doesn't Guarantee a Ready Interface

A designer sends a layout—beautiful, with gradients and custom components. The developer opens it and realizes: the button is 36pt, the tap target is 20pt. On an iPhone SE, it's physically impossible to press with a thumb. The bottom sheet covers content when the keyboard appears. Navigation is built against the native iOS model. Apple will reject the app, or users will leave within a week—depending on how lucky you get with the review.

We have been designing mobile UX/UI for over 5 years and have seen hundreds of such situations. During this time, we have designed and helped launch 30+ mobile apps—from fintech products to social networks. You don't need to guess whether the design will pass App Review or Google Play—we embed platform requirements from the first screen. We'll assess your project in one day, contact us.

Mobile UX/UI is not an adaptation of web design. It is a separate discipline with specific platform constraints: safe area, touch gestures, UIViewController lifecycle, Activity state management.

Why Can't You Ignore Human Interface Guidelines and Material Design 3?

Apple HIG and Google Material Design 3 are not aesthetic recommendations. They are documented user expectations formed by years of using system applications. Expectations confirmed by user experience research on mobile platforms (User experience design).

HIG defines: minimum tap target 44×44 pt, safe area insets for notch and Dynamic Island, standard gestures (swipe back on iOS, back gesture on Android 10+). Ignoring safe area is a common mistake. safeAreaLayoutGuide in UIKit and safeAreaPadding in SwiftUI exist precisely for this. A designer who doesn't set safe area margins in Figma guarantees a bug during development.

Material Design 3 introduced Dynamic Color—the color scheme is generated from the user's wallpaper via MaterialTheme.colorScheme in Jetpack Compose. An app that ignores dynamic colors on Android 12+ looks out of place. This is not critical for niche products but is noticeable in mass-market apps.

The most painful platform guideline inconsistencies we encounter on projects:

• Custom navigation on top of system navigation. iOS users expect swipe back from any point on the left edge of the screen. A custom NavigationController without interactive gesture breaks this. Android users expect the system back button—a custom back button in the left corner does not fully replace it.
• Modal windows instead of navigation push. Bottom sheets are appropriate for actions, not for navigating content.
• Missing haptic feedback. UIImpactFeedbackGenerator on iOS is not decoration but part of the interface response. Buttons, swipes, and confirmation actions without tactile feedback feel broken.

Table: Comparison of iOS and Android UX/UI Requirements

How to Get the Most Out of Figma?

The Figma Variables API has changed the workflow. Design tokens—colors, typography, radii, spacing—are stored as variables and exported directly to code via figma-tokens or style-dictionary. This eliminates manual value transfer and desynchronization between design and implementation. Practice shows: Figma Variables speeds up asset handoff to development by 2–3 times compared to static frames, and using design tokens reduces code transfer errors by 60%.

Auto Layout with wrap and spacing between elements allows building components that behave like flex containers. A developer opens a component and sees not a static artifact but a description of behavior at different content sizes.

Component Properties—variants, boolean toggles, instance swaps—enable building a full design system right in Figma. A button with 4 states (default, hover, pressed, disabled), 3 sizes, and 2 icon variants is one component, not 24 frames.

Figma Prototype with Variables allows creating an interactive prototype with real state: showing how the screen changes with different variable values. This is no longer just a "clickable layout" but a full UX testing tool.

How to Benefit from Prototyping and UX Testing Before Development?

The most expensive mistake in a mobile product is to develop a feature, release it, and discover that users don't understand how it works. A Figma prototype at the testing stage costs zero development hours. Redoing a finished screen costs days. Testing a prototype before development begins reduces the number of fixes by 80%.

For usability testing, we use Maze (task testing on a prototype—the user goes through a scenario, we get heatmaps and mis-click rates) or direct sessions via UserTesting. Key metrics are task completion rate and time on task, not "like/dislike."

A/B testing on mobile is harder than on web: the App Store doesn't allow UI changes without an app update. Therefore, it's important to test hypotheses on a prototype before release, not through production experiments. According to research, fixing a bug found on a prototype costs 10 times less than after production release. And average task completion time increases by 40% after proper UX optimization during prototyping.

Why Are Animations Critical for Interface Perception?

Animations in mobile apps are feedback. An element doesn't appear instantly—it transitions to the desired state over 200–350 ms. This gives the brain context to understand what happened.

• iOS: withAnimation in SwiftUI, UIViewPropertyAnimator in UIKit for interactive animations with interruption capabilities. Spring animations with dampingRatio are the basis of most Apple system transitions.
• Android: AnimatedVisibility, animateContentSize, Crossfade in Compose. MotionLayout for complex scenes with multiple transformations.
• Flutter: AnimationController + Tween, Hero animations between screens, Lottie for After Effects exports. Lottie is especially effective for onboarding illustrations and empty states.

The key constraint is 16 ms per frame (60 fps) or 8 ms (120 fps on ProMotion devices). Animations must run on the GPU via CALayer/RenderThread, not on the CPU via layoutSubviews. Profiling via Core Animation instrument in Xcode is a mandatory step before releasing animated screens.

Why Is Accessibility Not an Optional Feature?

VoiceOver on iOS and TalkBack on Android are used by up to 15% of users—this statistic is confirmed by accessibility research described in Accessibility (Wikipedia). In absolute numbers for a large app, this is thousands of people. Additionally, App Store rejections due to accessibility occur, though rarely.

Minimum checklist:

• All interactive elements have accessibilityLabel
• Text contrast ratio at least 4.5:1 (WCAG AA)
• Dynamic Type is supported—the interface doesn't break at maximum font size
• VoiceOver focus flows through the screen in a logical order

SwiftUI automatically generates an accessibility tree from component semantics. UIKit requires manual setup of accessibilityTraits, accessibilityHint, and grouping via shouldGroupAccessibilityChildren.

What Does the Work Include?

The UX/UI design deliverables include:

What Is the Process and Timeline?

Design goes through stages: research and competitive analysis → user flows and wireframes → design system → UI layouts → prototype → testing → handoff to development.

Timeline estimates:

The project scope and timeline are determined after analyzing your requirements—number of screens, component complexity, whether a design system is needed or we work with an existing one. Get a consultation for your project—contact us for a preliminary assessment. Order a complete mobile app design—we'll assess your project in one day and propose the optimal work scope.