Panels & Views

Navigate the UND interface -- chat, history, tasks, and settings

Unreal Neural Director provides a dockable editor panel with multiple views for interacting with the AI, managing tasks, and configuring the plugin.

Opening the Panel

Open UND from the editor menu:

Window > Neural Director > Unreal Neural Director

The panel opens as a standard Unreal dockable tab. Drag it to any position in your editor layout -- it persists across editor sessions.

Panel Layout

The UND panel has three main areas:

  • Top toolbar -- View switcher (Chat, History, Task Board, Settings) and active task title
  • Center area -- The active view's content
  • Bottom strip -- Agent profile selector, provider profile selector, and status bar with context usage

The toolbar lets you switch between four views, each described below.

Chat View

The primary view for interacting with the AI. This is where you send messages, watch the AI work, and review results.

Message List

The chat view displays a scrollable list of messages (up to 200 visible by default, configurable via MaxMessagesInView). Messages come in several types:

  • User messages -- Your input, shown on one side of the chat
  • Assistant messages -- AI responses with full markdown rendering
  • Tool call messages -- Collapsible blocks showing which tools the AI invoked and their results
  • System messages -- Status updates, warnings, and context information

During streaming, AI responses update at 30fps for smooth real-time output. Once the response completes, the raw text is replaced with a fully rendered markdown view supporting headings, code blocks with syntax highlighting, tables, lists, block quotes, and inline formatting.

Approval Queue

When the AI calls a tool that requires approval (based on your approval mode and auto-approval rules), an inline approval widget appears in the chat. It shows:

  • The tool name and a description of what it will do
  • An optional diff preview (for file edits) with animated expand/collapse
  • Four action buttons: Allow, Deny, Always Allow, Always Deny

Choosing "Always Allow" or "Always Deny" creates a persistent auto-approval rule in your settings, so you will not be asked about the same tool pattern again.

Multiple approval requests queue up and are processed in order.

Typing Indicator

While the AI is generating a response, an animated typing indicator appears at the bottom of the message list.

Suggestion Chips

When the chat is empty (no active conversation), the view displays suggestion chips -- quick-start prompts you can click to begin a common task.

Token Usage Bar

At the top of the chat view, a context utilization bar shows how much of the model's context window is currently in use. This helps you understand when the conversation might need condensing. The bar can be toggled via the bShowTokenUsageBar setting.

Input Area

The input area at the bottom of the chat view is a multi-line text editor with several features.

Basic Input

Type your message and press Enter to send (or click the Send button). Use Shift+Enter for line breaks within a message. While the AI is working, the Send button becomes a Cancel button to stop the current operation.

/Slash Commands

Type / to open the slash command popup with autocomplete:

CommandDescription
/newtaskStart a new conversation task
/clearClear the current chat view
/condenseAI-powered conversation summarization -- replaces older messages with a compact summary to free up context space
/compactAlias for /condense
/helpShow available commands and usage tips
/rulesLoad project-specific rules from a .undrules file

@Mentions

Type @ to open the mention popup with autocomplete. Mentions attach additional context to your message:

MentionDescription
@/path/to/file.cppAttaches the file's content to your message
@/path/to/folder/Attaches a directory listing
@https://...Fetches a web page and attaches its text content
@problemsAttaches current compiler errors and warnings
@gitchangesAttaches the output of git diff HEAD
@asset:Game/BP_MyActorAttaches UAsset metadata (class, properties, components)
@screenshotCaptures the current viewport as a PNG and attaches it for vision-capable models

History View

The History view lists all past conversation tasks. Click any entry to restore that conversation in the chat view.

Each entry shows:

  • Task title (derived from the first message or set manually)
  • Creation timestamp
  • Number of messages and tool calls

Task Board View

The Task Board provides a structured view of active and completed tasks, including their TODO lists and delegation status.

TODO Lists

Each task can have an associated TODO list, managed by the AI through the update_todo_list tool. The TODO list sidebar shows:

  • Pending items
  • In-progress items
  • Completed items

This is especially useful for complex multi-step tasks where the AI breaks down the work into discrete steps and tracks progress.

Delegation

When using the Orchestrator agent profile, the Task Board shows the delegation hierarchy -- which sub-agents were spawned for which sub-tasks, their status, and their results.

Settings View

The Settings view provides a full property editor for all UND configuration. It uses the standard Unreal IDetailsView interface, so it looks and behaves like other editor settings panels.

Settings are organized into categories:

  • Provider Profiles -- Configure AI providers, models, and API keys
  • Agent Profiles -- Customize agent behavior, system prompts, and tool access
  • Approval & Security -- Auto-approval rules, protected paths, ignore patterns
  • Knowledge & Search -- Knowledge injection, Auto-RAG, BM25 indexing
  • Checkpoints -- Conversation snapshot configuration
  • MCP Servers -- External tool server management
  • UI -- Token bar, tool details, message limits

For a full reference of all settings, see Settings.

Knowledge Browser

The Knowledge Browser panel lets you view and edit the knowledge modules that UND injects into the AI's context.

Two knowledge sources are listed:

  • UE Engine -- Built-in documentation modules covering 31 Unreal Engine systems. These are read-only and ship with the plugin.
  • Project -- Auto-generated modules describing your project's structure. These are created by the Project Scanner at editor startup and stored in Saved/UND/ProjectKnowledge/.

Knowledge Editor

Click any module to open the Knowledge Editor, which shows:

  • Module metadata (keywords, triggers, priority, token estimate)
  • Full markdown content
  • Edit controls (for Project modules)

Project modules can be edited to refine what the AI knows about your project. You can also create new project knowledge modules through the AI using the save_project_knowledge tool.

Agent Profile Editor

Accessible from the Settings view, the Agent Profile Editor lets you customize each agent profile:

  • System prompt template with variable substitution
  • Allowed and denied tool lists
  • Delegation configuration

MCP Server Editor

Also accessible from Settings, the MCP Server Editor manages external Model Context Protocol server connections:

  • Server name, command, and arguments
  • Enable/disable toggle
  • Connection status and health indicators