Changelog

Replace the barely-visible indeterminate ProgressBar + static "Thinking..." text

with an animated "Thinking." → "Thinking.." → "Thinking..." cycle (400ms per

frame) driven by a timer in AiChatMessageViewModel. The animation starts when

the message enters Loading state and stops when streaming/ready begins. Uses

italic styling for visual distinction from actual response content.

Desktop 1.63.0 → 1.63.1.

Duncan can now execute real actions (create tasks, events, notes, etc.) when users

ask in chat. The chat system prompt includes a dynamic intent catalog built from all

active IIntentProvider plugins. The AI responds with [ACTION] blocks containing

intent_id and slots, which are parsed and executed via the new ExecuteDirectAsync

method on IIntentEngine — bypassing the suggestion system for immediate execution.

Confirmation or error messages appear inline in the chat.

AI tray tabs (Chat/Intents/History) now show active state using the view-toggle

pattern with Classes.active bindings driven by IsChatTab/IsIntentsTab/IsHistoryTab

computed properties. Panel visibility is now VM-driven via IsVisible bindings,

removing the imperative ApplyTabVisibility code-behind.

Cloud token budgets bumped: Medium 800→1200, Long 2000→2500, MaxSentences(Long)

30→40 to accommodate intent-aware responses with JSON action blocks.

SDK 1.63.0→1.64.0 (new IIntentEngine.ExecuteDirectAsync interface method).

Desktop 1.62.2→1.63.0.

The shell-ai RAG chunk was too sparse for Duncan to explain its own

features. Expand it with detailed interaction instructions, provider

configuration details, response length behavior, and AI memory.

Add a new shell-intents RAG chunk documenting the full intent lifecycle

(signal → classification → slot extraction → execution → confirmation),

listing every available intent across all 14 plugins with descriptions

and example natural language queries.

Update APP_CONTEXT.md with a comprehensive AI & Intent System section

including a full intent catalog table, RAG explanation, and content

suggestion documentation.

This ensures Duncan can answer questions like "what are intents?",

"how do I interact with you?", and "what can you do?" with accurate,

context-aware responses from the RAG index.

Introduces ShellContentProvider implementing IIndexableContentProvider to

index shell-level features into the RAG vector index. This allows Duncan

to answer questions about global shortcuts, Universal Search, Knowledge

Graph, Dashboard, AI services, Info Panel, Cloud Sync, themes, speech

input, reminders, security, and the plugin ecosystem — without needing

plugin-specific context. Registered via CapabilityBroker so RagIndexService

discovers it alongside plugin providers during full index.

• Fix DispatcherPriority.Layout → Render (Layout doesn't exist in Avalonia)
• Revert auto-note creation from intent suggestions — intents should only

show action buttons, not auto-create notes. Auto-note stays in the

DatasetInsightOrchestrator for user-initiated insights only.

• Add Target icon (Habits plugin) and CurrencyDollar icon (Finance plugin)

to IconData with GetIcon switch cases