AIRI Fork: Maintainer Integration Checklist (Indexed) ​

This document enumerates independently mergeable features and subsystems implemented in this fork. Each item is indexed for precise selection during review.

1. Chat & UX Improvements ​

I. ACT-driven chat bubble styling (token-based visual states) II. Context usage UI (token counter + visual meter) III. Configurable send key (Enter / Ctrl+Enter) IV. Persona-based auto-titling for memory summaries V. Context-width inheritance per provider/model VI. Atomic session rebuild (context compaction with recent retention) VII. Unified journaling preview (inline carousel feed)

2. Character System Enhancements ​

I. Per-character LLM configuration overrides (model, temp, tokens) II. AIRI card JSON format (full config export, no user data) III. Background bundling with card export IV. Dynamic export snapshots (outfit + expression capture) V. SillyTavern PNG import/export compatibility VI. Duplicate-safe import handling (auto-renaming)

3. ACT / Expression Pipeline ​

I. Full ACT token parsing pipeline (chat → expression/animation) II. Heuristic emotion mapping system (keyword → archetype) III. Custom expression key mapping (user-defined bindings) IV. Dynamic expression resolution (VRM expressionMap fallback) V. VRMA-trigger support via ACT tokens VI. Blended transitions (lerp-based multi-emotion fade) VII. Live2D parity (stable baseline manager)

4. VRM Animation System ​

I. Expanded VRMA preset library (24 animations) II. Per-character animation palette selection III. Idle animation randomizer ("hairball" cycle) IV. ACT-triggered animation playback V. Performance priority system (interrupt + resume idle) VI. User-uploaded VRMA support VII. VRM settings panel refactor

5. Scene & Background System ​

I. Character-scoped persistent backgrounds II. Transparent rendering pipeline (VRM + Live2D) III. Image journal → background bridge IV. AI-triggered background generation + application V. Background export with character card VI. Photo mode (composite capture with countdown) VII. Selfie-based preview system (dynamic card thumbnails)

6. Memory System ​

I. Short-term memory summaries (daily context blocks) II. Long-term journal (IndexedDB persistent storage) III. Unified retrieval system (long-term → short-term fallback) IV. Per-character memory isolation V. Immutable daily summaries

7. Artistry (Image Generation) ​

I. Native ComfyUI API integration (no middleware) II. Replicate provider integration (cost-aware UI) III. NanoBanana provider support IV. BYOW (custom ComfyUI workflow mapping) V. Workflow templates + presets VI. Prompt placeholder system (, ) VII. Auto-archival to image journal VIII. Artistry enable/disable (global + per-character) IX. Prompt stripping when disabled X. Interactive gallery widget (metadata + actions)

8. Vision System ​

I. Dedicated Vision store (separate from chat LLM) II. Full request handoff to VLM for image turns III. Drag/drop + paste image input IV. Image-aware chat history V. Multi-provider support (local + cloud VLMs)

1. Provider Integrations ​

I. OpenRouter (LLM + onboarding integration) II. Deepgram STT integration III. AWS Polly TTS integration IV. Local Whisper + Kokoro pipeline (on-device) V. Chatterbox TTS system (preset + transformation pipeline) VI. Streaming support (DeepSeek / GLM reasoning-delta) VII. Gemini Live streaming pipeline (audio queue, marker parsing, grounding UI)

10. Proactivity & Sensors ​

I. OS telemetry injection (window, app, idle) II. Activity history tracking III. System metrics (CPU/GPU, volume, time) IV. Tool-aware proactivity (dynamic tool access) V. Session milestone triggers

11. Desktop UI (Control Island) ​

I. Glassmorphic floating UI system II. Mutual exclusion (single active island) III. Emotion picker UI IV. Profile switcher integration V. Animation cycle controls VI. ScrollLock mic toggle VII. Manual mic mode (no VAD) VIII. Resource status island IX. Transcription feedback toasts X. Hover timestamps XI. UI icon standardization

12. MCP System ​

I. MCP discovery (filesystem + GitHub + curated) II. MCP management UI (settings-integrated dashboard) III. Per-tool enable/disable (titration) IV. Tool readiness monitoring V. Config templates + setup guidance VI. Canonical path resolution

13. Wardrobe System ​

I. Schema-driven outfit system II. Base vs overlay layering logic III. Interactive outfit builder (state snapshot + preview) IV. Control Island wardrobe integration V. Outfit persistence in character cards

14. Platform / Stability Improvements ​

I. Window interaction throttling (IPC protection) II. Secure CORS bypass (Electron main process) III. Environment guardrails (Node/pnpm constraints) IV. Identity-guarded model switching V. Tray position restore VI. VRM animation stability improvements VII. Electron sandbox enablement VIII. MCP config path stabilization

15. Onboarding System ​

I. Terminology abstraction (Consciousness / Speech / Hearing) II. Easy Mode (OpenRouter + Deepgram auto-config) III. Advanced Mode (manual provider selection) IV. Multi-step onboarding orchestrator V. Polymorphic UI components for setup flows

16. Live2D Parity ​

I. Unified 3-panel configuration layout II. Expression mapping (hold-to-bind) III. Compact UI mode for side panel IV. Full persistence via AiriCard