Spatial Module Technical Specifications

Every scene is a tree of nodes (objects, lights, cameras, audio sources, particle systems) with a mesh, light, camera, group, portal, terrain, water, particle, audio, trigger, collision, navmesh, annotation, ui_panel, avatar_spawn, or spawn_point type. Scenes carry a type of their own — generic, digital_twin, product_viewer, vr_environment, ar_experience, architectural, or training — which the editor and renderer use to pick sensible defaults.

Core operations exposed via the API: create / read / update / delete / publish / list scenes; add / update / move / delete nodes within a scene; fetch the full node tree recursively, or a flat node list for bulk operations.

Bulk fetch optimization: batched node fetches are capped per call (with pagination) to keep large scenes — thousands of objects — responsive rather than causing a single unbounded query.

Publishing flow: publishing a scene emits a platform event that other modules (Commerce, Smart Spaces) can subscribe to, so a published scene shows up live wherever it's referenced.

A meeting tracks its scene, template, status (scheduled → active → ended/archived), host, and participants. Each participant carries their avatar state, device info, audio state, hand-tracking data (per-hand position + recognized gestures), optional eye-tracking data, and world-space position/rotation — plus host/presenter flags.

Meetings also carry their own spatial audio zones (named, positioned, with a radius, a max-participant count, and an optional private flag) and shared whiteboards (canvas data plus authorship/last-modified metadata).

Integrations:

• Collaboration — session management, participant roster
• Avatar — full-body avatar state per participant
• Spatial audio — 3D positional audio per zone
• Gesture / hand tracking — hand-tracking input → recognized gestures
• Hand / eye data — raw device input from WebXR
• Gestures — recognized library (point, grab, wave, thumbs_up)

Meeting lifecycle: create (scheduled) → start (active) → join (adds a participant) → end (ended) → archive (archived).

Presence data tracks each user's status (active / away / focused / idle), cursor position (both viewport and world-space), current selection, and focus target (a node, the viewport, or a UI element) — including camera position for viewport focus.

Core features:

• Cursor presence — See other editors' cursors in real time with name labels
• Selection sync — All participants see which nodes are selected
• Object locking — Prevent simultaneous edits to the same node
• Undo/redo replay — Operations replay across all participants' undo stacks
• Conflict resolution — CRDT-based (Yjs), no central authority needed
• History — Operation history with replay capability

Locking: locking a node returns who holds the lock and when it was acquired; attempting to lock an already-locked node returns the current holder instead of granting a conflicting lock.

Yjs integration: each scene is backed by a Yjs document — the scene graph as a shared map keyed by node ID, with child ordering as a shared array. Concurrent edits merge automatically under CRDT semantics, with no central coordinator required.

AeionClaw integrates with MetaHuman-grade procedural animation. Given a personality profile and emotional state, it drives full-body animation procedurally — blend weights, gestures, locomotion commands, and lip-sync — without hand-authored animation clips.

Big Five personality → body language:

• Openness → gesture frequency
• Conscientiousness → posture rigidity
• Extraversion → movement amplitude
• Agreeableness → eye contact duration
• Neuroticism → micro-movement jitter

Personality maps to a MetaHuman MotionDNA preset — "friendly_neighbor," "professional," "energetic," and similar archetypes.

Emotion → facial expression: each emotion (happy, sad, angry, fearful, surprised, disgusted) maps to a weighted combination of facial action parameters (mouth, brow, eye, jaw), scaled by intensity.

Also supported: live lip-sync driven directly from an audio stream, gaze targeting (a world point, the camera, or a conversation partner), and locomotion commands (walk, run, idle, turn, teleport).

Runtime integration: the animation bridge and the multi-user scene synchronization layer both attach at character initialization time, so procedural animation and multiplayer sync work together inside any published scene.