Skip to main content

Performance, Testing, and Risks

Performance Characteristics and Constraints​

Strong Areas​

  • Quest-specific texture formats reduce memory pressure versus desktop float textures
  • Raymarch shader is simplified for mobile
  • Foveated rendering is enabled
  • Progress UI and placeholder brain hide some load latency

Expensive Operations​

  • Full VRDF file reads into memory
  • CPU-side voxel unpacking
  • Runtime creation of Texture3D objects
  • Material and texture rebind on every modality switch
  • Aggressive unload and GC on Android during reload

Known User-Facing Risk​

The README and architecture both point to a consistent bottleneck:

  • Android reloads and modality changes can stutter or stall

This is structurally believable given the current synchronous decode and texture recreation path.

Testing and Observability​

Automated Testing​

The Unity Test Framework package is installed, but there is effectively no real custom automated test suite in the repository.

Manual Test Utility​

Assets/Scripts/UI/NetworkSmokeTest.cs is a manual smoke-test script that:

  • performs cleartext and HTTPS GET requests
  • tests websocket connectivity

This is useful for diagnostics but is not a substitute for automated tests.

Logging​

The project relies heavily on Debug.Log, Debug.LogWarning, and Debug.LogError. Several key systems are verbose:

  • VolumeDVR
  • PusherClient
  • GeminiVoiceInterface
  • PorcupineWakeWordListener

This is helpful for diagnosis, but the logging strategy is ad hoc rather than centralized.

Current Technical Debt and Risks​

Architectural Debt​

  • No custom assembly boundaries
  • High scene coupling through inspector references
  • Some static global state
  • Mixed naming quality and language conventions
  • Feature code and sample content live in the same asset tree

Functional Debt​

  • Reload path is synchronous and expensive
  • Some XR interaction features are present but only partially integrated into the main scene
  • Build profile configuration is inconsistent across assets
  • Security posture is intentionally permissive
  • Test coverage is minimal

Naming and Consistency Debt​

Examples visible in code:

  • Ochestrator directory spelling
  • Utiils directory spelling
  • converationURL field spelling
  • VolumeClipControllerr.cs

These do not break functionality, but they raise maintenance cost and reduce clarity.

Continue Reading​