Aura: Cortex Multimodal Inference and Dashboard Design

Cortex Multimodal Container and Dashboard

In the systematic architecture of Aura v11.0.0, to enhance the AI OS’s perceptual depth and human developers’ observability, we introduced two completely new and independent core components: Aura Cortex and Aura Dashboard.

1. Aura Cortex: The Brain Cortex Model Sandbox

Early architectures typically relied on a single cloud LLM or had model inference hardcoded into the logic. The birth of Aura Cortex announces the system’s full embrace of local multimodal computing streams.

1.1 Dedicated Local Inference Container

aura-cortex is designed as a dedicated local multimodal inference sandbox. It interfaces directly with local model assets (such as GGUF, Safetensors weights) stored in the root models/ directory.

1.2 Functions and Roles

Ultra-fast Response: Provides sub-second local inference for the Interaction system’s intent sampling and feedback rendering.
Privacy Shield: For physical data or user intents involving privacy, the system can completely digest it locally within the Cortex without invoking cloud APIs.
Multimodal Support: It is not just for text (LLM), but also serves as the computing hub for the system’s Vision and Speech (STT/TTS) capabilities, acting as the true “sensory cortex” of the system.

2. Aura Dashboard: The System’s Visual Observatory

A powerful AI-Native OS is like a complex black box. How can we let developers see the “gears” turning inside? This is the responsibility of aura-dashboard.

2.1 Interactive Panel Based on Modern Web Technologies

Unlike the underlying logic of Aura, which is heavily written in Rust, aura-dashboard is an independent monitoring and management panel built with HTML/JS/CSS. It connects via secure WebSocket APIs provided by the system kernel to comprehensively display the system’s internal state.

2.2 Core Display Dimensions

Process Supervision Graph: Real-time visualization of the microservice status and health of the four autonomous systems.
Causal Clock Pipeline: Graphically renders the routing path of trace_id. Developers can clearly see how a request is triggered from Interaction, planned through Inference, and finally lands in Execution.
Value and Evolution Curves: Intuitively displays the state of the system’s Variational Free Energy (VFE) and the parameter tuning trajectory within the Evolution system.

3. Conclusion

Cortex endows Aura with profound local perception capabilities, while the Dashboard unfolds the blueprint of the system’s complex operations to humans. The combination of these two makes Aura not just a powerful execution machine, but a highly transparent, observable, and “sentient” organism.

Produced by Dark Lattice Architecture Lab.