Skip to content

02-architecture-overview.md

Latest commit

 

History

History
64 lines (49 loc) · 2.34 KB

02-architecture-overview.md

File metadata and controls

64 lines (49 loc) · 2.34 KB

🏗️ Architecture Overview

NeurOps is built on a distributed, event-driven architecture that separates hardware connectivity, telemetry processing, and AI intelligence.

🗺️ System Map

graph TD
    subgraph Infrastructure ["1. Infrastructure & Chaos Layer"]
        Sim1[Redfish Endpoint 1]
        Sim2[Redfish Endpoint 2]
        Sim3[Redfish Endpoint 3]
        Proxy[Chaos Management Proxy]
    end

    subgraph Intelligence ["2. Intelligence Layer"]
        Sight[Neurosight Collector]
        PubSub[Google Pub/Sub]
        BQ[(BigQuery Warehouse)]
    end

    subgraph Interaction ["3. Interaction Layer"]
        ADK[NeuroTalk AI Agent]
        UI[Streamlit UI]
    end

    Sim1 & Sim2 & Sim3 --> Proxy
    Proxy --> Sight
    Sight --> PubSub
    PubSub -.-> BQ
    ADK -- Tools --> Sight
    ADK -. Tools .-> BQ
    UI -- Chat --> ADK
Loading

층 Layers Breakdown

1. Infrastructure & Chaos Layer

  • Redfish Endpoints: Standardized hardware interfaces speaking the DMTF Redfish protocol. NeurOps includes high-fidelity emulators for validation, but is designed to target real data center hardware directly.
  • Chaos Management Proxy: A production-grade middleware that intercepts telemetry for advanced chaos engineering, fault injection, and traffic shaping.

2. Intelligence Layer

  • Neurosight Collector: The heartbeat of the system. It scales horizontally to poll thousands of endpoints, apply statistical anomaly detection, and batch data for cloud ingestion.
  • Google Cloud Data Stack: Uses Pub/Sub for real-time messaging and BigQuery for long-term telemetry storage and trend analysis.

3. Interaction Layer

  • NeuroTalk (AI Agent): Powered by Google ADK and Gemini. It is "tool-aware," meaning it knows how to query both live status and historical data to answer complex natural language questions.
  • Streamlit Dashboard: A high-visibility interface for monitoring server health and chatting with the assistant.

🛠️ Tech Stack

  • Linguistics: Python 3.12 (standardized on mylab venv).
  • Frontend: Streamlit.
  • APIs: FastAPI / Uvicorn.
  • AI: Google ADK / Gemini 3 Flash.
  • Infra: Docker Compose.
  • Cloud: Google Cloud (Pub/Sub, BigQuery).

Important

To understand how data moves through these layers in real-time, see the Project Flow Guide.