MaintenanceAI 👁️

MaintenanceAI is a next-generation AI-powered Application Management Services (AMS) Copilot and Global OSINT telemetry dashboard. It acts as a hyper-intelligent Site Reliability Engineer (SRE), instantly ingesting millions of logs across distributed enterprise systems, dynamically visualizing active incidents, and generating real-time Root Cause Analysis (RCA) via generative AI models.

🌟 Key Features

1. High-Throughput Rust Ingestion Engine
   • Built on Axum and asynchronous Rust, the ingestion API handles OpenTelemetry (OTLP) Protobufs, AWS CloudWatch logs, Azure Monitor feeds, and generic webhooks, instantly normalizing them into a unified JSON format and publishing them to Redis Streams (or Kafka).
2. Real-Time Generative RCA (Python AI Worker)
   • A background Python daemon continuously tails the Redis log queue using a massive global rolling time-window buffer.
   • When a crash or anomaly occurs, it gathers context across disparate microservices and dispatches it to Ollama or Gemini to generate a cascading "Event Timeline Analysis" identifying exactly how a failure in one service cascaded into another.
3. Interactive AMS AI Dashboard (Vue 3)
   • A fully responsive enterprise dashboard showcasing live log streams and critical service degradations.
   • Features a Full-Screen AI Copilot module (FastAPI powered) that allows SREs to chat directly with their logs and render generative data visualizations (AG-UI Protocol).
4. Global OSINT Intelligence Dashboard
   • An independent mapping application leveraging CartoDB DarkMatter that polls open-source global APIs (NASA FIRMS active fires, USGS Earthquakes, ADS-B Flight Transponders, SANS DShield).
   • Cloud Vision Integration: Intercepts live global CCTV feeds and processes them through the qwen3-vl:235b-instruct multimodal vision model to detect physical world anomalies (e.g., traffic congestion, accidents) and streams those alerts back into the central AMS AI Engine.
5. Enterprise-Scale Simulators
   • Bundled with powerful Python chaos-monkeys (enterprise_simulator.py) that mock a massive Amazon-scale architecture (Storefront, Payment Gateways, Fulfillment Kiva robots, Zero Ledger accounting) and randomly inject sophisticated deadlocks, network timeouts, and x509 certificate failures.

---

🏗️ Architecture Overview

The system is highly decoupled and composed of five main layers:

graph TD
    A[Enterprise Simulators / OSINT Telemetry] -->|POST 8080| B(Rust Ingestion Backend)
    B -->|Publish| C[(Redis Stream / Kafka)]
    C -->|Subscribe| D[Python AI Worker]
    D -->|HTTP| E[Ollama / Gemini Cloud]
    D -->|Push RCA| C
    F[Vue Dashboards] -->|GET 8001| G(FastAPI Copilot Backend)
    G -->|Read RCA| C

Loading

---

🚀 Quickstart (Docker)

The fastest way to experience MaintenanceAI is via the bundled multi-stage Docker container, which compiles the Rust server, the Vue frontend assets, and provisions the Python environment automatically.

Prerequisites

• Docker & Docker Compose
• An API Key from Ollama Cloud (or a locally running LLM).

1. Configuration

Rename or copy config.yaml.sample to config.yaml in the root directory and securely insert your API keys.

2. Build & Run

From the repository root, run the bundled compose file:

docker-compose up --build -d

3. Access the Platforms

Once the container boots, it exposes the following distinct interfaces:

• AMS SRE Dashboard: http://localhost:3001
• Global OSINT Dashboard: http://localhost:3000
• Copilot / Vision API (FastAPI): http://localhost:8001
• Rust Ingestion API: http://localhost:8080

(Note: Upon startup, the enterprise_simulator immediately begins firing mock traffic and injecting catastrophic system failures. Open the AMS Dashboard and watch the AI instantly deduce the root causes!)

---

💻 Manual Setup & Development

If you wish to modify the code or run components natively:

1. The Rust Ingestion Server

# Ensure Redis is running locally on port 6379
cargo build
cargo run

2. The Python AI Worker & API

cd ai_worker
pip install -r requirements.txt
# Export your key for local shell dev
export OLLAMA_API_KEY="your-key"
# Start the background worker
python3 worker.py &
# Start the FastAPI Copilot backend
uvicorn api:app --host 0.0.0.0 --port 8001 &

3. Vue Dashboards

# Run the AMS Dashboard (Port 5174)
cd dashboard
npm install
npm run dev -- --port 5174

# Run the OSINT Dashboard (Port 5173)
cd ../osint_dashboard
npm install
npm run dev -- --port 5173

4. Trigger the Chaos

cd otel_simulator
pip install -r requirements.txt
python3 enterprise_simulator.py

---

🧠 Configuration Details

The central config.yaml governs the behavior of the AI heuristics:

• queue.type: Switch between redis or kafka buffering.
• heuristics.context_window_size: Determines how many chronological logs the AI sees prior to a crash (Default: 50).
• heuristics.trigger_levels: Which log severities immediately dispatch to the LLM (e.g., ERROR, FATAL).

Screenshots

🛡️ License

MIT License. See LICENSE for more details.