🚗 AutoNexus — AI-Powered Predictive Vehicle Maintenance Backend

A production-grade FastAPI backend that ingests real OBD-II sensor data from 81+ vehicles and runs a multi-agent AI system to predict component failures before they happen.

My role: Backend lead & integration owner — FastAPI server, database design, master agent orchestrator, and integration of ML models, frontend, and notification agents built by teammates.

---

What It Does

• Predicts vehicle failures 7–90 days in advance using an XGBoost model (94% accuracy)
• Coordinates 8 specialized AI agents — diagnosis, engagement, scheduling, feedback, manufacturing insights, UEBA security, and more
• Sends time-aware alerts via Twilio voice calls, SMS, and email (no midnight calls — 9 AM–8 PM only)
• Generates RCA/CAPA PDF reports for fleet-wide failure pattern analysis
• Monitors for anomalous behavior via UEBA (Isolation Forest) security agent

---

Tech Stack

• Framework: FastAPI 0.100+
• Database: SQLite (dev)
• ML: XGBoost, Scikit-learn, SHAP
• Agents: LangChain / LangGraph orchestration
• Notifications: Twilio Voice + SMS, SendGrid email
• Auth & Security: Role-based access control, UEBA anomaly detection
• Docs: Auto-generated Swagger UI at /docs

---

Quickstart

1. Clone & install

git clone https://github.com/Rashmivid/Autonexus-backend.git
cd Autonexus-backend
pip install -r requirements.txt

2. Configure environment

cp .env.example .env

Edit .env with your credentials:

DATABASE_URL=sqlite:///./autonexus.db      
TWILIO_ACCOUNT_SID=your_sid
TWILIO_AUTH_TOKEN=your_token
TWILIO_PHONE_NUMBER=+1xxxxxxxxxx
SENDGRID_API_KEY=your_key

SQLite works out of the box with no extra setup — recommended for local development.

3. Seed the database

curl -X POST http://localhost:8000/admin/seed

4. Run the server

uvicorn main:app --reload

Visit http://localhost:8000/docs for the full interactive API.

---

Key Endpoints

Method	Endpoint	Description
GET	/vehicles	List all vehicles with health status
GET	/vehicles/{id}	Vehicle details + sensor readings
POST	/vehicles/{id}/analyze	Run full AI agent analysis
POST	/vehicles/{id}/notify	Trigger time-aware Twilio alert
POST	/vehicles/{id}/book	Schedule a service appointment
GET	/vehicles/{id}/feedback	Retrieve feedback survey
GET	/manufacturing/insights	Fleet-wide RCA/CAPA report
GET	/security/ueba	UEBA anomaly detection status
POST	/admin/seed	Seed database with OBD-II data
POST	/admin/randomize	Randomize sensor readings for demo

Full docs at /docs (Swagger) or /redoc (ReDoc).

---

Agent Architecture

MasterAgent (Orchestrator)
├── DataAnalysisAgent     — Sensor threshold analysis, anomaly detection
├── DiagnosisAgent        — XGBoost failure prediction (94% accuracy)
├── EngagementAgent       — Time-aware voice/SMS/email notifications
├── SchedulingAgent       — Urgency-based appointment booking
├── FeedbackAgent         — 15-question post-service survey
├── ManufacturingInsightsAgent — Fleet RCA/CAPA PDF reports
└── UEBAAgent             — Isolation Forest security monitoring

---

Dataset

Real OBD-II sensor data — DOI: 10.35097/1130

• 81 vehicles, 1,200+ samples
• 6 sensor features: brake temp, oil pressure, engine temp, tire pressure, brake fluid, mileage
• Hosted on Hugging Face: divyanshi-02/autonexus-p2-ml-models

---

Project Structure

Autonexus-backend/
├── main.py                  # FastAPI app, startup, routing
├── models.py                # SQLAlchemy ORM models
├── agents/
│   ├── master_agent.py      # Orchestrator
│   ├── diagnosis_agent.py   # ML prediction
│   ├── engagement_agent.py  # Notifications
│   ├── scheduling_agent.py  # Booking
│   ├── feedback_agent.py    # Survey
│   ├── manufacturing_agent.py
│   └── ueba_agent.py        # Security
├── routers/
│   ├── vehicles.py
│   ├── admin.py
│   └── manufacturing.py
├── load_vehicles_from_OBD.py  # Data seeding
├── requirements.txt
└── .env.example

---

Part of a Larger System

This repo is the backend. The full AutoNexus system also includes:

• Frontend: React/Vite dashboard (separate repo)
• ML models: Hosted on Hugging Face

---