Skip to content

MaskedMan-code/LogiSentinel

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

15 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

🛡️ LogiSentinel: Enterprise Fleet Diagnostics AI

Live Demo Track 1 Stack

Transforming 600-page technical manuals into instant, actionable intelligence—without compromising enterprise data privacy.

🚨 The Problem: Uptime vs. Compliance

When a commercial fleet vehicle breaks down, every minute costs money. However, building AI tools for mechanics introduces two massive enterprise risks:

  1. Data Leaks: Mechanics naturally input contextual PII (driver names, phone numbers, exact breakdown locations) into chat prompts, sending sensitive corporate data directly to third-party cloud LLMs.
  2. Hallucinations: Generic AI models often confidently guess diagnostic procedures, leading to dangerous or expensive repairs.

💡 The Solution: LogiSentinel

LogiSentinel is a secure, Retrieval-Augmented Generation (RAG) platform designed specifically for commercial fleet maintenance. It acts as a middleman between the mechanic and the AI, strictly grounding answers in proprietary OEM manuals while actively sanitizing all data before it leaves the enterprise network.


🔒 Enterprise Guardrails (Track 1 Focus)

LogiSentinel was built from the ground up for Agent Security & AI Governance. It does not allow direct client-to-LLM communication.

  • 🛡️ The Privacy Shield (PII Interception): All prompts are routed through a custom Python/FastAPI proxy. An active regex engine scrubs sensitive data (Phone Numbers, SSNs, Addresses) before the payload reaches Google Gemini.
  • 📋 Immutable Audit Logging: Every interaction, including the specific Vehicle ID accessed and the exact PII elements redacted, is logged in the backend for compliance review.
  • 🎯 Zero-Hallucination Grounding: The AI is restricted to answering based only on the active vehicle manual. If a diagnostic code is not in the electrical manual, the AI is programmed to refuse the answer rather than guess.

🏗️ System Architecture

1. Frontend (The Mechanic Interface)

  • Framework: React + Vite
  • Language: TypeScript
  • Styling: Tailwind CSS (Dark Mode optimized for low-glare garage environments)
  • Hosting: Vercel

2. Backend (The Security Proxy)

  • Framework: Python + FastAPI
  • Security: Custom Regex Data Sanitization & Rate Limiting
  • Hosting: Render

3. Intelligence Layer (The Engine)

  • Model: Google Gemini (1.5/2.5 Flash)
  • Ingestion: Google Generative AI File API (Pre-indexed 600+ page PDF manuals)

🧪 Live Demo Instructions

Want to test the security guardrails yourself?

  1. Visit the Live Dashboard.
  2. Navigate to the AI Diagnostics tab.
  3. Select any fleet vehicle from the dropdown.
  4. Try to trick the system into leaking data by typing:

    "Driver John Doe is stranded. Call him at 555-123-4567. What causes a P0171 code?"

  5. Watch the Privacy Shield successfully intercept the phone number while still delivering the correct diagnostic repair procedure.

Built for the Transforming Enterprise Through AI Hackathon - May 2026

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors