fixity — Intelligent Predictive Maintenance Support Agent

Inspiration

Critical infrastructure like EV chargers face constant reliability challenges that lead to expensive downtime. Predictive analytics can identify failures before they occur, but there remains a major gap between machine-generated insights and human technician execution in the field.

fixity bridges this gap. It is an intelligent support agent designed to transform predictive maintenance alerts into an ultra-fast, accurately triaged repair workflow for on-site technicians.

What It Does

1. Prioritized Ticket List

The application surfaces predictive telemetry alerts sorted by urgency and failure probability.

Technicians immediately know:

What requires attention
The optimal repair order
Which failures are most critical

2. Context View

Selecting a ticket reveals the telemetry trace — the exact sensor anomalies that triggered the prediction.

Features:

Interactive line charts
Temperature trends
Pressure variations
Historical anomaly visualization

This gives technicians full situational awareness before arriving on-site.

3. AI-Generated Checklist

A dynamic repair checklist is generated using Retrieval-Augmented Generation (RAG) based on:

Charger model
Error code
Telemetry context

The checklist:

Is cached for performance
Tracks completion state
Updates ticket status automatically

4. AI Support Chat

Technicians can interact with fixity during repairs through an embedded AI assistant.

Capabilities:

Maintains conversation history per ticket
Uses full ticket context for multi-turn reasoning
Supports hands-free interaction:
- Voice input (Web Speech API)
- Camera/photo input

5. Guided Troubleshooting (RAG)

fixity includes 24 mock repair manuals covering:

ABB
Tritium
ChargePoint
Tesla chargers

The system retrieves:

Diagnostic procedures
LOTO (Lock Out/Tag Out) safety protocols
Model-specific repair steps

6. Ticket Completion Workflow

When all checklist items are completed → ticket automatically transitions to completed
Unchecking an item → ticket reverts to in_progress

This ensures workflow integrity without manual state updates.

How We Built It

Backend (Deployed on Render)

Runtime: Python 3.10+
Framework: FastAPI + Uvicorn
AI / LLM: Google Gemini (gemini-3-flash-preview) via LangChain
RAG Pipeline: LangChain + ChromaDB Vector Store
Embeddings: Google Embeddings (gemini-embedding-001)
Validation: Pydantic v2
State Management: Basic MySQL dt

Frontend (Deployed on Vercel)

Framework: React 19 + TypeScript
Build Tool: Vite
Styling: Tailwind CSS v4
Routing: React Router v7
Charts: Recharts (Telemetry Visualization)
Animations: Motion (Framer Motion)
Icons: Lucide React

Business Impact

Maximize Uptime: Pre-emptive repairs happen before a driver encounters a broken charger, preserving customer trust.
Reduce Operational Costs: By guiding techs efficiently and diagnosing the exact part needed before the truck rolls, we minimize time-on-site and eliminate costly "return trips."