Vigil

Inspiration

Field technicians spend their days troubleshooting equipment across wide territories — alone, no desk, no second opinion. Work orders arrive by email. Manuals live in binders. Answers require stopping, unlocking a phone, and typing with dirty gloves. AI exists, but it's stuck on a screen.

What it does

Vigil is a wearable AI assistant platform that gives field technicians hands-free access to AI through smart glasses and a ring controller. It uses a two-layer architecture: Vigil Core, a lightweight REST API backend, powers thin platform clients on the Even Realities G2 glasses and R1 ring. The result is an always-on AI assistant that works in environments where pulling out a phone isn't an option.

How I built it

Vigil is split into two layers by design.

Vigil Core is a Node.js REST API deployed on Railway, with GitHub-triggered auto-deploys. It handles session context, intent routing, and AI response generation. The API is intentionally thin and hardware-agnostic — any client that can make HTTP requests can use it.

The Even Hub plugin (v0.1.0) is the first platform client. It connects the Even Realities G2 smart glasses and R1 ring to Vigil Core via the Even Hub SDK, translating ring gestures and voice input into API calls and rendering responses as heads-up display text.

For work order integration, I identified ServiceNow as the ingestion path — Vigil Core will query it for active work orders, surface them contextually, and allow voice-driven status updates.

The entire stack is built and iterated solo, running on real hardware in an active field environment across a 150-mile service territory.

Challenges I ran into

The Even Hub SDK is lightly documented for third-party developers, so a significant portion of early work was reverse-engineering the plugin interface and understanding the data flow between the glasses, the R1 ring, and a custom backend. Getting reliable context persistence across short voice interactions — without over-burdening the glasses' limited display — required rethinking how sessions are managed at the API layer.

Accomplishments that I'm proud of

Shipping a working, deployed AI backend that auto-deploys from GitHub and is already integrated with real smart glasses hardware — solo, with no prior SDK documentation to lean on.

More than the technical output: Vigil is the first AI tool I've used that actually fits into my workday. I'm a senior maintenance technician covering a 150-mile territory. My hands are dirty, my phone is in my pocket, and I'm under equipment. Vigil works there.

What I learned

The hardest part of wearable AI isn't the AI — it's the interaction model. Designing for a two-line heads-up display and a single-button ring forces a level of UX clarity that screen-based apps never require. Every response has to earn its words.

What's next for Vigil

v0.2.0 targets live ServiceNow work order read/write via voice, proactive context surfacing, and a companion mobile app for configuration. The platform model means Vigil Core can power any wearable client — the glasses are the first, not the only.

Built With

• claude
• even-realities-sdk
• iot
• node.js
• railway
• rest-api
• servicenow
• smart-glasses
• voice-interface
• wearable-ai