Retail & Operations · Business

Recording everything. Understanding nothing — until now.

A computer-vision platform for SOP compliance and operational monitoring, with a real, working AI scoring engine for uploaded video today and live multi-camera streaming on the roadmap. This paper separates the two clearly.

The problem

Only a small fraction of standard operating procedures ever get gauged, because human spot-checks don't scale — hygiene, PPE, queue time, and staff presence all depend on someone watching, and mostly nobody is. Operators need automated, always-on compliance evidence they can hand to a regulator or a franchise auditor, not another camera feed nobody reviews.

The concept

Visionairy is designed to turn camera footage into scored, explainable compliance evidence — hygiene and PPE detection, queue and wait-time analytics, staff presence, and cleaning/safety workflow checks — without a human reviewing every frame. Each detection maps to a named parameter with a defined scoring threshold and a written remediation note, so a flagged issue tells an operator exactly what to fix.

Available today

A real, working multimodal AI scoring pipeline runs on uploaded video:

  • A working compliance-scoring engine — an uploaded clip is analyzed by Gemini's multimodal model against a structured parameter library, returning an overall score, category scores, and timestamped discrepancies with severity and a remediation note.
  • A seeded parameter library covering hygiene, food-safety, and other checks, each editable through an add/edit parameter workflow.
  • A multi-tenant data model (Firebase Auth + Firestore) with per-tenant zones, alerts, and configuration.
  • A full onboarding, alerts, sources, and reporting UI for evaluating the workflow end to end.

Upcoming features

The always-on, multi-camera promise depends on infrastructure planned as the next build phase:

  • Live RTSP/FTP camera ingestion via an edge-node agent — today, sources can be registered in the UI but no live stream connects.
  • Webhook delivery to Slack, Teams, or PagerDuty — currently a UI toggle without an outbound call.
  • Real usage-driven reporting analytics, replacing today's illustrative report figures.
  • Enforced trial-minute quotas, and a server-side proxy for the AI key currently bundled client-side.

Who it's for

  • QSR, retail, and hospitality operators needing evidence-backed SOP compliance.
  • Manufacturing and operations teams tracking safety and cleaning workflows.
  • Franchise and multi-site operators who need a regulator-ready compliance report.

Platform and stage

React 19 + Vite client, Firebase (Auth + Firestore), Gemini multimodal video analysis. Stage: MVP Live — the core video-scoring engine is real and functioning on uploaded clips; live camera streaming and alert delivery are the near-term roadmap.

This is the text version of a concept paper — the fully designed original is available as a download.

Want these insights applied to your systems?

Talk with our team about architecture, AI readiness, and product strategy.

Talk to Our Team