Visual Intelligence

Two distinct things share this name; pick the right layer before writing code:

• System Visual Intelligence (iOS 26+) — the user invokes the system camera/visual-search experience (Camera Control entry point), and your app's results appear in it. This is an App Intents integration using the VisualIntelligence framework's data types. This is the headline iOS 26 feature.
• On-device computer vision — your own in-app camera/photo analysis with Vision (text, barcodes, faces, pose, Core ML classification) and VisionKit (Live Text, Visual Look Up, DataScannerViewController, subject lifting). Stable since iOS 11/16.

The deep API reference — every Vision request, the full VisionKit surface, SwiftUI camera plumbing, and the corrected iOS 26 integration — lives in references/guide.md. This file is the decision and discipline layer: read it first, open the guide for specifics.

Dials

Set these explicitly at the start; they change what "correct" means.

1. LAYER — system-vi (appear in iOS 26 Visual Intelligence via App Intents; needs iOS 26) · in-app-cv (your own Vision/VisionKit analysis; iOS 11/16+) · both.
2. MATCHER — how system-vi turns the captured frame into results: coreml (on-device model/embeddings, private, offline) · server (upload features/image, richer catalog, needs network + privacy disclosure) · vision-builtin (Vision classification/Visual Look Up, no custom model).
3. RESULT_SHAPE — single-entity (one AppEntity type) · union (@UnionValue enum over several entity types, e.g. product + collection).

When to use