Imported from luxonis/oak-examples (tutorials/camera-demo/AGENTS.md). Install upstream with npx skills add luxonis/oak-examples --skill camera-demo. Copyright stays with the author.

AGENTS.md

Summary

This is the repository reference for the simplest multi-sensor camera preview tutorial. Use it when you need a baseline that publishes every connected camera stream with minimal processing.

Use This Example When

• You need a minimal camera-preview tutorial.
• You want to see every connected camera sensor on one device.
• You need a simple encoded-stream reference without inference.

Do Not Use This Example When

• You need stereo depth rather than raw camera previews.
• You need per-sensor custom resolutions or per-sensor logic.
• You need host-side OpenCV processing instead of Visualizer topics.

Quick Facts

• Category: tutorials/camera-demo
• Shape: script+standalone-service
• Primary task: stream every connected camera sensor to the Visualizer
• Entrypoint: main.py
• Standalone path: backend-run.sh and oakapp.toml
• Frontend: none
• Runs on: RVC2 peripheral, RVC4 peripheral, and RVC4 standalone packaging
• Requires: at least one connected camera sensor
• Input: all connected device sensors discovered at runtime
• Output: one Visualizer topic per sensor socket name
• Models: none
• Visualizer / UI: DepthAI Visualizer via dai.RemoteConnection

Read First

• README.md
• main.py
• utils/arguments.py
• backend-run.sh
• oakapp.toml

Architecture

• main.py enumerates device.getConnectedCameraFeatures() and creates one dai.node.Camera per sensor.
• Each sensor output is capped at 1920x1080 if the native resolution is larger.
• Every stream is encoded as H264_MAIN before being sent to the Visualizer.
• Topic names are the raw socket names such as CAM_A, CAM_B, and CAM_C.

Constraints

• This tutorial always publishes all discovered sensors; there is no CLI to select a subset.
• High-resolution sensors are intentionally limited to 1080p in main.py.
• The output is optimized for Visualizer preview, not for file recording or custom host processing.

Related Examples

• tutorials/camera-depth: use this when you need stereo depth instead of plain previews
• tutorials/play-encoded-stream: use this when you want to focus on encoded video playback paths
• streaming/on-device-encoding: use this when you need encoded recording rather than live preview topics

Validation

• Run: python3 main.py
• Success looks like: the Visualizer shows one topic per connected sensor and all streams respond to q
• Common failure meaning: the device has no usable camera sensors, the user expects only CAM_A, or the encoded preview path is mistaken for a raw host-processing example