From One Short Command to Cameras You Can Act On.
Seven steps take a computer at your site from a box on a shelf to Detections, recordings and alerts you manage from AiServa. Most of it is choosing, not configuring.
How Do I Pair an AI Vision Bot?
- 1
Choose Add AI Vision Bot
In the AI Vision Bots app on AiServa, choose Add AI Vision Bot. AiServa shows one short command for Mac, Linux or Windows, with a single use code.
- 2
Run the command on the site computer
Paste it into a terminal on the computer at your site. The installer checks the system and any GPU, installs the AI Vision Bot as a boot service and connects out to AiServa.
- 3
Confirm the six digit approval code
The computer shows a six digit code. Confirm it on AiServa. Only then does the computer join your organisation as an AI Vision Bot.
- 4
Add cameras
Search the site network for ONVIF cameras or type an RTSP address. Choose TCP, UDP or auto, and sign in. The password is sealed to the AI Vision Bot.
- 5
Run Detections
Pick a model and its classes, set thresholds, zones and lines, and run it on a camera. Each model on a camera is one Detection.
- 6
Add rules and recording
Choose how each camera records and which rules act on its events: notify, webhook, MQTT, relay, clip or ask AI.
- 7
Watch and act
Open the live wall, scrub the timeline, review events with snapshots and clips, and watch counters grow. Every setting stays in sync.
What does the computer need?
- Operating systemWindows, macOS or Linux, x86-64 or ARM64.
- GPUOptional. NVIDIA, Apple and DirectML GPUs are used when present, otherwise the CPU.
- NetworkReach the cameras on the site network and make outbound connections to AiServa. No inbound ports.
- DiskEnough for the recording retention you choose. A reserve is always kept free.
See the Pairing and the Video Path in 3D
Watch the steps play out: the site computer pairs, cameras stream to it, events go to AiServa and live video goes straight to the viewer. Drag, pan and zoom to look around.
What Happens to Each Frame? Camera to Action in Five Steps.
A fast detector looks at every frame. Heavier models are only called when something is worth explaining, so you get intelligence without paying GPU prices on every pixel.
Detect
Every frame is scanned for the objects the use case needs.
Recognize
Objects get a class, a track ID, a pose, a reading or a code.
Analyze
Counts, zones, dwell time, speed, payload and trends are worked out.
Process
Events are confirmed over time so a single odd frame never raises an alarm.
Inform
People and systems are told: WhatsApp, dashboards, webhooks, MQTT, relays.
Show Us One Camera. We Will Show You What AI Can See.
AI Vision Bots is available at selected sites. Send a short clip from your site and one sentence on the problem, and the VYROX AI team will tell you what a Detection could do there.