Concepts

How the EyePop.ai platform fits together

EyePop.ai turns images, video, and live streams into structured predictions. You describe what to analyze, and the platform runs the models and returns JSON.

Core building blocks

• Pop — a reusable vision pipeline. Combines one or more models or abilities into a workflow (detect → crop → OCR, and so on).

• Ability — a single preconfigured vision task (detect, classify, OCR, describe). Browse ready-made ones in the Abilities Hub.

• Source — one piece of media handed to a running Pop: an uploaded file, an HTTP URL, an RTSP camera. Each carries its own options for frame rate, region of interest, and motion gating.

Ability inference is metered in Compute Units and sessions are metered per minute; see Pricing for plans and rates.

Where Pops run

• Cloud — hosted and managed by EyePop. Start in Development with the SDK, then promote to a persistent Deployment for production.

• On-Premise — the runtime inside your own infrastructure, on CPU, NVIDIA GPU, Jetson, or Qualcomm hardware.

Both are covered in Deploying.

Next steps

• Quickstart — the shortest path to a first prediction

• Pops — build a pipeline from one or more abilities

• Sources and Options — every way to submit media, and the options that shape how it is processed

• SDKs — call EyePop from Python or Node

• Agent Skills — let your coding agent drive the CLI and SDKs for you