Skip to main content

Create your first pipelines

A pipeline is a flow that takes input, processes it, and produces an output. In first-use terms, it is how a camera image becomes an event, notification, stored image, or Home Assistant action.

This page is only a first pass. The full pipeline reference lives in Pipelines.

Start from the camera​

The easiest first pipeline usually starts from a camera. Camera screens may offer presets or helpers that generate a pipeline for common cases.

Use those helpers when available. They keep the first graph simpler and reduce configuration mistakes.

A simple first graph​

A useful first pipeline can be:

  1. Read a camera image or stream frame.
  2. Detect motion or objects.
  3. Restrict the result to one or more areas.
  4. Store a useful image or crop.
  5. Send a notification or expose a result.

Do not start with many branches, many cameras, or custom models. Validate one simple result first.

Use areas to reduce noise​

Areas are one of the best ways to make pipelines useful.

Examples:

  • person in Sidewalk may be normal;
  • person in Backyard may be important;
  • vehicle in Driveway may matter only if it stopped;
  • motion near Pool may need a different rule.

If your pipeline is noisy, improve areas before adding more logic.

Keep outputs modest​

For the first test, choose one output:

  • store an image;
  • send a notification;
  • show an event;
  • update a Home Assistant entity or automation later.

Avoid storing too many images or running many high-frequency flows until you understand storage and CPU usage.

When to use a processing server​

Use the main Toposync instance for the first simple flow. Consider a processing server later when:

  • CPU usage is high;
  • you run multiple cameras;
  • you add vision models;
  • the Home Assistant add-on runs on low-power hardware;
  • you want a stronger machine to handle camera or AI work.

See Processing server on Linux and macOS when you are ready.

Next: Next steps.