For the complete documentation index, see llms.txt. This page is also available as Markdown.

Deployment

Persistent, production-ready sessions with pre-loaded models

A Deployment is a persistent session: a dedicated cloud runtime that stays active with your Pop's models pre-loaded, so requests begin processing immediately — no startup or warm-up delay.

Use one for production workloads that need low latency or run continuously: live camera monitoring, real-time alerting, interactive apps, and 24/7 analytics. For occasional or batch work, a transient Development session is more cost-effective — see Cloud for the full comparison.

Create a Deployment

A Deployment is created ahead of time, and its Pop is bound at creation rather than passed from the SDK. Once created it has a stable session UUID that clients attach to.

Create and manage Deployments with the EyePop CLI. eyepop create deployment creates one and prints the session UUID; open the CLI section in this site's navigation for the complete command reference.

Attach from the SDK

Pass the Deployment's session UUID when you connect. The models are already loaded, so you can run inference right away — you do not pass a Pop from the SDK — it is set when the Deployment is created. Change it later with eyepop patch deployment.

See Python Configuration or Node Configuration for the language-specific connection options.

The UUID can also come from the EYEPOP_SESSION_UUID environment variable.

Next steps

  • Development — the transient session to build against first

  • Pops — define the pipeline a Deployment serves

  • CLI Resources — open the CLI section in this site's navigation to create, patch, and delete Deployments

Last updated