Deploying a Model

This guide explains how models are deployed and executed in Acklix. It focuses on deployment mechanics, lifecycle, and constraints.

What Is a Deployment

A deployment is a versioned, executable instance of a model. Models themselves are configuration objects and cannot be used for inference until they are deployed.

Each deployment is immutable. Any change to model configuration requires creating a new deployment.

Deployment Lifecycle

  1. A model configuration is validated.
  2. A deployment record is created with a version number.
  3. Tool bindings, secrets, and services are resolved.
  4. The deployment is marked active.
  5. The deployment becomes available for inference.

Preconditions

A deployment can only be created if all dependent resources referenced by the model are available.

  • All required secrets must be configured
  • Referenced services must be connected
  • Tool schemas must be valid

After Deployment

Once active, a deployment can be referenced by its deployment_id in inference requests.

Deployments are the only executable targets for:

  • Inference API
  • Conversations
  • Interfaces (Web, WhatsApp, Email)
  • Automations and scheduled tasks

Versioning

Multiple deployments may exist for the same model. Each deployment has an independent lifecycle and version number.

Existing deployments are not affected when a new deployment is created from an updated model.