Skip to main content
Leapter turns business logic, such as the rules for a price, a risk score or a decision on an application, into blueprints. People draft a blueprint with AI, read it as a diagram, test it and approve it. When it runs, no AI model takes part: the same inputs always give the same outputs. This page explains the words the rest of this guide uses.

​
Built in the Lab, run on a runtime cluster

Leapter has two parts, and the runtime cluster guide is about the second one:
  • The Lab is Leapter’s web app. People build, test, review and release their logic there. The Lab Agent, the AI assistant in the Lab, drafts and changes blueprints from a description in plain language.
  • A runtime cluster runs the released logic for other systems. An order system, an AI agent or an n8n workflow sends a request to the address of a blueprint and gets the result back.
A test in the Lab only tries the logic out, and the runtime cluster does not record it. A call to a runtime cluster decides a real case, and the cluster records it as a run: when it ran, which release ran, the input and the output. An AI agent can call a blueprint, but the AI agent does not compute the result. The blueprint does, the same way every time. So you can repeat a past decision with the same inputs and the same release, and get the same result.

​
Blueprints

A blueprint is one piece of business logic: its inputs, its outputs and the steps in between, shown as a diagram that can be run. For example, the blueprint Pizza Pricing takes a pizza size, the toppings, the crust and the day of the week, and returns the price. A blueprint has no AI in it. Its steps are calculations, decisions, decision tables, loops and calls to other blueprints, so it gives the same answer for the same inputs every time.

​
Projects

A project groups blueprints that belong together, for example the blueprints of one pricing or loan process. In the Lab, you create, open, release and share projects. Every blueprint in a project is public or private. Other systems can call a public blueprint once the project is released and deployed, and its inputs and outputs are what they send and get back. A private blueprint is only called by other blueprints of the same project.

​
Releases

A release is a version of a project that never changes once it is created. Callers never run the version people are still editing: they run a release. You give each release a name when you create it, for example v1.0; the Lab suggests one. You can keep working on the project while a release runs. A runtime cluster only runs releases. It has one or more environments, for example one for trying things out and one for production. Deploying a release to an environment makes its public blueprints callable there, for callers that hold a key. Runtime clusters, environments and releases explains environments, deploying and keys.

​
What to do next