Skip to main content
Here, we define some terms used in Openlayer.

​
Workspace

When you create an account, you are given a workspace. A workspace is your home in Openlayer, where all your projects live and where your team collaborates. Your plan applies to the whole workspace.

​
Project

A project lives inside a workspace. It represents an AI system or initiative and houses its data, tests, and compliance work. You can create multiple projects in your workspace. A project can use offline testing, observability, or both.

​
Offline testing

Offline testing is also called development mode. You test each version of your AI system on a dataset you curate, usually as part of CI/CD, to catch regressions before you ship. Refer to the Offline testing overview for details.

​
Observability

Observability is also called monitoring mode. Openlayer receives your production data, runs your tests on it at a regular cadence, and notifies you when tests fail. Refer to the Observability overview for details.

​
Tests

Tests codify expectations for your AI system and data. Offline tests run on the datasets you push in development mode. Monitoring tests run on production data from your data sources. Refer to the Tests overview for details.

​
Data source

A data source brings production data into a project. It is either a stream of requests from your app, sent with the SDKs or OpenTelemetry, or a table in your warehouse. Many teams use one data source per environment, such as staging and production. In the REST API, SDKs, and CLI, a data source is called an inference pipeline. For example, you send data with an inference_pipeline_id.

​
Trace

A trace is the full record of one request your AI system handles. It holds the inputs, outputs, intermediate steps, and metadata. You see traces on the Data page of your project. Refer to Instrument your code for details.

​
Request

A request is one trace sent to Openlayer. Usage and plan limits count requests: each trace counts as one request.

​
Session

A session groups related traces, such as the turns of a chatbot conversation. You identify it with a session_id. Refer to Track users and sessions for details.

​
Scorer run

A scorer run is one LLM-judge test, such as LLM-as-a-judge, evaluated on one request. Usage counts scorer runs per year.

​
Framework

A framework is a set of governance requirements, such as the EU AI Act or ISO 42001, that you activate for some or all of your projects. Openlayer provides pre-built frameworks, and you can build a custom one. Refer to the Governance overview for details.

​
Rule

A rule is one requirement inside a framework. Platform rules are satisfied automatically as your team uses Openlayer. Evidence-based rules need a document or link as proof.

​
Gateway

The Openlayer Gateway is a single endpoint between your apps and your LLM providers. It enforces budgets, model access, and guardrails on every request, and traces each request to your project. Refer to the Gateway overview for details.