> ## Documentation Index
> Fetch the complete documentation index at: https://docs.r3al.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Models and deployment

> Understand the difference between a registered model and a running endpoint.

A registered model is an immutable description of an artifact and its loading configuration. A connected runtime is the service that loads it and answers requests. A demo session is a time-limited period in which configured managed compute is enabled.

## Available today

Register an externally trained artifact, have R3AL configure a compatible runtime, and use that connection through chat and the API. The platform verifies model identity before reporting Ready. Ending a demo stops compute without deleting the registered model or its stored artifacts.

## Next capabilities

Self-service training, candidate comparisons, release promotion, rollback controls, and continuous improvement are still being integrated. Registering a model does not automatically promote it to production, and chat feedback does not automatically retrain it.

Before a production rollout, validate the task on held-out data, confirm the processing boundary, and agree on serving capacity and operational requirements with R3AL.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.