AI Model Registry

AI Model Registry is a Drupal-native catalog of AI provider and model metadata, stored as configuration entities. It gives governance, routing, budgeting and evaluation modules a single, editable source of truth for what each model is and how it behaves — so they never have to hard-code provider details.

The problem it solves

As soon as more than one module needs to reason about AI models — which provider runs where, what a call costs, whether a model may be used for regulated data, when a model reaches end-of-life — that knowledge gets duplicated and hard-coded. AI Model Registry centralises it in one config-entity catalog that site builders can edit and that other modules can query.

What each model records

• Capabilities — what the model does (chat, embeddings, and so on).
• Data residency — where the model runs and where data goes.
• Training use — whether the provider may train on your inputs.
• Lifecycle and end-of-life — deprecation and EOL signals.
• Risk level and a local-deployment flag for on-prem models.

Features

• Config-entity catalog — models are exportable, version-controllable configuration, editable through a full admin UI.
• Normalized repository service — ai_model_registry.model_metadata_repository exposes all() and find(provider, model), keyed by provider and model id.
• Pluggable adapters — a ModelMetadataAdapter plugin type lets you feed metadata from other sources (a provider API, a pricing feed) behind the same interface; a config-entity adapter ships by default.
• Ships with seed models for common local and hosted providers.

Requirements

• Drupal 10.3+ or 11

Standalone — no paid provider or extra dependency is required.

Who uses it

Companion modules consume the registry optionally: AI Policy Gateway for residency, risk and budget decisions; AI Grounding and AI Index Health for model metadata.

Installation

composer require drupal/ai_model_registry

Enable the module and manage models under the AI administration section.