fidzu : cms

Authors: Will Huggins, Jeremy Chinquist

Drupal AI 1.3.0 is now available, delivering the largest feature update since the module's initial release. This version focuses on three areas that organisations told us matter most:

• Keeping AI safe and accountable
• Making AI useful for content teams
• Giving technical teams the visibility they need to operate AI in production with confidence

And because Drupal AI is open source, you stay in control of your models, your data, and your infrastructure.

AI Guardrails: stay in control of every AI interaction

Trust is the foundation of responsible AI adoption. Drupal AI 1.3.0 introduces AI Guardrails, configurable checks that run before or after any AI request.

Teams can define rules to control how AI interacts with their content and data. For example, they can block sensitive data from leaving their organisation, filter harmful responses, or enforce compliance policies. All of this can be configured without writing code.

Guardrails work across all AI operations in Drupal. This gives security and compliance teams a single place to oversee how AI is used across the site.

In practice, this means AI governance becomes part of the platform itself, rather than something teams must manage separately.

One-click AI for editors, right where they work

AI should meet editors in their workflow, not force them into a separate tool.

Drupal AI 1.3.0 introduces Field Widget Actions, one-click AI workflows attached directly to content fields.

Editors can now:

• Generate images from text descriptions and attach them as media
• Create structured FAQs from existing content
• Extract addresses from images using AI and Google Places
• Plot data from a CSV into a chart field
• Generate audio summaries with text-to-speech
• Auto-fill metadata such as telephone numbers and office hours from unstructured text
• Evaluate whether the content is ready for publication and update its moderation state

Each action is backed by a customisable workflow allowing site builders to tailor AI behaviour to their organisation's needs and editorial standards without writing code.

A smarter, more flexible chatbot

Editors often need quick assistance while working on content. The built-in Drupal AI chatbot has been expanded to support this directly within the editing experience.

Open the chatbot via slide-in or full-screen mode, allowing the editor to choose the interface that best fits their workflow.

The chatbot receives page context, meaning it understands the content currently being viewed or edited.
Editors can ask questions such as:

• Make this title more engaging
• Summarise this article for social media

Because the chatbot receives the current page context, its responses relate directly to the content on the screen.
Custom loading messages replace generic spinners, allowing site builders to provide clearer feedback while AI requests are processed.

Better tools for developers and site builders

As organisations begin building AI-powered features in Drupal, developers need tools that simplify configuration and reduce boilerplate.

Drupal AI 1.3.0 includes a set of reusable form elements designed specifically for AI workflows.
These components provide consistent interfaces for working with AI providers, prompts, and structured outputs.

The new elements include:

• Markdown editor with WYSIWYG, diff view, and Word paste compatibility
• Mentions autocomplete, allowing users to type @ or { to insert prompt variables
• Provider Selector, a unified interface for selecting AI providers
• JSON Schema editor with validation for structured output configuration
• Chat History viewer that mirrors the conversational interface of modern AI playgrounds

Configuration files have also been improved. Prompts are stored in a human-readable format rather than single-line strings, making code reviews easier and reducing merge conflicts when teams collaborate on AI workflows.

New AI operation types: rerank, summarize, detect

Drupal AI's provider-agnostic architecture continues to expand with three additional operation types.

Rerank reorders search results or document lists based on relevance to a query. This is particularly useful in retrieval-augmented generation workflows.

Summarize uses lightweight summarization models instead of full language models, reducing cost and processing time.

Object Detection identifies objects in images using traditional machine learning models. The AI Validations module already uses this to verify image content automatically.

All operation types work with compatible providers, allowing organisations to change models without rewriting integration code.

Production-grade AI observability

Running AI in production requires visibility into how systems behave. Drupal AI's observability module exports spans, traces, and metrics through OpenTelemetry, the industry standard for application monitoring.

Teams are able to connect this data to platforms such as Datadog, Grafana, Sentry, or any OpenTelemetry-compatible system.

This allows engineering teams to monitor AI agent decisions, track usage and cost, and audit AI interactions across their sites.

Combined with the exclude-tags feature for logging, organisations also gain fine-grained control over what information is recorded and what remains private.

A maturing platform: consolidation and clear direction

Drupal AI 1.3.0 also simplifies parts of the platform.

AI Translate gives way to TMGMT (Translation Management Tool), aligning AI-assisted translation with Drupal's standard translation workflow.

Field Widget Actions now provide a flexible framework for AI-assisted editorial tasks. AI Content Suggestions remains available as a contrib module for teams that want to continue building on that approach.

AI Validations will use the Object Detection operation type going forward, while still allowing different validation modules to build on the same abstraction.

These changes simplify the core architecture while leaving room for contrib modules and alternative implementations.

Closing the Loop

AI in a CMS brings practical challenges around governance, editorial workflows, and production visibility.

Drupal AI 1.3.0 shows that these problems can be handled directly within Drupal.

AI can be governed, integrated into everyday publishing workflows, and monitored in production as part of the platform.

Get started

• Documentation and guides: https://www.drupal.org/project/ai
• Download Drupal AI 1.3.0
• Join the conversation: #ai on Drupal Slack

Drupal AI remains open source and provider-agnostic, allowing organisations to integrate AI capabilities while maintaining control over their infrastructure, data, and model choices.

Full release notes

This post highlights the major additions. For the complete list of changes, bug fixes, and improvements in 1.3.0, see the full release notes on drupal.org.

Read more articles about Drupal AI.