Drupal AI 1.2.0-alpha1 Released and Open for Testing

Drupal AI 1.2.0-alpha1 has been released as an early preview of the forthcoming 1.2.0 feature set, inviting site builders and developers to begin testing its new capabilities. Marcus Johansson shared the announcement in a detailed article emphasized that upgrade paths will arrive during the beta phase, with additional enhancements planned before the final release.

The Field Widget Actions module now adds configurable buttons to any entity form, allowing editors to generate or refine field content through AI suggestions. Editors may request categories, tags, summaries or titles and then accept or retry results; this functionality integrates with AI Content Suggestions, AI Automators and AI Agents to support everything from simple presets to fully autonomous workflows.

Prompt Library support enables modules to ship context-aware AI prompts as configuration. Third-party extensions can bundle best-practice prompts or share them across projects via reusable recipes. This approach standardizes context engineering and eliminates guesswork when setting up AI behavior.

AI Content Suggestions have been extended to Blocks and Taxonomy Terms as well as Nodes, broadening AI-driven assistance across more Drupal content types. Site authors may now leverage automated metadata creation and text generation on a wider range of entities.

A new mocking library captures real AI provider responses for offline replay during development and testing. Automated test suites will run faster and incur lower costs by simulating provider interactions without external network calls. The AI core foundation now accepts additional file types, including PDFs and video formats, to enable advanced vision and document processing use cases.

Three new AI Automator types simplify common content tasks: image alt text generation for accessibility, image filename rewriting for SEO and summary generation for Text with Summary fields. A unified abstraction layer for vector databases means recipes work with any supported service, such as Milvus, Postgres or Pinecone, without vendor-specific configuration.

Documentation is now versioned alongside module releases so users access the correct guides for each major or minor version of Drupal AI. A provider helper abstraction simplifies creation of OpenAI-compatible integrations, making it easier for contributors to support providers like Mistral, Anthropic and Ollama.

For full release notes, detailed configuration examples and versioned documentation, visit the Drupal blog post.