The architecture of digital experience platforms is undergoing a fundamental shift. We are moving rapidly from passive content repositories to active, intelligent systems capable of autonomous execution and deep contextual understanding. We're pleased to share another step forward today: the 1.0 stable release of the Gemini Provider module for Drupal is here.
Developed and maintained by Omedia and sponsored by Pantheon (a Gold Member of the Drupal AI initiative), this release does more than "add text" generation capabilities to your site. It integrates the multimodal capabilities of Google's Gemini models directly into the Drupal AI abstraction layer.
From Alpha to Stable
The 1.0.0-alpha1 was a proof of concept — a settings form, a text-to-text plugin, and a connection to Google's endpoints. Useful, but modest in scope.
Beta 1 added embeddings support, Drupal Recipes compatibility, CKEditor AI Assistant integration, and initial vision features. Stable release required addressing key upstream changes: updating the Gemini SDK, migrating all vector generation to the gemini-embedding-001 model following Google's deprecation of legacy embedding models (e.g., embedding-001), and adding all Gemini multi-model capabilities.
By RC-1, the module had achieved full feature parity with Gemini's multimodal API.
The stable release locks that in.
One Provider. The Entire Multi-modal Spectrum.
The Gemini Provider dramatically simplifies the architectural complexity inherent in achieving true multi-modality. Currently, the effort is addressing the need to combine multiple specialized providers—one for text, another for voice synthesis, a third for vector storage, and so on. The Gemini Provider eliminates this complex stitching process entirely.
The 1.0.0 stable release covers the full spectrum:
Text & Conversation
The Gemini Provider utilizes the latest iterations of Gemini models, specifically gemini-2.5-flash and gemini-2.5-pro, to handle complex conversational flows with streaming support. A critical feature in the 1.0 stable release is the support for strict structured JSON outputs based on schema definitions.
This capability transforms Gemini from an unpredictable text generator into a deterministic data transformation engine. Within the context of Drupal's AI layer, this means that unstructured content can be reliably parsed into specific Drupal data model according to a strict JSON schema. This ensures data integrity and allows the AI to interact with Drupal's structured content model without the risk of formatting errors or "AI slop".
Vision and Image
Through AI core module layer, the vision capabilities of Gemini are integrated directly into Drupal’s media and content workflows. This includes both text-to-image generation, where images are created based on prompts and attached to media entities, and vision capabilities for image analysis. This is particularly impactful for accessibility and SEO, as it allows for the automatic generation of WCAG-compliant alt text for uploaded images.
In an agentic context, Gemini’s vision models could be used to analyze design mockups or screenshots to generate functional code components (read Drupal Canvas integration here). This design-to-code workflow addresses the traditional friction between design tools like Figma and the Drupal implementation, as the AI can "see" the visual requirements and map them to Drupal's SDCs (single directory components). Furthermore, vision features enable the extraction of structured data from binary files, such as extracting addresses from photos of documents and plotting them onto a chart or map field.
Audio Processing
The Gemini Provider introduces advanced audio capabilities, including seamless speech-to-text transcription with automatic language detection and high-definition text-to-speech synthesis. The TTS feature utilizes five specialized voice profiles, which provide granular control over the tone, pace, and emotional expression of the generated audio (Kore, Puck, Charon, Fenrir, and Aoede).
Within the Drupal AI layer along with AI agents and other tools within the ecosystem, these voices can be used to generate audio summaries of long articles or provide interactive conversational support for users. Because these voices are steerable via natural language prompts, site builders can configure an agent to "speak in a formal tone" for administrative messages or "speak slowly and clearly" for educational content.
Semantic Search and Retrieval-Augmented Generation
Semantic search represents a fundamental shift in how users discover content. Unlike keyword-based search, which relies on literal matches, semantic search understands the underlying meaning and intent of a query. The Gemini Provider facilitates this through high-throughput vector generation using gemini-embedding-001.
These embeddings serve as the foundation for RAG pipelines, which are designed to ground AI responses in verified organizational data. By indexing Drupal content into vector databases, the system ensures that when an LLM answers a query, it draws from the site's "Single Source of Truth" rather than relying on its internal training data. This effectively eliminates the risk of hallucination in professional contexts, providing cited and factual answers grounded in the organization’s knowledge base.
Function Calling and Agentic Workflows
The most transformative capability in the 1.0 release is the support for agentic workflows via the Tools API and native Function Calling. This allows Gemini to act as an autonomous orchestrator that can trigger custom Drupal functions based on natural language instructions.
The CMS stops being a passive repository and starts acting as an intelligent agent. A single natural language instruction — "Read this policy document, summarize the changes, translate into Spanish, and set the moderation state to Needs Review" — can trigger a coordinated, multi-step Drupal workflow without manual intervention.
Get Started
The provider is available, traditionally via composer package on drupal.org:
# you can use composer directly if installed on host machine, or if you use DDEV, run composer command through it
ddev composer require drupal/gemini_provider
After package is installed, you need to enable module, and configure provider with Gemini API Key and then further configure it through settings page at `/admin/config/ai/providers/gemini`.
# with DDEV you can enable module with drush
ddev drush en gemini_provider
After module is enable, you can integrate Gemini with any kind of AI workflow that is supported in Drupal with AI module or any of its sub-modules, including AI Agents most importantly, and others.
A huge thank you to the team at Omedia for the development work, and to our partners at Pantheon for sponsoring the work and making it possible to deliver stable 1.0 version of Google Gemini for Drupal.