Drupal AI Release Digest: Gemini Embeddings, Mautic Tools and Views Agent Fixes

Three Drupal AI projects released updates between 27 and 29 September 2026, with different consequences for existing installations. Gemini Provider 1.1.0 adds image embeddings and updates model defaults for new installations, AI Mautic 1.0.0 reaches its first stable release while retaining a beta MCP dependency, and AI Agents Views 1.0.0-beta3 fixes crashes and ineffective operations in natural-language Views configuration.

The releases require different responses. Gemini Provider can affect model configuration and how AI Search content is chunked and indexed. AI Mautic requires a database update for existing installations and introduces permission-aware MCP tool discovery. AI Agents Views remains experimental despite fixes intended to make agent-driven configuration changes more reliable.

Gemini Provider Changes Embeddings and Model Defaults

Gemini Provider 1.1.0 adds image embeddings through gemini-embedding-2, allowing an image to be embedded alone or together with text. The release also moves the default models used by new installations to current Gemini models and improves error handling for common image and audio file types. Existing sites retain the models already stored in their configuration rather than automatically receiving the new defaults.

That distinction matters because some models configured by Gemini Provider 1.0.x are no longer available under the same conditions. The release notes tell administrators to review their choices under the AI module's default-model configuration after upgrading. Sites using gemini-embedding-2 with AI Search should also re-index their content. When no explicit chunk size is configured, the provider reports an 8,192-token input limit for the new embeddings model instead of 2,048 for the older text-only models, so newly indexed content can be divided differently. Sites that already specify their own chunk size are unaffected by that particular change.

Gemini Provider is not covered by Drupal's security advisory policy. The stable release designation therefore describes release maturity, not eligibility for coordinated Drupal Security Team advisories.

AI Mautic Reaches 1.0 With a Beta MCP Dependency

AI Mautic 1.0.0 is the project's first stable release. Its MCP submodule now works with MCP Server 2.0.0-beta5 and later after an earlier compatibility problem left clients without available tools. Tool discovery also respects the permissions of the connected account, so clients see only operations that account may use. The Mautic agent now replies in the language of the user's message even when information returned from Mautic is in another language.

Existing AI Mautic installations have an upgrade step beyond updating the package. Maintainers instruct users to run drush updb, which applies a post-update that gives an existing Mautic agent the new language instruction unless that instruction has already been customised. The MCP submodule requires MCP Server 2.0.0-beta5 or later, with Composer rejecting earlier versions. Developers extending its tool plugins should also note that the relevant plugin classes have moved to a new namespace.

The dependency leaves an important maturity distinction. AI Mautic itself has reached 1.0.0, but its MCP integration still depends on a beta MCP Server release. Teams using that path should retest tool discovery and permissions separately from the general AI Mautic upgrade. The AI Mautic project is currently outside Drupal's security advisory policy.

AI Agents Views Fixes Failures in Experimental Automation

AI Agents Views 1.0.0-beta3 addresses a different class of problems. The experimental module lets agents create and modify Drupal Views from natural-language requests, while the new beta fixes cases where incorrect display identifiers, malformed configuration and invalid tool input could cause crashes or fail to make the intended change. It also corrects operations that could be ineffective on non-default displays and adds validation around several configuration changes.

The beta resolves 27 issues and adds a Views Agent development-assistant submodule, along with expanded documentation and example prompts. Its system prompt has also been restructured to give the agent clearer guidance around reading and modifying Views configuration. These changes address known failure modes without changing the project's experimental status. The project documentation continues to tell users to review any View produced by the agent before relying on it in production.

The DropTimes previously covered the initial Views agent as an experimental approach to configuring Views through natural language. Beta3 advances that workflow with validation, reliability and development-assistance changes rather than moving it into production-ready status.

Different Releases Require Different Upgrade Actions

Upgrade requirements therefore differ across the three releases. Gemini Provider users should review configured models and re-index AI Search when adopting gemini-embedding-2 under the conditions described by the release. AI Mautic users should run database updates and retest permissions, particularly when MCP Server is involved. AI Agents Views users should rerun existing prompts and configuration workflows after beta3 while continuing to review agent-generated changes before relying on them.