Daily Digest

2026 Capabilities Progress

1. Page generation — Canvas is advancing the underlying infrastructure needed for AI-driven page assembly, including new HTTP APIs for managing page regions [175][197] and active test development for the page builder and template builder agents [192][201][208]. The next milestone is completing agent test coverage to a standard that supports a stable release; the risk is that scope and quality expectations for AI features have not been formally defined, which could slow prioritisation decisions.
2. Context management — This is the most active capability today. The Context Control Center (CCC) is approaching its release candidate, with a completed increase to maximum context items [5][17], a new extensions discovery page in draft [4][20], pre-release documentation merged [14], and a compatibility update to AI Core 1.4 under review [1][19]. A risk worth noting: two UI redesign tasks are temporarily unassigned [2][3], and the AI Core 1.4 compatibility update has surfaced a functional error under investigation. CKEditor context injection has been deliberately deferred to a later release [7][161].
3. Background agents — No direct progress today.
4. Design system integration — Canvas work on component data integrity — specifically handling what happens when a component's content limits are reduced — is nearing completion [190][198], which is a prerequisite for reliably building with design system components at scale. The HTTP API for page regions also lays groundwork for programmatic component management [175][197]. Progress is steady but incremental; no major milestones were reached today.
5. Content creation and discovery — The AI 1.4.0 release shipped today [152][165][170], which includes a standardisation of how AI writing prompts are stored in CKEditor [156][173] — a foundational improvement for consistent AI-assisted drafting. A behaviour issue where non-streaming AI calls are silently converted to streaming could affect features like the AI search block and should be monitored as a risk [160]. The Anthropic provider update to surface Claude Haiku 4.5 as a model option [139] also broadens the tools available for content generation tasks.
6. Advanced governance — The context listing page is being updated to show workflow state [13], which contributes to audit visibility for AI-managed content. This work is in progress but currently unassigned. No other direct governance progress today.
7. Intelligent website improvements — Demo 6 (Smart Personalization Blocks) and Demo 5 (AI-Driven Content Generation) are both active issues within the AI CMS Demo System [113][114], which is the primary vehicle for demonstrating and validating this capability. These remain in planning rather than delivery. No shipped progress today.
8. Multi-channel campaigns — No direct progress today. The initiative's marketing workstream is being reorganised into a dedicated project space [133], which may improve coordination capacity over time, but this is an operational change rather than a product capability advance.

Context Control Center (CCC)

Summary

The Context Control Center project had a productive 24-hour period, with one improvement shipped and several others moving steadily toward completion ahead of the planned release candidate.

The one confirmed delivery was an increase to the default and maximum number of context items that can be applied at once -- work tracked in Update max context items in general settings -- which gives users more flexibility when combining multiple AI guidance rules in a single task. This was contributed by Ahmad Khalil (ahmad-khalil-imagex) and reviewed and merged by Kristen Pol (kepol).

Significant effort this period went into user interface improvements. Kristen Pol (kepol) opened a new draft that surfaces an "extensions" page, helping users discover optional capabilities they can unlock by installing additional components, as tracked in CCC optional feature suggestions in UI. Redesigns of the context listing and detail pages are in progress but temporarily unassigned following a contributor's absence, as noted in Update context listing page per new design and Update context view page per new design. A large internal code quality cleanup is also under review via Clean up CCC inconsistencies and DX issues before rc1 (part 2).

One notable scope decision was made: the ability to inject context into the rich-text editor (CKEditor) has been deferred out of the 1.0 release, as it depends on a fix in an upstream module not expected until a later version. This is a deliberate and documented trade-off, not a surprise. The compatibility update to align with the latest AI Core release, tracked in Bump AI Context to AI Core 1.4, is in review but has surfaced a functional error that Kristen Pol (kepol) is investigating.

Pre-release documentation updates were completed and merged, as covered by Pre-rc1 CCC docs updates, keeping the project on track for a clean release candidate.

How can I help on this project?

• Backfill contributor capacity: Two UI redesign tasks are currently unassigned due to a contributor being unavailable. If your organisation has front-end contributors who could step in this week, that would directly unblock progress toward the release candidate.
• Prioritise upstream dependencies: The CKEditor context injection feature is blocked by work in a separate upstream project. Executive awareness or advocacy within the broader AI initiative could help accelerate that dependency.

View project on GitLab: ai_context

Drupal AI Initiative

Activity in the past 24 hours has been dominated by a significant housekeeping effort: the marketing workstream has been formally separated into its own dedicated project space. Jeremy Chinquist (jjchinquist) led a large-scale migration, closing and redirecting well over 60 marketing-related issues from the main project into the new ai_initiative_marketing home. This reduces clutter in the core project and gives the marketing team a cleaner, more focused place to work. A related open merge request, improve: Sprint management improvements + remove marketing templates, accompanies this change and is currently awaiting review.

On the events front, the Enterprise AI Summit (28 September 2026, Rotterdam) saw visible momentum. Martin Anderson-Clutz (mandclu) produced initial design concepts for the Enterprise AI Summit - Call for Sessions LinkedIn visual asset, with early feedback coming in from Kristen Pol (kepol). Separately, active work continued on a blog post announcing the Summit's keynote speaker, with Paul Johnson (pdjohnson) and Dan Stratton (dan-at-zoocha.com) collaborating on imagery during the period. A visual identity for the Summit is also in progress, with Scott Lawrence (slawrence10) due to present work imminently.

One operational gap was formally acknowledged: Improve communication with partners remains open, flagging that partner organisations currently have only a single point of contact, which means event and sponsorship outreach sometimes reaches the wrong person. Christoph Breidert (breidert) has assigned this to Kristen Pol (kepol) to address.

How can I help on this project?

• Unblock partner communications: The initiative has identified that its partner contact structure is limiting effective outreach. A leadership decision on how partner organisations should nominate and manage multiple contacts would help resolve this quickly.
• Accelerate Summit sponsorship: The call for sponsors for the Enterprise AI Summit has a deadline of 31 May 2026. Direct outreach to your own network or introductions to relevant organisations could meaningfully increase sponsorship uptake before that date.

View project on GitLab: ai_initiative

Anthropic Provider

Activity over the past 24 hours was light but focused, with one contribution under review. Carlos Ospina (camoa) opened a merge request, Add Haiku to ChatJsonOutput capability filter regex, which addresses a gap in model availability. Without this fix, Claude Haiku 4.5 -- a newer and potentially more cost-effective model option -- does not appear in selection menus for features that require structured data output. The change ensures users can choose from the full range of available models when configuring AI-powered functionality. The merge request is open and awaiting review.

Overall, this is a steady, incremental improvement consistent with the module's current development stage. There are no blocking issues or concerns to flag at this time.

How can I help on this project?

• Encourage your team to prioritise reviewing open merge requests promptly, so contributions from the community do not stall and momentum is maintained.
• Consider whether broader testing coverage across different model configurations would benefit the project, and whether dedicated time or resource can be allocated for that work.

View project on GitLab: ai_provider_anthropic

AI Agents Test

Summary

Activity over the past 24 hours centred on a single feature request: the ability to Run more than one test group in Bulk - Test Collections?. This feature would allow users to select and run multiple test groups at once, rather than executing them one at a time, making it significantly easier to evaluate AI model performance across a broad set of scenarios in a single action.

István Csáki (csakiistvan) picked up this issue during the reporting period, carried out a hands-on review against a working environment, and marked it as ready to be merged. This is a meaningful step forward -- the issue moved from open review to approved status within the day, suggesting good momentum on a capability that would meaningfully improve how teams use this module for testing and model comparison.

As an alpha-stage module, having open items and ongoing review activity like this is entirely normal and reflects a healthy, active development cycle.

How can I help on this project?

• Encourage relevant team members or partners to prioritise final review and sign-off so this bulk testing capability can be merged and made available promptly.
• Consider whether dedicated testing time or environments can be made available to support broader evaluation of AI models as new features land.

View project on GitLab: ai_agents_test

AI (Artificial Intelligence)

The headline activity this period is the release of two new versions. Version 1.4.0 shipped, followed immediately by version 1.3.6 as a parallel maintenance release, with Abhisek Mazumdar (abhisekmazumdar) and Artem Dmitriiev (a.dmitriiev) coordinating the tagging and verification process. The 1.5.x development line has also been opened, signalling active forward momentum.

Alongside the releases, a meaningful improvement to the CKEditor integration landed: the AI CKEditor: use prompt entities instead of custom configurations change, contributed by Sven Decabooter (svendecabooter), standardises how AI writing prompts are stored and managed within the editor. This unblocks a related feature that will allow the editor to draw on contextual guidelines specific to each content type.

A behaviour issue was identified where non-streaming AI calls made in certain execution contexts are silently converted to streaming, potentially breaking features like the AI search block. This is being investigated actively. A separate fix is in progress to prevent broken forms when no AI provider is configured for automator fields, improving the setup experience for editors. The team has also flagged that the CHANGELOG.md is outdated and needs a longer-term solution, either automation or removal.

The longer-range architectural planning around adopting Symfony AI as the underlying platform continues in the background, sized as a major multi-release undertaking.

How can I help on this project?

• The upgrade path for sites when submodules are removed is an open policy question needing a decision -- executive sponsorship of a clear, communicated migration strategy would help unblock several downstream issues.
• The Symfony AI integration is a large architectural investment spanning multiple releases. Leadership endorsement and potential resourcing of dedicated time for this work would accelerate progress.
• The documentation site defaulting to the correct version (Reset the documentation to point to 1.4.x as default) is unassigned -- encouraging a contributor or team member to own this would improve the new user experience immediately.

View project on GitLab: ai

Drupal Canvas

Activity over the past 24 hours was focused on expanding the module's capabilities, restoring a broken build pipeline, and advancing several ongoing workstreams around multilingual support, AI features, and data integrity.

Delivered

Two changes landed during the reporting period. First, wotnak's work on the Expose HTTP API for managing Page Regions feature was merged, enabling page layout regions to be managed programmatically through the command-line tool (Workbench) rather than only through manual UI interactions. This is a meaningful step toward automating and scripting site configuration. Second, Ignacio Sánchez Holgueras (isholgueras) resolved a broken automated build pipeline caused by an upstream security release in a third-party dependency, restoring the team's ability to run continuous integration checks. This was a swift turnaround and keeps development velocity intact.

In Progress

Several significant workstreams are advancing in parallel. Multilingual support is seeing concentrated effort, with work underway on previewing content in non-default languages, handling translation deletion, and showing which languages already have translations in the language switcher. These are collectively building out a coherent translation experience within the editor. On the AI side, test coverage for the page builder and template builder agents is progressing, with a new fixture support issue opened and an MR already submitted by AKHIL BABU (AkhilBabu). A data integrity fix to handle component upgrades when content limits are reduced is also nearing completion under Ignacio Sánchez Holgueras (isholgueras), with Wim Leers (wimleers) reviewing and pushing final improvements.

A newly noted concern from Wim Leers (wimleers) on the recently merged regions API is worth monitoring: a follow-up comment raised questions about technical debt in the implementation. Two small follow-up issues have been opened to address caching and configuration dependency gaps, which is a healthy and expected part of iterative development at this stage.

A reported issue where Canvas occasionally fails to load due to a truncated JavaScript file -- potentially related to a caching layer on the user's server -- remains open and unassigned.

How can I help on this project?

• Prioritise multilingual resourcing: Translation support is spanning multiple parallel workstreams. Ensuring dedicated reviewers are available to unblock these issues would accelerate delivery.
• Clarify AI feature scope: AI agent testing is expanding rapidly. A leadership decision on the intended scope and quality bar for AI features before stable release would help the team prioritise effectively.
• Triage the loading issue: The intermittent load failure affecting at least one production environment has no assigned owner. Directing someone to investigate would reduce the risk of it affecting other users.

View project on GitLab: canvas

FlowDrop

Summary

The past 24 hours saw a focused burst of delivery from Shibin Das (shibindasn), culminating in the release of version 1.5.0. The work centred on two meaningful improvements for end users and the teams that build on FlowDrop.

First, the module now tracks the full lineage of processing pipelines -- meaning that when a workflow triggers further workflows, the system can trace messages back to their original source. This makes debugging and auditing multi-step automation flows significantly easier, giving teams clearer visibility into how data moves through their processes.

Second, improvements to how messages are paginated through sessions will make it more practical to navigate large volumes of conversation or event history, supporting use cases where users need to scroll back through earlier records or jump to the most recent activity.

The project has also moved to require a current version of Drupal core, keeping FlowDrop aligned with the broader platform and ensuring it benefits from up-to-date security and performance foundations.

There were no open issues or merge requests active in this period, suggesting the team is in a focused heads-down delivery phase ahead of and following this release.

How can I help on this project?

Consider whether there are internal teams or partners who could begin testing the 1.5.0 release and provide real-world feedback. Encouraging early adoption or a structured pilot would help surface any gaps before wider rollout. If the project needs broader visibility in the Drupal community, leadership advocacy at events or through organisational channels can help grow the contributor base.

View project on GitLab: flowdrop

Tool API

Current Activity

Activity over the past 24 hours has been light, focused on a single compatibility improvement. Work is underway on Add contextdefinitionoverrides to the ToolPluginBase, which addresses an integration issue between the Tool API and the broader AI module. In practical terms, this fix ensures that tools built on this module behave correctly and consistently when used alongside the AI module, preventing unexpected results when multiple tools are running at the same time.

The issue is assigned to Marcus Johansson (Marcus_Johansson) and has been marked as ready for review. Abhisek Mazumdar (abhisekmazumdar) followed up today to confirm whether a review can proceed. The underlying tasks are noted as complete, so the next step is for a reviewer to sign off and move this forward. No merge requests or commits landed in this period.

Overall, this is steady, incremental progress consistent with an early-stage module. There are no blocking issues to flag at this time.

How can I help on this project?

• Ensure a reviewer with relevant expertise is available to assess the open work so it does not stall waiting for sign-off.
• Consider whether additional contributor capacity could help accelerate the review and merge cycle as the module matures toward a stable release.

View project on GitLab: tool

References

[1] Bump AI Context to AI Core 1.4 (ai_context, Issue)

[2] Update context view page per new design (ai_context, Issue)

[3] Update context listing page per new design (ai_context, Issue)

[4] CCC optional feature suggestions in UI (ai_context, Issue)

[5] Update max context items in general settings (ai_context, Issue)

[7] Inject context into ai_ckeditor requests (ai_context, Issue)

[13] Update context listing to show workflow state (ai_context, Issue)

[14] Pre-rc1 CCC docs updates (ai_context, Issue)

[15] Clean up CCC inconsistencies and DX issues before rc1 (part 2) (ai_context, Issue)

[17] Issue #3586212: Updated max items to 10. (ai_context, MR)

[18] Issue #3586192: Clean up code inconsistencies. (ai_context, MR)

[19] Issue #3586233: Bumped AI version to 1.4 (ai_context, MR)

[20] Issue #3586229: Add extensions page under settings. (ai_context, MR)

[22] Enterprise AI Summit - Call for Sessions (ai_initiative, Issue)

[38] Enterprise AI Summit - Call for Sponsors (ai_initiative, Issue)

[113] Demo 6: Smart Personalization Blocks (ai_initiative, Issue)

[114] Demo 5: AI-Driven Content Generation (ai_initiative, Issue)

[123] Look and feel for Enterprise AI Summit Rotterdam (ai_initiative, Issue)

[133] Marketing of the ai_initiative receives a dedicated ai_initiative_marketing Drupal project to reduce complexity and streamline workflows. (ai_initiative, Issue)

[137] Enterprise AI SUMMIT - Blog post announcing keynote speaker (ai_initiative, Issue)

[138] improve: #3586488: Sprint management improvements + remove marketing templates (ai_initiative, MR)

[139] Issue #3566246: Add Haiku to ChatJsonOutput capability filter regex (ai_provider_anthropic, MR)

[152] Release 1.4.0 (ai, Issue)

[156] AI CKEditor: use prompt entities instead of custom configurations (ai, Issue)

[160] non streaming broken (fiber branch) (ai, Issue)

[161] Add event hook to ai_ckeditor to allow context injection (ai, Issue)

[165] Issue #3586456: Resolve "Release 1.4.0" (ai, MR)

[166] Issue #3586456: Resolve "Release 1.4.0" (ai, MR)

[167] Issue #3586456: Resolve "Release 1.4.0" (ai, MR)

[168] Issue #3586456: Resolve "Release 1.4.0" (ai, MR)

[169] Issue #3586456: Resolve "Release 1.4.0" (ai, MR)

[170] Issue #3586456: Resolve "Release 1.4.0" (ai, MR)

[171] Issue #3586456: Resolve "Release 1.4.0" (ai, MR)

[173] #3549657 AI CKEditor: use prompt entities instead of custom configurations (ai, MR)

[175] Expose HTTP API for managing Page Regions (canvas, Issue)

[176] Create a preview mode for non-default languages (canvas, Issue)

[179] Create Route for Delete Translation (canvas, Issue)

[180] Gitlab CI broken because of upstream Symfony security releases that are only available in the `drupal/core-recommended`'s development branch, not in tags (canvas, Issue)

[188] [PP-1] Indicate translation availability per language in the Canvas language switcher dropdown (canvas, Issue)

[190] Handle change of cardinality (fewer/more values allowed): upgrade component instances with more values than cardinality allows by dropping those values (acceptable data loss) (canvas, Issue)

[192] Canvas AI: Create tests for Canvas AI Page builder Agent and Canvas AI Template Builder Agent (canvas, Issue)

[195] #3585452 Create preview for translations (canvas, MR)

[196] #3590572 Indicate translation availability per language in the Canvas language switcher dropdown (canvas, MR)

[197] feat(Internal HTTP API): #3589801 Expose HTTP API for managing Page Regions (canvas, MR)

[198] Issue #3587024: Handle decrease of cardinality (canvas, MR)

[201] Issues/3591591: Add fixture support for layout to use in tests (canvas, MR)

[208] #3591516: Canvas AI Restrict component catalog to canvas_test_sdc components when canvas_ai_agents_test is installed (canvas, MR)