Daily Digest

Daily Digest – 4 August 2026

2026 Capabilities Progress

  1. Page generation — Canvas began work to allow the component agent to run in non-looped mode driven by the browser [58][73], which will support interactive page assembly workflows. The module advanced work to support entity types beyond nodes [59][71][72], broadening the scope of content that can be generated from AI prompts.
  2. Context management — The Context Control Center began work to harden access controls and trust model documentation [93][115], improved conditional context behavior when AI decisions are disabled [105][121], and began work to clarify scope eligibility and limits in the UI [103]. Token budget management advanced with work to persist token counts and display budget usage [96], and work began on a priority ranking field to influence context selection [110].
  3. Background agents — Work began on a setup-guardrails agent skill that would allow agents to configure site guardrails autonomously [11], and the Scheduler extension was added to the CCC Extensions page to support time-based context publishing [101][119]. Progress on agent streaming capabilities [45] and status stream exposure [36] continued but remains incomplete.
  4. Design system integration — Canvas enabled pattern editing after creation [64][84], removed the Lupus CE API dependency in favor of a dedicated Canvas Headless endpoint [57][76], and advanced work to allow Canvas on any entity type with configurable displays [59][71][72]. The auto-save system was strengthened to generate stable hashes for change detection [65][88].
  5. Content creation and discovery — A reranking bug affecting chunk text extraction was identified and assigned [14], and a progressive indexing issue that can drop tail chunks was updated [124]. Translation capabilities advanced with work on fixes for media entity translation errors [125][127][129] and work on support for auto-translating referenced taxonomy terms [126][128], plus LiteLLM provider translation support was merged [131].
  6. Advanced governance — Canvas resolved a critical configuration import issue that prevented component version updates from deploying correctly [63][83], unblocking deployment workflows. Agent serialization was fixed to handle images in chat history [43][53], and work continued on conditional subcontext governance when AI decisions are disabled [105][121].
  7. Intelligent website improvements — No direct progress today.
  8. Multi-channel campaigns — Canvas advanced headless content delivery by implementing a dedicated API endpoint [57][76] and began work on headless previews for content templates [70], which supports rendering content across multiple presentation channels from a single source.

Shipped

  1. BigPipe chat rendering issue resolved in amazee.ai provider — The amazee.ai provider issue preventing chat under Fiber-based rendering has been closed [35].
  2. Agent session restoration no longer crashes when images are present — AI Agents can now reliably save and restore conversations containing file attachments, fixing a stability issue that prevented production use [43][53].
  3. Canvas patterns can now be edited after creation — A merged fix allows editing stored patterns, addressing a workflow limitation [64][84].
  4. Canvas configuration synchronization unblocked — A merged fix addresses drush config:import failures when component versions differ between environments [63][83].
  5. LiteLLM provider adds translation support and improved reliability — Merged changes introduce a translateText method and fallback logic for model retrieval [130][131].

Ongoing

  1. Chat Fundamentals architectural proposal awaiting approval — A twelve-issue plan to adopt Model Context Protocol shapes for rich structured chat remains postponed pending maintainer decision, which will determine the 1.6.x roadmap [1][2].
  2. Context Control Center preparing for beta4 release — Work continues on access control hardening, token budget visibility, demo content preparation, and user interface clarifications ahead of the next release [93][94][96][102][103].
  3. Canvas support for entity types beyond nodes in development — Active work underway to enable Canvas on taxonomy terms and other content entity types with configurable displays [59][71][72].
  4. Media translation batch failures under investigation — Multiple fixes proposed for an error that prevents editors from translating media entities, blocking multilingual publishing workflows [125][127][129].
  5. Anthropic provider gaining token counting and citation features — Third phase of SDK migration adds opt-in capabilities for usage tracking, source citation, and conversation management [34].

AI (Artificial Intelligence)

The module received a major architectural proposal this period. A coordinated set of twelve issues was filed, collectively called the "Chat Fundamentals plan," proposing to adopt Model Context Protocol shapes as the canonical format for rich structured chat. The proposal would introduce content blocks, citations, progress indicators, suggested replies, and forms-within-chat capabilities. The entire set is currently in postponed state, awaiting formal maintainer approval via the approval issue. Once approved, this work would target a new 1.6.x branch, not the current 1.5.x release line.

Active development continued on existing features. Work advanced on a setup-guardrails agent skill to let agents configure site guardrails without manual UI work. The request tag registry feature received a draft implementation allowing modules to declare which AI request tags they emit. A fix for the chatbot closing when adding media remained in review. Several bug fixes progressed: a correction to prevent rendering empty configuration widgets reached ready-to-be-committed status, and an issue affecting chunk text extraction during reranking was identified and assigned.

Two compatibility issues were reported: the DeepChat block no longer works under BigPipe after upgrading to Drupal 11.4, and streaming content generation fails with Gemini 2.x models, producing 400 errors.

How can I help on this project?

Review and decide on the Chat Fundamentals proposal awaiting approval — this twelve-issue architectural plan will shape the roadmap for 1.6.x. Ensure the team has bandwidth to address the newly reported Drupal 11.4 and Gemini 2.x compatibility issues, as these affect current production usage.

Contributors: Abhinav Jha (ajabhinavjha30), Abhishek Dhariwal (adaptative), Bharat Kelotra (bharatkelotra), Joshua Fernandes (joshua1234511), Marcus Johansson (Marcus_Johansson), Ricardo Castañeda (cadence96), Rob Loach (robloach), Rodel Duterte (rduterte), Shivam Sen (shivamsen_12579)

View project on GitLab: ai

Drupal AI Initiative

Activity this period was limited to project coordination. A quality assurance meeting from February 2026 was closed, reflecting completion of historical administrative tasks. No feature development, bug fixes, or code integration occurred during this 24-hour window.

The initiative appears to be in a quiet phase with no active development or testing work visible in this reporting period.

How can I help on this project?

  • Assess whether the project has sufficient allocated resources to maintain development momentum
  • Clarify strategic priority and timeline expectations with the team to ensure alignment with organizational goals

View project on GitLab: ai_initiative

OpenAI Provider

A new capability was opened for review that would allow the OpenAI provider to be configured automatically from the command line. The setup-provider agent skill is designed to simplify initial setup by reading API credentials from an environment variable without exposing secrets directly. This reduces manual configuration steps and supports more streamlined deployment workflows, particularly in automated or team environments.

No issues were updated and no code was merged during this period. Development activity remains focused on this configuration improvement.

How can I help on this project?

  • Prioritize review and approval of the setup automation work to accelerate deployment ease for teams adopting the module.
  • Clarify whether simplified onboarding is a strategic priority, and if so, allocate resources to expand similar capabilities across other providers.

Contributors: Joshua Fernandes (joshua1234511)

View project on GitLab: ai_provider_openai

Anthropic Provider

A new merge request was opened to add token counting, citations, and context management capabilities to the Anthropic integration. This work represents the third phase of migrating the module to use Anthropic's native software development kit.

The enhancement introduces three opt-in features that give site administrators more control over how their AI conversations are handled: the ability to track and limit the size of conversations (token counting), automatic citation of source materials in responses, and tools to manage conversation length by compacting or editing chat history. All three features are turned off by default, ensuring existing deployments continue to work exactly as they did before.

The merge request is currently under review. No issues were updated during this period.

How can I help on this project?

  • Prioritize review and testing resources for the open merge request to accelerate delivery of new capabilities to users.
  • Clarify strategic priority for token management features if cost control or usage monitoring is a business objective.

Contributors: Carlos Ospina (camoa)

View project on GitLab: ai_provider_anthropic

amazee.ai Private AI Provider

The module reached an important milestone this period: a critical blocking issue preventing the chat feature from working on standard Drupal installations has been resolved. The problem affected sites using default page rendering optimizations, meaning the chat functionality was previously unusable in the most common deployment scenario. With chat() fails under Fiber-based rendering (BigPipe) — "stream handler closure" exception now closed, the provider can reliably deliver AI-powered chat completions to users across typical production environments. This fix removes a significant barrier to adoption and ensures the module performs as expected in real-world usage.

How can I help on this project?

  • Prioritize user testing in production-like environments to identify deployment issues earlier in the development cycle.
  • Consider allocating resources to accelerate the module toward a stable release if it supports strategic business objectives.

View project on GitLab: ai_provider_amazeeio

AI Agents

The AI Agents module made measurable progress on core stability this period. A critical bug that caused the system to crash when restoring an agent with images in its conversation history was resolved and merged. The fix for serialization errors ensures that agents can now reliably save and restore sessions containing file attachments, removing a barrier to production use.

Work is underway to address usability issues in the agent configuration interface. A missing close button on the tool settings popup has a solution in review, and a warning message when configuring default tools is being addressed.

Three new feature proposals were documented, all currently postponed pending maintainer approval of a broader architectural plan. These would enable agents to expose real-time progress updates, request user confirmation before executing sensitive actions, and preserve richer output from tools. While these features remain on hold, their specification work represents strategic planning for future capabilities.

Two new issues were raised regarding operational limits: file size controls and clarification of model selection behaviour for agents and sub-agents.

How can I help on this project?

Prioritise maintainer review of the pending architectural plan to unblock three proposed features. Clarify documentation or provide decision-making on whether file size limits and per-tool execution limits should be configurable at the agent level.

Contributors: AKHIL BABU (AkhilBabu) [1], Alexis Martínez (alexismmd), Bharat Kelotra (bharatkelotra), Marcus Johansson (Marcus_Johansson)

View project on GitLab: ai_agents

Drupal Canvas

Over the past 24 hours, Canvas achieved several significant improvements to workflow stability and feature completeness.

Two user-facing gaps were closed. Patterns can now be edited after creation; previously users had to delete and recreate them to correct mistakes or evolve designs. The Allow patterns to be edited feature was completed and merged. A data integrity issue that prevented standard deployment workflows was also resolved: `drush config:import` can fail to update a Component whose `active_version` changed was fixed, unblocking configuration synchronization across environments when component versions differ.

The headless architecture took a step toward independence. Canvas Headless now uses a dedicated endpoint instead of relying on the Lupus CE API, reducing external dependencies and giving the project tighter control over response contracts. This work (Replace Lupus CE API with a dedicated Canvas Headless endpoint) was merged.

Internal reliability improved through consolidated data normalization logic, strengthening the auto-save system's ability to generate stable hashes for detecting content changes.

Several issues remain open with active merge requests, including a view-mode preview rendering bug and ongoing work to support Canvas on entity types beyond nodes. A regression affecting older installations that stored certain metadata has been identified and requires an upgrade path.

How can I help on this project?

  • Prioritize review and merge of the view-mode preview fix to restore editing for non-full templates.
  • Support scoping and resourcing the upgrade path needed for the default-relative-url data migration to prevent further 500 errors on existing sites.

Contributors: **** (project_19391_bot_17a61b0edeafd5a3db3e8b2462b6c010), **** (project_19391_bot_cb443e60680723bba08a5351bf595251), Adam G-H (phenaproxima), AKHIL BABU (AkhilBabu), Ben Mullins (bnjmnm), Christian López Espínola (penyaskito), Dave Long (longwave) [2], David Bravo (davidbp), Feliksas Mazeikis (f.mazeikis) [1], Harumi Jang (hooroomoo), Ignacio Sánchez Holgueras (isholgueras), Jakub Piasecki (zaporylie), Lauri Timmanee (lauriii), wotnak [1]

View project on GitLab: canvas

Context Control Center (CCC)

The Context Control Center continued work toward its first stable release this period, with a focus on hardening the product before beta4 and clarifying its subscription model for users.

Three significant merge requests landed. Work to add Scheduler to the CCC Extensions page was completed, making scheduled publishing and unpublishing of context items visible to administrators. A merge request that polishes developer APIs and internal documentation was merged, improving clarity for developers extending the module. The largest change was to improve conditional subcontext disabled-state behavior and admin UX, ensuring that when conditional AI decisions are turned off, the relevant context items are excluded from selection and editors receive clear guidance about why those items are inactive.

Two discussions concluded. A review of whether to add priority to context items to affect ranking ended, with the decision to proceed captured in a follow-up implementation issue. A separate discussion about integrating content reviews based on context was closed, with the goal of avoiding CCC changes for version 1.0.

Planning issues were opened to prepare for the next sprint, including a pre-release codebase review for beta4 and a discussion to review context selection ranking factors before 1.0. Work continued on hardening access checks and documenting the trust model, with a merge request already in review. Refinement of the user experience continues, including preparation of demo content to show how the module works and efforts to clarify scope eligibility and limits in UI help.

Several issues were created to simplify the module's configuration model. One proposes to remove the max items context limit and rely solely on the token budget, reducing overlapping controls. Another will persist token counts and show budget usage in the UI, helping site builders understand when they are nearing capacity. A third addresses a bug where sites enabling Dynamic Entity Reference after CCC installation can experience errors due to missing field storage.

How can I help on this project?

Prioritize user-experience and documentation review capacity before the release candidate. Ensure the team has time budgeted to finalize demo content and help text refinements. If scope eligibility decisions require product sign-off, schedule that review early in the next sprint to avoid blocking the stable release.

Contributors: Ahmad Khalil (ahmad-khalil-imagex), Juan Correa (jucs7), Kristen Pol (kepol) [3], Rob Loach (robloach), Tekla Aivazashvili (aivazashvilit)

View project on GitLab: ai_context

AI translate

The AI translate project saw focused activity on two features that will improve the translation workflow for content editors on multilingual sites.

Work began on a fix for a translation failure affecting media entities. When editors attempt to translate images or other media items, the system currently encounters an error that prevents the batch operation from completing. A solution has been proposed to resolve this blocking issue.

In parallel, a new capability is under development to auto-translate referenced taxonomy terms when translating content. Currently, editors must manually translate each referenced term separately, adding time and effort to multilingual publishing workflows. The proposed enhancement would automatically translate taxonomy terms and other linked content entities during the main translation batch, streamlining the editorial process. Two merge requests have been opened for this feature.

Both pieces of work are in the early review stage and have not yet been integrated into the module.

How can I help on this project?

  • Ensure code review capacity is available to evaluate and merge the pending translation fixes, particularly the media entity issue which blocks a key workflow.
  • Consider resourcing for user acceptance testing of the referenced entity auto-translation feature once merged, to validate time savings for editorial teams.

Contributors: Carlos Romero (o0Charly0o), szato (odi)

View project on GitLab: ai_translate

LiteLLM AI Provider

The LiteLLM AI Provider module made meaningful progress in the last 24 hours, with two improvements merged to strengthen reliability and expand functionality.

A fix to enhance model retrieval with fallback for admin key access was merged, improving the module's ability to retrieve available models even when configuration challenges arise. This makes the system more resilient during setup and administration.

Additionally, support for translation was added, enabling the provider to handle text translation tasks. This expands the range of use cases the module can support, making it more versatile for multilingual content workflows.

Behind the scenes, several refinements were made to improve how the module detects model capabilities, handles caching, and validates configuration settings. These changes help ensure the module operates smoothly and efficiently in production environments.

How can I help on this project?

  • Ensure sufficient testing resources are available to validate translation features across different language pairs and content types.
  • Support community outreach to gather feedback from early adopters on real-world use cases and pain points.

Contributors: Carlos Romero (o0Charly0o), Dan Lemon [3]

View project on GitLab: ai_provider_litellm

AI Image Classification

Two installation issues were reported affecting users attempting to deploy the recipe on Drupal CMS sites. The first issue, Error when installing on Drupal CMS, involves a failure when installing through the administrative interface, preventing the recipe from being added to sites using the Haven template. The second, Error when applying recipe via Drush in Drupal CMS, blocks command-line installation, which is a common deployment method for development teams.

These issues prevent new users from successfully adopting the recipe, which limits the module's reach and delays delivery of automated image classification capabilities to content teams. Both issues remain unassigned and require investigation to determine root causes. As this is an early-stage module, some installation friction is expected during initial adoption, but resolving these blockers is important for expanding the user base and gathering broader feedback.

How can I help on this project?

  • Allocate developer time to investigate and resolve the two open installation issues, which are currently preventing user adoption.
  • Consider prioritizing testing resources to validate the recipe across different Drupal CMS configurations before broader release.

View project on GitLab: ai_recipe_image_classification

AI Content Classification

No development activity occurred during this period. Two installation issues were reported when attempting to use the recipe with Drupal CMS. The first involves a validation error related to a dependency (Error when applying recipe via Drush in Drupal CMS), while the second concerns a failure during the automated installation process through the administrative interface (Error when installing on Drupal CMS). Both issues prevent users from successfully deploying the recipe in Drupal CMS environments, which limits its practical adoption. These are typical integration challenges for a developing module and do not indicate systemic problems, but they do block immediate usability for the target platform. Resolution of these compatibility issues would enable smoother deployment and testing by early adopters.

How can I help on this project?

  • Allocate developer time to investigate and resolve the two reported Drupal CMS installation blockers, which are preventing the recipe from being deployed by users.
  • Consider establishing a testing environment for Drupal CMS compatibility to catch integration issues earlier in the development cycle.

View project on GitLab: ai_recipe_content_classification

References

[1] Approve the Chat Fundamentals (MCP-shaped rich chat) issue set for 1.6.x (ai, Issue)

[2] [Meta] Rich structured chat: adopt MCP content and interaction shapes (ai, Issue)

[3] (Optional) Typed context references on user messages via input-side content blocks (ai, Issue)

[11] Add a setup-guardrails agent skill to configure guardrails, guardrail sets, and global guardrails (ai, Issue)

[14] `AiReranker::extractItemText()` should prefer chunk text over configured indexed fields (ai, Issue)

[34] Issue #3607960: Add token counting, citations, and context management (ai_provider_anthropic, MR)

[35] chat() fails under Fiber-based rendering (BigPipe) — "stream handler closure" exception (ai_provider_amazeeio, Issue)

[36] Expose the agent status stream as MCP-shaped progress to any chat consumer (ai_agents, Issue)

[43] AiAgentEntityWrapper::fromArray() fatals when the serialized chat history contains an image (ai_agents, Issue)

[45] Use streamed chat for agents (ai_agents, Issue)

[53] AiAgentEntityWrapper::fromArray() fatals when the serialized chat history contains an image (ai_agents, MR)

[57] Replace Lupus CE API with a dedicated Canvas Headless endpoint (canvas, Issue)

[58] Canvas AI: Run the component agent in the dev chat with setLooped(FALSE), driven by the browser (canvas, Issue)

[59] [META] Allow Canvas to be used on any content entity type (bundle), as long as it has certain entity type characteristics (canvas, Issue)

[63] `drush config:import` can fail to update a Component whose `active_version` changed (canvas, Issue)

[64] Allow patterns to be edited (canvas, Issue)

[65] Strengthen AutoSaveManager to generate stable hashes (canvas, Issue)

[70] feat(Headless): #3591893 Enable headless previews for content templates (canvas, MR)

[71] feat: #3518272 support all entity types with configurable displays for content templates (canvas, MR)

[72] #3498525 Allow entity types other than node; add feature flag module for taxonomy terms. (canvas, MR)

[73] Issues/3591884: Run component agnet with setlooped false. (canvas, MR)

[76] feat(Headless): #3591861 Use the new Canvas content API endpoint in the headless SDK (canvas, MR)

[83] #3591905: `drush config:import` can fail to update a Component whose `active_version` changed (canvas, MR)

[84] #3591851: Allow editing stored patterns (canvas, MR)

[88] refactor(Conflict resolution): #3591785 Strengthen AutoSaveManager to generate stable hashes (canvas, MR)

[93] Harden CCC context item access checks and document the authoring trust model (ai_context, Issue)

[94] CCC beta4 codebase review (pre-release) (ai_context, Issue)

[96] Persist context-item token counts and show token-budget usage in CCC UI (ai_context, Issue)

[101] Add Scheduler to CCC Extensions page (ai_context, Issue)

[102] [Meta] Prepare demo content and demo context items for the CCC to show how it works (ai_context, Issue)

[103] Clarify scope eligibility, ranking, and limits in UI help and documentation (ai_context, Issue)

[105] Improve conditional subcontext disabled-state behavior and admin UX (ai_context, Issue)

[110] Add context item ranking priority field and use for ranking selection logic (ai_context, Issue)

[115] Issue #3586356: Harden CCC context item access checks and document the authoring trust model. (ai_context, MR)

[119] Issue #3586353: Add Scheduler to CCC Extensions page. (ai_context, MR)

[121] Issue #3586292: Exclude conditional subcontext items when AI decisions are disabled. (ai_context, MR)

[124] Progressive indexing can drop tail chunks (ai_search, Issue)

[125] InvalidArgumentException: Invalid translation language (und) when translating media entities (ai_translate, Issue)

[126] Auto-translate referenced taxonomy terms (and other translatable entities) when translating a node (ai_translate, Issue)

[127] Resolve "InvalidArgumentException: Invalid translation language (und) when translating media entities" (ai_translate, MR)

[128] feat: Add batch translation for referenced translatable entities in AiTranslateController (ai_translate, MR)

[129] fix: #3585532 Translation language (und) when translating media entities. (ai_translate, MR)

[130] fix: enhance model retrieval with fallback for admin key access (ai_provider_litellm, MR)

[131] feat: add translateText method and support for translation in LiteLlmAiProvider (ai_provider_litellm, MR)

Period: 4 August 2026 to 5 August 2026  |  Generated: 5 August 2026 09:18 GMT

To view the full AI prompts used to generate this digest check this file: 24h_2026-08-04_prompts.txt

Looking for older digests? Browse all digests here.