ai 1.0.0-alpha8

Features:

• Add chunker checker to AI Search (AI Search)
• Add generic interface to override and add models + Model metadata to provider (AI Core)
• Make RAG Action of full entity a single chat call instead of multiple (AI Search)
• Provide a mechanism to combine vector search with traditional search (AI Search)
• Allow change to View Mode in RAG Action (AI Search)
• Add pre-chunk modification method in Search API AI (AI Search)
• Provide integration with Pinecone (AI Search)
• Implement chat with image for ai_automators (AI Automators)
• Add a GUI for external moderation (AI External Moderation)
• Add streaming support for Mistral (Mistral Provider)
• Add disposable Automator bundles (AI Automators)
• AI API Explorer for Translate Text (AI API Explorer)
• Add image Chat input possibility with Ollama (Ollama Provider)
• Add log pruning to AI Logging (AI Logging)
• [ai_translate] more flexibility (AI Translate)
• Translation operator type (AI Core)
• Make AI Logs Entity Bundleable and Fieldable (AI Logging)
• Consider moving sub-module specific composer.json requirements into suggest rather than require (AI Core)
• Reintroduce the 'Reformat' Function from OpenAI Module into AI CKEditor (AI CKEditor)

Tasks/Improvements:

• Add provider list documentation (Documentation)
• Check OpenAI connection on setup (OpenAI Provider)
• Add documentation for AI Search module (Documentation)
• Allow Milvus VDB Provider to filter on meta data (AI Search)
• Change how users select `tokenizer chat model` on AI Search / Search API server (AI Search)
• Document how you receive streaming vs none streaming output (Documentation)
• Search API Index fields need more user control, UI should be updated to make it clearer how the embedding content is prepared (AI Search)
• Chunked embeddings should have a unique ID (AI Search)
• Make AI Assistant Form more user friendly (AI Assistants API)
• Validate that the Milvus metric type matches that of the collection when using Milvus (AI Search)
• Minor Search AI fixes (AI Search)
• Add a way to get the chat system role (AI Core)
• Add System Role to chat logging (AI Logging)
• Update config schema for the AI settings (AI Core)
• Allow Pinecone VDB Provider to filter on meta data (AI Search)
• Refactor Search API integration to pass more of the work to the VBD Provider (AI Search)
• Document Translation Type (Documentation)
• Add a global JSON decoder for both streaming and none streaming responses. (AI Core)
• Document setChatSystemRole & getChatSystemRole (Documentation)
• Rename data name to ai_automators (Documentation)
• Switch to .cspell-project-words.txt to reduce merge conflicts (AI Core)
• AI Assistant full entity should convert to markdown for the LLM (AI Assistants API)
• Move AI Assistant to use JSON service (AI Assistants API)
• Move AI Automators to use JSON service (AI Automators)
• Switch to Json::decode() from Drupal core (AI Core)
• Remove default model selection, make the user choose (or provide a sensible default) (AI Core)
• Update hook for Search AI refactoring (AI Search)
• Add Search API details about Zilliz/Milvus status on the Index View page (AI Search)
• Add Readme (Documentation)
• Write documentation for AI CKEditor (Documentation)
• Add so model settings is actually overridable (AI Core)
• Ai Assistants Should be experimental (AI Assistants API)
• Add common models and dimensions for Embeddings info (AI Core)

Bugs:

• Combined vector & traditional search should only allow processor plugin where applicable (AI Search)
• ai_external_moderation: ModeratePreRequestEventSubscriber TypeError (AI External Moderation)
• Class Drupal\ai_translate\Drush\AiTranslateCommands extends unknown class Drush\Commands\DrushCommands (DevOps)
• Defaults for extraAdvancedFormFields in taxonomy.php (text manipulation and find similar tags) (AI Automators)
• Fix more references to metadata in code that is actually contextual content (AI Search)
• AI CK Editor sparkle should be black instead of yellow for contrast (accessibility) (AI CKEditor)
• Ai Translate - translateContent (AI Translate)
• Saving index pages other than Fields wipes the Fields indexing options (AI Search)
• AI translate settings form does not save entity reference settings (AI Translate)
• CK Editor AI Assistant throws Bad Request error when using AWS Bedrock Provider (AI CKEditor)
• AI Chatbot causes forms to not send. (AI Chatbot)
• AI Assistants have to much context (AI Assistants API)
• [error_message] does not get filled (AI Assistants API)
• Fix issue with embedding strategy configuration (AI Search)
• Ai Automators doesn't works with Queue/cron (AI Automators)
• nvalidArgumentException: "type" is an invalid render array key. (Huggingface Provider)
• Use natural instead of neutral (OpenAI Provider)
• Zilliz queries get 403. (Milvus VDB Provider)
• Problem with translation of multiple formatter fields turning HTML into string (AI Translate)
• AI Assistant fails when pure JSON is returned (AI Assistants API)
• Vector DB Explorer, show results on wrong place in Gin (AI API Explorer)
• AI Automators LLM Model doesn't show (AI Automators)
• [ai_translate] Error: Call to a member function hasTranslation() on null (AI Translate)
• RAG doesn't work without session on (AI Assistants API)
• isZilliz check in MilvusV2 is not correct (Milvus VDB Provider)
• Avoid crashing the AI settings page if the default provider is not fully configured (AI Core)
• AI Search Few Shot Learning is wrong (AI Search)
• When adding assistant the provider config form is empty resulting to an error (AI Assistants API)
• Log cron deletion options should not require auto-logging enabled (AI Logging)
• Deprecated function: explode(): Passing null to parameter #2 ($string) of type string is deprecated in Drupal\ai_translate\Form\AiTranslateForm->buildForm() (line 112 of AiTranslateForm.php (AI Translate)
• Warning : Undefined array key "indexing_options" dans Drupal\ai_search\Plugin\EmbeddingStrategy\EmbeddingBase->buildBaseMetadata() (AI Search)