AI Engine – The Chatbot and AI Framework for WordPress

Historial de cambios

3.3.1 (2026/01/04)

• Update: Removed unused files and code to reduce clutter and improve performance.
• Info: Wishing you a Happy New Year 2026 ☀️🎉 No doubt this year will be full of surprises again, let’s continue riding the AI wave together! 😊

3.3.0 (2025/12/29)

• Add: Added title filter to the Embeddings table.
• Add: Added filter to all model dropdowns.
• Add: Enabled custom placeholder support and model-aware credit customization via new and updated filters for developers.
• Add: Introduced prompt caching for Anthropic and a free-text dimension input for non-OpenAI embeddings.
• Add: Added dimension mismatch warnings for Pinecone and Qdrant to help diagnose configuration issues.
• Add: Included “speech_recognition” support in the embed.js system.
• Fix: Resolved multiple embedding issues, including OpenRouter parameter compatibility, Pinecone Serverless URLs, and vector store file uploads.
• Fix: Improved compatibility with multilingual plugins, placeholder processing, and post content handling through dedicated filters.
• Update: Performed an end-of-year cleanup by removing expired database migrations, deprecating outdated Claude 3 Opus models, and tidying up legacy debug and commented code.

3.2.9 (2025/12/17)

• Add: Added support for the new gpt-image-1.5 model, works quite well! 😊
• Fix: Restored reliable image editing (added missing file attachment handling, separating the edit mode prompt from templates, and preventing double dots in generated filenames).
• Fix: Corrected Mistral model filtering so available models now appear and work as expected.
• Fix: Resolved a UI crash when selecting OpenRouter providers that use non-standard model data.

3.2.8 (2025/12/12)

• Add: Support for GPT-5.2 models (including gpt-5.2, gpt-5.2-chat-latest, and gpt-5.2-pro).
• Add: Support for placeholders in chatbot instructions.
• Add: MWAI_SSL_VERIFY constant so you can override SSL verification.
• Add: Files Manager in Dev Tools (+ improved file cleanup, locally and remotely at OpenAI and Anthropic).
• Update: Increased the maximum embeddings results limit from 100 to 1000.
• Update: Simplified file “purpose” options to just “analysis” or “generated” for clearer file handling.
• Update: Enhanced the Embeddings Query Mode and Dev Tools.
• Fix: Resolved a UI crash when selecting OpenRouter or other environments with non-standard models.
• Fix: Chroma integration now auto-creates the database and collection when needed.
• Fix: Prevented a rare crash in the Discussions tab.
• Fix: Clarified and improved error messages.

3.2.7 (2025/12/02)

• Add: Support for OpenRouter embedding models.
• Update: Friendlier onboarding.
• Update: Improve default environments with better validation.
• Update: Clearer warnings for misconfigured setups.
• Update: Enhanced the default CSS for Custom Theme.
• Fix: A few issues with Qdrant.

3.2.6 (2025/11/29)

• Fix: Corrected the expected embedding model display for Qdrant and Pinecone.
• Fix: Corrected spacing between uploaded files/images and chatbot messages.
• Fix: Improved file handling in the Responses API to safely check file types.
• Fix: Resolved issues with mwai_forms_submit_params.
• Update: Refined the Discussions tab.
• Update: Added the –mwai-iconSize CSS variable to make the icon size adjustable.

3.2.5 (2025/11/26)

• Update: Redesign the Knowledge settings section.
• Fix: Prevent DOM elements from being selected behind the chatbot while dragging it.
• Fix: Ensure Qdrant and Pinecone correctly use their intended embedding environment.
• 🎵 Discuss with others about Ai Engine on the Discord.
• 🌴 Mantennos motivados con una pequeña reseña aquí. ¡Gracias!
• 🥰 If you want to help us, check our Patreon. Thank you!
• 🚀 Haz clic aquí para votar por las características que más quieres.

3.2.4 (2025/11/23)

• Add: Introduced a Mime Types selector (for Upload Files in Chatbots).
• Update: Added a new PHP API section and defined the REST API section. AI Engine is extremely easy to use with workflow automation tools like n8n, Make, Zapier, etc, so try it out! ✨
• Update: Renamed the “Context” section to “Knowledge & Context” and “Tools” to “Tools & Capabilities”.
• Update: Better links to tutorials and documentation.
• Fix: Corrected the display of model names in Query Logs (such as GPT-4.1 and GPT-5.1).

3.2.3 (2025/11/18)

• Add: Added support for the new gpt-5.1 model and included a ‘none’ option for reasoning effort settings.
• Add: Support for Chroma Cloud embeddings with the Qwen3 model.
• Fix: Enhanced error detection for Chroma responses.
• Fix: Improved accuracy of audio transcription in AI Forms.
• Fix: HTML Blocks are now in the scrollable conversation.
• Fix: Corrected MinScore of 0 to be recognized as a valid score.
• Fix: Fixed bulk sync process to correctly include vectors with model or env mismatches.
• Fix: Realtime transcript is now scrollable.
• Fix: Resolved the Embeddings Sync task to run properly with admin rights.
• Fix: Fixed undefined array key warnings in finetunes handling.
• Update: Removed deprecated fields from the embeddings section in AI Forms.
• Update: Minor enhancements in the UI.

3.2.1 (2025/11/11)

• Add: IP validation for image uploads/downloads to prevent SSRF.
• Add: Security measures to prevent PHAR deserialization.
• Add: Custom Theme CSS for Cross-Site .
• Add: New ‘Region’ field for Azure (Realtime API needs this).
• Add: New filters (mwai_ai_embeddings_query and mwai_ai_feedback_query).
• Add: Filter to allow customization of the embeddings rewrite prompt.
• Update: MeowCommon is now MeowKit.
• Update: Optimized the “Push All” feature for embeddings (now much, much faster).
• Fix: Corrected prompt mode to utilize the selected environment.
• Fix: Correct language detection in suggestions.
• Fix: Ensure tasks return success even when handlers are missing.
• Fix: Resolved issues with the embeddings sync task.
• Fix: Got rid of a few warnings.

3.1.6 (2025/10/24)

• Fix: Resolved MCP SSE worker exhaustion issue by handling agent cancellation signals.
• Add: Added Claude Haiku 4.5.
• Update: Enhanced the Import Embeddings task.
• Fix: Corrected the handling of implicitly nullable parameters.
• Fix: The Cleanup Files task now accurately deletes files based on their expiration dates.
• Fix: Minor bug fixes and improvements.

3.1.4 (2025/10/19)

• Add: New MCP tool (wp_get_post_snapshot) to optimize data retrieval for complex post types.
• Fix: Security hardening for REST API endpoints to prevent unauthorized access.
• Fix: Avoid malformed UTF-8 issues.

3.1.3 (2025/10/15)

• Add: Support for MCP with ChatGPT: https://meowapps.com/chatgpt-wordpress-mcp/.
• Update: Made the Public API easier to understand and to test.
• Update: Enhanced MCP and REST API sections.
• Update: Optimized the embeddings Push All/Sync process (much, much faster).

3.1.2 (2025/10/10)

• Add: Realtime Hold-to-Talk feature now includes a Replay button.
• Add: Support for Azure OpenAI Realtime.
• Add: Resolutions and pricing for Google models.
• Add: GPT-5 Pro, GPT Realtime Mini, Sora 2, Gemini Images.
• Add: Multi-file support for chatbots and forms, both on the UI-side and the models-side.
• Update: Embeddings Sync has been migrated to the Tasks System.
• Update: Updated Anthropic and Gemini models.
• Fix: Image and video pricing.
• Fix: Hotfix applied to address missing category column in tasks DB.
• Fix: Realtime API calling now parameterless functions correctly.
• Fix: Fixed TypeError in embeddings list.
• Fix: Improved safety with SSRF protection for image downloads.
• Fix: Resolved various UI/UX issues to improve user experience.

3.1.1 (2025/09/30)

• Add: Integrated Mistral AI into the plugin.
• Fix: Resolved an undefined array key “type” warning in Chroma.
• Fix: Corrected a database column name error in the task cleanup query.
• Fix: Fixed a layout issue in the Gutenberg editor.
• Fix: Resolved a crash in template creation when a user cancels the name prompt.
• Update: Optimized task refresh logic.
• Update: Included “reply” as an argument in mwai_ai_feedback.
• Update: Improved Google models retrieval logic.
• Update: Removed debugging logs, empty code blocks, etc.

3.1.0 (2025/09/17)

• Add: New ‘Hold to Talk’ mode for Realtime chatbot.
• Add: Tasks Manager (task scheduling, multi-step tasks, progress tracking, retry logic, etc).
• Fix: Resolved issues with Cross-Side to support renamed plugin folders.
• Fix: Hotfix to ensure proper removal of all database tables during uninstall.
• Fix: Corrected cross-origin CORS headers not being sent for allowed domains.

3.0.9 (2025/09/10)

• Update: Improved the UI of the Forms Editor.
• Update: Redesigned the PDF Import feature for a more streamlined workflow.
• Update: Enhanced Embeddings UI with sortable titles, a smarter Sync button, and title preservation.
• Update: Enhance many parts of the UI, UX, better explanations, and more.
• Fix: Corrected PDF import chunking.
• Fix: Resolved PDF Import to correctly utilize custom titles.
• Fix: Improved PDF Import process to use Admin API instead of Public API.
• Fix: Addressed cache issues with lazy-loaded chunks by adding dynamic versioning.
• Fix: Cross-Site feature blocking same-origin requests.
• Fix: Streaming issue where empty strings caused [Object] to appear.
• Fix: Pinecone metadata error by defaulting the ‘type’ field to ‘manual’.
• Fix: Resolved issues with bulk processes.

3.0.6 (2025/08/29)

• Add: Image Upload for Realtime Chatbot.
• Add: New ‘gpt-realtime’ model.
• Add: Prompt Mode for chatbots to support OpenAI’s new Prompts feature with prompt IDs.
• Update: Redesigned the Embeddings user interface with improved mode selection, icons, and standardized buttons.
• Update: Redesigned the Template component UI, cleaner layout, and icon-only buttons.
• Fix: Hotfix for orphan embeddings being created when editing from AI Search mode.
• Fix: Missing list styles in chatbot themes.

3.0.4 (2025/08/23)

• Add: Support for Responses API with Azure OpenAI.
• Add: Download button for generated images in chatbot.
• Update: Better color picker with alpha support.
• Update: Many subtle UI improvements.
• Fix: Fixed Site Editor compatibility.
• Fix: Hotfix for center positioning of popup chatbots.
• Fix: Resolved JavaScript errors in Content Generator, Images Generator, and Playground.

3.0.2 (2025/08/21)

• Update: Default token limits increased to 4096 and context length to 16384, with improved token info display.
• Add: Forms Editor for creating shortcodes, compatible with Classic Editor, Elementor, and more.
• Update: Enhanced the Timeless Theme with subtle visual improvements.
• Update: Added Slide and Fade animation effects with improved timing.
• Update: More refined window animations and timing synchronization.
• Fix: Support for ‘message’ type shortcuts in chatbots to enable simpler message delivery.
• Fix: Hotfix for header dragging issue on OSX in admin preview to prevent positioning problems.
• Fix: Resolved outdated URL issues.
• Fix: Center feature bug in Themes.
• Fix: Numerous minor CSS issues for a cleaner appearance.

3.0.1 (2025/08/17)

• Add: Support for Code Interpreter tool in OpenAI chatbots.
• Update: Enhancements to the themes.
• Update: Window Animation now supports Zoom, Slide, and Fade (select in UI, or via CSS classes mwai-animation-zoom, mwai-animation-slide, mwai-animation-fade).
• Fix: Fixed Magic CSS Generator.
• Fix: Resolved missing parameter in Qdrant.

3.0.0 (2025/08/13)

• Add: Brand new Terminal mode with SF Mono font, zoom animations, improved cursor behavior, fullscreen support, and more.
• Add: Custom CSS section for all themes for easy styling tweaks.
• Add: New clean square theme template for a fresh look.
• Update: Enhanced Discussions module and all themes with cleaner layouts, better gradients, and new styling options (border color, font family, etc.).
• Update: Popup chatbots now have smooth zoom animations, mobile-specific fixes, floating close button, and better trigger positioning.
• Update: UI Builder now disables Container and Header options when Popup is not enabled.
• Fix: Preserved HTML attributes in the “No Credits” message for proper styling.
• Fix: Corrected shortcode ID generation by ignoring system/visual-only parameters.
• Fix: Improved footer rendering, tools, alignment, and empty footer handling.
• Fix: Adjusted GPT-5 model usage calculations to handle dated variations correctly, fixed reasoning payload handling, and ensured JavaScript in HTML Blocks API executes properly.
• Fix: Fixed drag-and-drop file upload visual feedback for accepted/rejected states.

2.9.9 (2025/08/07)

• Add: Support for GPT-5 model with reasoning and verbosity parameters.
• Add: Windows compatibility support for MCP script.
• Add: Support for MCP Servers in AI Forms.
• Add: Cross-Site feature enabling chatbots to be embedded on external websites.
• Add: Modular UI Builder with customizable containers, headers, and footers, including MacOS-style components.
• Add: Chatbot window width controls, center positioning option, and auto-open delay for better UX.
• Add: Center (centerOpen) option in popup settings for centered chatbot windows.
• Fix: Hotfix for drag-and-drop file handling error.
• Fix: Resolved minor bugs and tiny enhancements.

2.9.8 (2025/08/01)

• Add: Support for Chroma (another Vector DB) has been introduced for both Cloud and self-hosted instances.
• Add: Multi-Upload feature has been added to chatbots for easier file management (currently only for developers, will be released to all users soon).
• Fix: Hotfix for discussion REST routes being accessible when the discussions module was disabled.
• Fix: Prevented pdf.worker.min.js from being accidentally deleted during production builds.
• Fix: Resolved PHP errors caused by duplicate property declarations and improved type safety for returned_price.
• Fix: Various other fixes and improvements throughout AI Engine.

2.9.6 (2025/07/29)

• Fix: Addressed environment validation issues in Realtime chatbot to ensure reliable operation.
• Fix: Solved GDPR consent issues in popup modal chatbots sharing the same botId.
• Add: clearCookies() in MwaiAPI.
• Update: Enhanced IP address display by truncating hashed IPs for privacy.
• Fix: Updated audio transcription to support local file paths and refined the test suite for better reliability.
• Fix: Secured file listing and deletion endpoints by adding user ownership checks and session-based tracking for guest users.
• Fix: Prevented cron_discussions from running as a guest and hitting usage limits.
• Fix: Ensured image uploads are properly saved to discussions.
• Fix: Made Event Logs operate independently of the Client Debug.
• Fix: Resolved query debug logs to only display when the Queries Debug setting is enabled.

2.9.5 (2025/07/22)

• Update: Changed the AI Form Container block to output HTML directly instead of nesting shortcodes. That enhances AI Forms compatibility.
• Fix: Resolved a security issue related to SSRF by validating URL schemes in audio transcription and sanitizing REST API parameters to prevent API key misuse.
• Fix: Corrected a critical security vulnerability that allowed unauthorized file uploads by adding strict file type validation to prevent PHP execution.
• Implemented IP address hashing when Privacy First mode is active to enhance user privacy without disrupting discussions.

2.9.4 (2025/07/19)

• Update: Improved information messages related to vector stores.
• Fix: Centralized fallback logic in simpleFastTextQuery now automatically uses the default model to prevent silent failures.
• Fix: Resolved fatal error when OpenAI Vector Store was set as default without a store ID.
• Fix: Assistant environment detection now automatically identifies the correct environment when using assistants from non-default setups.
• Fix: Corrected undefined method error related to logging.
• Update: Display errors as part of the conversation with options to copy, delete, or retry.

2.9.3 (2025/07/17)

• Add: New Database Optimization feature in Dev Tools to improve plugin speed by adding indexes and removing old logs and discussions.
• Add: Better errors when encountering issues with OpenAI Responses API, Vector Store, or multiple functions.
• Add: simpleFileUpload feature to the Simple API for easier file handling.
• Add: OpenAI Vector Store as a new embeddings environment type for seamless integration with the file_search tool.
• Fix: Corrected the embeddings API to ensure proper vector creation and fixed related issues.
• Fix: Refined logging system and removed debug logs for cleaner operation.
• Fix: Replaced hardcoded model list with dynamic API capability detection.

2.9.2 (2025/07/11)

• Add: Google embeddings are now live—only relevant environments show up, and we’ve built in safeguards against dimension mismatches.
• Add: A handy metadata bar in Discussions shows start date, last update, and message count. Plus, you can tweak its look via settings or our new PHP filters.
• Update: Embeddings sync now pops up a sleek NekoModal (goodbye alerts!), with clear stats on what’s updated, added, up-to-date, or errored—and even backend action logs.
• Update: API error messages got friendlier and more helpful.
• Update: Tables everywhere got a makeover—cleaner layouts, raw model names, better spacing, and clearer dimensions.
• Fix: Cron jobs no longer crash when no default environment or model is set.
• Fix: Chatbot module warnings when params or chatId were missing have been squashed with extra validation.
• Fix: System-logs no longer throw “undefined ‘sort’ key” warnings.
• Fix: Embeddings sync now clears the WP post cache so content changes are always detected, logs checksums for debugging, and adds a read-only Env ID field.
• Fix: Model/dimension mismatches are only checked if you’ve opted in, and the active Env ID is now visible in the UI.
• Fix: Custom chatbots keep their own embeddings environment instead of inheriting the default.
• Fix: Google model names are deduped and cleaned up (suffixes now in parentheses), with the newest versions listed first.

2.9.1 (2025/07/08)

• Fix: Corrected guest user authentication by fixing the strpos check in session validation and making the start_session endpoint publicly accessible for proper guest login support.
• Add: Introduced options for Embeddings Search supporting multiple search methods.
• Add: New simpleFastTextQuery API endpoint.
• Update: Optimized vector search queries to accurately handle exclusion terms.

2.9.0 (2025/07/07)

• Add: Persisted templates in Content Generator, Images Generator, and Playground.
• Update: Improved clarity in tables and selects.
• Fix: Limited PHP session starts to AI Engine’s REST endpoints and added session status checks.
• Fix: Improved error handling for dynamic function additions via mwai_ai_query filter and ensured proper JSON encoding for mwai_ai_feedback objects in Chat Completions API.
• Fix: Enhanced CSV/JSON import error handling with detailed validation, specific error messages, and helpful examples in a modal dialog.
• Fix: Corrected guest user display in Discussions and Insights tables.

2.8.9 (2025/07/05)

• Fix: Pinecone vector listing now correctly handles variable index dimensions by generating matching zero vectors instead of using a hardcoded size.
• Fix: Pinecone index name extraction has been improved to accurately process hyphenated index names, and a dimension mismatch indicator has been added next to the AI Environment title in embedding settings.
• Add: simpleTranscribeAudio method now provides a consistent API for audio transcription using the default AI audio settings.
• Fix: functions_list now exclusively retrieves snippets of the “function” type from Code Engine to improve accuracy.

2.8.8 (2025/07/04)

• Add: Customizable Languages section in Settings > Others for easier language management.
• Add: Query tracking in usage statistics with accuracy indicators for token counts and pricing sources.
• Fix: Code Engine now correctly uses function names instead of snippet names for better clarity.
• Fix: Resolved PHP warning when syncing vectors caused by missing ai_embeddings_dimensions key.
• Fix: Fixed UI errors related to embedding model selectors, including conditional display, dimension validation, and UI improvements.

2.8.7 (2025/07/02)

• Add: Smarter token and nonce management for longer, more reliable sessions.
• Fix: Hotfix resolves issues with event logs display and enables proper event logging under Streaming mode for realtime chatbots.
• Add: Quick Test buttons for AI and embedding environments to verify API connections.
• Update: Default embeddings set to text-embedding-3-small.
• Fix: Prevents duplicate Media Library entries and stops unwanted auto-saving in the Images Generator.
• Add: Web Search support for Google Gemini models.
• Fix: XSS vulnerability hotfix in chatbot shortcode.
• Fix: Hotfix for unquoted userId in SQL queries to prevent database errors.
• Update: Enhanced usage metrics, cleaner model names, and more.
• Fix: Improved multiple function call handling and resolved reset issues for compatibility with OpenAI, Claude, and Google.
• Add: Initial limits for Realtime (still a bit tricky, but progressing!).
• Update: Improved PDF import with different chunking modes.
• Update: Improved the default CSS theme.
• Update: General refactoring with many small improvements and fixes.
• Update: Better debugging and error handling all around.

2.8.4 (2025/06/18)

• Add: History Strategy for Responses API chatbots. This allows chatbots to maintain a history of interactions, including images being modified or generated.
• Add: AI Want support for more blocks, like tables, lists, headers, etc.
• Add: Tools support (web_search, image_generation) with the Responses API.
• Add: Edit Mode for Images, with mask support.
• Update: Streamlined MCP logging, improved documentation.
• Fix: Duplicate Media Library entries.
• Fix: PDF worker URL (sorry about that, guys).
• Fix: Vision support for Responses API.
• Fix: Hotfix for security issues related to MCP.
• Info: More fixes and improvements which are too numerous to list here. We will have another round of improvements coming within 3-4 days. Don’t hesitate to leave a review and mention something you would like to see improved or fixed.

2.8.3 (2025/06/07)

• Add: Support for the new OpenAI Responses API (function calling, vision, feedback, MCP) – enable it in Settings when you’re ready.
  Add: Vector-Aware Search – override the default WordPress search with either AI-generated keywords or Embeddings for sharper results.
• Agregar: Importación de PDF – sube un PDF, retoca el tamaño de los trozos y crea automáticamente incrustaciones en un solo flujo.
• Add: Embeddings stream event that shows when external context is pulled in.
• Add: Stream-events viewer now appears under each chatbot whenever Client Debug is on (and streaming is enabled).
• Add: Claude MCP support, MCP-server picker in Chatbots, per-category Show Details buttons, and copy-to-clipboard endpoint fields; you can even create or edit plugins via MCP.
• Agregar: operador “No contiene” para bloques condicionales de IA.
• Add: Extensible context menus in Discussions through new MwaiAPI filters.
• Agregar: dev-notes.md repleto de consejos para desarrolladores que quieran ampliar el AI Engine.
• Update: Replaced NekoCollapsableCategories with NekoAccordions and refreshed the whole Settings navigation.
• Actualización: Depurador de transmisión renombrado a ChatbotEvents, redacción unificada y mensajes de estado más claros.
• Fix: Function-execution mapping, duplicate result events, and double fires in React.
• Corrección: Los estados de error y de entrada están ahora totalmente aislados por instancia de chatbot.
• Corrección: Botón de vaciado y lógica de restablecimiento en el visor de eventos.
• Seguridad: Parcheado un potencial vector de inyección MCP.
• Varios: Muchas pequeñas optimizaciones, correcciones de errores tipográficos/traducción y comentarios más limpios.

2.8.2 (2025/05/23)

• Agregar: Nuevos modelos Claude 4 para mejorar las capacidades de la IA.
• Update: Settings reorganized for improved usability, and Statistics and Embeddings renamed to Insights and Knowledge for better clarity.
• Corrección: Corrección de problemas de transmisión y adición de un filtro para personalizar el intervalo de actualización de los debates.
• Actualización: Interfaz de usuario más coherente y fácil de usar en los generadores de contenido e imágenes, incluido un modal limpio para el generador de imágenes y un bloque de chatbot personalizado mejorado.
• Add: Multi-condition support for AI Forms with new operators, and better handling of required fields within conditional containers.

2.8.1 (2025/05/03)

• Corrección: Resolución de un problema con el uso del modelo DALL-E en plataformas Azure.
• Agregar: Permitir la anulación de la clave API a través del parámetro de consulta para la integración de OpenRouter.
• Agregar: Filtrar mwai_discussions_refresh_interval para controlar la frecuencia con la que se actualiza la lista de debates.

2.7.9 (2025/04/30)

• Agregar: Compatibilidad para gpt-image (el último modelo de OpenAI).
• Add: Support for MCP (Model Context Protocol). Check the tutorial! It’s awesome, but remember it’s a beta feature.
• Corrección: Evitar algunas advertencias y avisos.
• Corrección: Compatibilidad con WPs sin estado.
• Fix: Added Row Actions in Pages.
• Actualizado: Gracias a la nueva API de OpenRouter, ahora es posible obtener precios exactos sin problemas.
• Actualizado: Optimizado el tamaño del lote del chatbot.

2.7.6 (2025/04/15)

• Agregar: Agregados los modelos GPT 4.1, y establecido el 4.1 Nano como nuevo modelo por defecto.
• Agregar: La privacidad como primera opción para establecer la cantidad de datos personales al mínimo.
• Agregar: Manejar las propiedades de array para llamada a función.
• Agregar: Ahora se puede modificar la extensión para chatbots y formularios.
• Agregar: Agrega un filtro en el desplegable de modelos si hay más de 16 modelos.
• Agregar: Los precios precisos con OpenRouter se pueden activar agregando MWAI_OPENROUTER_ACCURATE_PRICING a su wp-config.php, y poniéndolo a true. Esto agregará 1-2 segundos al tiempo de respuesta.
• Actualizado: Solo definir MWAI_TIMEOUT si no está definido todavía, que permite que sea anulado.
• Corrección: Un formulario de IA sin entradas debería ser siempre válido.
• Corrección: Dar más información cuando un upsert Pinecone falla.
• Corrección: Mejoras para Google Gemini. Ahora funciona con llamadas a funciones. ¡Agradecimientos especiales a Anaheim!
• Corrección: Evitar un fallo silencioso con Pinecone cuando se agrega una barra oblicua a la URL del servidor.
• Corrección: Evitar que Meow_MWAI_Query_Parameter bloquee WordPress por completo.

2.7.5 (2025/03/12)

• Agregar: Introducida la compatibilidad con Claude 3.7.
• Agregar: Se han actualizado los precios y se ha agregado la compatibilidad con la vista previa de gpt-4.5-, o1-mini y o3-mini.
• Corrección: Los asistentes OpenAI (herramientas) ahora se pueden actualizar sin sobrescribir los ajustes existentes.
• Corrección: Evitaba que los formularios AI sobrescribieran elementos HTML sin un atributo data-default-value.
• Corrección: Los campos obligatorios de los formularios de IA ahora se cumplen estrictamente, incluso si no se mencionan en la indicación.
• Corrección: Resueltos los problemas con los campos a través de selectores que no se guardan o se vuelven a aplicar en la carga.
• Actualización: Se han agregado comprobaciones adicionales para garantizar que el almacén se procesa antes de adjuntarlo a un hilo en los asistentes de OpenAI.
• Actualización: Mejorada la compatibilidad para asistentes que trabajan con formularios y subidas de archivos individuales, aunque el almacén de vectores de OpenAI sigue teniendo fallos.
• Corrección: Se solucionó un problema con los formularios de IA y los tipos MIME—OpenAI solo es compatible con imágenes, pero Anthropic maneja bien los PDF.
• Actualización: Sustituido set_max_sentences por set_max_messages para mayor claridad.
• Actualización: los enlaces del chatbot ahora se abren siempre en una nueva pestaña para mejorar la experiencia del usuario.
• Nota: ahora se espera que funcione la llamada a funciones, pero Gemini sigue sin ser fiable.
• Eliminar: No más estados de OpenAI, ya que su feed RSS está en desuso.

2.7.4 (2025/01/26)

• Agregar: Soporte para modelos de Perplexity.
• Agregar: MwaiAPI trabaja con AI Forms (ai.formReply filter, forms, getForm).
• Actualización: establecer API Azure para la versión 2024-12-01.
• Actualización: Aplica correctamente las herramientas y etiquetas de visión en modelos de OpenRouter.
• Corrección: Agrega de nuevo el shortcode relacionado con las estadísticas.
• Corrección: Regresa a la resolución por defecto si el modelo no es compatible con la resolución.
• Información: Nueva extensión para DeepSeek: https://meowapps.com/products/deepseek/.

2.7.3 (2025/01/16)

• Agregar: Realtime ahora es compatible en Debates y Consultas (los costos se calculan en función de los tokens devueltos por OpenAI).
• Actualización: Una forma más inteligente de manejar la sincronización de las incrustaciones. Se borran si no tienen contenido, y solo se sincronizan si es necesario.
• Actualización: OpenAI ahora utiliza max_completion_tokens en lugar de max_tokens.
• Corrección: Con los Formularios de IA, el Selector/Radio no podía mostrar ninguna opción seleccionada por defecto.
• Corrección: o1 no funcionaba correctamente con el generador de contenidos (el problema estaba relacionado con los Max Tokens).
• Agregar: En el Generador de Contenidos, los Max Tokens son ahora opcionales.
• Corrección: Evitar que el Bloque Chatbot se bloquee al cambiar a Personalizado.
• Agregar: Un botón para “Ejecutar tareas” en las “Herramientas de desarrollo”.

2.7.2 (2025/01/05)

• Add: Realtime Audio Chatbot (Pro Version)! It works very well, including with function calling. Try it out! But be careful, those models are quite pricey. AI Engine doesn’t handle the statistics yet (Queries / Discussions tabs).
• Agregar: Nuevo atributo ‘className’ para los Shortcuts API.
• Corrección: Los selectores de los formularios de IA ahora recuperan correctamente el contenido de los divs.
• Corrección: El bloque de chat en modo personalizado no funcionaba correctamente.

2.6.9 (2025/01/01)

• Información: ¡Feliz Año Nuevo a todos, hermanos y hermanas de IA! 🎉
• Agregar: Soporte para o1 (si tienes acceso a él).
• Actualización: Manejar el streaming con o1 (dependiendo del modelo exacto).
• Información: La vista previa de o1 y o1 Mini no funcionan con Instrucciones o contextos (incrustaciones, etc.). Esta es una limitación de OpenAI. IA Engine se encargará de esto tan pronto como OpenAI lo permita. Sin embargo, funciona con o1.
• Agregar: soporte o1 en la sección Utilización.
• Agregar: Opción para la característica de autotitulación de los debates.
• Add: Option to modify the Header Subtitle for the Timeless Theme (“Discuss with”).
• Agregar: La opción de ignorar límites de palabra (= ¡espacios!) en la sección de seguridad.
• Agregar: Tres nuevas funciones JS para la API Chatbot: getBlocks, addBlock y removeBlockById.
• Agregar: Nuevo campo de subida en los formularios de IA. Trabaja con imágenes, archivos, audios (un archivo por formulario por ahora).
• Agregar: Nuevo tipo de ‘callback’ para los atajos. Requiere una función ‘onClick’.
• Agregar: Se han agregado los marcadores de posición {CATEGORY}, {CATEGORIES}, {AUTHOR}, y {PUBLISH_DATE} para las indicaciones de incrustaciones.
• Agregar: Memoria local para los formularios de IA (compruebe los parámetros del bloque de envíos).
• Agregar: Botón de reinicio para formularios de IA.
• Actualización: La característica de autotitulación de los debates es ahora más fiable.
• Actualización: Se permite HTML en las Instrucciones.
• Actualización: Mejores instrucciones en el editor de datos de Finetunes.
• Update: Tiny enhancements on the Timeless Theme.
• Agregar: Botón para restablecer el uso (pestaña Escritorio).
• Corrección: Problema relacionado con los bloques HTML iniciales (el bloque GRDP podía reiniciar todos los bloques).
• Corrección: La actualización automática de los debates solo funciona si la pestaña del navegador está activa.
• Corrección: El uso de chatId a través de la API provocaba un fallo.
• Corrección: Mejora de la API interna en torno a las adquisiciones.
• Corrección: Los Asistentes no se comportaban correctamente cuando una función de llamada no devolvía el resultado esperado.
• Corrección: El contexto en las consultas (módulo de estadísticas) está ahora disponible para los asistentes.
• Corrección: Gestionar mejor los sitios web RTL.
• Corrección: si el entorno es por defecto, entonces el Modelo también es por defecto (para Chatbots y Formularios de IA).