AI Power Version History

- **Added**: Ollama embedding support.
- **Added**: Ollama image analysis support.
- **Added**: New Google embedding model: `gemini-embedding-2-preview`.
- **Fixed**: Google embedding batch inputs now process all provided text items instead of only the first one.
- **Fixed**: REST embeddings `task_type` handling for Google now preserves valid uppercase task types (e.g. `RETRIEVAL_QUERY`).
- General bug fixes and performance improvements.

- Further performance improvements in the Chatbot module.

- Performance improvements in the Chatbot module.
- Added gpt-5.4 to the OpenAI recommended model list.

- **Fixed**: On-page chatbots no longer auto-switch to fullscreen on small screens; popup chatbots still fullscreen only when opened.

- **Added**: A shared Prompt Library for Content Writer, Content Assistant, and Automated Tasks.
- **Fixed**: OpenAI store lists now update right away in AI Forms and Automated Tasks.
- **Fixed**: Generated images now save correctly in Content Writer.

- **Fixed**: OpenRouter image selection now shows only models that can actually create images, and Content Writer/Automated Tasks no longer report image success when no image is created.
- **Fixed**: OpenRouter now correctly creates the requested number of content images in Content Writer and Automated Tasks.

- **Fixed**: Webhook call issue in chatbot rules.

- **Fixed**: Chatbot no longer sends `temperature` to unsupported OpenAI GPT-5 chat-series models (e.g., `gpt-5.2-chat-latest`).
- **Improved**: AI Forms reliability with autosave and unsaved-work protection to reduce accidental data loss.

- **Added**: New maintenance tools in Settings (backup/import, restore point, cache cleanup, sync all models).

- **Improved**: Added option to enable web search by default for chatbots.
- **Added**: Image editing support for Image Generator (OpenAI, Google, OpenRouter models).
- **Improved**: Simplified image generate/edit flow in the shortcode (`mode="generate|edit|both"` with default mode and switch).
- **Improved**: Better image upload experience for edit mode (drag & drop, preview, clear errors).
- **Improved**: One-click **Edit** from image history to reuse images without re-uploading.

- **Improved**: Migrated OpenRouter to the Responses API.
- **Added**: OpenRouter embeddings support in chatbot, content writer, automated task and ai forms.
- **Added**: OpenRouter web search support in chatbot and ai forms.
- **Added**: OpenRouter image analysis support in chatbot.
- **Added**: OpenRouter image generation support in chatbot, content writer, automated task and image generator module.

- **Improved**: OpenAI streaming now sends small status updates to keep streams stable.
- **Improved**: Increased default SSE timeout for AI Forms and Content Writer (can be changed with `aipkit_stream_timeout` filter).

- **Improved**: Moved assistant settings from Dashboard into the Content Writer module.
- **Improved**: UI/UX improvements for the Content Assistant, the post list assistant modal, and the “Add to Vector Store” modal.
- **Fixed**: Chat logs pagination now renders correctly.
- **Fixed**: Reasoning effort options now align with OpenAI model rules (e.g., minimal mapped to low).

- **Improved**: Automated Tasks UI polish.

- **Improved**: Chatbot custom themes.