* memory optimization for large media galleries

+ Adds support for OpenAI Sora 2 video generation models + Implements streaming response for all text based models (Local, Cloud, Networked)

+ Adds support for OpenAI GPT 5.1 & GPT Image 1.5 models * Fixed issue where image capture and selection buttons were not properly shown for all the models that support Vision input.

* Fixed templating prompt dialog from re-appearing while typing the same prompt.

* Now your device won't go to sleep when an ai is actively generating a response(iOS)

- Fix issue with unlocking private conversations when biometrics are unavailable - Fix template prompt detection to not display when typing a comma, or curly bracket in your prompt. Now it will only show it when actually pasting text with those characters. - Fix conversation titles only being the first character of your prompt - Fix delete single conversation ability. - Fix default model to use the bundled local model + Adds clarification about cloud models on the API Settings view.

+ Adds in app media gallery viewer with slide show support and bulk management options + Adds Drag and drop support for adding images or text files to a user prompt or dragging images out of a conversation to anywhere on your mac + Adds timestamps to each message context menu so you can see when that message was added to the conversation * Images generated in Ari will now have the prompt stored in the images metadata automatically

+ Adds in app documentation + Adds prompt draft persistence so you can jump between conversations and pick back up on your prompt ideas any time. + Adds support for automatically setting up and configuring OpenCode + Add support for prompt templating + Add ability delete individual messages in a conversation + Adds support for repeating a prompt up to 1000 times in a row. + Adds support for side by side conversations each running prompts in parallel + Adds prompt queue controls to pause or stop a prompt template or repeat run on a conversation + Adds indication to conversation list for conversations with new messages from running in parallel to the current open conversation + Adds indication about which model was used for each response and allow you to quickly switch to that specific model for your next prompt. * Fixed the conversation list ordering to properly be ordered by most recently opened. * Updates the Conversation list to use tabs to switch between Private and Normal Conversations * Makes Delete all button only apply to the current selected tab - Fixes formatting with share conversation feature

+ Adds private conversation passphrase management section to the API Settings view. * Improved Conversation Sync reliability

+ adds support for Claude Sonnet 4.5 + adds support for private conversations that are encrypted on disk using an iCloud synced secret + adds support for batch export of generated images in a conversation + adds a share conversation feature to share any conversation via either a convenient html or image file. + adds support for ollama vision input * improved reliability of conversation sync

* Fixes an issue that caused improper Draw Things Image Generation results because the proper model name was not being sent with the request.

* Fixes image selection not always showing your selected images.

* Optimized cloud prompt guard to not be so restrictive

+ Adds support for asking questions about multiple images at one tine. * Bug fixes and improvements.

+ Adds support for grok 4 fast * Fix issues where sending to large of a prompt to a cloud model was being limited incorrectly. GPT-5 and other cloud models can now properly handle complex prompts. * Fixed message history context so that you can have a true conversation with any modal where the AI remembers what was already discussed. * Updated local model engine to latest version to enable more reliable responses.

+ Liquid Glass UI * Fixed an issue where Conversation Mode would not properly record the users prompt. * Improved iCloud conversation sync reliability. * Bug Fixes and Improvements

+ Added ability to select sections of the ai responses. So when you only need a couple pieces of the response now you can select just those pieces. * automatically trim whitespace and new lines from the api keys to ensure the api keys can be entered properly.

+ Added support for chatting with Ollama self hosted models + Added support for generating images with DrawThings + Added support for image and text file input + Added support for chatting with OpenAI models + Added support for chatting with Grok Models + Added support for chatting with anthropic Claude models + Added support for downloading custom local models. + Added local on device spoken conversation mode with pause detection and auto resume + Switched default local model to be Gemma 3 4B

* Switched to using Qwen3

* Fixed a bug that caused conversation messages to not load for newly created conversations you re-open from the conversation history menu before re-launching Ari Helper.

* switched to an uncensored model to remove some political bias * Backend Dependencies were updated * Implemented DRY principles to reduce the repetitions of R1 Models.

Overall just making Ari look its best. * Improved Previous Chats animations and spacing. * Improved spacing around code block in the responses. * Improved spacing in header bar. * Added fallback to store Previous Chats on device only if iCloud is unavailable.

+ Code Syntax Highlighting + Code Block Copying + Previous Chats Selection menu with Delete All Button and Delete Context Menu + Cross device syncing via iCloud Drive

* Tuned the LLM parameters to improve the responses from DeepSeek R1 On Device.

* Changed AI Model to DeepSeek R1 1.5B