OpenClient: Private AI Chat

Version 1.7.5

Hi there! We've got some great new features for you in this update. • Your favorite chat model just got some talented teammates! Models that support tools can now call on specialists to analyze photos or create images, without switching models or starting another chat. Ask about a picture, dream up something new, or revisit an image from earlier in the conversation. It all happens in the same chat, and you can see which specialist is helping along the way. • You're in control of the team. Choose your vision and image generation specialists in the new Image and Vision section in Models. Use one compatible model for both jobs, mix and match, or choose None to turn off either kind of help. If your chat model already understands or creates images, it keeps doing that job itself. • Your assistant's toolbox is now in your hands! Open the new Tools section in Settings to see what each built-in tool does and turn it on or off. Everything is listed alphabetically in your language, and OpenClient remembers your choices. • Got a picture you'd like to change? Image-generation models with vision now accept one or more reference images for editing on iPhone, iPad, and Mac. Trying again keeps your original prompt and references, so you don't have to set everything up from scratch. • We've given your creations some extra care, too. Images stay in the conversation if you stop a reply after they arrive, and regenerating just the text keeps the image without requesting another generation. We've improved image-only replies, image format handling, and references in long chats and imported backups. If an analysis request is too large for your specialist, clearer guidance helps you adjust it. Private chats still keep images only for the session. • Your conversations just got easier to read. We've brought your messages and the model's replies closer together, with each reply's time, details, and actions in a compact row. Less empty space, more conversation in view. • Better formatting for your replies! Markdown tables now display correctly even without separators at their outer edges. Footnotes appear as numbered notes at the end of the message, and escaped characters are handled correctly inside tables, too. • Small details make a difference, too. You can now see subscripts and superscripts, like H₂O or x², even inside links and alongside bold or italic text. Their size follows your text size on iPhone, iPad, and Mac. • Finding a chat on iPad is now more convenient. Search directly from the top of the sidebar and see matching conversations on the right as you type. Open a result to jump into that chat and dismiss the keyboard; your search stays there for later. Prefer the top bar? The search button is still available there, too. • Voice setup is a little easier as well. If you haven't selected a model, or your previous choice is no longer available, text-to-speech picks the first available model and speech-to-text selects Apple speech recognition. Both work even when there's only one option. We've also included smaller fixes and improvements to keep things running smoothly. Thanks for your continued support and for helping us build the best possible LLM client together. Remember: you can suggest and vote on new features from the Feedback section in Settings.