• Long chats open at the latest message and no longer go blank after you send. • A new reply layout without bubbles. Choose it in Settings › Appearance › Reply Style. • Refined themes: better contrast, consistent corners and a soft light behind the chat. Hermes has a calmer blue. • Mermaid charts, SVG images and HTML previews are drawn right in the chat, in your theme's colors. • Options the model writes become buttons: tap one to reply. • While the model thinks, one line shows what it is working on. • Select just part of a reply to copy it. • Conversation too long for the model? The app summarizes it and carries on. • Pick a picture for the chat background. • Temporary chats no longer lose earlier replies.

• Follow-up suggestions: after each answer, the app shows questions to keep the conversation going. Tap one to send it. • Suggestions are saved with the chat, the same ones your server's web app shows. • You can turn them off in Settings › Chat › Follow-up suggestions. • On iPad, the conversation sits in a reading column in the middle of the screen instead of stretching edge to edge. • While an answer streams in, the chat follows it to the end instead of stopping halfway on its own. • Web search in voice mode: when a question needs current information, the model searches and answers, and the sites it used appear under the reply. It is on by default; you can turn it off in Settings › Voice & models › Search the web. • Models that write their search command out as text no longer show it or read it aloud: the app runs the search instead. • If you leave the app or lock your iPhone in the middle of a reply, it is no longer lost: when you come back, the app asks again and the full answer appears in the same place. • In voice mode, replies are shown formatted, with headings, lists and bold, instead of raw text. • Search the web, in the chat, now lists the sites it found instead of the search text. • Follow-up suggestions and chat titles also work with slow servers. • Thanks to SaintChris for the idea and the detailed write-up on GitHub.

• You can now listen to the messages you sent, not just the replies. • When the chosen voice can't play — say, a server with no speech engine set up — the chat tells you why and your device's voice reads the message instead of staying silent. • The model menu lists only the models you can chat with; image, video, speech, transcription and embedding models sit apart, each labelled with what it does. • Provider errors (Gemini, OpenRouter and others) show up as a readable sentence instead of raw code or a generic warning. • Gemma models answer again: when the provider refuses the system instruction, the app folds it into your message and retries. • The app opens faster when the server is slow or down. • Workspace and Model names group models by type.

• Voice mode now answers in your language — it used to answer in Portuguese, whoever was speaking. • The interface is now fully in the language you picked: the message field, dates and error messages no longer show up in Portuguese. • On-device transcription rebuilt: whisper.cpp 1.9.4 and NVIDIA Parakeet, with models tuned for 21 languages. • Your recording is saved before it is transcribed — if the app closes mid-way, it offers to recover it on the next launch. • The model catalogue opens folded: pick a language and the sizes come grouped. • The app tells you when a model does not fit your device's memory, and large models get more memory on supported iPhones. • A new Diagnostics screen, and a button to report a problem by email.

• The photo you attach now shows up in the conversation — and the model finally receives it. • A new chat opens with the last model you picked. • The model's reasoning is shown alongside the answer again. • The chat list no longer stops at the first 60. • If a save fails, the conversation is kept on your device and you can send it later. • Sign in with SSO or LDAP, when your server offers it. • Works with Open WebUI 0.11.

• Edit a question you already sent — the original and its answer are kept. • Regenerate a reply, or try again with a different model. • Move between versions with the arrows under the message. Nothing is deleted, not even branches created in the web version.

• Tools from your Open WebUI now work inside a chat — pick which ones with the wrench icon. • Voice follows the app language: read-aloud and dictation no longer always speak Portuguese. • Neural voice in 6 languages, up from Portuguese only. • Nicknames for models — no more giant IDs on screen. • Manage storage: see and delete downloaded voices and Whisper models. They no longer ride along in your iCloud backup either. • Add your own Whisper model by URL (https). • Permission prompts translated into all 44 app languages.

• Compatible with the latest Open WebUI: replies created in the browser now show up in the app, and nothing is lost when you continue the conversation on your phone. • Web search works again, now with cited sources and a progress indicator. • The open conversation refreshes itself when something arrives from the web version. • Stability fixes for reply streaming.

• 9 new languages — the app now speaks 44. • History of your generated images, with full screen and zoom. • Language-specific Whisper voice models. • Fixed: messages and images created on the web now show up in the app.

Improvements and fixes.

Version 1.2 brings 34 app languages: - Automatic device-language detection. - Right-to-left support (Arabic, Persian, Urdu, Pashto). - A new in-app language switcher in Settings. - Improvements and fixes.

Version 1.1 brings full multilingual support: - The app now automatically detects your iPhone's language. - Added English, Japanese, Hindi, and Bengali (in addition to Portuguese). - New language switcher in Settings > Language to change it anytime. - Various improvements and fixes.