This update lets you use your own OpenAI-compatible server and any model ID your provider offers, and shows each provider’s official logo. - Connect an OpenAI-compatible server over HTTPS: enter its address and exact model ID, with an optional API key. MILO asks your permission before sending it your prompts. - Pick a suggested model or type any model ID your provider offers, in chat or Settings. Your choice is kept after restarts and updates. - With web search on (Brave, Tavily or TinyFish in Settings → Tools), ask MILO to search the web in your message, for example “search the web for…”. It no longer searches on its own. - Providers show their official logos in onboarding, the model picker, chat, History, Settings and Help, sized to match your text. VoiceOver says which model filter is selected. - When a request fails, Settings, chat and Siri, including Siri in CarPlay, tell you whether to check your API key, model, account access or billing. Error messages no longer show the provider’s raw response. - Newer Claude models, including ones you enter by ID, use Claude’s default creativity setting. - If you use a cloud AI provider, open MILO after updating and allow AI Data Sharing again. Until then, Siri will ask you to open the app.

This update makes MILO easier to use with VoiceOver and larger text, keeps long chats working, and tightens how your keys and conversations are handled. - Optional web search can use current web results when you turn it on in Settings → Tools. - Conversation mode lets you choose Session (follow-ups remember context) or One-shot (each question stands alone). - VoiceOver now names every chat button, and text follows your Dynamic Type size. - Long conversations stay within the model’s context window instead of failing to send. - If local chat storage is damaged, MILO explains what happened and lets you reset — it no longer quits on launch. - API keys are saved more safely and stay on this device (they are not included in iCloud or device backups). - Chat content is no longer written to device logs. - Cloud AI is used only after you agree for the provider that will actually handle the request. - Checking the network no longer freezes the composer, and save errors are shown instead of failing silently. - Claude Opus 4.7 and 4.8 work again. - History search is faster, and the version in the sidebar matches the real app version.

• Added a clear AI Data Sharing permission screen before cloud AI requests. • Settings now explains what prompt/context data is sent and which third-party AI providers may receive it. • Updated Privacy Policy and Terms links for AI provider data use. • Removed bundled API-key migration; users stay in control with their own provider keys.