• Update name to Character Agent

• Characters can text you first: turn on check-ins, with quiet hours, Focus and a pause switch. • Message a character, continue the story or start a voice chat from Siri, Shortcuts, widgets and Control Center. • Branch a chat from any message: press and hold it, choose Branch from Here, and try a different direction while the original stays as it was. • Let characters know only what you allow, like the weather, your calendar or your steps, and have them set alarms and reminders. • Lock chats with Face ID, and see every alarm, message and check-in characters have set in Upcoming Events. • Long replies keep their pace to the very end, and several crashes are fixed.

• Talk to your characters: tap the mic, or start hands-free voice chat. • Hear replies read aloud with Apple voices, or your own OpenAI, ElevenLabs and other voice services. • Voice chat works from the Lock Screen and with AirPods presses. • Replies keep generating when you leave the app, and you'll get a notice when they're ready.

• Optimize lorebook defaults in non-sillytavern prompt engines • 30% more usage in all subscriptions • "What changed" box pops up for new version on launch, update box pops up for new updates

• Added FREE TRIAL AI Subscription options in-app. • Added Chat Launchers: Each card can have different settings. • Improved theme management: default to transparent themes, default to character card as background, prevent duplicate default themes.

• Agentic chat & character management: chat with AI to update chats and characters, or manage them (delete duplicates, summarize, etc.) • Improved cache preservation across the app • Fix some issues with tool use, normalize tool use across chats. • Add toggle for post-history message in system prompt (story settings): modern models can handle this, and preserve cache. Toggle to get default sillytavern post-history instruction location (breaks next message cache each round)

- Improve Settings & Subscription UX - Burst meter no longer shows > available weekly meter when weekly meter is low -

- Optimized HTML rendering - Chat import now complies with Character.AI and Tavo export formats - Optional "lazy load" toggle for long chat performance

- Add support for alternate greetings. Long press character card to use it. - Save partial generations: now you can stop, unfocus the app, lose signal, or kill the app and still see a partial generation. - Hide "think" xml blocks. For now this is hardcoded. General regex support is still in "planned". Use "Chat Debug Mode" to show them.

- Bugfixes: fix slowdown with long markdown messages, reduce repetitive renders of elements, remove extra message type detections, fix bad markdown element causing crash in new markdown renderer.

- fix: lorebooks are triggered by system message - fix: top chat bar properly accepts transparency - fix: swipe and immediate stop no longer results in a hidden message.

- Upload images in the chat. OpenRouter auto-detects image upload capable models, otherwise manually enable. - Correct swipe animation transparency to 100%, optimize swipe animation performance

• multiple chats can now receive replies at the same time • New llm model and provider features: • advanced model parameters • model profiles • custom model assignments for memories and titles • Fix: duplicate character if typewriter animation is enabled, during buffering • Fix: view jumping after long messages and some tool calls

- Filter CRLF to \n to bypass swift-markdown parser bug; likely mainly affects importing microsoft chat exports from certain platforms or windows systems.

- Improve chat scrolling smoothness and prevent jumping, especially around markdown and tool calling.

- Fix streaming content bug for iOS 26 and some iOS 18 cases. - Improve UI scroll performance and animations - Add more chat import compatibility - Upgrade markdown renderer: perf + ui - Improve tool prompt\ - Disable html by default for new users - Typewriter animation fixes

- message formatting bugfixes for deepseek-chat, deepseek-reasoning, and GLM 4.6 - fix: markdown detection for bold, italics, and some whitespace formatting cases - fix: show system prompt in sillytavern mode when chat is reloaded - fix: post history instruction system message was not showing up in the initial user message - fix: MCP preserves querystrings in url (e.g. for auth) - add small deadzone to swipe left/right on messages to avoid accidental swipes - remove animations from scrolling and simplify scrolling to improve reliability, includes improving scrolling when using deepseek models

- Added MCP Client for use with any MCP service.

- (internal) collect crash reports

- Chat import/export beta is now live. Import tested to & from SillyTavern. - Swipe animations were often not registering, this was a regression and now fixed. - Fixed: chat message word limit "off" value was not being persisted due to logic error.

- Improvements to navigating in the chat history - Add alert to "manually re-associate" any character card if a character is deleted and added back with another name. - Automatically re-associate characters when deleted and added again with the same name.

- added html in chats - improve keyboard dismissal animation in chats - ensure active chat can be returned to from home page

- Add extensive new user setup flow for Gemini API and character cards, as well as quickstart messages and links for new users. - Add import character card by URL, this works on mime type, but doesn't have any specific logic, so you need to have the URL of the png or json file. Otherwise download PNG/JSON and import from downloads. - Add "summarize and continue" chat feature which will bring up a new chat with your summary. - Replace character management view with character management menu. Home screen is now the character view. - Persist chat memory associations rather than dynamically allocating the global memories each message - Now when you start a chat, the checked memories are permanently associated with it (this will get a setting on long press in the future) - Improve typewriter animation and prevent markdown rendering for normal text objects, improving performance on older devices. - Only use prefixes on system messages until they can be incorporated into preset settings. - Manage partial responses well in message regeneration - Fix issue where themes were not installing for new users.

- New home view with tag filters - Preset Importing - use your favorite creator's SillyTavern prompt presets (use the chat completions json only, not the smaller prompt presets json) - Some bug fixes related to theme, persona, and provider switching - Align raw message contents to SillyTavern (name prefixing for user, character, and system messages & merge messages) - Improved swipe and added swipe animation - Fixed issue with cancelled/errored text generations causing messages to disappear: missing messages from past conversations will also now return on view reload - fixed stale persona name in character card display view - increase llm request timeout to 300 seconds - add "off" setting for typewriter animation on far left of slider

- Themes are free. Theme bugfixes. - Default chat theme with background - Improve settings UX