v0.10.0

This is a cumulative update for all 0.10.0-beta builds.

Features:

• Internationalization (i18n)

  • ChatterUI now supports translation files for the primary UI.

  • Supported languages:

    • en
    • zh
    • ru
    • in

  • You can override the language in Settings > Set Language.

  • Refer to #559 if you want to contribute your own language!

  • Disclaimer: A lot of machine translation was used, so translations may not always be contextually correct.

• Author Notes - Allows you to inject guidance text at a specific depth. Notes can be stored per chat, character, or globally.

• Lorebooks - Adds a popular RAG system that uses keyword searches to add additional context and descriptions.

  • Adds a new Data Source path in the navigation for Lorebooks and future Data Source features.
  • Lorebooks can be imported from JSON files or created manually.

• Character Links - You can now link specific configs to load when a character is loaded. This covers:

  • User cards
  • Sampler configs
  • Formatting configs
  • Connection configs

• Custom API Fields - You can now add custom fields to API requests.

  • You can enable this in Connections > Settings Dropdown > Enable Custom Fields. This toggles the input field for custom definitions in the connections editor.

  • Each value can be defined as a JSON key (nestable with .), separated from its value with =.

  • Example:

    preferences.security = "Low"
    preferences.privacy.allowRead = false
    

  • These values will be added directly to your request body:

    {
      // ... rest of body
      "preferences": {
        "security": "Low",
        "privacy": {
          "allowRead": false
        }
      }
    }

• Added a thinking budget for local mode.

• Local Mode: Prompt and generation times are now shown in chat bubbles.

• Added a token count estimator in chat. This will not work on older chats until a new message is sent.

• Updated to the latest llama.cpp.

• Added more valid Markdown quote types.

• Quote Markdown now supports more quote types.

• Added a Jump to End button after scrolling up for a while.

• Chat Input: The keyboard is now dismissed when pressing send.

Changes:

• Reworked the Connections screen.

  • You can now clone a connection.

• Reworked the Model List screen.

• Character Cards without images can now be exported as JSON.

• Image picker now uses the native media library instead of a file system picker.

• Autoformatter now cleans quote types into ASCII quotes.

• Migrated to a new LaTeX renderer.

• Added exit animations to BottomSheet components.

• Removed many redundant rerenders.

• Improved styles for certain components.

• Prefills for Claude are no longer supported due to deprecation.

• Under the hood:

  • Chat states are now managed completely differently UI-wise and should no longer hog massive amounts of memory for long chats.
  • The chat window is no longer tied to a Zustand state, removing many redundant functions where state and storage updates had to be optimistically updated.
  • The context builder has been revamped and may cause issues.
  • Updated to Expo SDK 56 with improvements to UI animations and startup time.
  • Migrated to react-native-vector-icons.
  • Context Menus have been reworked.

Fixes:

• Fixed a crash caused by a workletized function in BottomSheet.
• Fixed macros not being applied in Author Notes.
• Fixed creating a new character-level Author Note also creating a new chat-level note.
• Fixed think tags being improperly parsed.
• Added <thinking> to valid reasoning tags.
• Fixed reasoning token limits not working in local mode.
• Fixed text filters not applying properly.
• Fixed various issues with image Markdown.
• Fixed AvatarViewer image aspect ratios not updating when first imported.
• Fixed incorrect tags on exported character card PNGs, making them unreadable by other CCv2-supported apps.
• Fixed Unicode quotes crashing chats.
• Fixed an uncaught error in connection model parsing.
• Fixed RTL streaming for specific languages.
• Print Context will no longer print media data, which was causing crashes.
• Chat auto-generated names should no longer include reasoning in Remote mode.
• Fixed being able to press send twice in a chat during generation.