IMPROVEMENTS
• Added a Share button to the action bar under every completed assistant response

BUG FIXES
• Fixed images generated by models that return base64 data URIs in the files array not showing in the app
• Fixed "Show follow-up suggestions" setting not hiding already-stored suggestions when turned off
• Fixed tapping citation links and web links in assistant messages doing nothing — taps now correctly open the in-app browser
• Fixed very long assistant messages (a single giant paragraph or flat list) rendering completely blank
• Fixed very long user messages rendering blank in the chat bubble — long text is now split into chunks to stay within rendering limits
• Fixed hamburger menu button not responding to taps on its center and left side when the drawer was closed
• Fixed newly opened chats landing mid-conversation instead of at the bottom when Markdown finishes rendering after the initial scroll settles
• Fixed a memory leak where notification observers accumulated each time a chat was reconfigured and were never released after closing a chat
• Fixed rendered message text reappearing after the view was reset — the markdown subject no longer replays its last value to new subscribers
• Fix stale loading placeholder when Markdown content changes during parse
• Fixed duplicate message views staying blank when two views requested the same Markdown parse simultaneously — all waiters are now notified when a shared parse finishes

WHAT'S NEW
• Added a "Compacting context…" status indicator that appears while the server is compressing a long conversation before generating a response
• Added "Rename" and "Insert Path into Chat" actions to the terminal file browser context menu

IMPROVEMENTS
• Model selector in the navigation bar now expands dynamically
• Added activity spinner next to conversations in the sidebar while a generation is in progress, including while waiting for the first token in new chats

BUG FIXES
• Fixed auto-scroll stopping early during long streaming responses — the scroll pump now stays active until content fully renders, not just for a fixed window after streaming ends
• Fixed hamburger menu becoming inaccessible while a response is streaming
• Fixed short conversations opening with messages not fitting on the screen
• Fixed keyboard not dismissing when tapping the empty area of the chat in short conversations
• Fixed chat input bar being hidden behind the keyboard in landscape mode when the terminal file browser is open
• Fixed ask-user tool call marking response as finished early while waiting for you to answer a question

IMPROVEMENTS
• Tightened conversation row spacing in the sidebar.
• Removed the duplicate update notification icon from the top of the sidebar

BUG FIXES
• Fixed tool call cards showing a checkmark immediately when the model writes the call request
• Fixed responses finishing prematurely during long-running tool calls — the recovery timer no longer treats an empty task list as "done" when there is still an in-progress tool call waiting for a result.
• Fixed responses being marked as finished when returning from the background too quickly

WHAT'S NEW
• Added custom photo picker

IMPROVEMENTS
• Allow OpenWebUI's new streaming architecture that now sends tokens as `response:completion` delta events which the app was silently discarding.

BUG FIXES
• Fixed assistant messages not appearing when sending new messages after clearing chat history or continuing responses
• Fixed streaming responses appearing choppy with visible pauses between chunks — responses now flow smoothly at a constant rate
• Fixed TTS related issues with sentence breaking.
• Fixed app icon showing a badge number after every message even when the response finished while you were still in the app — the badge now only appears when you go to the home screen while a response is still streaming.

WHAT'S NEW
- Added in-app browser: tapping links and citation chips now opens a built-in browser instead of leaving the app.
- Added "Open links in app" toggle in Settings → Chat → Chat Behavior to disable the in-app browser.
- Added API key management to Privacy & Security settings — generate, copy, and revoke your server API key directly from the app.
- Added human-in-the-loop tool approval: a "Tool Permissions" entry appears in the + attachment menu to switch between Full access and Ask for approval.
- Added ask-user question cards: models can now pause mid-response to ask you multiple-choice questions with an optional free-text answer field.

IMPROVEMENTS
- Failed tool calls now show a red X instead of a green checkmark.
- Profile pictures that fail to load now fall back to the default avatar instead of showing a broken placeholder.
- Typing @ in a channel now lists members in alphabetical order.
- Attaching a file and sending before the upload finishes now queues the message and sends automatically once the upload completes.
- Added OpenSERP as a web search engine option in Admin → Web Search.
- Admin Documents: added Knowledge File Retention, CSV Shape Summary, and Metadata Max Value Characters settings.
- Admin Interface: added Task Model Generation Parameters editor with individual controls for all parameters.
- Admin Interface: added Default Interface Settings editor for configuring server-wide defaults for all new chats.
- Admin Integrations: Terminal server connections now have Enable in Chats, Enable in Automations, and Scope settings.
- Admin Groups: added Import Members button — import group members from a CSV file containing email addresses.
- Automations: added Channel Destination option so automation responses can be posted to a specific channel.

BUG FIXES
- Fixed server update changelog to use GitHub since the webui API does not provide the current update changelog.