oxproxion
io.github.stardomains3.oxproxion
Zapstore _@zapstore.dev Republished from GitHub / F-Droid by the Zapstore main account.
First release: Sep 23, 2025, 123 total releases.
Most recent release: Sep 26, 2026.
Appears in 2 app stacks.
0 sats / 0 zaps received in the past year.
Sats Received
Underlying data available via MCP: app_zaps, app_releases.
Zap Count
Underlying data available via MCP: app_zaps, app_releases.
Releases
- Sep 26, 2026 2.2.5### Changed - Updated some dependencies ### Added - Implement crash-recovery draft persistence for active chat - Continuously persist the current chat to a rolling draft file - Restore draft on cold start so chats survive OS kills, force-stops, and reboots
- Sep 20, 2026 2.2.4- Added support for Nativ, a Mac-only inference server. https://github.com/Blaizzy/nativ It supports the transcription endpoint in addition to text/vision. - Added Quit button for Transcription preset for when recording. - Updated some dependencies. - Found a way to bring back code highlighting in the normal chat interface
- Sep 13, 2026 2.2.3- Dependencies: Updated some dependencies. - Brave Search: Migrated brave_search to LLM Context API (supports pre-extracted content and up to 50 results). - Brave News: Replaced news filtering with new brave_news tool via /res/v1/news/search.
- Sep 6, 2026 2.2.2* Build Improvements: Updated Gradle and key dependencies. * Transcription Enhancements: * Optimized the copy button within the chat interface. * Improved transcription preset workflows: You can now edit transcriptions directly in the display. * Replaced auto-copy with a manual copy button. * Added "Try Again" and "Quit" buttons for better control.
- Aug 24, 2026 2.2.1- build.gradle.kts change - PDF input optimizations And From 2.2.0: - Breaking Change: Syntax highlighting in native TextView has been removed to support AGP 9+ compatibility. HTML view remains fully supported. - Dependencies updates - Action buttons on chat bubbles now scroll horizontally if needed.
- Aug 21, 2026 2.2.0- Breaking Change: Syntax highlighting in native TextView has been removed to support AGP 9+ compatibility. HTML view remains fully supported. - Dependencies updates - Action buttons on chat bubbles now scroll horizontally if needed.
- Aug 2, 2026 2.1.104- Networking improvements - Removed auto copy to clipboard for watermark transcription. Instead, a copy button appears for a few seconds. - Updated a couple dependencies.
- Jul 28, 2026 2.1.103- Updated a couple of dependencies. 🎙️ Voice Transcription & Audio Improvements • Fixed Network Reliability: Resolved unexpected end of stream errors on local transcription requests by using a dedicated, unpooled HTTP client with automated retries. • Hardened Touch Gestures: Prevented parent scroll views from stealing touch events while holding the watermark recording icon. • Audio Hardware Safeguards: Ensured MediaRecorder resources are always safely released to prevent microphone locks and added a minimum duration threshold to discard accidental taps. • Cache Management: Improved temporary audio file cleanup on upload completion and app startup.
- Jun 14, 2026 2.1.94- A user requested a light theme, so you can now select that in settings. This is an initial usable version of it. - Added an option to be able to use the Speech-to-Text model in settings for general voice transcription. If you have an STT model inputted and this setting on, press and hold the watermark in the main chat screen while you talk. When you lift your finger, your speech is transcribed, copied to the clipboard and entered into the prompt box. - Added a delay tool.
- Jun 12, 2026 2.1.93- Added an option in settings for an alternate STT model for long-pressing the mic icon that is in the extended dock. Set your preferred cloud or LAN(oMLX) STT model in Settings. - Long press the clear button in the extended dock to copy prompt box text to clipboard. - OpenRouter fetch function now fetches all image generation models available too.
- Jun 11, 2026 2.1.92- Updated dependencies - Added audio transcription support for those models on OpenRouter and oMLX. Audio Transcription Models: - Support for audio transcription models for OpenRouter and oMLX(Other LAN model servers don't support it as far as I know.) - Click the mic icon on the main chat screen to load your audio file and then click send button. - Only returns the transcript. These models do not chat. - For a transcription model on oMLX, toggle the Transcription toggle in the edit model screen for it to be allowed to attach an audio file. - Different models may support a different set of audio file formats.
- Apr 30, 2026 2.1.90- Updated some dependencies. - Fixed: Ollama thinking mode not respecting "Disable Reasoning" setting
- Apr 25, 2026 2.1.89- Some fixes for the Chat Memory feature introduced in v2.1.88
- Apr 25, 2026 2.1.88- Added option in settings for to allow the user to control context by limiting amount of past messages sent in chat.
- Apr 24, 2026 2.1.87- Tools management now has its own screen. Access from long press tools button or the new button in the Settings screen. - If you didn't select the Download/oxproxion folder for the workspace for when asked initially (and therefore tools aren't working), you can select again via a link in the Help screen.
- Apr 23, 2026 2.1.86- Fix for asking for permissions for the new set_sound_mode tool from v2.1.85
- Apr 21, 2026 2.1.84- Button settings are now above the switch settings in the Settings screen. - Reworked the Prompt Library screen/functioning. Tapping the item itself takes you back to main chat screen and will enter that saved prompt into the prompt box. Copy button will take you back to main chat screen but only puts the saved prompt in your clipboard.
- Apr 19, 2026 2.1.82- Fixed/Improved delete_files tool - Added oMLX as LAN model server option.
- Apr 18, 2026 2.1.81🗺️ Native Google Maps Navigation You can now ask the AI to start turn-by-turn navigation directly. This tool integrates with the Google Maps app to provide a hands-free transition from chatting to driving. • Turn-by-Turn Navigation: Launch Google Maps instantly by providing a destination name, address, or coordinates. • Custom Travel Modes: Support for Driving, Walking, Bicycling, and Public Transit. • Route Preferences: Optionally tell the AI to avoid tolls, highways, or ferries. • Intelligent Defaults: Defaults to driving mode unless specified otherwise, and automatically handles address encoding for better accuracy. 🛠️ Technical Improvements • Enhanced Tool Support: Added the start_navigation tool to the internal toolset. Example commands to try: • "Navigate to the nearest Starbucks avoiding tolls."
More…
• "Start walking directions to the Empire State Building." • "Drive to 1004 N Spruce St, Ogallala, Nebraska." - Apr 18, 2026 2.1.80- More fixes for when using HTTPS with LAN models: llama.cpp ejecting/loading models
- Apr 16, 2026 2.1.79- Fix for correcting prompt and context menu grammar/spell check for when using HTTPS. - Added setting option to show a border animation when new message comes in.
- Apr 16, 2026 2.1.78- Added Tools: copy_file and edit_file.
- Apr 16, 2026 2.1.77- Fixed AI Memory in Tool Chains: Resolved an issue where the AI would "forget" the results of earlier tool calls during complex, multi-step tasks. The AI now retains full context of all files read and actions taken throughout the entire chat.
- Apr 15, 2026 2.1.76- Fix for getting a suggested chat title when using llama.cpp server with https. - Added Plus Code Converter tool (Convert between geographic coordinates (Lat/Long) and Plus Codes). - Optimized Brave Place Search tool so the AI knows it can use named places in addition to coordinates.
- Apr 14, 2026 2.1.75🚀 Notable Changes - Android 17 HTTPS Preparation: In anticipation of Android 17 restricting cleartext HTTP traffic (android:usesCleartextTraffic="true"), we've upgraded LAN networking to fully support secure HTTPS connections. - Self-Signed Certificate Support for llama.cpp: You can now run your local llama server with HTTPS using self-signed certificates (via --ssl-key-file and --ssl-cert-file). The app will securely connect to your local server without certificate validation errors. - Isolated Network Transports: Introduced a dedicated, isolated network client specifically for LAN connections. This ensures that your public internet traffic (OpenRouter, etc.) remains strictly secured by standard SSL validation, while only local LAN traffic is permitted to bypass self-signed certificate checks. 🛠️ How to Use: If you are running a local llama.cpp server, you can now generate a self-signed certificate and key on your host machine and start the server with HTTPS: bash
More…
./llama-server -m model.gguf --ssl-key-file key.pem --ssl-cert-file cert.pem Enter in the app the correct Https LAN IP address and the app will automatically handle the secure HTTPS connection to your local IP. - Apr 13, 2026 2.1.73- New Tool: Get Location — AI can now fetch your precise GPS/Network location, returning Plus Code, coordinates, map links, and accuracy. Requires Location permission. - New Tool: Brave Nearby Places — AI can search for local businesses/POIs using your current location (e.g., "nearest coffee shop"). Returns name, address, phone, website, hours, and ratings. Requires Brave API key.
- Apr 12, 2026 2.1.71- Optimizations to the correct prompt, spell check and suggest chat title functions.
- Apr 9, 2026 2.1.70- For llama.cpp models, you can switch reasoning off/on for models that support: chatTemplateKwargs enable_thinking parameter. No advanced reasoning parameters though. - Added "Is Free Model" switch for adding/editing a model.
- Apr 7, 2026 2.1.69- Changed "Select Model" screen to "Your Models" screen name. - When you add or edit a model from that screen all toggles show for to enter correct properties of the model(vision, reasoning, LAN, image gen) - When you edit a model that is also the currently active chat model, any edits to model coapabilities will reflect when you go back to the chat screen.
- Apr 5, 2026 2.1.68- Updated Commonmark dep. - Added optional Brave Search Tool. Enter your Brave Search API key in the settings and the option for the tool will appear when long-pressing tools button. Allows for local or cloud models to have Brave search results injected into context.
- Apr 1, 2026 2.1.67- Added support for google/lyria-3 models on OpenRouter. It works just for streaming so that gets locked in. It can take a minute or so to download the song, but eventually a snackbar appears when done.
- Mar 28, 2026 2.1.66- Updated ktor dependency. - For llama.cpp models : Tap the eject icon to the right of the model name to unload, tap the power on button to load the model.
- Mar 25, 2026 2.1.65- Added Hermes Agent for a model you can chat with on your lan. To use Hermes Agent with it: - Use Hermes Agent v0.4.0 (v2026.3.23) or newer on your computer - Enable the API server Add to `~/.hermes/.env`: Environment Variables | Variable | Default | Description |
More…
|----------|---------|-------------| | `API_SERVER_ENABLED` | `false` | Enable the API server | | `API_SERVER_PORT` | `8642` | HTTP server port | | `API_SERVER_HOST` | `127.0.0.1` | Bind address (localhost only by default) | | `API_SERVER_KEY` | _(none)_ | Bearer token for auth | You would need to set API_SERVER_HOST to 0.0.0.0 and set a API_SERVER_KEY to be able to use it with oxproxion on your LAN. Then enter the ip endpoint with port and API_SERVER_KEY you chose for the LAN setting in oxproxion. The model will populate as "hermes-agent" in the LAN models list. Add it to your list and start chatting with the Hermes Agent on your computer. Use at own risk. See docs: https://github.com/NousResearch/hermes-agent/blob/main/website/docs/user-guide/features/api-server.md Will not act like Telegram, etc in that it can't message you whenever. It is just for back and forth chat. "List all the files on my MacBook's Desktop", "Check out example.com and tell me what it says" etc.. Also it doesn't accept image input, nor tools from oxproxion(Hermes Agent uses its own tools on the computer you have it on.) Check out Hermes Agent docs for all it can do: https://hermes-agent.nousresearch.com/ - Mar 19, 2026 2.1.59- Added a little color to new chat screen icon. - Even if user has the volume keys to scroll setting on, when TTS is playing, the volume buttons will work for setting volume.
- Mar 16, 2026 2.1.57- Fixed the issue of the Settings buttons having been kicked off-screen because I added the new Tools button.
- Mar 13, 2026 2.1.50- Updated coreKtx and activityKtx dependency. - Fixed a bug where edits to a model weren't saving its capabilities correctly. - In the interest of keeping the users Download folder tidy, files created by the app are now still in Download folder, but in their own folder entitled "oxproxion" there.
- Mar 8, 2026 2.1.49- Enhancements so that OpenRouter image generation models named with bytedance, black-forest-labs, and sourceful can work with the app. - If the user has Fossify File Manager installed, any function that creates a file in the Download folder will now show a Snackbar (instead of a toast) with a button that opens the folder directly in Fossify File Manager.
- Mar 8, 2026 2.1.48- Fixed new Nano Banana not having the aspect ratio options available for generated images. - Enhanced the "Fix with oxproxion" function in overflow menu on selected text so you can approve/disapprove the fixed text before it is replaced.
- Mar 7, 2026 2.1.47- Added long-press function to both "Back to App Button" and "Copy/Back to App Button". This will clear the chat before returning to the previous app. - Added a Save to File button for the AI response bubble. Press it and a dialog will appear asking for the name and extension of file to be saved. Good for if you ask the AI to return ONLY the code and no explanation for HTML, JavaScript, Kotlin, etc. Works even if AI responds with the code in a code block.
- Mar 6, 2026 2.1.46- Added an indicator near the model name on the main chat screen to indicate if the model is LAN or cloud based. - Updated README.md to clarify oxproxion can be used with local models served on a LAN.
- Mar 4, 2026 2.1.45- Updated the `ktorClientAndroid` dependency. - Added the “Fix with oxproxion” function. This option appears in the overflow menu of other apps (that allow it) when input text is selected. It sends the selected text to your currently chosen model for grammar, spelling, punctuation, and capitalization corrections, then replaces the original text with the corrected version. - Edited and expanded the filter bars on the model‑selection list screen to provide finer granularity when narrowing down models.
- Mar 1, 2026 2.1.44- Added a setting preference: Auto Back after receiving a Share. If on, when you share text to a share target of the app, it will automatically take you back to the app you shared it from. New Notification Button Option: - New setting to choose between "Open" or "Dismiss" button in AI response notifications - "Copy" instantly copies the response text (markdown-stripped) to clipboard and dismisses the notification - Toggle found in Settings → choose your preferred quick action when receiving AI responses - If you use the Copy option, you can still dismiss the app from the notification by swiping it away. - Copy button is especially helpful if you are using oxproxion for spelling/grammar correction from a share to it from another app.
- Mar 1, 2026 2.1.43New Notification Button Option: - New setting to choose between "Open" or "Copy" button in AI response notifications - "Copy" instantly copies the spoken text (markdown-stripped) to clipboard and dismisses the notification - Toggle found in Settings → choose your preferred quick action when receiving AI responses - If you use the Copy option, you can still open the app from the notification by tapping anywhere on it other than one of the buttons. - Copy button is especially helpful if you are using oxproxion for spelling/grammar correction from a share to it from another app.
- Feb 28, 2026 2.1.42- Added LaTeX support for math expressions in the HTML AI response view.
- Feb 27, 2026 2.1.41- You can now long-press the HTML button on a individual AI response to save it as HTML in your downloads folder. - Added Atkinson Hyperlegible Next and Atkinson Hyperlegible Mono fonts. - You can now increase/decrease the font size of the AI and User message bubbles. Tap the new Font resizing button in the menu, which appears increase/decrease font size buttons on the main screen. Tap the check button when done. Your new size is saved automatically.
- Feb 27, 2026 2.1.40- Added a setting to show/hide citations
- Feb 23, 2026 2.1.38Fixes for last additions which were: - Added "Extend Top Bar" setting option that allows the reasoning, web search, streaming, conversation, presets, and setting buttons to be on chat screen all the time below the top bar for convenient access. Also includes a Home button to bring to Android Home screen.
- Feb 23, 2026 2.1.36- Added "Extend Top Bar" setting option that allows the reasoning, web search, streaming, conversation, presets, and setting buttons to be on chat screen all the time below the top bar for convenient access. Also includes a Home button to bring to Android Home screen.
- Feb 18, 2026 2.1.35- Moved the new Back and Copy/Back buttons (from last release) to the upper area of the screen.
- Feb 17, 2026 2.1.34- Added an option in settings to include OpenRouter Middle-Out Message Transforms for context management. (https://openrouter.ai/docs/guides/features/message-transforms) - Added two buttons that temporarily appear when the user comes to the app from a notification or a share to the Presets share-target: "Back to App" and "Copy/Back to App." These buttons appear above the menu button and allow for a faster return to the previous app. The "Copy" function copies the last response and then returns the user to the previous app. This feature is especially helpful when using oxproxion for spelling and grammar fixes from other apps.
- Feb 15, 2026 2.1.33- Updated a bunch of dependencies - A little more robustness to contract input box when doing a back gesture as it sometimes was getting stuck in expanded state.
- - Optimized the functioning of the recently added "Expand Input on Focus" setting/option.
- Jan 5, 2026 2.1.32- Optimized the functioning of the recently added "Expand Input on Focus" setting/option.
- - Optimized the functioning of the recently added "Expand Input on Focus" setting/option.
- - Change to timeout now takes effect on next AI request rather than next restart. - Added a setting/option to expand prompt box on focus.
- - Added scrollbars to prompt box to appear if prompt is larger than what the box can show. - A user of the app can now save a chat to an EPUB file. Note: Chats with large tables will not render well in EPUB readers.
- - Added a setting for the user of the app to change the timeout length for the AI request.
- - A user of the app now has the ability to edit an AI message. - Updated a dependency.
- - Updated the llama.cpp model list function for its new format
- - Added ability to have web search on/off for a preset. - Added a setting to turn off auto-off web search after each use. - Updated some dependencies.
- - After tapping a collapse button, scroll buttons/progress indicators will reflect proper state.
- - Added a copy button target for Prompt Library items(tapping on item also copies). After copy, user is now taken directly back to main chat screen. - Fix: Timeout reduced to 10 seconds for screen that lists LAN models. Backing out stops the request and won't crash the app. - Corrected some small button spacing issues.
- - If the user prompt is large, the bubble will auto collapse. Tap the collapse button to see it fully. - Added a collapse/uncollapse button to the AI bubble. This is manually done. Tap to collapse, tap again to uncollapse. - Memory and compute optimizations for scrolling in a chat.
- - Settings Changes: Moved less-often used settings into a new Settings screen. - Added a scroll progress bar optional setting. It gives you an indication where you are in the chat.
- - Typo/fixes in readme/help. - Optimized HTML for when you save a whole chat. Those fun little code block copy buttons will be present and functional. - Networking optimization.
- - Added a Prompt library page for power users. Long-press Presets button to access. It is a library you can make of prompts you might want easy and quick access to. Tap the add button to make a new one. See Help screen for more info. - For the contexual buttons that appear for the entire chat in the menu, long-pressing the Markdown button there will save the whole chat to a .txt file in Downloads.
- - STT from notification now won't speak markdown. - Other notification optimizations. - Style and typo fix. - Variable Substitutions: Use `{{oxdate}}` (yyyy-MM-dd), `{{oxtime}}` (HH:mm:ss), `{{oxdatetime}}` (ISO), or `{{oxhdt}}` (human-readable) in prompts/system messages. Auto-replaces with current date/time on send. - Updated Help screen.
- - Added ability to increase or decrease font size in the HTML rendering screen of the AI response. - Optimized print rendering.
- - Optimizations and improvements to HTML rendering of AI response. - Added a preference to keep the screen from turning off while using the app.
- - With the new TTS from the notification, if speech finishes and the app is in the foreground, the notification will auto-dismiss.
- - Added an HTML button to AI responses. Tap and the response is opened in an Android Webview. This fixes issues with code blocks wrapping and adds a one-tap copy button for them. Print button is on this screen for paper or pdf printing.
- - Long press PNG button on a single AI bubble -> make a webp screenshot of the message instead. - Long press PDF button on a single AI bubble -> make jpg screenshot of the message. - Updated help screen. - Added Third-Party Licenses screen (tap in Help).
- - Squashed a bug related to the last version's added new features.
- - Optimized streaming quite a bit. If streaming is on, Markdown rendering takes place once the stream is finished, in order to minimize unnecessary processor work.
- - When the response comes in a short-duration color animation on the modal name plays to inform the user the AI has finished its response. - Styling change. - Scroll behavior change: If you scroll or touch the chat area after you send the prompt, but before the AI is done, it will not auto scroll to the new response, thus respecting that you want to look at another area of the chat. If you do not touch/scroll during this period, it will auto scroll to the new response once it comes in fully. - Scroll behavior change: If you have streaming enabled, the screen will auto-scroll until it is about to go off screen, then it will stop. If you touch or scroll manually during the incoming stream, auto-scrolling stops.
- - Networking enhancements.
- - You can long-press a scroll arrow and it will take you one screen length in the respective direction. - Mellowed out the "working..." animation.
- - Made action buttons on the messages a more visible color.
- - Made some Manifest changes and foreground service start code so as to hopefully make app work with Android 13.
- - Fixed a scroll regression.
- - Added a save to HTML file button for the whole chat. - Style change on Print look.
- - Changed cloud icon on button that opens OpenRouter models screen. - Changed Connectivity Service priority to high to guard against system shutting it down. - Some theme elements. - Updated a couple dependencies.
- - You can choose the aspect ratio of the generated image now also with google/gemini-3-pro-image-preview model - Further optimization of guard against tables not rendering properly if nested in a list.
- - Added a button to save whole chat as a .md markdown file in user's downloads folder. - Added a Print button to print the whole chat in letter-size or whatever page size user wants. Brings up Android print screen. Can print to PDf or printer. - Added a guard to prevent tables not rendering when included in a markdown list. - Further markdown optimizations.
- - Added a long-press to the Share button on AI response: share raw markdown. - Added Web Search button. * **Web Search Button**: Enables web search(model gets web search information) for **one response only** (auto-off, OpenRouter models only). Long-press to choose engine: [**Default** (native if available, else Exa), **Native only**, **Exa only**]. **Note**: Exa is OpenRouter's search provider. Native is server-side of provider(OpenAI, xAI, Anthropic, etc.) Check OpenRouter/provider's docs for pricing — can be expensive!
- - Added Noto Serif font. - Optimized Markdown rendering and text selection in the AI response bubble. - Added a Save to Markdown file button for the AI response. It saved the response as an .md file to the users download folder. - Updated dependencies.
- - Corrected links clickable.
- - Widened AI bubble full.
- - Citations now show title and url. - More reliable scroll on new prompt send. - Much better text selection. User can now see highlight in code blocks. Link menu pops up on long-press.
- - Scroll issue fix - Updated dependency
- - Fixed some styling issues. - Added guard to handle certain markdown that would cause a crash.
- - If you do the Preset Button above Send Button option, the one in menu will be gone. - Fix for occasional non-scroll when sending prompt. - Changed max token button icon.
- - A couple additions for enthusiasts: 1. Long press the Extend Dock button to show/hide a Presets button above the Send button in the main chat screen for quick/easy access to Presets. 2. Long press the TTS button of an AI response bubble to save an audio wav file of it to your downloads folder. (Generation usually done in seconds.)
- - Added Feature: If you make a preset titled "Digital Assistant"(case-insensitive), and have oxproxion as your system digital assistant in your Android settings(Settings->Apps->Default apps), this preset will be applied for when you use it as the system digital assistant.
- - Bug fix for when editing/saving a model and a fix for resending a prompt if user had changed a model.
- - Added **Attach Document(s) Button**: Click icon to attach **text-based only** files to send with your prompt. - Changed magic prompt corrector function to use model "qwen/qwen3-30b-a3b-instruct-2507"
- - Bug fixes
- - Added a "LAN" filter for your AI models list screen. - Also a note about the PDF uploads new function in the app: If the page has a white background the conversion may make that transparent, but this shouldn't be an issue with the vision model; It just may look different in the image preview you see.
- - Notifications changed. The "Connectivity Service Channel runs and you can hide that; go into your app settings in the system and go to notifications and you can toggle that notification off. ("Connectivity Service Channel") And the only notifications you will receive(if in the app you have the notification bell active) is if the app is back-grounded and there's either an error or you receive the response from the model. Notifications won't show if app is in the foreground. Notifications are auto-dismissed when returning to the app. Closing the app by swiping it away in Recents will shut down the service properly.
- - Removed stop button from notification as it was confusing. - Made notifications default to on for new installs as it is almost necessary for proper background functioning.
- - If you are in an existing saved chat that you have changed, when you save it again the existing title will prepopulate the title prompt. - Added an option for clearing the current chat on the presets list screen. - Added an option for clearing the current chat on the presets share target screen. - Updated a dependency.
- - Fixed LAN model issue: When using magic prompt correction (long-press menu button) or AI-generated chat title saving, the loaded local model is now used.
- - Added llama.cpp as a locally-served LAN AI chat option.
- - Small fix for preset drag-sort.
- - Remove an unnecessary dependency. - Styling fixes.
- - Fix for returning citations with Perplexity models when streaming is on.
- - Changed method of how the app changes fonts as the previous was occasionally restarting the activity. Some utility screens use the phones main system font as a result. - Added Merriweather Sans font
- - Added Merriweather font. - If you are using streaming, and do not scroll whilst a stream comes in, once it is done it will auto-scroll the response as high as possible for best visibility. If you do scroll whilst it comes in, it will not auto-scroll when finished.
- - Added 10 or so font options the user can choose from - Changed "thinking..." to "working..."
- - Tweak for "Thinking..." message being removed too soon before first streaming chunk arrives.
- - The chat won't scroll by itself when streaming is enabled.
- - Added ability to copy either Assistant message or whole chat as Markdown RAW. Do it by long-pressing the robot icon and Copy Chat buttons respectively.
- - Added dynamic hint text in chat input box to display selected system message title when not using the default, improving user awareness of active system prompts.
- - When you close out the app from recents(swipe up) the notification service of the app will close too. - Fixed typo
- - Added sorting (Alphabetical/Newest) and filtering (Type: All/Vision/Image Gen, Cost: All/Free/Paid) to the personal model picker screen for enhanced model browsing and selection.
- - Added a cost filter (All, Free, Paid) to the OpenRouter models screen, allowing users to easily browse and select models based on pricing.
- - Added a filter bar to OpenRouter models screen
- - Added option to use camera for image input. Tap image button for dialog to appear asking for image picker or camera. Long-Press image button to go straight to camera. - If using Nano-Banana, tap palette button to set aspect ratio of returned generated image.
- - Added expandable/collapsible items and drag-and-drop reordering to the System Messages list (except the Default System Message).
- - Added an "Audio Conversations" option. When enabled, Speech-to-Text automatically sends recognized prompts to the model, and responses are automatically read aloud via Text-to-Speech. - App can now be set as the phone's Assistant app. When called as that, Speech-to-Text automatically starts.
- - Moved messages buttons to below message to align with standard AI Chat app structure. - Added Edit and Resend buttons for user messages/prompts. These remove all subsequent messages so be mindful of that. - Resolved issue of screen flash when pressing the TTS button. - Notification if a saved chat is loaded.
- - Fixed PDF generation to use scaled images - Updated dependencies
- - Enhanced error handling for streaming - Bug fix - Increased timeout to 5 minutes