What's Changed
- feat: Organize World Books into groups and select a whole group in chat.
- feat: Add a full-width chat layout on desktop.
- feat: Give thinking and tool cards their own style.
- feat: Choose whether to expand all thinking and tool steps when exporting messages as images.
- feat: Keep individual images visible in chat while leaving them out of future requests to the model.
- feat: Set your own desktop shortcut for Find in Conversation.
- feat: Select or clear all assistant replies at once in the chat summary preview.
- feat: Choose whether the chat summary uses your instruction cards.
- feat: Attach text files directly in regular chats.
- feat: Let workspace tools open images from public URLs.
- feat: Let the assistant install skills when workspace file writing is enabled.
- feat: Drag to reorder models on desktop.
- feat: Choose which response speeds each OpenAI-compatible model offers. GPT-6.2 Astra and Sol now support Fast and UltraFast.
- feat: Read model limits and pricing from supported gateways.
- feat: Refresh provider balances on a timer or after a reply, with settings for each provider.
- feat: Add 4get and Keenable as web search options, and let you set custom search service icons.
- feat: Translate replies and text into Bangla (Bengali).
- feat: Add a shortcut to the Kelivo import toolbox for chats from ChatGPT, Claude, Chatbox, Cherry Studio, and more.
- perf: Start faster on desktop and use less memory for images and background chats.
- perf: Keep chats responsive while replies stream and while large conversations are deleted.
- perf: Export large backups faster and show clearer progress during export and restore.
- perf: Load search results for interactive chat cards faster after restarting the app.
- fix: Keep chat titles in sync and show the correct reply status when switching chats.
- fix: Keep text selected while scrolling and let you select part of an inline code snippet.
- fix: Restore Minimap search previews and show how many matches each message contains.
- fix: Fit Markdown table columns to their contents and restore readable inline code colors.
- fix: Open chats that used the built-in browser at the latest message and keep tool image previews in their original proportions.
- fix: Turn off chat image recall by default.
- fix: Retry with a smaller request when a provider says a chat summary request is too large.
- fix: Show filenames and their folders in workspace file lists.
- fix: Show stored files while attachment checks run and protect files that are still linked in chats.
- fix: Keep WebDAV and S3 backup settings and file exclusions across restarts.
- fix: Recover S3 backups when a temporary upload goes missing, and fix backup restores on Linux.
- fix: Restore Mermaid and SVG previews on Windows.
- fix: Restore Android file opening, reply status animations, and keyboard delete gestures when editing long messages.
- fix: Keep dictated text when system dictation returns a blank final result.
- fix: Save color profiles correctly when compressing images for chat.
- fix: Keep Codex model lists up to date, fix Claude Code request errors, and show the right reasoning options for Grok Composer.
- fix: Repair Gemini MCP tool calls and show models that were missing from Gemini-compatible servers.
- fix: Show custom model names in usage stats and chat summary settings, and group Devin and Pangu models correctly.
- fix: Ask before resetting a provider's balance settings.
- fix: Keep image editor buttons readable on narrow screens and make the desktop message box easier to focus.
Full Changelog: v1.3.6...v1.3.7