Changelog
This log tracks FlowDown version history and key changes.
4.1.2
- You can now toggle Apple Intelligence in Inference settings with availability hints; default chat and auxiliary models update automatically.
- Cloud models let you choose the OpenAI request format (
/v1/chat/completionsor/v1/responses), with automatic endpoint detection and format-aware routing. - Temperature presets now include a
Disabled (-1)option so you can defer randomness to the provider. - The cloud model editor adds quick actions (verify/export/duplicate) and safer endpoint, header, and body editors to simplify model maintenance.
3.10.11
Memory System
- Smart Recall: Automatically pulls in relevant memories for more personalized replies.
- Memory scope controls: Choose the sharing window (past day/week/month or latest 15/30 items).
Shortcuts Integration
- Quick actions: Launch FlowDown, send messages, and receive responses.
- Content classification: Classify text and images.
- Smart summaries: Turn long text into paragraphs or bullet lists.
- Conversation tools: Search chats, switch models, and create conversation links.
- Image analysis: Mixed image/text analysis (iOS 18+).
Interface Polish
- Swipe gestures: Swipe list items to favorite, delete, or copy conversations.
- Visual refinements: Smoother animations and clearer hierarchy.
3.10.10
- Resolved recent stability regressions and improved adoption of upstream changes for the complimentary model pool.
- Polished CloudKit sync heuristics to reduce redundant fetches when switching devices quickly.
- Tuned release packaging ahead of macOS/iPadOS notarization changes.
3.10.8
Starting with 3.10, minor versions now use semantic versioning instead of build numbers for clarity.
- Added adaptive parsing for new Pollinations AI endpoints; new models no longer require app updates.
- Introduced custom request body editors so each model profile can use advanced inference parameters.
- Replaced the custom context menu with the native platform menu for a consistent experience across iOS and macOS.
- Redesigned cloud sync scheduling to reduce conflicts while working across multiple devices.
3.8
- Cloud Sync: Full CloudKit rollout with unified indicators, status glyphs in the sidebar, and Settings → Data Control hooks for scoped sync.
- Data Management: Export/import entire workspaces as signed
.ziparchives and clear iCloud data without affecting local caches. - Model Library: Assign friendly nicknames to cloud models for easier switching.
- Performance & Logging: Faster deletion, smarter refresh pipelines, and a new Support logging console for troubleshooting.
3.6.419
- Early CloudKit build for advanced users. Upgrading migrates the database schema to V2 and cannot be rolled back.
- Paired with a cautionary GitHub release for manual installers.
3.6
- TestFlight rollout for iCloud sync, marked unstable with a data-loss warning; please back up before installing.
- TestFlight join link:
https://testflight.apple.com/join/StpMeybv.
3.5
- Added a long-press tool menu on the editor bar so you can adjust tool availability without opening settings.
- Celebrated National Day with a limited free offer (ended 2025-10-06); earlier customers were invited to request redeem codes for other apps.
3.4
- Debuted a new immersive conversation interface.
- Enhanced Apple Foundation Model support, especially alongside Apple Intelligence on supported devices.
- macOS desktop release briefly delayed because of third-party compilation issues; the fix followed shortly after.
3.2
- Added per-model temperature configuration for finer control over creativity versus precision.
- Introduced initial Apple Foundation Model support and refreshed the MLX dependency for better performance.
Legacy Releases (1.x–2.x)
Expand archived release history
iOS 2.9 / macOS 2.9.1 (370)
Fixes & Improvements
-
Editor and Rendering
- Fixed multiline equation rendering.
- Fixed pointer device text selection on iPad.
- Fixed a bug where streaming tasks were not cancelled properly after an error.
-
MCP Server
- Fixed a critical bug that blocked MCP server connections during edits.
- Fixed an MCP services bug that dropped required fields while connecting.
-
General UI
- Corrected sidebar behavior on iPad.
- Fixed menu placement in the model management page.
2.6 (357)
Fixes and Improvements
-
MCP Server Connection
- Fixed a critical issue that could prevent connection when editing MCP servers.
- Improved MCP connection management.
-
Task Management & Stability
- Fixed an issue where streaming responses were not cancelled properly after server errors.
- Resolved several threading issues to improve stability.
-
User Interface (UI)
- Corrected menu placement in the model management page.
- Fixed sidebar behavior on iPad.
2.2 (330)
New Features
-
Added MCP Server Support
- Add, manage, and edit MCP servers.
- Introduced
.fdmcpfiles with drag-and-drop and file picker support.
-
Editor and UI Enhancements
- The editor now auto-focuses after creating a new conversation.
- MarkdownView supports scrolling while selecting text.
- ToolHintView layout and response speed optimized.
-
Other
- Support for custom brand names.
Fixes and Improvements
- Bug Fixes
- Fixed editor focus on Mac Catalyst.
2.0 (324)
New Features
- Global Search: Search for keywords across all conversations.
Improvements
- Rewritten Markdown Text Engine: Rebuilt the core text engine with:
- Major performance gains.
- Better math rendering compatibility for Math Functions.
- Improved code highlighting without blinking.
Bug Fixes
- Fixed blinking highlights and table views during streaming.
- Fixed text copying while a response was streaming.
1.26 (262)
This update introduces several new features and addresses key issues to enhance your experience.
New Features
- AI-Powered Template Rewriting: Use AI to rewrite chat templates with extra instructions.
- Template Management:
- Rearrange chat templates with drag and drop.
- Export and open chat template files.
- Favorite Conversations: Mark conversations as favorites for quick access.
- Drag and Drop Export: Export
fdmodelfiles by dragging and dropping.
Improvements
- Sidebar Enhancements:
- Collapse the sidebar on macOS.
- Adjust sidebar width on iPad and macOS.
- Keep the sidebar persistently visible on iPad for easier navigation.
Bug Fixes
- UI: Fixed shadows in the chat view.
1.25 (250)
- Added streaming support and prevented rejection from FoundationModels (@at-wr).
- Improved iPad experience with Magic Keyboard (@at-wr).
- Set a default model identifier for new conversations when none is provided (@at-wr).
- Added custom decoding for CloudModel to handle optional properties (@at-wr).
1.24 (242)
- Included
@generableData Struct FoundationModels (@at-wr). - Fixed multiple crashes when creating a new chat from compressor (@at-wr).
1.24 (241)
- Added compress to new chat.
- Added rewrite action.
- Added redo for #16.
- Improved multi-line LaTeX support in prompts.
1.23 (231)
- Added support for custom headers in cloud model.
- Fixed escape key responsiveness.
- Fixed unwanted shake behavior.
- Fixed a bad merge that lost the HTTP-Referer header.
- Updated keyboard logic.
1.22 (223)
- Minor fixes and improvements.
1.22 (222)
- Minor fixes and improvements.
1.21 (210)
- Enhanced PDF import with dual modes (text extraction or image conversion).
- Improved error handling for large files.
- Added keyboard navigation shortcuts for quick chat switching.
- Added menu bar template integration for direct template access.
- Added template menu to Menu Bar (@at-wr).
- Added shortcuts to navigate between conversations (@at-wr).
1.20 (200/201/202)
Chat Template System 📋
- New reusable chat templates for faster workflows.
- AI-generated templates from existing conversations.
- Intuitive template editor with creation and deletion.
- Quick preset selection when starting new chats.
General Improvements
- Fixed missing entitlements for efficient iOS/iPadOS performance.
- Fixed on-device visual language model support.
- Added Gemma3 model support.
- Added HTML preview for generated code.
1.19 (190)
- Added new menu items.
- Added a welcome log for first-time app launch.
- Improved OpenRouter integration.
- Fixed auto-focus issues when creating new chats on Mac.
- Improved background processing.
- Fixed reasoning content not updating in some cases.
- Fixed logical errors related to model tools.
Thanks to the following contributors:
- @at-wr
- @Zach677
1.13 (130)
- Fixed keyboard shortcuts for copying text in code view.
- Fixed incorrect cursor display when selecting text.
- Fixed unresponsive drag gestures on the first and last lines of text.
- Fixed code syntax highlighting not applying.
- Fixed reasoning requests being added to auxiliary tasks.
- Added an option to use developer role when needed.
- The app can now be opened via
flowdown://protocol. - The app can now open model files in place.
1.11
- Fixed incorrect model identifiers when copying models.
- Fixed math content rendering.
- Fixed menu disappearing when right-clicking the conversation list.
- Fixed documents not being indexed after web search.
1.10 (101/102)
- Improved the visual task scheduler with an option to skip auxiliary tasks for vision-native models.
- Fixed editor lag when processing large images.
- Fixed editor inability to restore editable content.
- Improved auxiliary task structure and execution speed.
- Improved data flow for web search when tool calling is enabled.
- Fixed missing reasoning content in image recognition tasks.
- Added support for OpenRouter model inference tokens.
1.9 (90)
- Added preliminary math content support.
- Added LaTeX math formula display.
1.7 (70)
- Fixed Qwen3 local model support.
- Updated MLX model manifest structure for compatibility.
- Temporarily disabled local vision while waiting for upstream fixes.
- FlowDown is now fully open source so you can verify privacy protections yourself.
1.6 (60)
- Improved local model support.
- Fixed multiple known issues.
1.5 (50)
- Fixed issues when pasting content.
- Fixed web search phase indicator.
- Fixed wrong return key title in the input box.
- Fixed a bug where multiple tool calls were received simultaneously.
- Added support for connecting to locally hosted LLMs via OLLAMA and LM Studio.
1.4 (49)
- Fixed a crash when using the location tool with location services disabled.
1.4 (47/48)
- Fixed table border color updates.
- Fixed a crash when quickly deleting messages.
- Allowed sending empty messages when attachments are present.
- Added direct text selection.
1.3 (44)
- Fixed crashes with certain reasoning models.
- Fixed content unexpectedly showing as reasoning.
- Fixed chat list element positioning errors.
- Fixed endpoints that did not support usage statistics.
- Fixed unsaved content after crashes.
- Fixed code blocks that could not open the details page in the Hugging Face download screen.
- Fixed code blocks and tables not using correct information in edge cases.
- Fixed logical errors when interrupting inference.
- Fixed deleting error messages during retries.
- Disabled local model functionality on devices that do not support MLX.
- Adjusted the order and logic of several menus.
- Adjusted logic for deleting conversation content.
- Adjusted logic for editing conversation content.
- Added an option to send messages with Command + Enter.
- Added an option for long-press to create a new line.
- Added an option to save replies as images.
- Removed the temporary chat option.
1.2 (36/37/42)
- Added image export to save chat content as images.
- Added the ability to end conversations at any time.
- Improved web search with a deduplication algorithm for results.
- Fixed inconsistent text size rendering in supplement views.
- Refactored code and resolved multiple known issues.
1.1 (35)
- Fixed an issue where extracted image information was not stored correctly.
- Fixed an issue where the system prompt was not updated correctly.
- Fixed an issue where tool calls could trigger server errors.
1.0 (32/33)
First release to the App Store.
- Changed the app icon.
- Fixed the DeepSeek model not supporting out-of-order system prompts.
- Fixed retries not using the correct model.
- Fixed tool call failures by removing the
minLengthparameter unsupported by some models. - Fixed the model not using the latest time.
- Adjusted prompts used by the tool.
- Added an option for 64k context length.
1.0 (28)
- Cloud models now support tool use.
- Added a right-click/long-press shortcut menu to the model selector.
- Improved tool call status display, and some tools show call details.
- Added an option for 64K context.
- Fixed crashes caused by undo operations.
- Improved handling when the model is unavailable.
- Fixed permission issues and improved error messages.
- Optimized data export logic and logging.
- Improved the pre-dialogue check mechanism for stability.
1.0 (20)
- Fixed menu event responses.
- Fixed code view height mismatches.
- Fixed the model name not updating as expected.
- Improved the long-press interaction in the conversation list.
- Improved inference engine efficiency.
- Improved the URL opening mechanism and added security warnings.
- Reduced network crawler resource usage.
- Replaced widget icons.