What's Changed

• Add Phonon architecture and performance article by @Blaizzy in #740
• Fix model selector hover blinking by @Yavaren in #742
• Use lazy stacks for sidebar chat lists by @Yavaren in #743
• Bump version to 0.4.3 by @lucasnewman in #744

Full Changelog: v0.4.2...v0.4.3

What's Changed

• Export audio summaries and transcripts as PDF, DOCX, and TXT by @Blaizzy in #713
• Add the ability to read responses aloud by @lucasnewman in #711
• Make sure older RC updates see newer stable builds by @lucasnewman in #712
• Add decision model discovery by @Yavaren in #721
• Add decision inference and model loading by @Yavaren in #722
• Test decision requests and responses by @Yavaren in #723
• Add decision request composer state by @Yavaren in #724
• Add the decision model testing page by @Yavaren in #725
• Open decision model tests from Models by @Yavaren in #726
• Add configurable chat response timeout by @Yavaren in #729
• Fix OpenCode 2 launch compatibility by @Yavaren in #730
• Support browser uploads of original chat attachments by @Blaizzy in #715
• Recover from unknown tool calls by @Yavaren in #734
• Return current document state on revision conflicts by @Yavaren in #735
• Show confirmed work-pane saves in chat by @Yavaren in #736
• Bump version to 0.4.2 by @lucasnewman in #738
• Add Flux 3 and MOSS-Image image generation models by @lucasnewman in #739
• Fix PDF compaction replay and work-pane controls by @Blaizzy in #714

Full Changelog: v0.4.1...v0.4.2

What's Changed

• Add blog section and EmbeddingGemma 2 article by @Blaizzy in #672
• Improve wakeword efficiency by @lucasnewman in #671
• Fix signed history commits blocking website deployment by @Blaizzy in #673
• Fix branch input for signed website history commits by @Blaizzy in #674
• Website/restore glm 5 3 flash blog by @Lazarus-931 in #675
• reduce system prompt overhead by @Lazarus-931 in #654
• Fix/chat response metrics by @Lazarus-931 in #682
• Assemble chat prompts in cache-stable layers by @Lazarus-931 in #679
• Route read_file through the shared document extraction router by @Lazarus-931 in #647
• Add Retry to the Hugging Face Hub unavailable notice by @Lazarus-931 in #690
• Show a clear notice when a model download lacks disk space by @Lazarus-931 in #689
• Fix title bar double-click to zoom window by @Yavaren in #692
• Fix default Python MCP servers failing to start by @Yavaren in #695
• Diagnose/short output repro by @Lazarus-931 in #637
• Identify Nativ in Parallel Search requests by @georgeatparallel in #698
• Fix browser agent recovery from unavailable websites by @Blaizzy in #696
• Swift xet downloads by @Yavaren in #697
• Fix mcp lifecycle by @Yavaren in #703
• Keep chat compaction stable across work-pane updates by @Blaizzy in #669
• Keep agent commands responsive and stop processes on quit by @Blaizzy in #704
• Match compaction status to the Working glow by @Blaizzy in #668
• Add embedding model comparison and align blog benchmark visuals by @Blaizzy in #676
• Microphone exclusions by @Yavaren in #702
• Choose dictation paste destination when recording ends by @Yavaren in #705
• Fix compaction using actual context usage by @Blaizzy in #684
• Revert mlx-vlm build pin for the release by @Blaizzy in #708
• Show tool arguments as they stream into chat by @Blaizzy in #707
• Bump version to 0.4.1 by @lucasnewman in #709
• Fix signing order for notarization archive. by @lucasnewman in #710

Full Changelog: v0.4.0...v0.4.1

What's Changed

• Fix duplicate windows when reopening from Dock by @Yavaren in #632
• Fix bundled filesystem MCP access in project chats by @Yavaren in #618
• Use the main branch of mlx-vlm for prereleases by @lucasnewman in #640
• Preserve symlinked integration configs by @Lazarus-931 in #642
• Add shared chat work pane by @Blaizzy in #635
• Add per-chat Git worktrees and recovery by @Blaizzy in #645
• Bump version to 0.3.12rc1 by @lucasnewman in #644
• Fix chat transcript buttons ignoring mouse clicks by @Blaizzy in #648
• Keep Agent terminal output in the background by @Blaizzy in #651
• Remove folders from the sidebar by @lucasnewman in #652
• fix screenshot drag attachments by @Lazarus-931 in #643
• Treat NVFP4 as a quantization suffix when pairing drafters by @Lazarus-931 in #657
• Use AVCaptureSession for native microphone switching by @Yavaren in #638
• Save local hardware diagnostics automatically with a 1 GB history limit by @Blaizzy in #587
• Add drag-and-drop reordering for work-pane tabs by @Blaizzy in #663
• Sync new worktrees and name branches from the first prompt by @Blaizzy in #656
• Update wakeword model and operating point tuning by @lucasnewman in #665
• Add automatic conversation compaction to chat by @Blaizzy in #655
• Bump version to 0.4.0 by @lucasnewman in #666
• Add a splash screen for new features by @lucasnewman in #667

Full Changelog: v0.3.11...v0.4.0

What's Changed

• Fix duplicate windows when reopening from Dock by @Yavaren in #632
• Fix bundled filesystem MCP access in project chats by @Yavaren in #618
• Use the main branch of mlx-vlm for prereleases by @lucasnewman in #640
• Preserve symlinked integration configs by @Lazarus-931 in #642
• Add shared chat work pane by @Blaizzy in #635
• Add per-chat Git worktrees and recovery by @Blaizzy in #645
• Bump version to 0.3.12rc1 by @lucasnewman in #644
• Fix chat transcript buttons ignoring mouse clicks by @Blaizzy in #648
• Keep Agent terminal output in the background by @Blaizzy in #651

Full Changelog: v0.3.11...v0.3.12rc1

What's Changed

• Support dragging and dropping images onto the composer for image editing by @lucasnewman in #615
• Fix Discover capability filters hiding matching models by @Yavaren in #617
• issue 601 -> capability unification by @Lazarus-931 in #622
• Use model-specific sampling defaults for image generation by @lucasnewman in #625
• Show incomplete downloads so their cache can be finished or deleted by @Lazarus-931 in #627
• Preserve artifact references for reversible deletion by @Yavaren in #610
• Add safe artifact moves to the macOS Bin by @Yavaren in #608
• Connect artifact deletion and automatic Finder recovery by @Yavaren in #609
• Bump version to 0.3.11 by @lucasnewman in #628

Full Changelog: v0.3.10...v0.3.11

What's Changed

• Add Hey Nativ wake-word dictation by @Blaizzy in #567
• added CPU optimization by @Yavaren in #602
• Fix composer resize loop with always-visible scroll bars by @lizen in #614
• Optimize cross-window image session updates by @Yavaren in #605
• Add configurable summary and merge prompts with language selection by @Blaizzy in #604
• Trust the macOS keychain for Hugging Face downloads by @mattzcarey in #619
• Support image editing model capabilities by @Lazarus-931 in #620
• Bump version to 0.3.10 by @lucasnewman in #621

New Contributors

• @mattzcarey made their first contribution in #619

Full Changelog: v0.3.9...v0.3.10

What's Changed

• Avoid update errors during publish by @lucasnewman in #550
• Search through chats by @Yavaren in #539
• Show chat prefill progress in a Reading prompt status pill by @Blaizzy in #552
• Add PR contributor reminders for commit signatures and CI failures by @Blaizzy in #555
• keyboard shortcuts for tool callings in chat by @Lazarus-931 in #557
• Fix shortcut setup and prevent accidental recording by @Yavaren in #556
• Add Hugging Face, StepFun, and InternLM provider logos by @Lazarus-931 in #576
• Size models from the literal download byte count by @Lazarus-931 in #573
• Fix artifacts refreshing after image generation by @Lazarus-931 in #574
• Separate removed models in the dashboard model filter by @lizen in #559
• buffering tokens to streaming by @Lazarus-931 in #582
• Fix sleep recording by @Yavaren in #572
• Personalized memories by @Yavaren in #551
• removing redundent py files by @Lazarus-931 in #584
• removing unsused componenets by @Lazarus-931 in #585
• 1st of #575 by @Lazarus-931 in #577
• Add structured request failure diagnostics and version context by @Blaizzy in #588
• Bump version to 0.3.9rc1 by @lucasnewman in #590
• 2nd of #575 by @Lazarus-931 in #578
• Remove the cached model discovery toggle from the API by @Yavaren in #589
• remove unused code by @Yavaren in #595
• Connect to loopback when the server host is a wildcard address by @Lazarus-931 in #594
• 3rd of #575 by @Lazarus-931 in #580
• Chat streaming cadence and fade by @Yavaren in #598
• 4th of #575 by @Lazarus-931 in #581
• Detect Qwen-Image checkpoints as image generation models by @Lazarus-931 in #600
• Bump version to 0.3.9 by @lucasnewman in #599

Full Changelog: v0.3.8...v0.3.9

What's Changed

• Avoid update errors during publish by @lucasnewman in #550
• Search through chats by @Yavaren in #539
• Show chat prefill progress in a Reading prompt status pill by @Blaizzy in #552
• Add PR contributor reminders for commit signatures and CI failures by @Blaizzy in #555
• keyboard shortcuts for tool callings in chat by @Lazarus-931 in #557
• Fix shortcut setup and prevent accidental recording by @Yavaren in #556
• Add Hugging Face, StepFun, and InternLM provider logos by @Lazarus-931 in #576
• Size models from the literal download byte count by @Lazarus-931 in #573
• Fix artifacts refreshing after image generation by @Lazarus-931 in #574
• Separate removed models in the dashboard model filter by @lizen in #559
• buffering tokens to streaming by @Lazarus-931 in #582
• Fix sleep recording by @Yavaren in #572
• Personalized memories by @Yavaren in #551
• removing redundent py files by @Lazarus-931 in #584
• removing unsused componenets by @Lazarus-931 in #585
• 1st of #575 by @Lazarus-931 in #577
• Add structured request failure diagnostics and version context by @Blaizzy in #588
• Bump version to 0.3.9rc1 by @lucasnewman in #590

Full Changelog: v0.3.8...v0.3.9rc1

What's Changed

• Refine sidebar navigation, pinned chats, and bulk selection by @Yavaren in #512
• Safer voice activation key combo monitoring by @lucasnewman in #520
• Handle external model cache lifecycle by @Yavaren in #443
• Enable multiple application windows by @Yavaren in #436
• Add external model storage controls by @Yavaren in #444
• Fix model picker tooltip clipping by @Lazarus-931 in #500
• Rename existing typography system to LegacyTypography by @Yavaren in #522
• Stop tracking the generated Xcode scheme by @Lazarus-931 in #527
• Add optional cached model discovery to the API by @Yavaren in #531
• Recall the last prompt with the Up arrow by @Lazarus-931 in #524
• Add DeepSeek Harness (dsh) as a Nativ integration by @Lazarus-931 in #532
• added download progress container in installed models by @Yavaren in #481
• Add new Markdown layout engine and use paged scrolling for long chat transcripts by @lucasnewman in #529
• Improve audio engine error handling by @lucasnewman in #521
• Resolve exact model download sizes by @Yavaren in #533
• Add quote replies to chat messages by @Yavaren in #525
• Add support for using beta releases by @lucasnewman in #534
• add HF model support indicators in discovery page by @Yavaren in #478
• Fix dashboard token usage chart layout by @Yavaren in #327
• Move model reset below panel toggle by @Lazarus-931 in #501
• Bump version to 0.3.8rc1 by @lucasnewman in #535
• Add Return shortcut and hint to tool confirmations by @Lazarus-931 in #509
• Fix typo in settings by @lucasnewman in #537
• Move notification settings into Permissions section by @Yavaren in #536
• Add LWC MCP catalog entry by @JanYork in #540
• Restore GGUF filtering in model discovery by @Blaizzy in #543
• Bottom scroll button by @Yavaren in #542
• Add pasted text attachments to chat by @Yavaren in #541
• Add configurable spoken Return command for dictation by @Blaizzy in #546
• Constrain sidebar resizing and adapt Discover model filters by @Yavaren in #545
• Fix composer undo crash and attachment memory growth by @Yavaren in #548
• Bump version to 0.3.8 by @lucasnewman in #549

New Contributors

• @JanYork made their first contribution in #540

Full Changelog: v0.3.7...v0.3.8