CherryHQ/cherry-studio v2.1.9 on GitHub

Downloads (v2.1.9)

Platform	Architecture	Download
Windows	x64	Installer · Portable
Windows	arm64	Installer · Portable
macOS	Apple silicon (arm64)	DMG · ZIP
macOS	Intel (x64)	DMG · ZIP
Linux	x64	AppImage · DEB · RPM
Linux	arm64	AppImage · DEB · RPM

Release Notes

Cherry Studio 2.1.9 - Diagrams, File Pane, and Provider Updates

✨ New Features

[Diagrams] Mermaid diagrams now render with a faster engine in the background, match the app's light/dark theme, and open as previews from .mmd/.mermaid files.
[Agent Files] Open files in tabs with a breadcrumb, resize the tree, and configure autosave, indent, and word wrap in pane settings. Right-click to attach files or folders as context.
[Agents] Archive all conversations of an Agent in one action, with a confirmation showing the count. Delete conversation branches from the Branch Management panel.
[Web Search] Add Desearch and SearchApi as web search providers.
[Providers] Add QwenCloud with chat, image, reasoning controls and native web search. Rebrand DashScope to Qwen AI Platform with updated endpoints and icons. Add Anthropic prompt caching support for Vercel AI Gateway.
[Code Theme] New default “Cherry” code theme blends into the app background in light and dark mode.
[Thinking] Edit a message's reasoning content directly in the assistant message editor.
[Permissions] Enabling Full Access now shows a scope confirmation before taking effect.
[Tabs] The active conversation tab has a clearer selected outline and slightly bolder title.
[Notes] Open a newly saved note directly from the export success toast.

🐛 Bug Fixes

[Tooltips] Fixed tooltips being cut off at the edge of dialogs, such as the help tooltips in the Selection Assistant custom action dialog.
[Chat] Long silent reasoning phases no longer abort after 30 minutes; the idle bound is now 120 minutes. Fixed misleading error when an assistant's model no longer exists after an update.
[Tabs] Show the conversation name and assistant icon on tabs again. New tabs open next to the active tab; the plus button appends at the end.
[Navigation] Unified sidebar and tab selection style with clearer contrast in light and dark themes.
[Notes] Fixed a freeze when typing a diagram code block. New diagram blocks open in edit mode.
[Knowledge] Fixed rebuilds resetting chunking rules, imports timing out while waiting, interrupted conversions resuming after restart, and stuck “preparing” states.
[Browser] Preserve unsaved input and page state when following fragment links in local HTML files.
[File Links] Removed the hover reveal icon from chat file links; use the context menu to reveal a file.
[Web Search] Fixed DashScope models failing with “This model does not support enable_search”. Blacklist rules now exclude domains on Bocha and Querit.
[TTS] Kokoro Chinese text-to-speech now initializes correctly on Windows with the required Mandarin eSpeak data.
[MCP] Fixed built-in MCP server documentation links that opened a removed page.
[AiHubMix] Fix Claude chats failing with HTTP 400 when built-in tools include input examples.

⚡ Performance

[Knowledge] Deleting or reindexing items in large knowledge bases is much faster.
[Conversations] Fixed slow branch switching and message deletion in very long conversations. Loading earlier messages is faster.

What's Changed

fix(message-rules): drop duplicate tool parts sharing one id before replay by @cherry-coding[bot] in #22532
fix(html-artifact): stop the inline preview height from flipping on a fixed embed by @cherry-coding[bot] in #22512
feat(message-renderer): add an opt-in switch to keep intermediate assistant text in the response lane by @lianshuang-photo in #21086
fix(file-preview): enable copying text from DOCX previews by @WhiteZeek in #22548
feat(model-list): filter completely failed models by @Bruce-Yii in #21015
feat(tool-permission): add plain-language risk summary to approval dialog by @404-Page-Found in #20444
fix(knowledge): keep conversation save when a referenced file is missing by @gokay-ai in #20355
chore: version packages by @github-actions[bot] in #22557
fix(i18n): restore missing Italian tool risk translations by @DeJeune in #22558
perf(code-highlight): unify Highlights and virtualize large code previews by @DeJeune in #22469
fix(paintings): show quality and size controls for GPT Image 2.5 models by @cherry-coding[bot] in #22515
chore: version packages by @github-actions[bot] in #22561
fix(agent-tools): render the navigate card in Agent sessions by @cherry-coding[bot] in #22565
docs: remove duplicated install step in development guide by @pei711 in #22573
feat(composer): search skills and knowledge bases by metadata by @SiinXu in #22358
feat(provider-registry): support Kimi K3.1 Pro models by @ousugo in #22575
chore: version packages by @github-actions[bot] in #22590
fix(tool-disclosure): restore scroll chaining at panel boundaries by @tommyzhang100504 in #22591
fix(image-generation): honor a configurable submit timeout by @NatsUIJM in #22448
fix: close small correctness gaps found in a code sweep by @angri450 in #22593
perf(dialog): remove entrance scaling to avoid raster jank by @Wit-2k in #22597
hotfix(file-edit): resolve the conflict when dismissing the edit-conflict dialog by @ousugo in #22617
fix(voice): make maximum-length dictation recordings transcribe reliably by @angri450 in #22611
fix(e2e-regression): honour CHERRY_CDP_PORT when launching in branch mode by @cherry-qa[bot] in #22619
fix(api-gateway): never answer a failed generation with HTTP 2xx by @cherry-coding[bot] in #22342
feat(knowledge-ui): allow knowledge batch imports beyond 20 files by @EurFelux in #22089
feat(knowledge-ingestion): copy imported sources outside the base lock by @EurFelux in #22090
feat(knowledge-ingestion): let file items wait in preparing while copied in by @EurFelux in #22094
fix(knowledge-ingestion): backfill embeddings from the in-base copy, one indexing job per leaf by @EurFelux in #22454
test(e2e): add an opt-in in-job mock provider to the regression workflow by @cherry-qa[bot] in #22452
fix(code-viewer): keep measured wrapped row heights across width changes by @tommyzhang100504 in #22639
fix(import): make a failed conversation import side-effect free by @ousugo in #22642
fix(error-category): stop diagnosing a rejected request body as invalid AI content by @404-Page-Found in #22638
fix(updater): distinguish withheld feed from up-to-date result by @404-Page-Found in #20865
fix(file-storage): parent legacy save dialogs to caller window to avoid Unicode filename crash by @404-Page-Found with @Copilot in #20778
feat(knowledge-data-source): import a folder by drag-and-drop by @cherry-coding[bot] in #22656
fix(notes): bound generated note names and report the real rename outcome by @cherry-coding[bot] in #22654
feat(chat-files): surface a reveal action on chat file links by @cherry-coding[bot] in #22646
fix(renderer-styles): keep test files out of tailwind scanning by @sseymdoc in #22664
fix(hooks): stop post-checkout hang on fresh worktrees by @404-Page-Found in #22627
test(voice): assert scratch POSIX modes only where they are enforced by @angri450 in #22624
fix(chat): validate local image bytes before model dispatch by @404-Page-Found in #22626
hotfix(release): restore backport checkout and CI test cleanup by @DeJeune in #22671
fix(agent-tools): report a platform limitation for to_markdown on Windows ARM64 by @cherry-coding[bot] in #22635
fix(claude-code): pin the sub-agent model on the gateway route by @cherry-coding[bot] in #22649
fix(e2e-regression): drive the macOS 27 native panel and decouple fixtures from locale by @cherry-coding[bot] in #21866
fix(mcp): launch the configured stdio command when lookup is denied by @cherry-coding[bot] in #22598
fix(webdav-backup): preserve upload size rejection guidance by @Dante-dan in #22685
chore(release): sync v2.1.8 metadata by @kangfenmao in #22691
feat(provider-registry): add the DeepSeek V4 Flash off switch by @JuLuogo in #22629
feat(reasoning): allow editing a message's thinking content by @JuLuogo in #22628
chore: version packages by @github-actions[bot] in #22692
fix(web-search): send bare domains to Bocha and Querit site exclusion by @RaphaelFakhri in #22613
fix(file-edit): preserve autosave drafts and unblock file operations by @DeJeune in #22687
fix(proxy-settings): remove detailed proxy mode description by @DeJeune in #22701
feat(agent-permission): confirm scope before enabling Full Access by @404-Page-Found in #20807
fix(dashscope): send enable_search only to models that serve it by @cherry-coding[bot] in #22719
fix(i18n): point four renderer labels at keys the catalogs define by @mmporong in #22716
chore: version packages by @github-actions[bot] in #22723
fix(cherry-autonomy-tools): deny config mutations on headless turns in every runtime by @kovsu in #21351
feat(notes): open saved notes from success toast by @louisfy in #22720
fix(streaming-markdown): release completed fade animations by @DeJeune in #22735
fix(tabs): open new tabs next to the active tab by @DeJeune in #22733
fix(tabs): append tabs opened from the plus button by @DeJeune in #22749
refactor(async-coordination): drop duplicate queue and legacy async forwards by @kovsu in #22762
feat(provider-registry): add GPT-6.1 Terra support by @ousugo in #22711
chore: version packages by @github-actions[bot] in #22777
fix(mcp): accept extra fields in a legacy ping health result by @cherry-coding[bot] in #22783
refactor(browser-automation): remove tool adapters orphaned by #21704 by @kovsu in #22776
fix(renderer): handle stale assistant-bound model after update (#22571) by @404-Page-Found in #22765
feat(agent-sessions): add session rename tool by @ousugo in #22796
fix(web-search): preserve Querit page failures and scope the fetch probe by @cherry-coding[bot] in #22793
chore: version packages by @github-actions[bot] in #22814
feat(web-search): add SearchApi search provider by @sgan0420 in #22599
refactor(in-loop-compaction): cache the folded prefix token estimate by @xiong-wei1 in #22842
fix(knowledge-ingestion): count the file-processing wait from the job's start by @EurFelux in #22636
fix(knowledge-ingestion): cancel interrupted file conversions at restart instead of resuming them by @EurFelux in #22662
refactor(knowledge-ingestion): check file-processing results when the job settles instead of polling by @EurFelux in #22657
perf(code-highlight): unify Highlights and virtualize large code previews by @DeJeune in #22730
chore: version packages by @github-actions[bot] in #22816
fix(agent-tasks): broadcast task read-model changes after every committed write by @ousugo in #22874
feat(knowledge): let agents create knowledge bases and batch-add sources by @angri450 in #22647
perf(message-branch): page conversation history on path ids by @kovsu in #22899
perf(message-service): use parent_id index for the newest-leaf CTE by @kovsu in #22903
perf(knowledge-item): resolve batch selections and subtrees in bounded queries by @kovsu in #22901
refactor(message-tree): load the flow-canvas tree in one topic query by @kovsu in #22898
docs(selection-assistant): propose the agent window Spec by @EurFelux in #22827
fix(docs): escape table cells in the generated documentation index by @c020627 in #22914
fix(i18n): include attribute-bearing tags in the tag-placeholder check by @c020627 in #22913
fix(pi-runtime): send Vercel AI Gateway OpenAI requests to /v1 by @cherry-coding[bot] in #22916
docs(code-theme): propose Cherry default code theme by @DeJeune in #22910
feat(code-theme): add the Cherry default code theme by @DeJeune in #22911
fix(archive): re-read the All list when the page is opened (#22712) by @cherry-coding[bot] in #22943
feat(settings-search): keep the query box open and focused after clearing it by @cherry-coding[bot] in #22944
fix(mcp): repoint built-in MCP setup links to the current docs page by @cherry-coding[bot] in #22951
fix(citation): keep dots in the knowledge-citation preview text by @cherry-coding[bot] in #22955
fix(api-gateway): parse empty Chat Completions tool-call arguments as {} by @jaeminbird in #22957
fix(webview): freeze browser tabs hidden for five minutes by @DeJeune in #22950
feat(browser): make the tool-driven browser panel auto-open configurable by @cherry-coding[bot] in #22963
fix(automation): scope assistant automation REPL guidance to enabled capabilities by @cherry-coding[bot] in #22967
fix(poe): route bots by the chat endpoint they declare by @cherry-coding[bot] in #22969
feat(paintings): labelled prompt preset entry with in-page management by @cherry-coding[bot] in #22989
fix(proxy): keep loopback traffic off the user proxy by @superbigcup325 in #21030
fix(data-api): avoid overlapping timed-out reads by @Dante-dan in #20349
fix(data-api): format timeout test imports by @DeJeune in #23010
fix(renderer-styles): let message prose honor the configured global font by @cherry-coding[bot] in #22953
fix(profile-avatar): retire replaced image files by @focus-zyw in #21454
fix(provider-settings): confirm copies in place instead of with a toast by @EurFelux in #23001
feat(markdown): open knowledge-app URI schemes from message links by @cherry-coding[bot] in #22994
fix(pi-runtime): apply the persisted reasoning effort to Pi sessions by @cherry-coding[bot] in #22997
fix(conversation-surfaces): re-level the dark conversation surfaces by @kennyzheng-builds in #17538
fix(provider): survive legacy order keys in the startup registry sync by @ggbdpq in #21284
test(migration): make the v2 migration suite host-agnostic by @angri450 in #21775
feat(web-search): sync DeepSeek API keys from the model provider by @EurFelux in #23000
fix(import): import API-style Claude exports without silent message loss by @linxule in #18135
feat(usage-analysis): restore full-window total breakdowns by @ning-jl in #20347
fix(provider-registry): infer StepFun step-5-preview reasoning effort by @cherry-coding[bot] in #21869
revert(chat-files): remove inline reveal button by @DeJeune in #23019
chore: version packages by @github-actions[bot] in #23017
feat(skills): add enabled-only filter by @SiinXu in #19019
fix(error-category): classify embedded runtime connection codes as network by @cherry-coding[bot] in #22344
fix(aiCore): stop the Azure chat-mode web search crash (provider.tools is undefined) by @cherry-coding[bot] in #23014
feat(provider): add QwenCloud provider by @GeorgeDong32 in #19810
feat(provider): rebrand dashscope to Qwen AI Platform by @GeorgeDong32 in #19811
fix(error-category): stop reading a Groq service-tier denial as quota exhaustion by @cherry-coding[bot] in #22340
fix(web-search): point the Yuanbao API key link at the live Tencent console by @cherry-coding[bot] in #23042
fix(binary-manager): pin bundled uv 0.13.6 and reject unmet explicit targets by @cherry-coding[bot] in #23012
chore: version packages by @github-actions[bot] in #23040
refactor(ipc): retire legacy channels duplicated by IpcApi by @kovsu in #23039
fix(pi-runtime): upgrade Pi to 1.4.0 for Gemini thinking levels by @DeJeune in #23027
feat(provider-gateway): add Anthropic prompt caching support for Vercel AI Gateway by @404-Page-Found in #22926
fix(horizontal-scroll): avoid duplicate resize listeners by @DeJeune in #18941
Fix Windows Kokoro initialization and add Mandarin eSpeak data by @404-Page-Found with @Copilot in #23020
feat(web-search): add Desearch web search provider by @Chkhikvadze in #21987
chore: version packages by @github-actions[bot] in #23053
fix(packaging): declare macOS privacy usage descriptions and entitlements for local MCP servers by @cherry-coding[bot] in #23051
fix(knowledge): preserve chunking rules when restoring bases by @ousugo in #23063
fix(fileProcessing): rotate API keys over distinct entries by @c020627 in #21434
perf(backup): load S3, WebDAV, archiver, and cheerio on demand by @kovsu in #23062
feat(paintings): add Grok Imagine aspect-ratio controls for image editing by @cherry-coding[bot] in #23050
chore: version packages by @github-actions[bot] in #23069
fix(tabs): restore conversation names as Chat and Work tab titles by @NatsUIJM in #22722
perf(markdown): keep mermaid out of the default preset bundle by @kovsu in #23079
test(tabs): cover conversation titles in Electron e2e by @DeJeune in #23088
feat(chat-branches): delete a conversation branch from Branch Management by @cherry-coding[bot] in #23075
feat(agent-file-pane): add file tabs, breadcrumb, and pane settings by @DeJeune in #23064
fix(browser): preserve local HTML state during fragment navigation by @ousugo in #23058
feat(tabs): make the active conversation tab visually distinct by @cherry-coding[bot] in #23085
fix(stream-manager): raise the chat turn parsed-chunk idle bound to 120 minutes by @cherry-coding[bot] in #23073
docs(path-containment): propose collapsing duplicated main-process containment predicates by @kovsu in #23057
feat(mermaid): render diagrams with Merman in a worker and preview Mermaid files by @DeJeune in #23095
fix(rich-editor): stop diagram code blocks from freezing the notes editor by @DeJeune in #23101
fix(app-shell): unify navigation selection and contrast by @DeJeune in #23104
fix(preview-tests): restore main CI fixture and resize readiness by @DeJeune in #23105
feat(settings): show installed size in Environment Dependencies by @cherry-coding[bot] in #23086
feat(agents): archive all conversations of an Agent in one action by @cherry-coding[bot] in #23067
fix(tooltip): portal dialog tooltips beside the dialog content by @EurFelux in #22992
fix(aihubmix): backport #22364 to v2.1.9 by @DeJeune in #23123

New Contributors

@pei711 made their first contribution in #22573
@Wit-2k made their first contribution in #22597
@sseymdoc made their first contribution in #22664
@JuLuogo made their first contribution in #22629
@RaphaelFakhri made their first contribution in #22613
@mmporong made their first contribution in #22716
@sgan0420 made their first contribution in #22599
@xiong-wei1 made their first contribution in #22842
@jaeminbird made their first contribution in #22957
@linxule made their first contribution in #18135
@ning-jl made their first contribution in #20347
@Chkhikvadze made their first contribution in #21987

Full Changelog: v2.1.8...v2.1.9