v1.9.12

What's Changed

• fix: forward DeepSeek V4+ reasoning_effort for openai-compatible providers by @EurFelux in #15834
• fix(model): support Claude Fable series and Opus 5 in capability detection by @EurFelux in #15872
• fix(models): add Anthropic native /v1/models fetcher(target V1 branch) by @Raymond8196 in #15779
• fix(tools): prevent AccordionTrigger toggle when clicking approval buttons in collapsed group header by @LWWZH in #15839
• fix: use anthropicApiHost for NewAPI provider's anthropic endpoint by @Raymond8196 in #15934
• fix(agents): inject Cherry Studio identity headers into agent mode requests by @Raymond8196 in #15978
• fix: support downloading SVG data URL images blocked by CSP by @Raymond8196 in #15301
• fix(file): restore leading dot in ext for saveBase64Image and savePastedImage by @Raymond8196 in #14668
• fix(reasoning): correct MiniMax-M3 thinking control by @kangfenmao in #15991
• feat(models): add Kimi K2.7 Code support by @ousugo in #16017
• fix(ai-core): surface abnormal stream finish reasons as error chunks by @EurFelux in #16079
• fix(websearch): preserve full URLs and parse Highlights in ExaMCP results by @Raymond8196 in #15329
• fix: recover claude-code install when native binary is missing by @kovsu in #15369
• fix(aihubmix): derive Gemini baseURL from configured baseURL by @DDU1222 in #16371
• fix: use first configured key for claude code agents by @zhibisora in #16437
• fix: use fs.statfs for disk info to stop powershell process leak by @Raymond8196 in #16556
• feat(aihubmix): redirect provider links to inferera and follow configured base URL by @DDU1222 in #16606
• fix: hide CherryIN manual model add button by @zhibisora in #16703

Full Changelog: v1.9.11...v1.9.12

v1.9.11

What's Changed

• chore(models): prune legacy MiniMax models, keep M2.7 and M3 by @kapelame in #15580
• fix: encode gemini model list api key by @kangfenmao in #15575
• fix: infer custom provider model groups by @kangfenmao in #15577
• fix: clean up agent session messages on delete by @kangfenmao in #15531
• fix: oversized text in Gemini mini-program due to Electron UA detection by @LWWZH in #15389
• fix(reasoning): add MiniMax M3 thinking process support (v1.9.8) by @HUQIANTAO in #15513
• fix(anthropic): support Opus 4.8 adaptive thinking by @kangfenmao in #15532
• fix: use openPath instead of openExternal for HTML artifacts external browser by @Raymond8196 in #15644
• fix: avoid embedding dimensions request parameter by @eeee0717 in #15727
• fix: bind CherryIN OAuth flow to sender by @3em0 in #15410
• fix: upgrade mermaid to 11.15.0 to fix CJK parsing error in flowcharts by @LWWZH in #15391
• fix: agent task completed but task list still shows unchecked items by @LWWZH in #15358

New Contributors

• @3em0 made their first contribution in #15410

Full Changelog: v1.9.9...v1.9.11

v1.9.9

What's Changed

• fix: guard OV OCR availability CPU detection by @kangfenmao in #15538
• fix: respect topic naming setting for agent sessions by @kangfenmao in #15533
• feat(models): add MiniMax M3 to the system model registry by @kapelame in #15543

New Contributors

• @kapelame made their first contribution in #15543

Full Changelog: v1.9.8...v1.9.9

v1.9.8

What's Changed

• fix(ci): add v1 branch to workflow triggers by @GeorgeDong32 in #15446
• refactor: update tool components for improved styling and functionality

Full Changelog: v1.9.7...v1.9.8

v1.9.7

What's Changed

• fix: clarify OpenClaw external install prompt by @mturac in #15147
• fix: support Grok 4.3 reasoning effort in xAI responses by @ousugo in #15137
• fix(code): keep light theme tokens readable by @mturac in #15143
• fix(claudecode): apply [1m] context suffix for DeepSeek V4 Flash and MiMo V2.5+ by @DeJeune in #15162
• fix(agents): pass custom headers to claude code by @kovsu in #15185
• fix: launch opencode from package-local executable by @kovsu in #15188
• hotfix(model): align Gemini 3.x UI and sampling handling by @ousugo in #15204
• chore(gitignore): Ignore Antigravity CLI local files by @ousugo in #15201
• fix: add missing key prop to SendMessageButton in InputbarCore by @LGiki in #15233
• fix: correct provider ID for AIHubMix reasoning effort by @xxiaoxiong in #15283
• fix(provider): support StepFun as Anthropic-compatible provider by @lianshuang-photo in #15277
• hotfix(models): support capabilities for grok-build-0.1 by @ousugo in #15256
• fix(models): exclude Qwen max series from vision model detection by @xxiaoxiong in #15288
• fix(analytics): respect enableDataCollection setting in renderer by @EurFelux in #15318
• fix(aiCore): do not treat Gemini model via non-Gemini provider as native PDF by @octo-patch in #14329
• fix(websearch): align ExaMCP response parser with server field names by @zhibisora in #15331
• fix: add CherryIN DeepSeek 1m suffix by @zhibisora in #15352
• hotfix(ci): run GitCode sync on signing runner by @zhibisora in #15324
• chore: release v1.9.7 by @kangfenmao in #15362
• fix: keep note toolbar static at top during scroll by @LWWZH in #15349
• Fix: change Gemini Safety Settings to BLOCK_NONE - For main by @ous50 in #15303
• fix: refresh cached miniapp urls by @mturac in #15146

New Contributors

• @mturac made their first contribution in #15147
• @LGiki made their first contribution in #15233
• @xxiaoxiong made their first contribution in #15283

Full Changelog: v1.9.6...v1.9.7

v1.9.6

What's Changed

• fix(knowledge): preserve HTTP URLs in knowledge base documents by @Raymond8196 in #14983
• fix(ai): thread idle timeout handle to stream chunk adapter by @GeorgeDong32 in #15056
• fix(ci): improve GitCode sync reliability by @zhibisora in #15063
• fix(image): convert assistant image blocks to base64 for multi-turn editing by @Raymond8196 in #15011
• fix(agent): invalidate provider cache on settings changes by @404-Page-Found in #14596
• fix: recognize ctrl by uniKey value by @kovsu in #15002
• hotfix(aiCore): prevent crash when model.provider not found (#14999) by @DeJeune in #15001
• fix: align provider model list actions by @kangfenmao in #15078
• fix(agents): use task name as session name for cron task sessions by @Kingzi49 in #15053
• fix: resolve ENOENT in OpenMinerU preprocessing caused by MinerU ZIP structure change by @suujonH in #15061
• fix: clean up OAuth tokens when deleting MCP server by @xr843 in #14083
• fix: exclude deepseek-r1 models from function_calling capability by @404-Page-Found in #15102
• fix: auto-expand MCP tool approval card so action buttons are immediately visible by @LWWZH in #15080
• hotfix: add environment variable for claude code provider management by @kovsu in #15089
• fix(agents): preserve updatedAt timestamps when creating new agents by @LWWZH in #14675
• fix(aiCore): Qiniu PDF fallback for GPT-5.4 by @404-Page-Found in #15090
• fix(code-tools): handle Codex CLI reserved provider IDs by @GeorgeDong32 in #15068
• chore: release v1.9.6 by @kangfenmao in #15116

New Contributors

• @Kingzi49 made their first contribution in #15053
• @suujonH made their first contribution in #15061

Full Changelog: v1.9.5...v1.9.6

v1.9.5

What's Changed

• fix(search): prevent empty activeTools when web search is the only tool by @TakotsuboChen in #14752
• hotfix(vertex): vertex model list fetch and improve service account setup by @ousugo in #14611
• fix(messages): prevent outer scrolling in horizontal multi-model layout by @Fu-Jumping in #14768
• fix(reasoning): use enable_thinking param for SiliconFlow DeepSeek/Zhipu models when reasoning_effort is none by @Raymond8196 in #14782
• feat(claw): add timeout_minutes parameter to cron tool by @xvweirong in #14707
• hotfix(gateway): bypass @ai-sdk/gateway schema for Vercel model listing by @DeJeune in #14772
• fix(api-server): strip trailing /v1 from Anthropic SDK baseURL by @Raymond8196 in #14818
• hotfix(feishu): use emoji reaction as typing indicator by @DeJeune in #14799
• hotfix(i18n): update default assistant and topic names on language switch by @Raymond8196 in #14607
• fix(anthropic): support Claude Opus 4.7 by @DeJeune in #14349
• hotfix(models): update DeepSeek default models to V4 by @ousugo in #14877
• fix(NewApiPage): restore generated image re-editing by @404-Page-Found in #14500
• Feat(AIhubmix): painting module add gpt-image-2 to AiHubMix provider by @DDU1222 in #14903
• hotfix(models):detect routed DeepSeek V4 model slugs for reasoning effort by @ousugo in #14887
• hotfix(models): support hosted Gemma 4 thinking mode by @ousugo in #14793
• fix(code-tools): disable opencode built-in auto-update check by @Raymond8196 in #14925
• hotfix(models): correct Tongyi model icon matching by @ousugo in #14922
• fix(agent): allow CherryIN OpenAI-protocol models in Agent model picker by @Azure12355 in #14811
• hotfix(backup): stage restored data into .restore dirs to fix silent data loss on macOS by @DeJeune in #14781
• fix: sanitize subprocess call in run_eval.py by @orbisai0security in #14842
• hotfix(agents): remove stale mcp__browser__* references from agent prompt by @DeJeune in #14815
• fix(backup): clean up selection hook before restore relaunch by @kangfenmao in #14961
• hotfix(agents): show all sessions and agents in sidebar (#14854) by @DeJeune in #14954
• hotfix(agents): report 1M context for DeepSeek V4+ pro on official host by @DeJeune in #14965
• chore: release v1.9.5 by @kangfenmao in #14963
• hotfix(naming): sanitize non-ASCII characters in provider name for env vars by @Raymond8196 in #14915

New Contributors

• @TakotsuboChen made their first contribution in #14752
• @xvweirong made their first contribution in #14707
• @Azure12355 made their first contribution in #14811
• @orbisai0security made their first contribution in #14842

Full Changelog: v1.9.4...v1.9.5

v1.9.4

What's Changed

• hotfix(models): add DeepSeek V4+ model support with reasoning effort by @SiinXu in #14551
• hotfix(copilot): github copilot model fetch by @ousugo in #14566
• hotfix(models): add MiMo V2.5 model support by @ousugo in #14557
• hotfix(deepseek): forward reasoning effort for DeepSeek V4+ via Claude endpoint by @DeJeune in #14572
• fix(models): add vision and reasoning_effort support for mistral-small-2603 by @GeorgeDong32 in #14541
• hotfix(models): disable temperature and top_p for Kimi K2.6 by @DeJeune in #14580
• hotfix(agents): drop @cherry/browser MCP auto-injection by @DeJeune in #14571
• hotfix(image-generation): fix gpt-image-2 / gpt-image-1.5 failures and stuck pending placeholder by @DeJeune in #14578
• fix(gateway): pass user apiKey to createGateway when fetching models by @Raymond8196 in #14605
• fix(code-tools): support new-api providers and endpoint type routing for OpenCode by @GeorgeDong32 in #14511
• fix(images/edits): default to image/png when file has no extension by @LWWZH in #14616
• chore: release v1.9.4 by @kangfenmao in #14681
• fix(CodeBlock): do not recognize Unix-style paths as clickable links on Windows by @LWWZH in #14709
• fix(deps): bump @ai-sdk/deepseek to 2.0.30 by @DeJeune in #14718
• fix: suppress DeepSeek DSML tags in web search results by @kangfenmao in #14721
• fix: parse DeepSeek DSML tool-call markup into AI SDK tool calls by @DeJeune in #14747
• fix(deps): omit max_tokens for non-Anthropic models in @ai-sdk/anthropic by @DeJeune in #14749
• fix: refresh openrouter provider patch by @zhibisora in #14750

New Contributors

• @LWWZH made their first contribution in #14616

Full Changelog: v1.9.3...v1.9.4

v1.9.3

What's Changed

• fix(quick-panel): reset stale state on panel close to prevent model deselection by @GeorgeDong32 in #14378
• chore(deps): bump form-data from 2.3.3 to 4.0.5 by @dependabot[bot] in #14441
• fix: use cli-wrapper.cjs to launch claude-code instead of native binary by @Raymond8196 in #14430
• hotfix: Custom params dropped by CherryIN/NewAPI — respect model.endpoint_type by @zhibisora in #14409
• fix(security): allow vscode/cursor/zed deep-links in openExternal by @beyondkmp in #14437
• hotfix: Support case-insensitive file extensions for drag-and-drop uploads when editing messages by @ELDment in #14416
• fix(agents): preserve session instructions in Soul Mode by @DeJeune in #14424
• fix(api-server): rotate comma-separated provider API keys by @DeJeune in #14346
• hotfix: disable native structured output for AiHubMix/NewAPI Anthropic models by @DeJeune in #14376
• fix(feishu): handle image messages from IM channel by @DeJeune in #14421
• fix(useInstalledSkills): keep agent skills list stable while toggling by @404-Page-Found in #14472
• fix: prevent empty baseURL/region string in Bedrock provider config by @Tsudrat in #14425
• fix(ai-core): keep native tool loops going by @404-Page-Found in #14481
• fix(mcp): delegate connect timeout to SDK to restore slow SSE servers by @DeJeune in #14348
• hotfix(ai-sdk/openai): patch @ai-sdk/openai to support gpt-image-2 by @DeJeune in #14488
• fix: prevent outer scrolling in horizontal multi-model messages by @Fu-Jumping in #13964
• fix: handle symlinks during backup copy by @kangfenmao in #14471
• fix: limit builtin web search usage by @kangfenmao in #14466
• fix: align CherryIN seed apiHost with migration 183 (.cc) by @zhibisora in #14528
• fix(models): add gpt-image-2 model icon by @DeJeune in #14539
• chore: release v1.9.3 by @kangfenmao in #14523

New Contributors

• @zhibisora made their first contribution in #14409
• @Tsudrat made their first contribution in #14425
• @Fu-Jumping made their first contribution in #13964

Full Changelog: v1.9.2...v1.9.3

v1.9.2

What's Changed

• fix(openclaw): persist language selection in overview page by @404-Page-Found in #13992
• fix(agents): set NODE_PATH so spawned Claude Code process can resolve @img/sharp by @beyondkmp in #14183
• fix(agents): sync agent settings changes to active session by @DeJeune in #14205
• fix(skills): support agent-authored skills via skills tool init/register by @DeJeune in #14184
• fix: update dify knowledge retrieval payload by @kovsu in #14246
• feat(agents): add "Add More Skills" button to agent skills settings by @kennyzheng-builds in #14258
• fix: limit topic naming retries and timeout by @kovsu in #14265
• feat(skills): per-agent skill enablement with workspace symlinks by @DeJeune in #14247
• chore(deps-dev): bump axios from 1.13.2 to 1.15.0 by @dependabot[bot] in #14293
• chore(deps-dev): bump lodash from 4.17.21 to 4.18.1 by @dependabot[bot] in #14294
• chore(deps): bump form-data from 2.3.3 to 4.0.4 by @dependabot[bot] in #14295
• feat(i18n): add Vietnamese (vi-VN) language support by @kennyzheng-builds in #14279
• fix: sync builtin agent prompt updates to sessions by @kovsu in #14304
• fix(knowledge): support Ollama knowledge embeddings by @404-Page-Found in #14172
• 🤖 Daily Auto I18N Sync: Apr 17, 2026 by @github-actions[bot] in #14321
• fix(agents): set Cherry Assistant default avatar to 🍒 by @DeJeune in #14333
• hotfix: Custom params not passed to Gemini API when using NewAPI/AiHubMix by @ELDment in #14352
• fix(search): support phrase search and fix CJK matching in history search by @Raymond8196 in #14225
• fix(export): add obsidian:// to allowed protocol whitelist by @rfslzy in #14366
• fix: Fix Ollama model list loading when metadata contains null families values by @lancediarmuid in #14364
• fix(analytics): respect data collection setting for all tracking events by @Yuerchu in #14390
• fix: update node-abi for Electron 41 packaging by @kangfenmao in #14403
• chore: release v1.9.2 by @kangfenmao in #14406

New Contributors

• @ELDment made their first contribution in #14352
• @rfslzy made their first contribution in #14366
• @lancediarmuid made their first contribution in #14364
• @Yuerchu made their first contribution in #14390

Full Changelog: v1.9.1...v1.9.2