LobeHub Changelog
v1.x
Features
- Share page improvements and pg17 docs update.
What's fixed
- Group builder not set true edit data.
Features
- Add the fork tag show in community detail page.
What's fixed
- Slove the agentbuilder install market tools not work.
What's fixed
- Fix cron job issue, fix share single message.
What's improved
- Update i18n.
What's fixed
- Fix update memory tools, resolve server version check issue for desktop app, slove the descktop use offical endpoint mcp not use stdio.
What's improved
- Improve popover trigger styles and component consistency.
What's fixed
- Library cannot nav.
What's fixed
- Show fallback title for custom assistant in chat messages, webhook user service compatibility for old nextauth users.
What's improved
- Refactor search model implement.
What's fixed
- Fix add message and improve local system tool.
What's fixed
- Broadcast tools calling and improve auto scroll.
Features
- Update the discover page sort, add haveSkill、mostUsage params.
What's improved
- Update share action bar.
What's fixed
- Slove group member plugin is lost & not use the plugins.
Features
- Support history context auto compress.
Features
- Add the agent/group profiles page the states and forked by tag.
Features
- Optimize profile editor.
What's fixed
- Add cron pages enables change should reload the state.
What's fixed
- Prevent recently viewed items from shrinking.
What's fixed
- Docker deploy REDIS_URL check, fix sub task issue.
What's fixed
- Fixed when windows withd low the protal will resize.
What's fixed
- Fix page selection not display correctly.
What's fixed
- Login success callback url error.
What's fixed
- Surface streaming errors during mid-stream pulls.
What's fixed
- Page content switch mismatch.
Features
- Remove NextAuth.
What's fixed
- Fix favorite refresh bug and group topic refresh issue.
What's improved
- Migrate AI Rules to Claude Code Skills.
What's fixed
- Fix auto scroll.
What's fixed
- When use market group, the group sys role was not used.
What's fixed
- Fixed the agent group builder tools excaution edge case crash, fixed the group topic copy not right.
What's improved
- Move plugin store button outside scroll container.
What's fixed
- Add advace config back in agent/group profiles.
Features
- Remove Clerk authentication code.
What's improved
- Improve auto scroll and group profile.
What's fixed
- Slove the agents header switch agents the lobeAI not show problem.
What's fixed
- Fixed the sandbox tools call when error should use right callback.
Features
- Add server version check for desktop app.
What's improved
- Update og.
Features
- Skill setting page and skill store.
What's improved
- Move vercel-react-best-practices skills to .agents directory.
What's fixed
- Fix group broadcast trigger tool use, fix local system tools.
What's fixed
- Updata cron job ui & fixed commnuity pagenation goto error.
What's fixed
- Fix memory schema, update the agentbuilder tools not always use humanIntervention.
Features
- Support agent group unpublish agents.
What's fixed
- Fix tool argument scape and improve multi task run.
Features
- Add platform-aware download client menu option.
Features
- Update the sandbox preinstall libs in sys role.
What's fixed
- Fix multi tasks no summary issue.
What's fixed
- Improve e2e server and complete i18n resources.
What's fixed
- Slove the agent group editor not focus in editdata area.
What's fixed
- Fix multi agent tasks issue.
What's fixed
- Sloved the old removeSessionTopics not work.
Features
- Support client tasks mode.
What's fixed
- TypewriterEffect not refreshing on language change.
Features
- Add the agents and agents group fork feature.
What's improved
- Optimize profile settings skeleton screen.
What's fixed
- Fix Topic component causing stack overflow and freezing the app, simplify updater config logic, slove the nuqs error in commnuity agent group page.
What's fixed
- Slove commnuity user avatarUrl is wrong, should update others in profile.
What's fixed
- Fix parallel tools calling race issue.
What's fixed
- When use trpc client should include the credentials cookies.
Features
- Add the cloudEndpoint & Klavis Tools Call in Excuation Task.
Features
- Improve desktop onboarding window management and footer actions.
What's fixed
- Fix server agent task run with headless, internlm provider base url and homepage.
Features
- Improve the agentbuilder systemRole.
What's improved
- Change the /community/assistant to /agent routes.
What's improved
- Refactor market sdk into market servers.
What's improved
- Update i18n.
What's fixed
- Fix group sub task execution.
What's fixed
- Fixed the createGroupWithSupervisor function test, slove when use copy & install group from market, the member system Role is lost.
What's fixed
- Fix supervisor id issue.
What's improved
- Improve auto scroll and loading hint.
What's improved
- Improve operation hint and fix scroll issue.
What's improved
- Fix left panel on group page.
What's fixed
- Try to fix group supervisor id not sync successful.
What's fixed
- Fix supervisor group prompt.
What's fixed
- Fix topic messages display error when switch topic quickly.
What's fixed
- Fix switch skill in home.
Features
- Add agent group publish into market & use market group agents in lobehub.
What's improved
- Improve todo list.
What's improved
- Migrate Next.js navigation APIs to React Router for SPA.
What's improved
- Use fallbackData to prevent useActionSWR auto-fetch.
What's improved
- Improve agent loading state.
What's fixed
- Click lobe ai topic trigger create new agent.
What's fixed
- Fix internal editor onTextChange issue and add test case.
What's fixed
- Fix page content mismatch when switch quickly.
Features
- Improve group prompt context engine and fix group supervisor response issue.
What's fixed
- Prevent auto navigation to profile when clicking topic.
Features
- Update the agent profiles tools check & agentbuilder tools & publish to market button.
What's improved
- Remove the old lobehub plugins.
What's fixed
- Slove the settings/profile change but not refresh the profiles.
What's improved
- Fix UI issues with tooltip wrapping and dropdown type.
What's improved
- Update readFile content.
What's fixed
- Fix group ux and memory retriever.
What's improved
- Add MiniMax-M2.1 and GLM-4.7 for Qiniu provider.
What's fixed
- Fix new topic flick issue, fix thread portal not open correctly.
Features
- Update the community user layout action button, update the cron job visiual way.
Features
- Improve PageEditor header UX with DropdownMenu and i18n support.
Features
- Improve group profile builder, update model definitions and sync i18n.
What's improved
- Update i18n.
What's fixed
- Refresh sidebar after sendAsGroup and add E2E tests.
Features
- Improve baseline alignment for tool items.
What's improved
- Improve cron jobs style.
What's fixed
- Fix duplicate agent and group, Fix Windows desktop build error with macOS native module, force plain text paste in ChatInput editor.
What's fixed
- Handle will-prevent-unload event to allow app quit.
What's improved
- ModelParse support to get model settings params & Customize extendParams UI.
What's improved
- Rename chat folder to agent.
Features
- Update the cron patterns fields values.
What's fixed
- Fix memory search context.
What's improved
- Remember page agent panel width.
What's improved
- Refactor page and notebook document usage.
What's fixed
- Fix auto add group member crash.
What's improved
- Update i18n.
Features
- Add the agent cron job.
What's improved
- Force gpt-5.2 use response api.
Features
- Generate agent_cron_jobs in agents.
What's improved
- Improve Tools popover component structure and fix UI consistency.
Features
- Add the twitter lobehub skill.
Features
- Change the klavis Linear to LobeHub oauth Linear.
What's fixed
- Topic renaming input focus issue in context menu.
What's improved
- Update i18n.
Features
- Add browser compatibility detection and fallback page, add the lobehub market tools servers.
What's fixed
- Correct BrandTextLoading position after removing SSG CSS-in-JS injection.
What's fixed
- Update desktop onboarding privacy description.
What's fixed
- Fix edit rich render codeblock.
What's fixed
- Update mobile topicRouter import path to lambda directory.
What's fixed
- Add separate border-radius for bottom-right corner on macOS 26 Chrome.
What's fixed
- Allow zero-byte files and add business hooks for error handling.
What's improved
- Change all market routes & api call into lambda trpc client call.
What's fixed
- Fix callback url error during signin period.
What's fixed
- Fix editor modal and refactor ModelSwitchPanel.
What's improved
- Convert glossary from JSON to Markdown table format.
What's fixed
- Resolve desktop upload CORS issue.
What's fixed
- Restore getBounds mock in Browser test beforeEach.
What's fixed
- Resolve BaseUI dropdown compatibility issue.
Features
- Update the sandbox export files & save files way.
What's fixed
- Fix editor modal when Markdown rendering off.
What's fixed
- Restore window position safely.
What's fixed
- Update CI bun version to v1.2.4, when the document filetype is agent/plan, not show the saveinto docs button.
What's improved
- Update i18n.
What's fixed
- Add lost like button in discover detail page.
What's fixed
- Use configured embedding provider instead of hardcoded OpenAI.
What's fixed
- Auto jump to group.
What's fixed
- Slove the old agents open profiles error problem.
What's fixed
- Fix data inconsistency in ai provider config.
Features
- Add new provider Xiaomi MiMo.
What's improved
- Update i18n.
What's improved
- Refactor and fix model runtime initialize.
What's fixed
- Restore window resizable before hard reload in desktop onboarding.
Features
- Add work path for local system.
What's fixed
- Filter empty assistant messages for Anthropic API.
What's fixed
- Support thoughtSignature for openrouter.
What's improved
- Remove client db and refactor test.
What's fixed
- Fix file upload issue.
What's improved
- Refactor to remove access code.
What's fixed
- Fix tool call message content missing.
What's improved
- Update i18n.
What's fixed
- Fix model edit icon missing.
What's improved
- Refactor to remove meta in message.
What's improved
- Update i18n.
What's improved
- Migrate to new DropdownMenuV2 and showContextMenu API.
What's improved
- Improve tools UI and fix Google schema compatibility.
What's improved
- Add Gemini 3 Flash & Doubao Seed 1.8 models.
What's improved
- Refactor oidc env to auth env.
What's improved
- Update i18n.
What's improved
- Improve loading and local-system render.
Features
- Brand new 2.0 ui for next.
Features
- Add a white waitlist in edge config env, add always show tools render in createPlan & createDoc tools, add batch tasks ui, add Bundle Analyzer workflow for detailed bundle size analysis, add business features support with new components and hooks, add business settings features with dynamic loading for Plans, Funds, Usage, Billing, and Referral tabs, add db and schema feature, add home page create group builder button, Add i18n UI locales and improve tool types, add like action in community detail, add memory implement, add subscription settings group with dynamic loading for Plans, Funds, Usage, Billing, and Referral tabs, add the market auth auto generate way, Add turbopack configuration support to CustomNextConfig, add user memory, agent builder, agent builder, agent builder and group builder, app ui page, brand new 2.0 ui for next, buildin some tools should save into docs, code-interpreter tool, code-interpreter tool, code-interpreter tool, desktop feature, enhance desktop onboarding with sign out and localization, enhance macOS desktop permissions and onboarding, enhance onboarding process by removing mode selection step and adding export functionality in advanced settings, file search feature, gtd create plan support streaming render, implement agent builder, implement builtin agents packages, implement memories package, implement Redis caching for presigned URLs in file proxy service, implement server data feature, include Subscription settings group in the Accordion component, Integrate bcryptjs for password verification in BetterAuth, integrate BrandingProviderCard and update Provider components for branding support, onboarding ui, page and knowledge base, rebranding total UI of app, refactor authentication handler to support dynamic loading of better-auth and next-auth, refactor desktop implement with brand new 2.0, rename codeinterpreter into lobe sandbox, server implement, support CMD K, support exec async sub agent task, support export and import topic JSON, support files upload in chat input, support notebook tool, support swr local cache, topic message swr cache, translate AI model descriptions to English, update agent builder ui, update create group chat use builder, update gtd tools( use editor & update metadata ), update user memory embedding model selection based on business features, user memory, user memory, user onboarding, when use usesend to create agent/group, the model should override by lobeAi, wrap ConversationArea and ModelSwitchPanel in TooltipGroup for enhanced UI.
What's improved
- Improve ExecTask and task message UI, improve gtd tool inspector and todo list, improve page document tool inspector UI, improve RunCommand Inspector, rebranding chat ui, refactor UI in features, rerun i18n, setting style, support streaming and display ui for group mode, support tool streaming and title custom render, update i18n, Update i18n microcopy, update ui.
What's fixed
- Agent profiles update, agent tools config set, editor placeholder, bump charts 3.0.4 to fix import es path, fix anthropic thinking budget, fix async task and improve tool style, fix default waitlist bug, fix delete agent group bug, Fix desktop test cases and refactor translations, Fix desktop test cases and refactor translations, fix gemini 3 model thinking issue, fix gemini 3 pro parallel tool use, fix gemini 3 thinking params, fix identity memory not working, fix supervisor flag, fix thread not working issue, fix when use branch topic,the branch index error problem, fixed the welcome card the create button not work, handle session invalidation on 401 error by logging out signed-in users, improve test infrastructure and mock configurations, locale resolve bug with ESM module loading, page agent editor, prevent redundant login redirect when already on auth pages, redis read json object, remove openapi pkg patch file, slove input editor on pause emit, slove swr mutate not work in Cache Provider, slove the group add member checkbox not work, slove the model select null problem, slove the mutate not work problem, slove when click agentbuilder should clean topic, slove when first call thread, not show ai chat message, support retry error message and fix continueGenerationMessage, update contextMenu in group tools message, update OFFICIAL_URL to app.lobehub.com, update PlanTag link paths for subscription settings, update test snapshots for model description changes, when use agentbuilder the topic id should use new & clear topic….
Features
- Mobile native better auth support.
What's improved
- Refactor database schema.
What's fixed
- Request to gpt5 series should not with top_p, temperature when reasoning effort is not none.
What's improved
- Update i18n.
What's improved
- Update GPT-5.2 models.
What's fixed
- Fix CVE errors.
What's fixed
- Slove market oidc error.
Features
- Add Replicate image provider.
What's improved
- Update link handling in PlanTag component to use react-router-dom.
What's fixed
- Add smooth scroll to top on 'More' button click in Title component.
Features
- Support klavis mcp connector.
Features
- Betterauth username signin.
What's fixed
- Limit check-user response surface.
What's improved
- Update Spark X1.5 model.
What's fixed
- Fix React CVE issue.
What's fixed
- Missing init user after user creation.
What's fixed
- Udpate discover detail tools get & more link.
Features
- Optimize betterauth UX.
What's improved
- Unify retry logic to async-retry.
What's fixed
- Better-auth add apple sso icon and label.
What's fixed
- Remove apiMode param from Azure and Cloudflare provider requests, when desktop use contextMenu not work.
Features
- Support apple sso auth.
What's improved
- Refactor agent slug schema.
Features
- Email provider support resend.
What's fixed
- User email unique migration error.
Features
- Support market cloud endpoint mcp.
What's fixed
- Remove internal apiMode param from chat completion API requests.
What's fixed
- Drop user.phoneNumber and reuse user.phone.
Features
- Integrate better-auth admin plugin.
What's improved
- Update i18n.
What's fixed
- Update apiMode handling in ChatService to prioritize user preferences.
What's fixed
- Refresh custom AI provider on selection.
What's fixed
- Fix BetterAuth Unable to link account - untrusted provider.
What's fixed
- Betterauth public url auto detect from VERCEL_URL.
What's fixed
- Betterauth name should mapped to fullName.
What's fixed
- Unable to switch to default topic.
What's fixed
- Implement uniform callback URL for SSO providers.
What's fixed
- Add handling for content_part and reasoning_part events in fetchSSE.
What's fixed
- Filter out file with sourceType = file.
What's improved
- Update i18n.
What's fixed
- Better-auth fallback next-auth providers env.
What's fixed
- Align docker auth defaults and better-auth docs.
Features
- Support better-auth.
What's fixed
- Fixed the agent settings plugins pages error problem, improve topic item interaction and editing behavior.
What's improved
- Update i18n.
What's fixed
- Slove the publish to market the agent config error.
What's improved
- Add image aspect ratio and resolution settings for Nano Banana Pro.
What's fixed
- Try to fix “TypeError: Response body object should not be disturbed or locked”.
What's improved
- Update i18n.
What's fixed
- Showing compatibility with both new and old versions of Plugins.
Features
- Bedrock claude model thinking support.
Features
- Support nano banana pro.
Features
- Add Claude Opus 4.5 model.
What's fixed
- Fixed the topic link dropdown error.
What's fixed
- Fixed when desktop userId was change manytimes the aimodel not right.
What's improved
- Add Kimi K2 Thinking to Qwen Provider.
What's fixed
- Fix db migration snapshot not align with db schema, Separate agent file injection from knowledge base RAG search.
What's improved
- Add hyperlink to each topic & pinned agent, support ContextMenu on ChatItem.
What's improved
- Update i18n.
What's fixed
- Fixed the knowledge files cant open error.
What's fixed
- Fixed the pinned session not work.
What's improved
- Optimize nana banana pro error message.
Features
- Add nano-banana-pro model support and optimization.
What's improved
- Update i18n.
What's fixed
- Hide ai image config item in settings category.
Features
- Add new provider ZenMux & Gemini 3 Pro Image Preview.
Features
- Support bedrok prompt cache and usage compute.
What's fixed
- Gemini 3 Pro does not display thought summaries.
Features
- Refactor to use kb search tool.
What's improved
- Fix some translations.
What's fixed
- Fixed changelog pages and open again.
What's improved
- Update i18n.
Features
- Support Command Menu (CMD + J).
Features
- Add Security Blacklist for agent runtime.
What's fixed
- Provider settings button unable to redirect.
What's improved
- Update i18n.
What's improved
- Remove debug console logs and add loading state.
What's fixed
- Fixed the hydrated false problem.
What's improved
- Extract StatusIndicator component and improve tools display.
Features
- Support gemini 3.0 tools calling.
What's improved
- Fully support Gemini 3.0 model.
What's improved
- Refactor chat selectors.
Features
- Support user abort in the agent runtime.
What's fixed
- Slove discover pagination router.
What's improved
- Add Gemini 3.0 Pro Preview to Google Provider.
What's fixed
- Fix noisy error notification.
What's fixed
- Slove when logout always show loading.
What's fixed
- Fixed the discover page categray sider link error.
What's improved
- Delete /settings/newapi pages in nextjs build.
Features
- Support Interleaved thinking in MiniMax.
What's improved
- Update i18n.
Features
- Edit local file render & intervention.
Features
- Support parallel topic agent runtime.
What's improved
- Add model information for the Qiniu provider.
What's fixed
- Fix desktop user panel.
What's improved
- Remove language_model_settings and remove isDeprecatedEdition.
What's fixed
- The tool to fail execution on ollama when a message contains b….
What's improved
- Refactor to virtua.
Features
- Support to collapse message.
What's improved
- Update i18n.
What's improved
- Refactor package types.
Features
- Show orphaned tool message and support delete tool message.
What's improved
- Update i18n.
What's fixed
- Reduce threshold.
What's improved
- Update i18n.
Features
- Support DeepSeek Interleaved thinking.
What's improved
- Revert background style.
Features
- Add folder creation UI and clean up debug code.
Features
- Create Pages in Knowledge Base.
What's improved
- Refactor and support move locale file intervention.
Features
- Add GPT-5.1 models.
What's improved
- Fix approving render and improve Conversation style.
What's fixed
- Filter out reasoning fields from messages in ChatCompletion API.
What's improved
- Update ERNIE-5.0-Thinking-Preview model.
What's fixed
- Fix oidc accountId mismatch.
What's fixed
- Fix mcp server return image error.
What's improved
- Fix thread display.
What's improved
- Edge to node runtime.
What's fixed
- Fix reasoning issue with claude and Response API thinking.
What's fixed
- Abnormal animation of tokens.
What's fixed
- Fix missing messages when finish runtime.
What's improved
- Update i18n.
What's improved
- Update i18n.
What's fixed
- Don't include runtimeProvider in JWT for non-image operations.
Features
- Refactor to use agent runtime as the generation core and support branch mode.
What's improved
- Use react-router-dom change /chat page to spa mode.
What's improved
- Add sorting functionality for disabled models and model providers with tooltip support.
What's improved
- Refactor message create name.
What's fixed
- Model name display in the assistant panel disappears.
What's fixed
- Should install new version after quit this instance.
What's improved
- Enhance message router with service layer and comprehensive tests.
What's improved
- Refactor chat message model to speed up.
Features
- Support install sreamable http mcp server on web.
What's improved
- Refactor services to a more clean structure.
What's improved
- Add settings (jsonb) column to ai_models table.
Features
- Display assistant message in group.
What's improved
- Improve lab style.
What's fixed
- Fix send message.
What's fixed
- Fix oidc auth timeout issue on the desktop.
What's improved
- Improve oidc layout style.
What's improved
- Remove NEXT_PUBLIC_SERVICE_MODE env and use server by default.
What's improved
- Improve built-in client OIDC user flow.
What's fixed
- Fix regex ReDoS.
What's improved
- Remove deperated code.
What's improved
- Remove client service.
What's fixed
- Fix image prompt form.
What's improved
- Add padding to TopicList component.
What's improved
- Smoothed model descriptions in ko-KR locales.
What's improved
- Remove dalle builtin plugin.
What's improved
- Update i18n.
Features
- Upgrade to Next 16.
What's improved
- Migrating Firecrawl to v2.
What's improved
- Update i18n.
What's improved
- Add new bedrock model support, add pricing info for Azure GPT-5 series models.
What's fixed
- Hide marketplace link from Plugin List when market disabled, OIDC error when connecting to self-host instance, only include input_fidelity parameter for gpt-image-1..
Features
- 2.0 next init.
What's improved
- Update i18n.
What's improved
- Add MiniMax-M2 model.
What's improved
- Pre render ModelSwitchPanel, The error details of the connectivity check lead to a layout problem.
What's improved
- Adjust modal setting form styles for improved layout and responsiveness, Unzip file when uploading in knowledge base [LOB-500].
What's improved
- Improve provider modal height when creating custom provider.
What's improved
- Update i18n.
Features
- Use env to control clerk allow origin feature.
What's fixed
- Loadmore not work & navbar not show in pwa.
What's improved
- Improve local system tools render.
What's improved
- Improvement for Agent Team After Alpha Launch [LOB-517].
What's improved
- Allow removal of top_p and similar request parameters.
What's improved
- Change discover page from RSC to SPA to improve performance.
What's improved
- Fix model runtime cost calculate with CNY.
What's improved
- Update i18n.
What's improved
- Refactor context engine.
Features
- Add PDF export functionality to share modal.
What's improved
- Add knowledge base mansory layout [LOB-496], improve rich text link display.
What's fixed
- Ignore abort signal errors in TRPC client, slove when pwa user info have code cannot be viewed in full.
Features
- Add ComfyUI integration Phase1(RFC-128).
What's fixed
- Pass threadId to messages in sendMessageInServer.
What's improved
- Show message author in minimap.
What's improved
- Solve when desktop the sider agent list too long.
What's improved
- Update i18n.
Features
- Support image generation for siliconcloud.
What's improved
- Refactor upload router into lambda and decide to remove it in V2.
What's fixed
- Fix response API tools calling issue.
What's fixed
- Fix topic fetch not correct in custom agent.
What's improved
- Improve welcome message.
What's fixed
- Automatic topic creation switch does not work.
Features
- Support Group Chat, Mention, and Multi-Agent Orchestration with feature flag.
What's improved
- Add Claude Haiku 4.5 model.
What's improved
- Improve update notification.
What's improved
- Add region support for Vertex AI provider.
What's fixed
- Fix duplicate tools id issue and fix link dialog issue.
What's improved
- Use different favicon.ico in dev mode.
What's fixed
- Update Claude workflows to use oauth token, vertext ai create image.
What's improved
- Add imagen model to vertex ai.
What's fixed
- Prevent Vertex AI JSON credentials from being split by comma.
What's fixed
- Fix mcp server connect issue and refactor web search implement, fix tools calling long name length >64 issue.
What's fixed
- Fix the Worker URL cross-origin issue.
What's improved
- Change the user chatItem maxWidth should use flex 1.
Features
- Add new setting for default image num.
What's improved
- Optimize OpenRouter modelFetch endpoint, update i18n.
What's fixed
- Fix input cannot send markdown.
What's improved
- Add more AWS regions, Update infini-ai models.
What's improved
- Add capability inference for web search, image output and video recognition in model parsing and update UI form items to support search, imageOutput and video abilities.
What's improved
- Improve search experience.
What's improved
- Add lab to support disable/enable rich text.
What's fixed
- Disable rich text in markdown editor.
What's fixed
- Add 'gemini-2.5-flash-image' to disabled models Thinking.
What's improved
- Add delete & regenerate hotkeys.
What's improved
- Update i18n.
What's improved
- Update i18n.
What's improved
- Add GPT-5 pro model.
What's improved
- Improve Korean translate.
What's improved
- Improve styles and fix tools calling condition.
Features
- Huanyuan text-to-image 3.
What's improved
- Update i18n.
What's improved
- Add promptfoo to improve prompts quality.
What's fixed
- Type not preserved when model is sorted.
What's improved
- Allow switching model type.
What's improved
- Update i18n.
Features
- Support double-click to open multi agent window on the desktop.
What's improved
- Nano banana support aspect_ratio.
What's fixed
- type not preserved when model is disabled or sorted.
What's improved
- Optimized extendParams UI, update i18n.
What's fixed
- Custom provider fails when client requests are enabled.
What's improved
- Refactor a ssrf-safe-fetch module.
What's fixed
- Fix frontend random API key config not work.
What's improved
- Add minimap to chat list for quick navigation.
What's improved
- Update i18n.
Features
- Add builtin Python plugin, add Claude Sonnet 4.5 model to AI chat models.
What's improved
- Update i18n.
What's fixed
- Refactor tools-engine and fix search token count.
What's fixed
- Fix input empty group name.
What's fixed
- Resolve qwen-image-edit imageUrls conversion issue.
What's fixed
- Add proxyUrl configuration for NEW API provider.
What's improved
- Update i18n.
What's fixed
- Slove setting proxy page with style error.
What's improved
- Enhanced Nvidia NIM chat experience, OpenAI models in AiHubMix use Responses API.
What's fixed
- Macos desktop sign.
What's improved
- Refactor all @/types in model runtime to @lobechat/types.
What's improved
- Move the ModelProvider to model-bank.
What's improved
- Enable thinkingBudget control for Vertex Gemini 2.5 models, update i18n.
What's improved
- Added AUTH_MICROSOFT_ENTRA_ID_BASE_URL routing.
What's fixed
- Fix non stream mode in OpenAI Response API.
What's fixed
- Fix missing provider in server message.
Features
- Support google video understanding.
What's improved
- Enhanced AkashChat experience.
What's fixed
- Update Responses search tool to web_search.
What's improved
- Use ID as name if provider name is empty.
What's improved
- Extend custom provider runtime options, Optimized modelFetch for Vercel AI Gateway, update i18n.
Features
- Qwen provider add qwen-image-edit model support.
What's fixed
- Fix oidc open direct issue.
Features
- Add scroll support for pinned assistants using ScrollShadow.
What's fixed
- Fix svg xss issue.
What's fixed
- Add qwen provider support for image-edit model.
What's improved
- Update SiliconCloud reasoning models.
Features
- Support Vercel AI Gateway provider.
What's fixed
- Fix azure ai runtime error.
What's improved
- Improve error handle with agent config, support .doc file parse.
What's improved
- Enable toggling search on/off via search button click & historyCount button.
What's improved
- Update i18n.
What's fixed
- Google stream error unable to abort request.
What's improved
- Fix discover plugin link.
What's fixed
- Fix open chat page with float link modal.
What's improved
- Update i18n, Update model configs.
What's improved
- Refactor message proccesser to the context engine.
Features
- ChatInput support resize.
What's fixed
- Improve OpenAIStream processing to emit usage data for chunks lacking choices.
What's improved
- Refactor model runtime folder structure and add more tests.
What's fixed
- Delete files should delete chunks、embedings、fileChunk.
What's improved
- Add CometAPI model provider and chat models, update i18n.
What's fixed
- Fix editor key handling.
What's fixed
- Fix Assistant List error message.
Features
- Add Math and TaskList to Editor.
What's fixed
- Revert V1 Mobile.
What's improved
- Refactor to remove edge runtime and add more tests.
What's fixed
- Fix ChatInput send command switch.
What's improved
- Update doubao-seed-1.6-vision models.
What's fixed
- Enhance NewAPI with environment variables and fix routers compatibility.
Features
- ChatInput support rich text and support parallel send.
What's improved
- Remove edge runtime.
What's fixed
- Fix mobile header title to loog not ellipsis.
What's fixed
- Not use branch topic when this topic is not save.
What's improved
- Update i18n.
Features
- Add NewAPI as a router provider for multi-model aggregation.
What's improved
- Make LobeNextAuthDBAdapter Edge Compatible.
What's improved
- Make LobeNextAuthDBAdapter Edge Compatible.
What's improved
- Update i18n.
What's improved
- Update the price of the o3 model in OpenRouter.
What's fixed
- Support base64 image from markdown image syntax.
Features
- Refactor to speed up send message in server mode.
What's fixed
- Fix socks5 proxy not work problem, fix virtuaso minheight was null.
Features
- Add nano banana Chinese prompt notify.
What's improved
- Update i18n.
What's improved
- Add upload hint for non-visual model.
What's improved
- Adjust ControlsForm component to adapt to mobile phone display.
What's improved
- Support new provider Nebius.
What's improved
- Remove base path.
What's improved
- Update i18n.
Features
- Rename Gemini 2.5 flash image to Nano Banana.
What's improved
- Remove webrtc sync feature flag.
What's fixed
- Update enableStreaming name.
Features
- Added support for Azure OpenAI Image Generation.
What's improved
- Update DeepSeek V3.1 & Gemini 2.5 Flash Image Preview models.
What's improved
- Update i18n.
What's improved
- Support non-stream mode.
What's improved
- Fix clerk scrollBox style, ModelFetcher support getting prices.
What's improved
- Fix chat session part switch theme issue.
What's fixed
- Correct totalOutputTokens calculation for XAI provider.
What's improved
- Add Grok Code Fast 1 model.
What's improved
- Support Gemini URL context tool.
What's fixed
- Add Content-Security-Policy env.
Features
- Add new provider AkashChat.
What's improved
- Move chat item into chat.
Features
- Ai image support Gemini 2.5 Flash Image, Support Gemini 2.5 Flash Image Preview in OpenRouter.
What's improved
- Update i18n.
What's improved
- Support html preview.
What's fixed
- Fix desktop route error.
Features
- Add gemini 2.5 flash image for vertex ai.
What's improved
- Update mistral model vision ability.
What's improved
- Move database to packages.
What's fixed
- Can't load custom provider config.
What's improved
- Move chain into @lobechat/prompts.
What's fixed
- Support Grok thinking models in AiHubMix, The 'stream_options' parameter is only allowed when 'stream' is enabled.
What's fixed
- Improve mcp tracing with user config.
What's improved
- Refactor prompts folder to the @lobechat/prompts pacakge.
What's improved
- Refactor const folder to a new package.
What's improved
- 重构 ArgsInput 组件.
What's improved
- Add Imagen 4 GA models, style improve auth sign in box loading.
What's improved
- Update Mistral AI models & Optimize many model providers fetching.
What's improved
- Adjust near bottom size on thinking scroll.
What's improved
- Improve Gemini error display with promptFeedback.
What's improved
- Support new GPT-5 Verbosity params.
What's improved
- Refactor model-runtime to a seperated package.
What's fixed
- Solve the cache problem caused by the same dom id when sharing pictures.
What's improved
- Add mask effect to thinking scroll, update i18n.
What's improved
- Improve thinking auto scroll style, Support session switch shortcut key, update i18n.
What's improved
- Add descriptions for the FLUX.1 Krea and Qwen Image.
Features
- Add GPT-5 series models.
What's fixed
- Missing languages it-IT, pl-PL, nl-NL.
What's fixed
- Optimize Gemini error message display & Filter empty messages.
What's improved
- Refactor trace type.
What's improved
- Fix provider setting page hydration error.
What's improved
- Add context menu for desktop, support different model tabs.
What's fixed
- Fix fail to fetch aihubmix model on client mode.
What's fixed
- Fix remote avatar broken in desktop again.
Features
- Support mcp plugin install from web.
What's improved
- Add Claude Opus 4.1 model, update i18n.
What's fixed
- Fix ollama model output without thinking.
Features
- Support gpt-oss in ollama provider.
What's fixed
- Provider config checker uses outdated API key.
Features
- Support 302ai provider.
What's fixed
- Break line for Gemini Artifacts.
What's improved
- Update models.
What's fixed
- When s3 files not exist , global files should delete.
What's fixed
- Aihubmix provider request headers.
What's improved
- Move types to separate package.
What's improved
- Update i18n.
Features
- Support aihubmix provider.
What's improved
- Support SenseNova V6.5 models.
What's improved
- Update Aliyun Bailian models.
What's fixed
- Fix oidc oauth callback pages 404.
What's improved
- Improve mcp plugin calling and display.
What's improved
- Add volcengine kimi-k2 model, Add Zhipu GLM-4.5 models.
What's fixed
- Moonshot assistant messages must not be empty.
What's fixed
- Fix desktop auth redirect url error.
What's improved
- Support Minimax T2I models.
Features
- Add support for Okta Authentication.
What's improved
- Open new topic by tap Just Chat again.
What's fixed
- Reorder AppTheme and Locale to fix modal i18n.
What's fixed
- Revert jose to ^5 to fix auth issue on desktop.
What's fixed
- Fix subscription plan tag display.
What's improved
- Add more OpenAI SDK Text2Image providers, update i18n.
What's improved
- Support more Text2Image from Qwen.
Features
- Implement API Key management functionality.
What's improved
- Fix setting window layout when in desktop was disappear.
What's improved
- Fix setting window layout size, update i18n.
What's improved
- Add Gemini 2.5 Flash-Lite GA model.
What's fixed
- Fix update hotkey invalid when input mod in desktop.
What's fixed
- Update convertUsage to handle XAI provider and adjust OpenAIStream to pass provider.
Features
- Support custom hotkey on desktop.
What's improved
- Add cached token count to usage of GoogleAI and VertexAI, fix desktop titlebar style in window, fix sub topic width in md responsive.
What's fixed
- Fix chat stream in desktop and update shortcut.
What's improved
- Update i18n.
Features
- Add Qwen image generation capabilities.
What's improved
- Update tray icon.
What's fixed
- Remove debug logging from ModelRuntime and async caller.
What's improved
- Add notification for desktop.
What's improved
- Modal list header sticky style.
Features
- Add image generation capabilities using Google AI Imagen API.
What's improved
- Fix lobehub provider /chat in desktop.
What's fixed
- Try fix authorization code exchange & pin next-auto to beta.29.
What's fixed
- Fix webapi proxy with clerk.
What's fixed
- Use server env config image models.
Features
- Refactor desktop oauth and use JWTs token to support remote chat.
What's fixed
- Desktop local db can't upload image.
What's fixed
- Fix page error when url is not defined in web search plugin.
What's fixed
- Fix apikey issue on server log.
What's fixed
- Chat model list should not show image model.
What's fixed
- Some ai image generation feedback issues.
Features
- support AI Image.
What's improved
- Update i18n.
What's improved
- Fix discover translation.
Features
- Add network proxy for desktop.
What's improved
- Support Hunyuan A13B thinking model.
What's improved
- Update i18n.
What's fixed
- Add vision support to Grok 4.
What's improved
- Add Kimi K2 model.
What's fixed
- Revert "💄 style: Open new topic by tap Just Chat again".
What's improved
- Support new Doubao thinking models, update i18n.
What's fixed
- Grok-4 reasoning model universal matching.
What's improved
- Open new topic by tap Just Chat again.
What's improved
- Update i18n.
What's improved
- Integrate Amazon Cognito for user authentication.
What's improved
- Add google search grounding for Vertex AI.
What's improved
- Replace utility-types with type-fest.
What's improved
- Fix: solve the loading was strange spin when switch show.
What's improved
- Add grok-4-0709 model from xAI, fix theme issue in desktop.
What's improved
- Implement data analytics event tracking framework.
What's fixed
- Fix locale hydration error in SSR.
Features
- Add MCP marketplace and mcp plugin one-click installation in desktop.
What's improved
- Add MCP_TOOL_TIMEOUT env and improve debug usage guide.
What's fixed
- Pin [email protected] to fix server error.
What's improved
- Files hello pages should scroll.
What's fixed
- Change the wrong github checkmodel name.
What's improved
- Update i18n.
What's fixed
- Wrong Gemini 2.5 Pro thinkbudget.
What's improved
- Migrate to @google/genai SDK for Google Gemini API and Vertex AI.
What's improved
- Update i18n.
What's improved
- Add DeepResearch models from OpenAI.
What's fixed
- Pin [email protected] to fix build error.
What's fixed
- Fix desktop chunk issue.
What's fixed
- Google Gemini tools declarations.
What's improved
- Optimized Gemini thinkingBudget configuration.
What's improved
- Add blockAds & stealth params for Browserless.
What's improved
- Update i18n.
What's fixed
- Remove unsupported parameters of Hunyuan.
What's improved
- Update i18n.
What's fixed
- Fix MiniMax-M1 reasoning tag missing.
Features
- Add Brave & Google PSE & Kagi as build-in Search Provider.
What's improved
- Update i18n.
What's fixed
- Correctly pass reasoning.summary.
What's improved
- Update model card for Gemini 2.5 Pro via OpenRouter.
What's improved
- Update i18n.
What's improved
- Update Gemini 2.5 Pro, Flash GA models. Add Gemini 2.5 Flash-Lite Preview model.
What's improved
- Add MiniMax-M1 model.
What's fixed
- Enhance the multi-display window opening experience.
What's fixed
- Improve chat selectors and enhance topic handling logic.
What's improved
- Add kimi-thinking-preview model from Moonshot.
What's fixed
- Correctly handle reasoning_effort.
What's improved
- Add Doubao Seed 1.6 model.
What's fixed
- Abort the Gemini request correctly & Add openai o3-pro.
What's improved
- Support web_search_preview & fix some bug form OpenAI Response API.
What's improved
- Transition animation switch.
What's fixed
- Update Gemini range of thinkingBudget.
What's improved
- Update pplx abilities tags, support vision.
Features
- Support google sso as auth provider.
What's fixed
- Restore reasoningEffort in setting.
What's improved
- Refactor branding info.
What's improved
- Refactor <think> & </think> handling.
What's fixed
- Fix openai default Responses API issue.
Features
- Support OpenAI Responses API mode.
What's improved
- Support OpenRouter Claude 4 reasoning.
What's fixed
- Fix client s3 getObject throw error.
What's improved
- Add support to azureopenai embedding.
What's improved
- Improve {{username}} placeholder variable, Update Gemini & Qwen models.
Features
- Support placeholder variables in prompts and input.
What's improved
- Support Vertex AI thought summaries.
What's fixed
- Some web search bugs.
What's improved
- Add openAI websearch and claude 4 to modelproviders.
What's fixed
- Correct deepseek R1 fc support display.
What's improved
- Add Volcengine & OpenAI-like Provider (e.g. oneapi) model fetch support.
What's improved
- Improve loading state.
Features
- Add more provider support for search & crawl.
What's improved
- Update modelscope models.
What's improved
- Rename the createOpenAICompatibleRuntime.
What's fixed
- Update the clerk middleware to support route protection.
What's improved
- Enable deploymentName for Aliyun Bailian.
What's improved
- Support web_search tool for MiniMax & Zhipu.
What's improved
- Use default deployment name when parseModelString doesn't contain deployment name.
What's fixed
- Disable LaTeX and Mermaid rendering in SystemRoleContent to prevent lag caused by massive rendering tasks when switching topics, fix DeepSeek new R1 Search error.
Features
- Support protect page.
Features
- Support ModelScope Provider.
What's improved
- + in the welcome message can be clicked to create an assistant.
What's improved
- Support 01.ai proxy url.
What's improved
- Enhanced reasoning_effort Slider Component, Update Hunyuan models & deepseek-r1-0528.
What's fixed
- Agent automatic completion meta not working error.
What's improved
- Support Web Search Tools and Beta Header from Anthropic.
What's fixed
- Close historySummary correctly, Enable thinking output only for supported Gemini thinking models.
What's improved
- Increase the history limit.
What's improved
- Update GitHub models.
What's fixed
- Cmd + click chat tab not open new tab.
What's improved
- Add fc ability to deepseek-reasoner model.
What's fixed
- Cant invoke the application after OIDC authorization in Windows 11.
What's improved
- Support adjust thinkingBudget in gemini 2.5 flash.
What's improved
- Correct model name SenseChat-5-1202.
What's fixed
- Auto sync theme mode in desktop.
What's improved
- Allow SliderWithInput to have no input limit.
What's fixed
- Fix chat header in the desktop.
What's improved
- Fix a few typos in the model tooltips.
What's fixed
- Fix draggable issue with agent header.
What's improved
- Support share single message.
What's improved
- Improve thread flicker when first-time loading.
What's improved
- Add gemini & hunyuan & Claude models.
What's fixed
- Fix message refresh 401 on desktop.
What's improved
- Add live search support for xAI.
What's improved
- Support Gemini 2.5 thought reasoning.
What's fixed
- User nickName & username selector in desktop.
Features
- Add claude 4 series.
What's fixed
- Fix missing email field to user, update agent config of client db will override old config.
What's improved
- Resolve InputNumber display overlap issue.
What's fixed
- 'top_p' is not supported with o4-mini, pin zustand version to avoid type error.
What's fixed
- Bump @lobehub/ui to 2.1.7.
What's improved
- Improve tools display.
What's fixed
- Fix desktop open issue on linux like Fedora42.
What's fixed
- Supported SenseNova v6 models correctly & update Gemini models.
What's improved
- Clean code with new antd api.
What's improved
- Support Doubao 1.5 Thinking Vision Pro model.
What's improved
- Update internlm model list, add series, update Spark X1 model list & fix build-in search params.
Features
- Support custom language and Mermaid Appearance.
What's fixed
- Fix oidc redirect url.
What's improved
- Refactor agent runtime to model runtime.
Features
- Add Qiniu Provider.
What's improved
- Update electron style on windows.
What's improved
- Improve smoothing on completion.
What's fixed
- Redirect unauthorized next-auth user to signin page.
What's fixed
- Fix config import issue in the desktop version.
What's fixed
- Fix streamable http url valid and refactor local files to local system.
What's fixed
- Fix window close issue and release Window/Linux beta.
What's fixed
- Fix nothing return when reset the client db.
What's improved
- Improve pdf and xlsx file content parser.
What's fixed
- Remove mcp client cache.
What's improved
- Upgrade anthropic sdk.
What's improved
- Add Qwen3 models for infiniai.
Features
- Support upload files direct into chat context.
What's improved
- Add reasoning tokens and token usage statistics for Google Gemini.
What's improved
- Add qwen3 for ollama.
What's fixed
- Fix desktop upload image on macOS.
What's fixed
- Fix changelog issue on desktop app.
What's improved
- Add new gemini & Mistral models.
What's improved
- Add Xiaohongshu crawler rules.
What's fixed
- Fix electron state init on window.
What's improved
- Remove doubao Provider.
What's improved
- Show Aliyun Bailian tokens usage tracking.
What's improved
- Fix init state of loading.
What's improved
- Add perf stat support for openai factory.
What's improved
- Add write file tool to local-file plugin.
What's fixed
- Fix desktop quiting with reopen window.
What's fixed
- Siliconflow requests with tools no longer force non-streaming.
What's improved
- Add windows control and tray.
What's improved
- Fix style issues.
What's improved
- Support Qwen3 thinking_budget param.
What's improved
- Allow copy/edit when generate error.
What's improved
- Add Gemini 2.5 Pro Experimental model, Add siliconflow Qwen3 & support thinking params, Arrange model tags neatly, Update mobile style and fix issues problem, update ollama checker.
What's fixed
- Embeddings Format Issue with OpenAI API when using Ollama.
What's fixed
- Fix stdio mcp server env issue.
What's fixed
- Fix custom provider and knowledge base crash.
What's improved
- Fix OpenRouter models config, improve mcp stdio custom field.
What's improved
- Add Qwen3 models.
What's improved
- Improve stdio mcp server install experience.
What's fixed
- Fix mcp server stdio spawn ENOENT in electron bundle, Fix mobile agent settings and textarea scroll, Remove gpt-4o-search-preview unsupported parameters.
What's improved
- Improve hotkey for delete messages.
Features
- Upgrade to lobe ui v2 with better design style.
What's fixed
- Fix cloud data sync.
What's improved
- Fix doubao-1-5-thinking-pro-vision endpoint calling.
What's fixed
- Fix desktop cloud sync issue.
What's fixed
- Fix desktop beta redirect uris.
What's fixed
- Fix cloud connection issue.
Features
- Support desktop release framework and workflow.
What's improved
- Support ernie-x1 & ernie-4.5 series from Wenxin.
What's improved
- Improve oidc OAuth workflow.
What's improved
- Improve categories selection via SearXNG.
What's fixed
- Fix oidc redirect urls.
What's fixed
- Pwa-install cause mobile infinity scroll.
What's fixed
- Improve stability of Cloudflare Workers AI.
What's improved
- Add avatar for server database upload to S3, removing SSO dependency for avatar management.
What's fixed
- Countries-and-timezones return invalid timezone.
What's fixed
- Fix hydration error.
What's fixed
- Fix openai tools calling.
What's improved
- Improve stdio mcp form, Update the id of some OpenAI models in OpenRouter.
What's improved
- Improve mcp server type select and fix refresh mcp manifest issue.
Features
- Support Streamable HTTP MCP server.
What's fixed
- Fix search prompt.
What's improved
- Show token generate performance.
What's fixed
- Together.ai fetch model list.
What's improved
- Add hotkey for clear chat messages.
What's fixed
- Lock nextauth to beta.25.
What's improved
- Support update tool calling params and trigger again.
What's fixed
- Gemini error Tool use with function calling is unsupported.
What's improved
- Update GitHub & Cloudflare models, Update Google models.
What's improved
- Allow folding SystemRole box.
What's improved
- Add SenseNova-V6 series & SenseChat-Vision support.
Features
- Support search1api as search provider.
What's improved
- Update 360 AI & Taichu & AI21 model list.
What's improved
- Add o3/o4-mini models.
What's improved
- Refactor to get user info from api.
What's fixed
- Update default file embedding model info.
What's improved
- Add step-r1 & glm-z1 model fetch support.
What's fixed
- Fix model support file and image logic.
Features
- Opening settings copyright i18n, regenerate migrate sql for new agent opening settings.
What's improved
- Opening message container add border.
What's fixed
- Opening questions should should be hidden when no setted, opening settings migration type error.
What's improved
- Make gpt-4.1-mini default model.
What's improved
- Add GPT-4.1 models.
What's fixed
- Fix Azure OpenAI unable to process image messages in local s3.
What's improved
- Refactor hotkey import to avoid db generate error.
What's fixed
- Fix grok-3-mini series calling.
What's fixed
- Fix langfuse intergation.
What's improved
- Update Grok 3 models.
What's fixed
- Remove Azure AI o3-mini unsupported parameters.
What's fixed
- Fix env for Tencent Cloud & remove deepseek-v3 series fc tag.
What's fixed
- Fix xAI API calling, not support stream_options.
Features
- Add wiki migrate instructions.
Features
- Add Keycloak SSO provider support.
What's improved
- Add time_range & categories support for SearXNG.
What's fixed
- Refactor ollama pull flow and model service.
What's improved
- Refactor the file service.
What's improved
- Improve public procedure implement.
What's improved
- Add ernie-x1-32k-preview support for Wenxin.
What's fixed
- Fix data not show correctly in 1.77.11.
What's improved
- Fix QVQ Max model, support default config for system agent and pre-merge some desktop code.
What's fixed
- Fix fetch issue in offline mode and make jina crawler first.
What's improved
- Add QVQ-Max model.
What's improved
- Auto refresh TokenTag count.
What's fixed
- Add SEARCH1API_CRAWL_API_KEY env.
What's improved
- Add desktop-release workflow and improve code.
What's improved
- Refactor the db to context inject mode.
What's improved
- Update branding.
What's improved
- Move general db models to database folder.
What's fixed
- Fix decrypt error with imported pg data.
What's fixed
- Fix export button and clean orphan agent.
Features
- Support pglite and postgres data export.
What's improved
- Supports OpenAI's latest voice model gpt-4o-mini-tts.
Features
- Add Hotkey Settings.
What's improved
- Add tencentcloud deepseek-v3-0324, support for parsing imageOutput, update models for siliconcloud & infiniai.
What's improved
- Update models info.
What's fixed
- Fix anthropic import issue.
What's fixed
- Update 'gemini-2.5-pro-exp-03-25' maxOutput and contextWindowTokens.
What's improved
- Update siliconcloud models.
Features
- Add reasoning content selector and update AutoScroll component, add Xinference provider support.
What's improved
- Add Gemini 2.5 Pro Experimental model, improve editing scroll experience.
What's improved
- Upgrade styles for Drawer.
What's improved
- Add hunyuan-t1-latest from Hunyuan.
What's improved
- Add reject pattern for browserless to boost crawl performance.
What's fixed
- Update create message loading issue.
What's fixed
- Remove Tooltip component in Topic in mobile mode.
What's fixed
- Set max_completion_tokens to undefined for Azure OpenAI.
What's fixed
- Fix wechat login error with next-auth.
What's fixed
- Upgrade next to 15.2.3 to fix CVE-2025-29927.
What's improved
- Clear previous model check result.
What's fixed
- CheckModel change clears other configs, update input and output prices.
What's improved
- Fix deepseek-r1-70b-online search tag missing from Search1API.
Features
- Add infini-ai provider, add Search1API provider with web search DeepSeek models.
What's fixed
- Fix agent chatConfig override issue.
What's improved
- Update shiki to v3.
Features
- Add Cohere provider support, add search1api crawler implementation for WeChat Sogou links.
What's fixed
- Allow historyCount to be set to 0.
Features
- Update db schema to add user_id for data export.
What's improved
- Support screenshot to clipboard when sharing.
What's improved
- Update Wenxin & Hunyuan model list.
What's fixed
- Fix claude 3.5+ models context max output.
What's fixed
- Fix knowledge base issue.
What's improved
- Update Vertex AI models.
What's fixed
- Fix google gemini output relative issue.
Features
- Support gemini image output in chat.
What's fixed
- The agent setting -edit_agent not work.
What's fixed
- Fix theme flicking.
What's fixed
- Fix crawl result for short content.
What's fixed
- Link jump in mobile terminal data statistics.
What's fixed
- Refactor the theme implement.
What's improved
- Support OpenRouter custom BaseURL.
What's fixed
- Update cvpr cvf url rules.
What's fixed
- Fix anthropic max tokens.
Features
- Support no-fc models like deepseek r1 with online search.
What's fixed
- Fix context cache control and model builtin search switch.
What's improved
- Support openrouter claude 3.7 sonnet reasoning.
What's fixed
- Fix mistral can not chat.
What's improved
- Add login ui for next-auth.
What's improved
- Refactor the agent runtime implement.
What's improved
- Add Qwen QwQ model.
Features
- Support Anthropic Context Caching.
What's improved
- Add Gemini 2.0 Flash model variations, add QwQ models.
What's fixed
- Fix litellm streaming usage and refactor the usage chunk.
What's improved
- Add epub file chunk split support.
What's improved
- Improve openrouter models info.
What's improved
- Refactor agent runtime to better code format.
What's fixed
- Fix custom ai provider sdk type.
What's improved
- Fix provider order.
What's improved
- Support to show token usages.
What's fixed
- Improve url rules.
What's improved
- Add build-in web search support for Wenxin & Hunyuan.
What's fixed
- Fix page crash with crawler error.
What's fixed
- Improve some crawl case.
Features
- Support web page crawl in the search.
What's improved
- Add gpt-4.5-preview for OpenAI.
What's improved
- Improve portal style.
What's improved
- Optimize smooth output.
What's fixed
- Fix fetch assistants plugin error.
What's fixed
- Update Claude sonnet 3.7 model ID.
What's improved
- Added eu-central-1 region for bedrock.
Features
- Add online search support for available providers.
What's improved
- Support parsing the search flag when parsing the model list, Update Gemini & Qwen models.
What's fixed
- Fix claude 3.7 sonnet thinking with tool use.
Features
- Support claude sonnet 3.7 thinking.
What's improved
- Update Gemini 2.0 search settings.
What's improved
- Add Claude 3.7 Sonnet and Haiku 3.5.
What's fixed
- Fix 0 search results with specific search engine.
What's fixed
- Disable fc for ds-v3 series.
Features
- Support application search with searchXNG.
What's fixed
- Fix citation=null issue in stream.
What's fixed
- Fix model settings config.
Features
- Support model-level search for Google/Qwen.
What's improved
- Update many models info.
What's improved
- Add custom proxyUrl support for Volcengine.
What's fixed
- Refine role assignment logic for specific Azure OpenAI models & Sensitive URL.
What's fixed
- Fix fetch on client check status display.
What's fixed
- Next-auth user id not found in create agent index.
What's fixed
- Fix image prompts with some user cases.
What's improved
- Add Volcano Ark models.
What's improved
- Refactor the plugin render style.
What's fixed
- Fix default agent loading.
What's fixed
- Fix a feature flag.
What's fixed
- Fix message roles for specific Azure OpenAI models.
What's fixed
- Add sambanova proxy url.
Features
- Support pplx search grounding.
What's fixed
- Azure AI env var configuration issue..
What's fixed
- Casdoor webhooks error.
What's improved
- Show sso providers for next-auth in profile page.
What's improved
- Improve perplexity models.
What's improved
- Improve error content and console error.
What's improved
- Add kimi-latest for Moonshot.
What's improved
- Improve serveral error code.
Features
- Support google vertex ai as a new provider.
What's fixed
- Try to fix pglite worker.
What's improved
- Sync chat limit.
What's improved
- Remove deprecated gemini models, update MiniMax models.
What's improved
- Add o1 vision metadata.
What's fixed
- Fix loading on not login for db.
What's fixed
- Fix agent config not load correctly.
What's fixed
- User feedback for empty/long group names in create/edit group modals.
What's fixed
- Fix model list issue in client mode.
What's improved
- Update Jina AI Provider name & model info.
Features
- Add SambaNova provider support.
Features
- Add volcengine as a new provider.
Features
- Add Azure AI as new Provider.
What's improved
- Fix mobile agent settings not show correctly.
Features
- Add Jina AI model provider support.
What's fixed
- Match o1 series models more robust in Azure OpenAI provider, set max_completion_tokens to null for Azure OpenAI.
What's fixed
- Fix ai provider description not show correctly.
What's improved
- Improve inbox agent settings.
What's fixed
- Fix inbox agent can not save config.
What's fixed
- Fix inbox agent edit way in the new mode.
Features
- Add configurable PDF processing method with Unstructured.
What's improved
- Improve mobile params style.
What's improved
- Add deepseek r1 distill models for qwen series.
What's fixed
- Avoid blank reasoning with OpenRouter.
What's improved
- Update openrouter model list and descriptions.
What's fixed
- Fix Azure OpenAI O1 models and refactor the Azure OpenAI implement.
Features
- Add vLLM provider support.
Features
- Add Nvidia NIM provider support.
What's improved
- Improve advanced params settings.
What's improved
- Improve model fetch behavior.
What's fixed
- Fix provider form api key.
What's fixed
- Fix api key input issue.
What's improved
- Support select check models.
What's fixed
- Fix model fetch for spark and fix the support of model reset.
What's improved
- Update model list.
What's fixed
- Fix not enable models correctly.
What's fixed
- Fix latex in thinking tag render.
What's fixed
- Fix ai model abilities issue.
What's fixed
- Fix tencent cloud api issue.
What's fixed
- Disable openrouter client fetch.
What's fixed
- Fix reasoning output for OpenRouter reasoning models like deepseek-r1.
Features
- Support tencent cloud provider.
What's improved
- Update i18n, update provider i18n.
What's improved
- Refactor the agent runtime test case.
What's improved
- Support mistral proxy url.
What's improved
- Refactor agent settings modal.
What's fixed
- Fix Aliyun deepseek-r1 reasoning parsing with oneapi, Support Aliyun deepseek-r1 reasoning.
What's fixed
- Fix language incorrect on page hydration.
What's improved
- Support Mermaid in Artifacts.
What's fixed
- Fix changelog issue.
What's improved
- Update github model list, Update openrouter model list.
What's improved
- Update Cloudflare models.
What's fixed
- Rewrite to local container in docker deployment mode.
What's improved
- Update ZeroOne models.
What's fixed
- Fix changelog modal.
What's fixed
- Fix changelog modal.
What's improved
- Update Gemini 2.0 models.
What's fixed
- Add Zhipu param limit, Fix translation in variants mode.
What's improved
- Add siliconcloud pro models.
What's fixed
- Fix static relative issues.
Features
- Refactor the auth condition in Next Auth.
What's fixed
- Fix home next auth error and update pnpm.
What's improved
- Update changelog cache and upgrade anthropic sdk.
What's fixed
- Fix next auth error.
What's fixed
- Try to fix next-auth issue.
What's fixed
- Fix /file/[id] 500 issue.
What's fixed
- Fix provider 500 issue.
What's improved
- Update edtion tag display and improve prerender.
What's improved
- Refactor model fetch method.
What's improved
- Add Aliyun deepseek-r1 distill models.
What's fixed
- Try to fix discover error.
What's improved
- Add siliconcloud models.
What's improved
- Add Cache, Metadata, FeatureFlag Viewer to DevPanel.
What's fixed
- Artifact Parsing and Rendering Bug Fix for Gemini 2.0 Flash.
What's improved
- Update model list, add reasoning tag.
Features
- Add reasoning tag support for custom models via UI or ENV.
What's fixed
- Fix deepseek-v3 & qvq model tag fetch error from SiliconCloud, fix model ability missing.
What's improved
- Add/Update Aliyun Cloud Models, update GitHub Models.
What's fixed
- Fix invalid utf8 character.
What's improved
- Update model locale.
What's fixed
- Fix o1 series calling issue.
What's fixed
- Bind the selected group name in the rename modal..
Features
- Add o3-mini support for OpenAI & GitHub Models.
What's fixed
- Fix parse of deepseek r1 in siliconflow provider.
What's improved
- Update perplexity models.
What's improved
- Update Fireworks check model and fix check error.
What's fixed
- Fix provider update issue.
What's fixed
- Optimize requests without historical messages.
What's fixed
- Fix can not stop generating.
What's fixed
- Fix ollama intergration checker and client fetch issue.
What's fixed
- Fix <think> tag crash with special markdown content.
What's improved
- Update siliconcloud models.
What's improved
- Support thinking for all non DeepSeek official api R1 models.
What's fixed
- Multiple deepseek-reasoner request errors.
What's fixed
- Support litellm reasoning streaming.
What's fixed
- Pin @clerk/[email protected] to avoid build error.
What's fixed
- Fix changelog locale not showing English.
What's fixed
- Fix discover ssr hydration error.
What's improved
- Remove use query.
What's improved
- UseMobileWorkspace use nqus to replace useQuery.
Features
- Support Doubao Models.
What's improved
- Improve thinking style.
What's improved
- Improve model pricing with CNY.
What's improved
- Add parallel_tool_calls support for Qwen, fix tag version and add provider changelog.
What's fixed
- Fix ollama Browser Request failed in PG mode.
Features
- Support display thinking for DeepSeek R1.
What's improved
- Fix model fetch match tag error & add Hunyuan model fetch support.
What's fixed
- Fix form input in provider.
What's improved
- Add HuggingFace Model: DeepSeek R1.
What's fixed
- Fix tts in new provider model.
What's improved
- Add new stepfun model.
What's fixed
- Fix debounce issue of provider config.
What's fixed
- Upgrade react-i18next to ^15.
What's improved
- Add gemini new model.
What's improved
- Improve discover model page.
What's improved
- Support model list with model fetcher settings.
What's improved
- Add ModelFetcher for supported providers.
What's improved
- Refactor [@nav](https://github.com/nav) layout and improve pin list style.
What's improved
- Improve code for ai provider.
What's improved
- Support assistant blacklist.
What's improved
- Improve error code.
What's improved
- Add deepseek r1 model.
What's improved
- Remove redundant payload remapping in client-fetch.
What's improved
- Refactor provider code.
What's improved
- Improve ai provider code.
What's fixed
- Fix hydration error.
What's fixed
- Fix api key in api key form.
Features
- Support new ai provider in client pglite.
What's fixed
- Improve validation for provider and model in parseFilesConfig, temporarily disable S3 client integrity check for Cloudflare R2.
What's fixed
- Gemini models HarmBlockThreshold.
What's improved
- Refactor some implement for the next performance improvement.
What's fixed
- Fix azure in new ai provider.
What's improved
- Add auth support for PROXY_URL.
Features
- Add lm studio provider, support to customize Embedding model with env.
What's improved
- Refactor Minimax with LobeOpenAICompatibleFactory.
What's improved
- Improve ai provider code.
What's fixed
- Fix pull models error in new ai provider.
What's improved
- Improve model config form modal.
What's fixed
- Fix enable_search parameter intro condition in Qwen.
What's fixed
- Support Gemini 2.0 HarmBlockThreshold.
What's fixed
- Fix some providers issues.
What's fixed
- Fix pin package manager to pnpm@9 for docker.
What's fixed
- Refactor dynamic import in RSC.
What's fixed
- Fix released at for undefined condition.
What's fixed
- Fix *_MODEL_LIST env in new provider.
What's fixed
- Revert officeparser.
What's fixed
- Fix GitHub and huggingface provider config unusable.
What's fixed
- Fix some ai provider known issues.
What's improved
- Update siliconcloud model list.
What's fixed
- Fix remark gfm regex breaks in Safari versions < 16.4.
Features
- Update Remark.
What's fixed
- Fix provider enabled issue.
What's fixed
- Add provider id validate.
What's fixed
- Fix model select not auto update and sort issue.
Features
- Brand new AI provider.
What's fixed
- Fix portal suspense error when first open.
What's improved
- Fix style warning in antd 5.23.0 and some error logs.
What's fixed
- Fix format short number.
What's improved
- Upgrade @clerk/nextjs to v6.
What's fixed
- Fix heatmap and manifest.
What's fixed
- Fix stats data query issue.
What's improved
- Fix zero-sized element in topic list.
What's fixed
- Fix topic mobile view ui error.
What's improved
- Refactor provider info and improve settings side bar loading.
What's improved
- Add o1 model in openai and openrouter models.
What's fixed
- Fix custom max_token not saved from customModelCards.
Features
- Add custom stream handle support for LobeOpenAICompatibleFactory.
Features
- Support white list for discover assistant.
What's improved
- Update deepseek V3 model.
What's fixed
- Fix fetch error in changelog modal.
What's improved
- Refactor tokens to contextWindowTokens.
What's fixed
- Fix o1Models list.
What's improved
- Improve loading brand.
What's improved
- Refactor sensenova provider with LobeOpenAICompatibleFactory.
What's fixed
- Fix image input on pglite.
Features
- Upgrade to next15 and react19.
Features
- Support thread in client pglite.
What's improved
- Move pglite to client service.
What's improved
- Refactor the client service to deprecated.
Features
- Support to use pglite as client db.
What's improved
- Refactor client mode upload to match server mode.
What's improved
- Add o1 model in GitHub models.
What's improved
- Add Gemini flash thinking model.
What's fixed
- Fix HUGGINGFACE endpoint url.
What's improved
- Upgrade react scan.
What's improved
- Seperate user keyVaults encrpyto from user model.
What's improved
- Refactor to use async headers().
What's improved
- Refactor layout props.
What's improved
- Improve home page loading for better UX.
What's fixed
- Fix pdf preview with capital ext.
What's fixed
- Fix GitHub model fetch.
What's improved
- Refactor the drizzle code style.
What's improved
- Refactor the data fetch with clientDB init check.
What's improved
- Improve page loading state.
What's fixed
- Fix discover locale with different default lang.
What's fixed
- Add unique keys to <ModelList> children.
What's improved
- Update models of Gitee AI provider.
What's improved
- Add new grok models.
What's improved
- Refactor file Url query in message model.
What's improved
- Support csv chunking.
What's improved
- Update locale.
What's fixed
- One of Gemini functionCall error.
What's fixed
- Fix claude first message can not be assistant.
What's improved
- Refactor the file model method.
What's improved
- Enable googleSearch Tool for gemini-2.0-flash-exp.
What's improved
- Refactor database file model to remove server env.
What's improved
- Add Gemini 2.0 Flash Exp model.
What's improved
- Update sql and types.
What's improved
- Refactor data importer to repos.
What's improved
- Clean the gpt-4-vision-preview models.
What's improved
- Refactor the clerk user service implement.
What's fixed
- Nullptr errors in NextAuth adapter.
What's improved
- Add GLM-4V-Flash from Zhipu.
What's fixed
- Fix pricing with 0 digit.
What's improved
- Update groq, add llama3.3, Upgrade lobe-ui.
What's fixed
- Fix wrong email linking in next-auth db adapter.
What's fixed
- Support request headers for chat.
What's improved
- Refactor async params route to adapt next15 breaking change.
What's improved
- Add gemini-exp-1206 model.
Features
- Add Higress ai model provider.
What's improved
- Refactor page params to adapt next15 breaking change.
What's fixed
- Fix typo of prompts.
What's improved
- Refactor the server db model implement.
What's improved
- Move schema and migration folder.
What's improved
- Add QwQ 32B Preview model.
What's improved
- Deprecated the current client mode code.
What's improved
- Add gpt-4o-2024-11-20 model.
What's improved
- Improve i18n.
What's improved
- Update ollama models.
Features
- Support ollama tools use.
What's improved
- Add QWEN_PROXY_URL support for Qwen, update model list, add qwq-32b-preview.
What's improved
- Add Google LearnLM model.
What's improved
- Add switch portal thread.
What's fixed
- Fix fallback behavior of default mode in AgentRuntime.
What's improved
- Improve thread i18n locale.
What's fixed
- Fix Qwen baseUrl calling.
Features
- Forkable Chat Mode.
What's improved
- Update the description translation of Gitee AI.
What's improved
- Refactor getLlmOptionsFromPayload from AgentRuntime.
What's fixed
- Fix fetchOnClient functional for Moonshot.
What's fixed
- Fix multi-turns tools calling.
What's improved
- Add gemini-exp-1121 model.
Features
- Add Gitee AI model provider.
What's improved
- Support to reset fetched models.
What's fixed
- Fix XAI_PROXY_URL env missing.
What's fixed
- Fix tool message display.
What's improved
- Refactor the main chat.
What's improved
- Refactor the default locale.
What's improved
- Add grok-vision-beta model, update Mistral model list, add pixtral-large-latest.
What's improved
- Support o1 models using streaming.
What's fixed
- Keyword search for chat history & sessions.
Features
- Add support InternLM (书生浦语) provider.
What's fixed
- Connection check logic.
What's improved
- Add gemini-exp-1114 model.
What's improved
- Move ChatInput to features/ChatInput.
What's improved
- genServerLLMConfig function, get *_MODEL_LIST from env.
What's improved
- Refactor the chat conversation implement.
What's improved
- Update some provider modellist & fix ai360 baseurl.
What's improved
- Fix Cloudflare Workers AI Sort.
What's improved
- Refactor languageModel & DEFAULT_LLM_CONFIG generate.
What's improved
- Update deepseek model.
What's improved
- Fix Windows always showing scrollbar.
Features
- Add support xAI provider.
Features
- Support Cloudflare Workers AI.
What's improved
- Add Sonnet 3.5 v2 inference model to BedRock.
What's fixed
- Fix summary range.
What's fixed
- Disregard remoteModelCards when showModelFetcher is disabled.
What's fixed
- Fix the display model of history summary.
What's improved
- Allow users to disable SSRF or set a whitelist.
What's fixed
- Fix topic summary field on server db.
Features
- Support compress history messages.
What's fixed
- Fix env typo of MS Entra ID.
What's fixed
- Disable model fetch for GitHub.
What's improved
- Move portal code to features folder.
What's improved
- Refactor and clean some code.
What's improved
- Update database fields.
Features
- Support export as markdown and JSON.
What's improved
- Add claude 3.5 haiku model.
What's improved
- Improve group topic styles.
What's fixed
- Remove the 'resetConversation' hot key tip.
What's fixed
- Fix /webapi/plugin/store server error.
Features
- Support group topics by time.
What's fixed
- If enable login and not signed in, return unauthorized error.
What's improved
- Disable chunking button for unsupported files.
What's fixed
- Fix duplicate key value violates unique constraint "slug_user_id_unique" when create inbox session.
What's fixed
- Fix MS Entra ID and Azure AD authorization.
What's improved
- Improve server log on chat api.
What's fixed
- Fix server Network connection lost error.
What's improved
- Refactor embedding as plain vector array.
What's improved
- Add fa-ir locale.
What's fixed
- Fix the artifacts interface not scrolling.
What's fixed
- Fix file image prompts in client mode.
What's improved
- Refactor the aiChat slice actions.
What's fixed
- Update zhipu param process.
What's fixed
- Remove PWA Install in Firefox and Arc.
What's improved
- Add Qwen2.5-72B-Instruct model on HF provider.
What's improved
- Improve user guide when user not login.
What's fixed
- Remove the 'resetConversation' hot key.
What's improved
- Remove SenseChat-Vision model, due to model limitation.
What's fixed
- Fix Huggingface API interrupting when the output exceeds 140 tokens.
What's fixed
- Fix page not switch when clicking on the pin assistant.
What's improved
- Refactor the plugin prompts to xml format.
Features
- experimentally support to pin assistant to sidebar.
What's fixed
- Fix the issue of the switch assistant portal not closing.
What's improved
- Update stepfun models.
What's fixed
- Fix modelList merge.
Features
- Support ZEN mode.
What's improved
- Update Google Model list, add gemini-1.5-flash-8b.
What's improved
- Refactor the Google Gen AI.
Features
- Add SenseNova (商汤) model provider.
What's improved
- Add *_MODEL_LIST env to all models, update Spark model id & display name.
Features
- Support system agent config.
What's improved
- Add bedrock claude-3.5-sonnect-v2.
What's improved
- Fix some custom branding detail.
What's improved
- Remove unused user tables.
What's improved
- Support plugin flag.
What's improved
- Improve error i18n.
What's improved
- Improve i18n.
What's improved
- Refactor cookie/headers to async mode.
What's improved
- Add new claude-3.5-sonnet model.
What's improved
- Move responsive to server utils folder.
What's improved
- Fix dynamic import in rsc layout.
What's fixed
- Fix azure-ad.
What's improved
- Update format utils and shared layout.
What's improved
- Update wenxin 4.0 turbo model to latest.
What's improved
- Add Ministral model, update Together AI model list, add function call & vision.
What's improved
- Add Llama 3.1 Nemotron 70B model & reorder some provider model list.
What's improved
- Refactor azure ad to ms entra id.
What's improved
- Update Fireworks AI model list.
What's improved
- Add Yi-Lightning model.
What's improved
- Add qwen vision model & update qwen2.5 72b to 128k for siliconcloud.
What's fixed
- Fix images not go in to chat context.
What's improved
- Reorder github model list & updata info & add new model.
What's improved
- Separate message slice and aiChat slice.
What's improved
- Support multi-windows for PWA.
What's fixed
- Allow use email as name in logto.
What's fixed
- Fix function calling issue, disable stream when using tools.
What's improved
- Fix artifacts render markdown.
What's improved
- Refactor agent runtime implement of stream and ZHIPU provider.
What's improved
- Refactor the jwt code.
What's improved
- Refactor the backend code for better organization.
What's improved
- Updata gpt-4o model info.
What's improved
- Update qwen vl model to latest.
What's improved
- Support yml in file chunk.
What's fixed
- Fix auto rewrite query when user message is too long.
What's improved
- Refactor text-to-image endpoint.
What's improved
- Move backend api to (backend) folder group.
What's fixed
- Fix txt-to-image api.
What's improved
- Support shadcn in Artifacts.
What's fixed
- Fix recharts deps in the Artifacts React Renderer.
What's improved
- Move most /api to /webapi.
What's improved
- Adjust Wenxin icon size.
Features
- Add wenxin model provider.
What's improved
- Update groq model list.
What's fixed
- Improve delete orphan chunks when delete files.
What's improved
- Add zhipu glm-4-flashx model.
Features
- Add Hunyuan(Tencent) model provider.
What's improved
- Add llama3.2 model for openrouter provider.
What's improved
- Add o1-preview and o1-mini model to github model provider.
What's improved
- Update google provider model info.
What's fixed
- MiniMax output long content interrupted by non-existent error.
What's improved
- Add function call for taichu_llm.
What's improved
- Add google gemini 1.5 002 series.
What's improved
- Disable taichu2.0 functioncall & default disable taichu2.0v model.
What's improved
- Update taichu provider info & add taichu vision model.
What's improved
- Add function call support for Stepfun.
What's improved
- Improve images display in chat messages.
What's improved
- Add support function call for 360AI, left sidebar has only assistants.
What's fixed
- Fix url config import after user state init.
What's fixed
- Add missing translations.
What's fixed
- Fix artifacts code language highlight.
What's improved
- Add spark max-32k model.
What's fixed
- Fix ollama model download panel.
What's improved
- Refactor to improve branding customization.
What's fixed
- Fix Content-Security-Policy.
What's fixed
- Casdoor webhooks providerAccountId not found.
What's fixed
- providerAccountId not exist in provider.
What's improved
- Improve i18n for discover and improve version check.
What's improved
- Improve i18n in discover.
What's fixed
- Pin [email protected] to fix code highlight.
What's fixed
- Try to implement better ssrf-protect.
What's improved
- Support webhooks for casdoor.
What's fixed
- Custom model initialization not taking effect error.
What's improved
- Add qwen2.5 math and coder model for siliconcloud provider.
What's fixed
- Fix a bug with server agent config when user not exist.
What's improved
- Delete siliconflow outdated model & disable stepfun functioncall.
What's improved
- Add siliconflow qwen2.5 model.
What's improved
- Refactor the tts route url.
What's improved
- Enable functioncall for stepfun models, Update qwen models.
What's improved
- Refactor the sitemap implement.
What's improved
- Updata qwen model info & add qwen2.5 & reorder provider list.
What's improved
- Add mistral provider new models.
Features
- Add Ai21Labs model provider, add Github Models provider, support native Artifacts just like Claude.
What's fixed
- Fix InterceptingRoutes in discover.
Features
- Add Discover Page.
What's improved
- Delete duplicate models in ollama.
What's fixed
- Fix a corner case of tools_call with empty object.
What's improved
- Rename artifacts to plugins in portal.
What's improved
- Add MiniCPM-V 8B model entries to Ollama model providers.
What's improved
- Update fullscreen loading style.
What's improved
- Delete "-" in deepseek displayname.
What's improved
- Update zhipu model info.
Features
- Support openai new OpenAI o1-preview/o1-mini models.
What's improved
- Support Google Model List.
What's improved
- Update siliconcloud model.
What's improved
- Remove brackets from model names with dates in OpenAI.
What's improved
- Default disable mistral provider useless models.
What's fixed
- Support webhooks for logto.
What's improved
- Support Environment Variable Inference For NextAuth.
What's fixed
- Qwen model param error.
What's improved
- Add model and provider desc and url.
What's improved
- Improve models and add more info for providers and models.
What's improved
- Optimize model token display method.
What's fixed
- Pin [email protected] to fix Internal error.
What's fixed
- Add LLM_VISION_IMAGE_USE_BASE64 to support local s3 in vision model.
What's improved
- Update Upstage model list.
What's improved
- Reorder the provider list, update spark check model to spark-lite & default disable useless model.
Features
- Add Fireworks AI Model Provider, Add Spark model provider.
What's improved
- Update CustomLogo.
What's improved
- Change empty content stream behavior.
What's fixed
- Fix /etc/resolv.confedit permission in docker image.
What's fixed
- Fix tools calling in some edge cases.
What's fixed
- Baichuan should not introduce freequency_penality parameters.
What's fixed
- Fix claude 3.5 image with s3 url.
What's fixed
- Gemini cannot input images when server database is enabled.
What's fixed
- Update baichuan param.
What's improved
- Add siliconcloud new model.
What's improved
- Update perplexity model list.
What's improved
- Fix title in about settings.
What's improved
- Improve branding implement.
What's improved
- Update model display name & Remove Qwen preview model.
What's improved
- Temperature range from 0 to 2.
What's improved
- Support anthropic browser request.
What's fixed
- Fix auth log.
What's improved
- Update Bedrock model list & add AWS_BEDROCK_MODEL_LIST support.
What's improved
- Add LLaVA 1.5 7B model in Groq.
What's fixed
- Fix typo in RAG prompt.
What's fixed
- Fix speed and rag prompt.
What's fixed
- Fix .PDF can not be chunked.
What's improved
- Fix provider disabled title style.
What's improved
- Stepfun default enabled model, update Groq model list & add GROQ_MODEL_LIST support.
What's improved
- Update Together AI model list.
What's improved
- Update Novita AI model info & add NOVITA_MODEL_LIST support.
What's improved
- Add *_MODEL_LIST for Qwen and ZeroOne, fix model info, update Claude 3.5 Sonnet maxOutput vaule.
What's improved
- Update Qwen and Gemini models info.
What's improved
- Update the sorting of each provider model.
Features
- Add Upstage model provider support.
What's improved
- Fix ms doc file preview, Update the sorting of each provider model.
What's improved
- Update Stepfun models info.
What's fixed
- Fix file relative chunks.
What's fixed
- Fix whisper-1 typo.
What's improved
- Improve chunk and file preview.
What's fixed
- Disable ChatGPT-4o Tools Calling.
What's improved
- Update Gemini models.
What's fixed
- No user name if Cloudflare Zero Trust with onetimepin.
What's improved
- Move model and provider icon components to @lobehub/icons.
What's fixed
- Improve aysnc error type.
What's fixed
- Fix agent setting.
What's improved
- Improve zhipu model config.
Features
- Supports Cloudflare Zero Trust login.
What's fixed
- Bypass vercel deployment protection, fix can send message on uploading files.
What's improved
- Update Qwen models.
Features
- Supports Authelia login.
What's fixed
- Feature flag knowledge_base doesn't affect ActionBar.
What's fixed
- Fix cannot clone agent when imported from client.
What's fixed
- Fix dayjs error in en-US language.
What's fixed
- Fix multi file upload dupicate.
What's fixed
- Session not found error on mobile.
What's fixed
- Fix tts file saving in server mode.
What's improved
- Update 01.AI models.
What's fixed
- Remove orphan chunks if there is no related file.
What's fixed
- Refactor and fix dalle.
What's fixed
- Improve s3 path-style url.
What's fixed
- Fix NEXT_PUBLIC_S3_DOMAIN error on Docker.
What's fixed
- Logout button not shown on mobile view when using nextauth.
What's improved
- Refactor s3 env and support path-style for minio.
What's fixed
- Fix clipboard copy issue and improve upload cors feedback.
What's improved
- Fix link style.
What's improved
- Hide settings in repo.
What's fixed
- Fix embeddings multi-insert when there is issues with async task.
Features
- Files and knowledge base.
What's fixed
- Fixed bedrock llama model id.
What's improved
- Update zhipu models.
What's fixed
- Fix topic scroll issue.
What's improved
- Refactor the SITE_URL to APP_URL.
What's improved
- Refactor the fetch method to fix response.undefined.
What's improved
- Add SILICONCLOUD_MODEL_LIST & SILICONCLOUD_PROXY_URL support for SiliconCloud.
What's improved
- Refactor PanelTitle and move commit from file uploading.
What's fixed
- Make S3 upload ACL setting optional.
Features
- Add 2 new models to openai provider.
Features
- Add SiliconCloud model provider.
What's improved
- Resize the image size in chat message.
What's improved
- Updated AWS bedrock model list.
What's fixed
- Fix import clerk AuthObject from public api.
What's improved
- Refactor server db schema for better code organize.
What's fixed
- Azure modelTag icon display.
Features
- Skip login page if only one provider exists.
What's fixed
- Add PROXY_URL in docker with proxychains-ng.
What's improved
- Fix aya, mathstral model tag icon & update ollama model info.
Features
- Add NextAuth as authentication service in server database.
What's improved
- Add Gemini 1.5 Pro Exp model.
What's improved
- Fix stepfun & baichuan model tag icon missing, update Perplexity models.
What's fixed
- Fix Mistral models calling & update model info.
What's improved
- Improve tools calling UI.
What's fixed
- Disable anthropic browser request.
What's improved
- Update 360GPT model (360GPT2 Pro).
What's fixed
- Fix create_session edit_agent feature flags and add more flags.
What's fixed
- Improve remote model list fetching for Novita AI.
What's fixed
- Update minimax models.
What's fixed
- Avoid baseURL being an empty string, resulting in incorrect client fetch.
What's fixed
- Fix dalle tools calling prompts to avoid content risk.
Features
- Enabled function calling on Deepseek models.
What's improved
- Fix file upload height.
What's improved
- Improve input file upload.
What's improved
- Updated Groq model list to include llama-3.1 and llama3-Groq.
What's improved
- Add new models to groq which are llama 3.1.
What's fixed
- Fix UNAUTHORIZED issue with clerk auth provider.
What's improved
- Fix the scrolling of the return result area of function calling.
What's improved
- Move server modules.
What's improved
- Add new model provider Novita AI.
What's fixed
- Content lost unexpectedly on Qwen provider when finish_reason is stop.
What's improved
- Add trpc query client with react-query.
What's improved
- Update Zhipu models (GLM-4-AllTools & CodeGeeX-4).
What's fixed
- Fix dayjs render on server.
What's improved
- Refactor the DragUpload.
Features
- Add gpt-4o-mini in OpenAI Provider and set it as the default model.
What's improved
- Added Gemma2 instead of outdated Gemma.
What's fixed
- Fix delete session group.
What's fixed
- Fix OpenAI deployment restrictions, fix cant duplicate assistant.
What's fixed
- Fix session not reorder after send message.
What's improved
- Improve brand url.
Features
- Spport qwen-vl and tool call for qwen.
What's fixed
- Save assistant info on blur.
What's improved
- Remove code related to gemini-pro-vision.
What's improved
- Add cloud promotion banner.
Features
- Add 360AI model provider.
What's improved
- Improve agent runtime code.
What's fixed
- Fix assistant meta change race issue.
What's improved
- Support disable clientFetch by default.
What's fixed
- Allow user to use their own WebRTC signaling.
What's fixed
- Automatic refresh when change language.
Features
- Add Taichu model provider.
What's improved
- Provider changes with model in model settings.
What's fixed
- Fix tool message order.
What's fixed
- Fixed mobile web page navigation issue with inbox assistant, support to disable clerk signup with feature flag.
What's improved
- Update deepseek max token.
What's fixed
- Improve tools calling error feedback when arguments are not correctly.
What's improved
- Fix tool message suspense loading.
What's fixed
- Allow builtin tools to trigger AI message.
What's improved
- Improve delete assistant message with tools.
What's fixed
- Clerk provider refreshes continously.
What's fixed
- Fix clerk appearance is not applied correctly.
What's improved
- Update ProviderAvatar for Baichuan & Stepfun.
What's improved
- Refactor the portal implement.
What's fixed
- Display issue when select default model in System Agent.
What's improved
- Refactor the portal file.
Features
- Add Baichuan model provider.
What's improved
- Add Gemini 1.5 stable version model & 2M context window.
What's improved
- Refactor to prepare for Chat Portal.
What's fixed
- Fix clerk UNAUTHORIZED error after long-time hang-up.
What's fixed
- Pin @azure/[email protected] to fix azure openai issue.
What's improved
- Refactor plugin state.
What's fixed
- -check_updates cannot be set by FEATURE_FLAGS.
What's fixed
- Fix azure tools calling.
What's improved
- Refactor format utils.
What's fixed
- Fix anthropic parallel tools calling.
What's improved
- Add AES-GCM footer and tooltip.
What's improved
- Refactor the server db implement.
What's fixed
- Fix incorrect baseURL for Groq in client mode.
What's improved
- Remove deprecated env.
What's fixed
- Create first-time user on server db.
What's fixed
- Ollama not enabled client fetch by default.
What's improved
- Fixed System Agent missing in mobile layout.
Features
- Supoort Claude 3.5 Sonnet.
What's fixed
- Fix to send image without text.
What's fixed
- Fix and improve tool calling.
What's fixed
- Fix auto avatar.
What's improved
- Fix phi3 icon display under OpenRouter.
What's fixed
- Typo in reset settings.
What's improved
- Update brand and improve docs.
What's fixed
- Fix hydration mismatch on macOS.
What's improved
- Remove some i18n text.
What's improved
- Fix footer being cut on wide screen.
What's fixed
- Fix onboard and auth for community version.
What's improved
- Add stepfun as a new provider.
What's fixed
- Fix clerk UNAUTHORIZED auth error.
What's fixed
- Openai key and openai proxy are invalid in feature flags.
What's improved
- Remove r2 cdn url.
Features
- Release LobeChat 1.0.
What's improved
- Bump version.
- Release LobeChat 1.0, closes #2897 (97bb377)
- Bump version, closes #2896 (6e2d03b)
- Update LICENSE to Apache 2.0
- update to 1.0.0