LobeHub
LobeHub
Back to Home
Changelog

Lobe UI v2 Design System and Desktop App Launch

LobeHub launches a refreshed visual design with Lobe UI v2 and officially releases the desktop app for Windows and macOS.

Lobe UI v2 Design System and Desktop App Launch

April brings two major changes you can see and touch: a complete visual refresh powered by the new Lobe UI v2 design system, and the official release of the LobeHub desktop app. Together they make daily usage feel more modern, more fluid, and better suited to desktop workflows.

A refreshed visual foundation

The interface now runs on Lobe UI v2. Every component has been reconsidered for clarity and consistency. Buttons, typography, spacing, and color palettes align across the entire application.

In practice, this means less visual noise and more predictable interactions. The sidebar is cleaner. Model tags are arranged more neatly. System roles can be collapsed to save space. Mobile layouts have been refined for better thumb reachability.

Desktop app with native integration

LobeHub is now available as a native desktop application for Windows and macOS. It includes system tray integration, native window controls, and keyboard shortcuts that feel at home on your operating system.

The desktop app matters most when you:

  • Want LobeHub available without keeping a browser tab open
  • Prefer native window management and alt-tabbing
  • Need system-level shortcuts for common actions
  • Work across multiple workspaces and want the app docked consistently

MCP, search, and SSO improvements

MCP protocol support has been enhanced with Streamable HTTP server support. Stdio MCP server installation is smoother, and environment variable parameters are now supported for more flexible server configurations.

Search adds Search1API as a new provider, and SearXNG now supports category and time-range selection for more targeted results.

SSO integration expands with Keycloak support, and the OIDC OAuth workflow has been refined for fewer friction points during authentication.

Model library updates

Coverage has expanded to keep provider evaluation and migration straightforward:

  • OpenAI: GPT-4.1 series, o3/o4-mini
  • Google: Gemini 2.5 Pro Experimental with reasoning token tracking
  • xAI: Grok 3 series
  • Anthropic, Mistral, Qwen, Ollama: latest model refreshes

Improvements and fixes

  • Performance stats: token generation throughput display, Aliyun Bailian token tracking, Google Gemini reasoning token statistics
  • Hotkey enhancements: shortcuts for clearing chat and deleting messages, plus full custom hotkey configuration
  • Tool calling: update tool-call arguments and re-trigger; local file plugins now support writing files
  • Error handling: copy and edit options available when errors occur
  • Web crawling: added crawler rules for Xiaohongshu

Credits

Huge thanks to these contributors:

@arvinxx @hezhijie0327 @Innei @nekomeowww @lobehub-team

Version details

Features

  • v1.84.0Upgrade to lobe ui v2 with better design style.
  • v1.83.0Support desktop release framework and workflow.
  • v1.82.0Support Streamable HTTP MCP server.
  • v1.81.0Support search1api as search provider.
  • v1.80.0Opening settings copyright i18n, regenerate migrate sql for new agent opening settings.
  • v1.79.0Add wiki migrate instructions.
  • v1.78.0Add Keycloak SSO provider support.

What's improved

  • v1.84.27Add reasoning tokens and token usage statistics for Google Gemini.
  • v1.84.26Add qwen3 for ollama.
  • v1.84.23Add new gemini & Mistral models.
  • v1.84.22Add Xiaohongshu crawler rules.
  • v1.84.21Remove doubao Provider.
  • v1.84.20Show Aliyun Bailian tokens usage tracking.
  • v1.84.19Fix init state of loading.
  • v1.84.18Add perf stat support for openai factory.
  • v1.84.17Add write file tool to local-file plugin.
  • v1.84.14Add windows control and tray.
  • v1.84.13Fix style issues.
  • v1.84.12Support Qwen3 thinking_budget param.
  • v1.84.11Allow copy/edit when generate error.
  • v1.84.10Add 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.
  • v1.84.6Fix OpenRouter models config, improve mcp stdio custom field.
  • v1.84.5Add Qwen3 models.
  • v1.84.4Improve stdio mcp server install experience.
  • v1.84.1Improve hotkey for delete messages.
  • v1.83.7Fix doubao-1-5-thinking-pro-vision endpoint calling.
  • v1.82.10Support ernie-x1 & ernie-4.5 series from Wenxin.
  • v1.82.9Improve oidc OAuth workflow.
  • v1.82.8Improve categories selection via SearXNG.
  • v1.82.5Add avatar for server database upload to S3, removing SSO dependency for avatar management.
  • v1.82.2Improve stdio mcp form, Update the id of some OpenAI models in OpenRouter.
  • v1.82.1Improve mcp server type select and fix refresh mcp manifest issue.
  • v1.81.8Show token generate performance.
  • v1.81.6Add hotkey for clear chat messages.
  • v1.81.4Support update tool calling params and trigger again.
  • v1.81.3Update GitHub & Cloudflare models, Update Google models.
  • v1.81.2Allow folding SystemRole box.
  • v1.81.1Add SenseNova-V6 series & SenseChat-Vision support.
  • v1.80.5Update 360 AI & Taichu & AI21 model list.
  • v1.80.4Add o3/o4-mini models.
  • v1.80.3Refactor to get user info from api.
  • v1.80.1Add step-r1 & glm-z1 model fetch support.
  • v1.80.0Opening message container add border.
  • v1.79.10Make gpt-4.1-mini default model.
  • v1.79.9Add GPT-4.1 models.
  • v1.79.7Refactor hotkey import to avoid db generate error.
  • v1.79.4Update Grok 3 models.
  • v1.77.18Add time_range & categories support for SearXNG.

What's fixed

  • v1.84.25Fix desktop upload image on macOS.
  • v1.84.24Fix changelog issue on desktop app.
  • v1.84.22Fix electron state init on window.
  • v1.84.16Fix desktop quiting with reopen window.
  • v1.84.15Siliconflow requests with tools no longer force non-streaming.
  • v1.84.9Embeddings Format Issue with OpenAI API when using Ollama.
  • v1.84.8Fix stdio mcp server env issue.
  • v1.84.7Fix custom provider and knowledge base crash.
  • v1.84.3Fix mcp server stdio spawn ENOENT in electron bundle, Fix mobile agent settings and textarea scroll, Remove gpt-4o-search-preview unsupported parameters.
  • v1.83.8Fix cloud data sync.
  • v1.83.6Fix desktop cloud sync issue.
  • v1.83.5Fix desktop beta redirect uris.
  • v1.83.4Fix cloud connection issue.
  • v1.82.8Fix oidc redirect urls.
  • v1.82.7Pwa-install cause mobile infinity scroll.
  • v1.82.6Improve stability of Cloudflare Workers AI.
  • v1.82.5Countries-and-timezones return invalid timezone.
  • v1.82.4Fix hydration error.
  • v1.82.3Fix openai tools calling.
  • v1.81.9Fix search prompt.
  • v1.81.7Together.ai fetch model list.
  • v1.81.5Lock nextauth to beta.25.
  • v1.81.4Gemini error Tool use with function calling is unsupported.
  • v1.80.2Update default file embedding model info.
  • v1.80.1Fix model support file and image logic.
  • v1.80.0Opening questions should should be hidden when no setted, opening settings migration type error.
  • v1.79.8Fix Azure OpenAI unable to process image messages in local s3.
  • v1.79.6Fix grok-3-mini series calling.
  • v1.79.5Fix langfuse intergation.
  • v1.79.3Remove Azure AI o3-mini unsupported parameters.
  • v1.79.2Fix env for Tencent Cloud & remove deepseek-v3 series fc tag.
  • v1.79.1Fix xAI API calling, not support stream_options.
  • v1.77.17Refactor ollama pull flow and model service.
LobeHub

Super-Agents on Your Team

Bring Super-Agents onto your team, starting today.