
From May through June, LobeHub focused on making repeated workflows easier to run. Prompt variables let you reuse prompt structures with different inputs. Claude 4 support brings the latest reasoning capabilities with integrated web search. Expanded crawling and file handling round out the update.
Reusable prompts with variables
You can now use placeholder variables in prompts and the input box. The same prompt structure can be reused with different dynamic values, reducing manual editing when you run similar tasks repeatedly.
In practice, this works well for:
- Templates that need a name, date, or topic inserted each time
- Structured analysis where only the source material changes
- Repeated workflows with consistent formatting but variable content
Variables are defined with simple syntax and filled in at runtime, so you spend less time retyping and more time getting results.
Claude 4 with web search and beta features
Anthropic Claude 4 reasoning models are now fully integrated. The implementation includes Web Search tool support and Beta Header compatibility, so Claude 4 can use the same reasoning and retrieval capabilities that power other advanced workflows.
This means you can:
- Run complex reasoning tasks with transparent chain-of-thought
- Combine reasoning with live web retrieval in a single conversation
- Access the latest Claude capabilities as they become available
Broader provider and content support
ModelScope joins the provider lineup, expanding the range of models available without changing how you work. Search and crawling support has been extended across more platforms for better content coverage.
File uploads now work directly in chat context, with improved parsing for PDF and XLSX files. This makes it easier to reference documents and spreadsheets within ongoing conversations.
Access control and security improvements
Protected page access has been added for sensitive workflows. Clerk middleware route protection has been improved to handle authentication edge cases more reliably.
Improvements and fixes
- Enhanced PDF and XLSX content extraction accuracy
- Improved file upload handling in chat context
- Better route protection for authenticated pages
- Provider-specific optimizations for Claude 4 token handling
Credits
Huge thanks to these contributors:
Version details
Features
- v1.93.0Support OpenAI Responses API mode.
- v1.92.0Support placeholder variables in prompts and input.
- v1.91.0Add more provider support for search & crawl.
- v1.90.0Support protect page.
- v1.89.0Support ModelScope Provider.
- v1.88.0Add claude 4 series.
- v1.87.0Support custom language and Mermaid Appearance.
- v1.86.0Add Qiniu Provider.
- v1.85.0Support upload files direct into chat context.
What's improved
- v1.93.3Refactor branding info.
- v1.93.2Refactor ` & ` handling.
- v1.92.3Support OpenRouter Claude 4 reasoning.
- v1.92.2Add support to azureopenai embedding.
- v1.92.1Improve {{username}} placeholder variable, Update Gemini & Qwen models.
- v1.92.0Support Vertex AI thought summaries.
- v1.91.3Add openAI websearch and claude 4 to modelproviders.
- v1.91.2Add Volcengine & OpenAI-like Provider (e.g. oneapi) model fetch support.
- v1.91.1Improve loading state.
- v1.91.0Update modelscope models.
- v1.90.4Rename the createOpenAICompatibleRuntime.
- v1.90.3Enable deploymentName for Aliyun Bailian.
- v1.90.2Support web_search tool for MiniMax & Zhipu.
- v1.90.1Use default deployment name when parseModelString doesn't contain deployment name.
- v1.88.23+ in the welcome message can be clicked to create an assistant.
- v1.88.22Support 01.ai proxy url.
- v1.88.21Enhanced reasoning_effort Slider Component, Update Hunyuan models & deepseek-r1-0528.
- v1.88.19Support Web Search Tools and Beta Header from Anthropic.
- v1.88.17Increase the history limit.
- v1.88.16Update GitHub models.
- v1.88.15Add fc ability to deepseek-reasoner model.
- v1.88.12Support adjust thinkingBudget in gemini 2.5 flash.
- v1.88.11Correct model name SenseChat-5-1202.
- v1.88.8Allow SliderWithInput to have no input limit.
- v1.88.6Fix a few typos in the model tooltips.
- v1.88.5Support share single message.
- v1.88.4Improve thread flicker when first-time loading.
- v1.88.3Add gemini & hunyuan & Claude models.
- v1.88.2Add live search support for xAI.
- v1.88.1Support Gemini 2.5 thought reasoning.
- v1.87.9Resolve InputNumber display overlap issue.
- v1.87.6Improve tools display.
- v1.87.3Clean code with new antd api.
- v1.87.2Support Doubao 1.5 Thinking Vision Pro model.
- v1.87.1Update internlm model list, add series, update Spark X1 model list & fix build-in search params.
- v1.86.1Refactor agent runtime to model runtime.
- v1.85.10Update electron style on windows.
- v1.85.9Improve smoothing on completion.
- v1.85.3Improve pdf and xlsx file content parser.
- v1.85.2Upgrade anthropic sdk.
- v1.85.1Add Qwen3 models for infiniai.
- v1.84.27Add reasoning tokens and token usage statistics for Google Gemini.
What's fixed
- v1.93.1Fix openai default Responses API issue.
- v1.92.3Fix client s3 getObject throw error.
- v1.92.0Some web search bugs.
- v1.91.3Correct deepseek R1 fc support display.
- v1.90.4Update the clerk middleware to support route protection.
- v1.90.1Disable LaTeX and Mermaid rendering in SystemRoleContent to prevent lag caused by massive rendering tasks when switching topics, fix DeepSeek new R1 Search error.
- v1.88.20Agent automatic completion meta not working error.
- v1.88.18Close historySummary correctly, Enable thinking output only for supported Gemini thinking models.
- v1.88.16Cmd + click chat tab not open new tab.
- v1.88.14Cant invoke the application after OIDC authorization in Windows 11.
- v1.88.9Auto sync theme mode in desktop.
- v1.88.7Fix chat header in the desktop.
- v1.88.6Fix draggable issue with agent header.
- v1.88.3Fix message refresh 401 on desktop.
- v1.88.1User nickName & username selector in desktop.
- v1.88.0Fix missing email field to user, update agent config of client db will override old config.
- v1.87.8'top_p' is not supported with o4-mini, pin zustand version to avoid type error.
- v1.87.7Bump @lobehub/ui to 2.1.7.
- v1.87.5Fix desktop open issue on linux like Fedora42.
- v1.87.4Supported SenseNova v6 models correctly & update Gemini models.
- v1.87.0Fix oidc redirect url.
- v1.85.9Redirect unauthorized next-auth user to signin page.
- v1.85.8Fix config import issue in the desktop version.
- v1.85.7Fix streamable http url valid and refactor local files to local system.
- v1.85.5Fix window close issue and release Window/Linux beta.
- v1.85.4Fix nothing return when reset the client db.
- v1.85.3Remove mcp client cache.
