LobeHub
LobeHub
返回
更新日志

快捷键自定义、数据导出与服务商扩展

LobeHub 新增快捷键自定义、数据导出功能,并扩展服务商支持,让日常使用更顺手、数据更可迁移。

快捷键自定义、数据导出与服务商扩展

三月的版本聚焦于日常体验的打磨。快捷键现在可以自定义,按你的习惯设置。数据导出让对话记录可以随身携带。新增服务商和优化的推理展示进一步完善了这次更新。

符合习惯的键盘快捷键

快捷键现已支持自定义。你可以将常用操作映射到自己觉得顺手的按键上,而不必去适应预设的键位布局。

同时新增了清除对话和删除消息的快捷键。如果你偏好键盘驱动的工作流,这些新增快捷键能减少在快节奏会话中伸手去够鼠标的次数。

数据导出用于备份与迁移

PgLite 和 PostgreSQL 后端现已支持数据导出。你可以备份对话记录,或在不同安装环境之间迁移数据,无需手动操作数据库。

这在以下场景尤其有用:

  • 归档重要的对话历史
  • 将设置迁移到新设备或服务器
  • 在云端存储之外保留本地备份

服务商与推理支持扩展

六个新服务商加入:Xinference、Cohere、Search1API、Infini-AI、PPIO。每个都集成到现有的服务商选择流程中,无需额外配置步骤。

推理模型现在拥有专用的内容选择器,Claude 3.7 和 DeepSeek R1 的思维链展示也得到了优化。推理内容在长对话中更易阅读和定位。

更完善的网页爬取能力

网页爬取现在能处理更多内容类型,特别支持 YouTube、Reddit 和微信公众号链接。短内容提取已优化,Search1API 提供了额外的爬虫实现以获得更好的覆盖。

模型库更新

本次更新带来多个值得关注的新增模型:

  • Google Gemini 2.5 Pro Experimental 和 Gemini 2.0 Flash 系列
  • Anthropic 上下文缓存支持
  • OpenAI gpt-4o-mini-tts 语音模型
  • DeepSeek-V3-0324
  • Hunyuan-T1-Latest
  • QwQ、QVQ-Max
  • 百度文心 ernie-x1-32k-preview

体验优化与修复

  • 界面改进:重构 Drawer 样式、优化编辑器滚动、支持截图分享至剪贴板
  • 搜索增强:在线搜索现支持非函数调用模型(包括 DeepSeek R1);为文心和混元增加内置网页搜索
  • 文件处理:新增 EPUB 分块支持,优化 PDF 处理
  • 性能提升:重构 Agent Runtime,优化数据库核心代码,修复 LiteLLM 流式统计
  • 稳定性修复:解决主题闪烁、知识库行为异常、微信登录等问题

Credits

感谢以下贡献者:

@AmAzing129 @arvinxx @hezhijie0327 @tjx666 @lobehub-team

版本详情

Features

  • v1.77.0Support pglite and postgres data export.
  • v1.76.0Add Hotkey Settings.
  • v1.75.0Add reasoning content selector and update AutoScroll component, add Xinference provider support.
  • v1.74.0Add infini-ai provider, add Search1API provider with web search DeepSeek models.
  • v1.73.0Add Cohere provider support, add search1api crawler implementation for WeChat Sogou links.
  • v1.72.0Update db schema to add user_id for data export.
  • v1.71.0Support gemini image output in chat.
  • v1.70.0Support no-fc models like deepseek r1 with online search.
  • v1.69.0Support Anthropic Context Caching.
  • v1.68.0Add new model provider PPIO.

What's improved

  • v1.77.16Refactor the file service.
  • v1.77.15Improve public procedure implement.
  • v1.77.14Add ernie-x1-32k-preview support for Wenxin.
  • v1.77.12Fix QVQ Max model, support default config for system agent and pre-merge some desktop code.
  • v1.77.9Add QVQ-Max model.
  • v1.77.8Auto refresh TokenTag count.
  • v1.77.7Add desktop-release workflow and improve code.
  • v1.77.6Refactor the db to context inject mode.
  • v1.77.4Update branding.
  • v1.77.3Move general db models to database folder.
  • v1.76.1Supports OpenAI's latest voice model gpt-4o-mini-tts.
  • v1.75.5Add tencentcloud deepseek-v3-0324, support for parsing imageOutput, update models for siliconcloud & infiniai.
  • v1.75.4Update models info.
  • v1.75.1Update siliconcloud models.
  • v1.75.0Add Gemini 2.5 Pro Experimental model, improve editing scroll experience.
  • v1.74.11Upgrade styles for Drawer.
  • v1.74.10Add hunyuan-t1-latest from Hunyuan.
  • v1.74.9Add reject pattern for browserless to boost crawl performance.
  • v1.74.3Clear previous model check result.
  • v1.74.1Fix deepseek-r1-70b-online search tag missing from Search1API.
  • v1.73.1Update shiki to v3.
  • v1.71.5Support screenshot to clipboard when sharing.
  • v1.71.4Update Wenxin & Hunyuan model list.
  • v1.71.1Update Vertex AI models.
  • v1.70.4Support OpenRouter custom BaseURL.
  • v1.69.5Support openrouter claude 3.7 sonnet reasoning.
  • v1.69.3Add login ui for next-auth.
  • v1.69.2Refactor the agent runtime implement.
  • v1.69.1Add Qwen QwQ model.
  • v1.68.11Add Gemini 2.0 Flash model variations, add QwQ models.
  • v1.68.9Add epub file chunk split support.
  • v1.68.8Improve openrouter models info.
  • v1.68.7Refactor agent runtime to better code format.
  • v1.68.5Fix provider order.
  • v1.68.4Support to show token usages.
  • v1.68.2Add build-in web search support for Wenxin & Hunyuan.

What's fixed

  • v1.77.17Refactor ollama pull flow and model service.
  • v1.77.13Fix data not show correctly in 1.77.11.
  • v1.77.11Fix fetch issue in offline mode and make jina crawler first.
  • v1.77.8Add SEARCH1API_CRAWL_API_KEY env.
  • v1.77.2Fix decrypt error with imported pg data.
  • v1.77.1Fix export button and clean orphan agent.
  • v1.75.3Fix anthropic import issue.
  • v1.75.2Update 'gemini-2.5-pro-exp-03-25' maxOutput and contextWindowTokens.
  • v1.74.8Update create message loading issue.
  • v1.74.7Remove Tooltip component in Topic in mobile mode.
  • v1.74.6Set max_completion_tokens to undefined for Azure OpenAI.
  • v1.74.5Fix wechat login error with next-auth.
  • v1.74.4Upgrade next to 15.2.3 to fix CVE-2025-29927.
  • v1.74.2CheckModel change clears other configs, update input and output prices.
  • v1.73.2Fix agent chatConfig override issue.
  • v1.72.1Allow historyCount to be set to 0.
  • v1.71.3Fix claude 3.5+ models context max output.
  • v1.71.2Fix knowledge base issue.
  • v1.71.1Fix google gemini output relative issue.
  • v1.70.10The agent setting -edit_agent not work.
  • v1.70.8Fix theme flicking.
  • v1.70.7Fix crawl result for short content.
  • v1.70.6Link jump in mobile terminal data statistics.
  • v1.70.5Refactor the theme implement.
  • v1.70.2Update cvpr cvf url rules.
  • v1.70.1Fix anthropic max tokens.
  • v1.69.6Fix context cache control and model builtin search switch.
  • v1.69.4Fix mistral can not chat.
  • v1.68.10Fix litellm streaming usage and refactor the usage chunk.
  • v1.68.6Fix custom ai provider sdk type.
  • v1.68.3Improve url rules.
  • v1.68.1Fix page crash with crawler error.
  • v1.68.0Fix search web-browsing display bug.
  • v1.67.2Improve some crawl case.
LobeHub

Super-Agents on Your Team

今天就把 Super-Agents 招进你的团队。