
LobeHub 在二月完成了史上最大规模的 AI 生态扩展。目标很简单:在不破坏现有工作流程、无需额外配置的前提下,让你能接入更多服务商和更强的模型。
服务商与推理能力扩展
本次新增 10+ 个主流服务商,覆盖全球与国内平台。你可以直接从现有的 LobeHub 设置中连接到更广泛的服务。
DeepSeek R1 现已完整支持,推理模型兼容性也扩展到 Claude 3.7 Sonnet 和 OpenAI o3-mini。这些模型会实时展示思维链,让你可以跟进结论的推导过程。跨服务商场景下 DeepSeek R1 的解析表现更加一致,推理内容易于阅读。
搜索能力重构
在线搜索已升级,集成 SearchXNG 和 Perplexity,支持深度网页爬取。Gemini 2.0 和 Qwen 系列模型现可在推理过程中使用原生搜索能力。
利用这些更新,你可以:
- 在对话中直接通过实时网页结果进行主题研究
- 对比不同服务商如何处理同一搜索查询
- 无需离开对话即可获取更深度的内容提取
模型库刷新
本次更新 50+ 个模型配置,保持你的选择与时俱进:
- OpenAI:gpt-4.5-preview
- Anthropic:Claude 3.7 Sonnet、Haiku 3.5
- Google:Gemini 2.0 系列优化
- 国内服务商:月之暗面、通义千问、MiniMax 最新模型
- 其他平台:Perplexity、Cloudflare、硅基流动等刷新
体验优化与修复
- 修复跨服务商的推理展示不一致问题
- 改进搜索结果格式,提升可读性
- 增强模型参数跨服务商同步
Credits
感谢以下贡献者:
版本详情
Features
- v1.67.0Support web page crawl in the search.
- v1.66.0Add online search support for available providers.
- v1.65.0Support claude sonnet 3.7 thinking.
- v1.64.0Support application search with searchXNG.
- v1.63.0Support model-level search for Google/Qwen.
- v1.62.0Support pplx search grounding.
- v1.61.0Support google vertex ai as a new provider.
- v1.60.0Add SambaNova provider support.
- v1.59.0Add volcengine as a new provider.
- v1.58.0Add Azure AI as new Provider.
- v1.57.0Add Jina AI model provider support.
- v1.56.0Add configurable PDF processing method with Unstructured.
- v1.55.0Add vLLM provider support.
- v1.54.0Add Nvidia NIM provider support.
- v1.53.0Support tencent cloud provider.
- v1.52.0Refactor the auth condition in Next Auth.
- v1.51.0Add reasoning tag support for custom models via UI or ENV.
- v1.50.0Add o3-mini support for OpenAI & GitHub Models.
What's improved
- v1.66.6Add gpt-4.5-preview for OpenAI.
- v1.66.5Improve portal style.
- v1.66.4Optimize smooth output.
- v1.66.1Added eu-central-1 region for bedrock.
- v1.65.2Support parsing the search flag when parsing the model list, Update Gemini & Qwen models.
- v1.65.0Update Gemini 2.0 search settings.
- v1.64.3Add Claude 3.7 Sonnet and Haiku 3.5.
- v1.63.1Improve plugin calling style.
- v1.63.0Update many models info.
- v1.62.11Add custom proxyUrl support for Volcengine.
- v1.62.7Add Volcano Ark models.
- v1.62.6Refactor the plugin render style.
- v1.62.4Add some error types.
- v1.61.5Show sso providers for next-auth in profile page.
- v1.61.4Improve perplexity models.
- v1.61.3Improve error content and console error.
- v1.61.2Add kimi-latest for Moonshot.
- v1.61.1Improve serveral error code.
- v1.60.8Sync chat limit.
- v1.60.7Remove deprecated gemini models, update MiniMax models.
- v1.60.6Add o1 vision metadata.
- v1.60.1Update Jina AI Provider name & model info.
- v1.57.1Fix mobile agent settings not show correctly.
- v1.56.3Improve inbox agent settings.
- v1.55.4Improve mobile params style.
- v1.55.3Add deepseek r1 distill models for qwen series.
- v1.55.1Update openrouter model list and descriptions.
- v1.54.0Improve advanced params settings.
- v1.53.12Improve model fetch behavior.
- v1.53.9Support select check models.
- v1.53.7Update model list.
- v1.53.0Update i18n, update provider i18n.
- v1.52.19Refactor the agent runtime test case.
- v1.52.16Support mistral proxy url.
- v1.52.15Optimized MaxToken Slider.
- v1.52.14Refactor agent settings modal.
- v1.52.11Support Mermaid in Artifacts.
- v1.52.8Update github model list, Update openrouter model list.
- v1.52.7Update Cloudflare models.
- v1.52.6Update ZeroOne models.
- v1.52.3Update Gemini 2.0 models.
- v1.52.2Add siliconcloud pro models.
- v1.51.14Update changelog cache and upgrade anthropic sdk.
- v1.51.9Update edtion tag display and improve prerender.
- v1.51.8Refactor model fetch method.
- v1.51.7Add Aliyun deepseek-r1 distill models.
- v1.51.5Add siliconcloud models.
- v1.51.3Add Cache, Metadata, FeatureFlag Viewer to DevPanel.
- v1.51.2Update model list, add reasoning tag.
- v1.50.5Add/Update Aliyun Cloud Models, update GitHub Models.
- v1.50.3Update model locale.
- v1.49.16Update perplexity models.
- v1.49.15Update Fireworks check model and fix check error.
What's fixed
- v1.67.2Improve some crawl case.
- v1.66.3Fix fetch assistants plugin error.
- v1.66.2Update Claude sonnet 3.7 model ID.
- v1.65.1Fix claude 3.7 sonnet thinking with tool use.
- v1.64.2Fix 0 search results with specific search engine.
- v1.64.1Disable fc for ds-v3 series.
- v1.63.3Fix citation=null issue in stream.
- v1.63.2Fix model settings config.
- v1.63.1Fix groq location request.
- v1.62.11Refine role assignment logic for specific Azure OpenAI models & Sensitive URL.
- v1.62.10Fix fetch on client check status display.
- v1.62.9Next-auth user id not found in create agent index.
- v1.62.8Fix image prompts with some user cases.
- v1.62.5Fix default agent loading.
- v1.62.4Fix hotkeys of open agent settings.
- v1.62.3Fix a feature flag.
- v1.62.2Fix message roles for specific Azure OpenAI models.
- v1.62.1Add sambanova proxy url.
- v1.62.0Azure AI env var configuration issue..
- v1.61.6Casdoor webhooks error.
- v1.61.0Try to fix pglite worker.
- v1.60.5Fix loading on not login for db.
- v1.60.4Fix agent config not load correctly.
- v1.60.3User feedback for empty/long group names in create/edit group modals.
- v1.60.2Fix model list issue in client mode.
- v1.56.5Match o1 series models more robust in Azure OpenAI provider, set max_completion_tokens to null for Azure OpenAI.
- v1.56.4Fix ai provider description not show correctly.
- v1.56.2Fix inbox agent can not save config.
- v1.56.1Fix inbox agent edit way in the new mode.
- v1.55.2Avoid blank reasoning with OpenRouter.
- v1.55.1Fix Azure OpenAI O1 models and refactor the Azure OpenAI implement.
- v1.53.11Fix provider form api key.
- v1.53.10Fix api key input issue.
- v1.53.8Fix model fetch for spark and fix the support of model reset.
- v1.53.6Fix not enable models correctly.
- v1.53.5Fix latex in thinking tag render.
- v1.53.4Fix ai model abilities issue.
- v1.53.3Fix tencent cloud api issue.
- v1.53.2Disable openrouter client fetch.
- v1.53.1Fix reasoning output for OpenRouter reasoning models like deepseek-r1.
- v1.52.15Fix lmstudio baseURL.
- v1.52.13Fix Aliyun deepseek-r1 reasoning parsing with oneapi, Support Aliyun deepseek-r1 reasoning.
- v1.52.12Fix language incorrect on page hydration.
- v1.52.9Fix changelog issue.
- v1.52.7Rewrite to local container in docker deployment mode.
- v1.52.5Fix changelog modal.
- v1.52.4Fix changelog modal.
- v1.52.3Add Zhipu param limit, Fix translation in variants mode.
- v1.52.1Fix static relative issues.
- v1.51.15Fix home next auth error and update pnpm.
- v1.51.13Fix next auth error.
- v1.51.12Try to fix next-auth issue.
- v1.51.11Fix /file/id] 500 issue.
- v1.51.10Fix provider 500 issue.
- v1.51.6Try to fix discover error.
- v1.51.3Artifact Parsing and Rendering Bug Fix for Gemini 2.0 Flash.
- v1.51.0Fix deepseek-v3 & qvq model tag fetch error from SiliconCloud, fix model ability missing.
- v1.50.4Fix invalid utf8 character.
- v1.50.2Fix o1 series calling issue.
- v1.50.1Bind the selected group name in the rename modal..
- v1.50.0Fix parse of deepseek r1 in siliconflow provider.
- v1.49.14Fix provider update issue.
- v1.49.13Optimize requests without historical messages.
