LobeHub
LobeHub
返回
更新日志

Gemini 图像生成与非流式模式支持 🎨

Gemini 2.5 Flash Image 图像生成、非流式响应模式,以及扩展的模型覆盖,让内容生成和接收方式更灵活。

Gemini 图像生成与非流式模式支持 🎨

八月到九月,LobeHub 围绕 Gemini 补齐了图像生成、响应模式和模型覆盖。目标很直接:让新能力在真实使用场景更易接入,同时保持服务商选择的开放性。

更多图像生成方式

本次新增 Gemini 2.5 Flash Image(Nano Banana)、Imagen 4 GA 支持,同时扩展 Nebius、AkashChat、BFL 等服务商。Azure OpenAI 也可处理图像生成请求。

全新的非流式响应模式适合需要完整响应后再展示的场景 —— 比如批处理、结构化输出,或期望同步结果的集成场景。

可用于:

  • 直接在对话中通过 Gemini 模型生成图像
  • 按使用场景在流式与完整响应模式间切换
  • 将图像请求路由到新上线的服务商

模型库更新

模型库新增 GPT-5 系列、Claude Opus 4.1、Grok Code Fast 1、DeepSeek V3.1,以及用于检索增强的 Gemini URL Context Tool。

体验优化与修复

  • 优化推理滚动遮罩效果
  • 新增会话切换快捷键
  • 改进移动端表单控件
  • 完善 Gemini 错误提示信息

版本详情

Features

  • v1.120.0Rename Gemini 2.5 flash image to Nano Banana.
  • v1.119.0Added support for Azure OpenAI Image Generation.
  • v1.118.0Add new provider AkashChat.
  • v1.117.0Ai image support Gemini 2.5 Flash Image, Support Gemini 2.5 Flash Image Preview in OpenRouter.
  • v1.117.0Ai image support Gemini 2.5 Flash Image.
  • v1.117.0Ai image support Gemini 2.5 Flash Image.
  • v1.116.0Add gemini 2.5 flash image for vertex ai.
  • v1.111.0Add GPT-5 series models.
  • v1.110.0Support mcp plugin install from web.

What's improved

  • v1.120.7Update i18n.
  • v1.120.6Add upload hint for non-visual model.
  • v1.120.4Adjust ControlsForm component to adapt to mobile phone display.
  • v1.120.3Support new provider Nebius.
  • v1.120.2Remove base path.
  • v1.120.1Update i18n.
  • v1.120.0Remove webrtc sync feature flag.
  • v1.118.8Update DeepSeek V3.1 & Gemini 2.5 Flash Image Preview models.
  • v1.118.7Update i18n.
  • v1.118.6Support non-stream mode.
  • v1.118.5Fix clerk scrollBox style, ModelFetcher support getting prices.
  • v1.118.4Fix chat session part switch theme issue.
  • v1.118.2Add Grok Code Fast 1 model.
  • v1.118.1Support Gemini URL context tool.
  • v1.117.1Move chat item into chat.
  • v1.117.0Update i18n.
  • v1.117.0Update i18n.
  • v1.116.4Support html preview.
  • v1.114.5Update mistral model vision ability.
  • v1.114.4Move database to packages.
  • v1.114.1Move chain into @lobechat/prompts.
  • v1.112.4Refactor prompts folder to the @lobechat/prompts pacakge.
  • v1.112.3Refactor const folder to a new package.
  • v1.112.2重构 ArgsInput 组件.
  • v1.112.1Add Imagen 4 GA models, style improve auth sign in box loading.
  • v1.111.11Update Mistral AI models & Optimize many model providers fetching.
  • v1.111.10Adjust near bottom size on thinking scroll.
  • v1.111.9Improve Gemini error display with promptFeedback.
  • v1.111.8Support new GPT-5 Verbosity params.
  • v1.111.7Refactor model-runtime to a seperated package.
  • v1.111.5Add mask effect to thinking scroll, update i18n.
  • v1.111.3Improve thinking auto scroll style, Support session switch shortcut key, update i18n.
  • v1.111.1Add descriptions for the FLUX.1 Krea and Qwen Image.
  • v1.110.4Refactor trace type.
  • v1.110.3Fix provider setting page hydration error.
  • v1.110.2Add context menu for desktop, support different model tabs.
  • v1.109.1Add Claude Opus 4.1 model, update i18n.

What's fixed

  • v1.119.1Update enableStreaming name.
  • v1.118.3Correct totalOutputTokens calculation for XAI provider.
  • v1.118.1Add Content-Security-Policy env.
  • v1.116.3Fix desktop route error.
  • v1.114.2Can't load custom provider config.
  • v1.113.3Support Grok thinking models in AiHubMix, The 'stream_options' parameter is only allowed when 'stream' is enabled.
  • v1.112.5Improve mcp tracing with user config.
  • v1.111.6Solve the cache problem caused by the same dom id when sharing pictures.
  • v1.110.7Missing languages it-IT, pl-PL, nl-NL.
  • v1.110.5Optimize Gemini error message display & Filter empty messages.
  • v1.110.2Fix fail to fetch aihubmix model on client mode.
  • v1.110.1Fix remote avatar broken in desktop again.
  • v1.109.1Fix ollama model output without thinking.
LobeHub

Super-Agents on Your Team

今天就把 Super-Agents 招进你的团队。