Repository navigation
Changelog
- Redesigned AI Chat: the title shows the AI service and mode, and tapping it opens a sheet to switch the service or choose Chat or Q&A. Messages have a cleaner look in light and dark themes, a new chat shows your first prompt templates, and the message box has a round send button that turns into Stop while a reply is written. Prompt templates open in a sheet that marks the ones that ask for input. A Jump to latest button appears when you scroll up during a reply.
- Added a LiteRT-LM AI service for running
.litertlmmodels (such as Gemma and Qwen) on the phone. Choose it as the provider when adding an AI service, pick the model file and, optionally, let it use the GPU; it falls back to the CPU when the phone or model doesn't support the GPU. It works in AI Chat and for AI prompts. It needs a 64-bit ARM phone. - Added Ask about a Photo: point the camera at something (or choose a photo), optionally type or say a question, and an on-device AI model answers as your language teacher and makes a word card in the language you are learning, with an example sentence. Speak the word, look it up in your dictionaries, add it to a word book with the photo, or ask follow-up questions. Choose your language, the language to learn and one of four teacher personas (Buddy, Nerdy, Secret Agent, Wizard), whose prompts you can edit. It needs a LiteRT-LM AI service with a model that takes images, such as Gemma 4 E2B or E4B. Open it from the home menu, or from the new icon in the search box, which shows once you have a LiteRT-LM service.
- You can download LiteRT-LM models, including Gemma 4 E2B and E4B for Ask about a Photo, from Hugging Face: https://huggingface.co/collections/litert-community/gemma-family
- Redesigned the AI Service and AI Prompt lists and edit pages in the style of the display font screens: numbered rows with each provider's icon, a details sheet with Edit, Move to top and Delete, drag handles in reorder mode, and card-style edit pages with help text and checks before saving.
- Fixed a crash when sorting the dictionaries of a large group by name.
更新日志
- 重新设计了“AI 对话”:标题显示当前 AI 服务和模式,点按标题可切换服务或选择“对话”“问答”模式。消息在浅色和深色主题下都更清爽,新对话会显示前几个提示词模板,输入框改用圆形发送按钮,生成回复时变为“停止”。提示词模板改在底部面板中打开,并标出需要输入的模板。回复过程中向上滚动时会出现“跳到最新”按钮
- 新增 LiteRT-LM AI 服务,可在手机上运行
.litertlm模型(如 Gemma、Qwen)。添加 AI 服务时选择它作为服务提供方,选择模型文件,并可选择使用 GPU;手机或模型不支持 GPU 时会自动改用 CPU。可用于 AI 对话和 AI 提示词。需要 64 位 ARM 手机 - 新增“看图学单词”:用相机对准一样东西(或选择一张照片),可先输入或说出问题,本机上的 AI 模型会以语言老师的身份回答,并用你正在学的语言生成带例句的单词卡。可朗读单词、在你的词典中查找、连同照片加入单词本,也可继续追问。可选择你的语言、要学的语言以及四种老师角色之一(小伙伴、学霸、特工、巫师),角色的提示词可以编辑。需要一个模型能读图片的 LiteRT-LM AI 服务,例如 Gemma 4 E2B 或 E4B。可从首页菜单打开;添加 LiteRT-LM 服务后,搜索框中也会出现新图标
- LiteRT-LM 模型(包括“看图学单词”所需的 Gemma 4 E2B 和 E4B)可从 https://huggingface.co/collections/litert-community/gemma-family 下载
- 重新设计了“AI 服务”和“AI 提示词”列表及编辑页面,与“显示字体”界面风格一致:带序号和服务提供方图标的列表、包含编辑、移到顶部和删除的详情面板、排序模式下的拖动手柄,以及带说明文字和保存前检查的卡片式编辑页面
- 修复了对大型词典分组按名称排序时闪退的问题
繁體
更新日誌
- 重新設計了「AI 對話」:標題顯示目前的 AI 服務和模式,輕觸標題可切換服務或選擇「對話」「問答」模式。訊息在淺色和深色主題下都更清爽,新對話會顯示前幾個提示詞範本,輸入框改用圓形傳送按鈕,產生回覆時變為「停止」。提示詞範本改在底部面板中開啟,並標出需要輸入的範本。回覆過程中向上捲動時會出現「跳到最新」按鈕。
- 新增 LiteRT-LM AI 服務,可在手機上執行
.litertlm模型(如 Gemma、Qwen)。新增 AI 服務時選擇它作為服務提供者,選擇模型檔案,並可選擇使用 GPU;手機或模型不支援 GPU 時會自動改用 CPU。可用於 AI 對話和 AI 提示詞。需要 64 位元 ARM 手機。 - 新增「看圖學單字」:用相機對準一樣東西(或選擇一張照片),可先輸入或說出問題,本機上的 AI 模型會以語言老師的身分回答,並用你正在學的語言產生附例句的單字卡。可朗讀單字、在你的字典中查詢、連同照片加入單詞本,也可繼續追問。可選擇你的語言、要學的語言以及四種老師角色之一(小夥伴、學霸、特務、巫師),角色的提示詞可以編輯。需要一個模型能讀圖片的 LiteRT-LM AI 服務,例如 Gemma 4 E2B 或 E4B。可從首頁選單開啟;新增 LiteRT-LM 服務後,搜尋框中也會出現新圖示。
- LiteRT-LM 模型(包括「看圖學單字」所需的 Gemma 4 E2B 和 E4B)可從 https://huggingface.co/collections/litert-community/gemma-family 下載。
- 重新設計了「AI 服務」和「AI 提示詞」列表及編輯頁面,與「顯示字體」介面風格一致:附序號和服務提供者圖示的列表、包含編輯、移到頂部和刪除的詳情面板、排序模式下的拖動把手,以及附說明文字和儲存前檢查的卡片式編輯頁面。
- 修復了對大型詞典分組按名稱排序時閃退的問題。
Release notes: https://github.com/Jimex/DictTango-Android/blob/main/ReleaseNotes.md