本次更新重点围绕插件生态展开,增加插件市场、更多的接口,同时增加qqbot对接等插件示范,并增加软件的稳定性。

增加

• 新增计划模式,计划准备阶段不会修改文件,在确定需求以及用户批准后,进行计划实现
• 升级世界书,支持国际化、变量系统、导入、搜索、刷新、角色卡绑定、注入位置设置和更完整的管理界面
• 新增 QQ Bot,可以直接在qq中发消息并接收到Operit的执行结果
• 增强外部调用能力,现在可以在同一局域网里面,通过其他设备的浏览器访问Operit聊天界面了
• 聊天输入增加@以及/来引用文件、工具包的能力
• 聊天历史支持消息定位预览、按时间戳跳转、消息收藏以及更稳定的长历史浏览
• 作品市场升级为项目/版本节点结构,支持输入需求,一键创建自己的插件,支持基于插件市场的插件二次修改,并重新发布作为该插件的新版本。
• ToolPkg 支持工作流模板、工作区模板、桌面小组件、WebView UI、聊天输入 Hook、聊天视图 Hook、Prompt Hook、XML 渲染插件和跨运行时 IPC
• Compose DSL 大幅增强,支持 slots、图片缩放、文件选择、WebView 控制器、重新渲染、文本输入同步和更多 Material 图标
• 新增 4Router、Mimo 等模型/服务入口,并支持 ToolPkg 注册 AI Provider,增加对deepseek v4的思考适配
• 文件工具新增更清晰的 create_file / edit_file,zip 工具支持严格的 include_root_directory 参数
• GitHub 登录支持外部 OAuth 流程
• 新增多个侧边栏和网页示例,包括账户本、模型站点、OpenCode、SillyTavern、Dino Runner 等,已发布于插件市场
• 聊天消息现在可以显示更完整的 token、耗时和消息时间信息,方便排查模型调用表现

修复

• 修复 various_search 视频搜索可能返回首页或无效链接的问题,现在会尽量返回真实目标链接
• 修复大聊天历史导入/导出容易占用过多内存的问题,聊天备份改用流式处理
• 修复长聊天记录下消息定位、滚动导航、插入总结和变体加载不稳定的问题
• 修复 Compose DSL 文本框聚焦、外部值更新和重新渲染时可能不同步的问题
• 修复市场页面返回后已安装状态、列表数据和 MCP 插件状态不刷新的问题
• 修复工具结果过长时直接挤占上下文的问题,最终工具结果会更克制地截断
• 修复 zip 与调试器文件工具中的路径引用和根目录打包行为问题
• 修复未绑定工作区统计不准确的问题,内部/外部工作区会分开识别
• 修复浏览器工具在 iframe、点击、会话上下文和动作 settle 场景中的若干不稳定问题
• 修复 TTS 播放,增加朗读的暂停/继续能力
• 修复工作区的刷新逻辑,现在预览html只会在文件有改动的时候刷新

优化

• 删除任务完成和等待标记,记忆总结改为按照数量和时间累计总结,删除错误的"记忆附着"行为说明
• 市场浏览、详情、管理、安装和发布流程更统一,优化市场体验
• 绘图类插件迁移到各自的插件存储目录,减少不同插件之间的文件互相影响
• 新增OpenAI Local,优化Llamacpp、mnn的本地工具调用方式
• 部分内置的包已删除并发布到市场,有需要可以自行下载
• 优化软件侧边栏,优化软件LOGO

This update focuses mainly on the plugin ecosystem. It adds the plugin marketplace, expands plugin interfaces, includes examples such as QQ Bot integration, and improves overall app stability.

Added

• Added Plan Mode. Files are no longer modified during planning; implementation starts only after the requirements are confirmed and approved by the user.
• Upgraded Worldbook with internationalization, a variable system, import, search, refresh, character-card binding, injection-position settings, and a more complete management interface.
• Added QQ Bot, allowing messages to be sent directly in QQ and Operit's execution results to be received there.
• Improved external access. Operit's chat interface can now be opened from a browser on another device in the same local network.
• Chat input now supports using @ and / to reference files and tool packages.
• Chat history now supports message-location previews, timestamp jumps, message favorites, and more stable browsing for long histories.
• Upgraded the Artifact Market to a project/version-node structure. Users can enter requirements to create their own plugins with one click, modify plugins from the marketplace, and republish them as new versions of those plugins.
• ToolPkg now supports workflow templates, workspace templates, desktop widgets, WebView UI, chat-input hooks, chat-view hooks, Prompt hooks, XML rendering plugins, and cross-runtime IPC.
• Compose DSL has been greatly enhanced with support for slots, image scaling, file selection, WebView controllers, rerendering, text-input synchronization, and more Material icons.
• Added model/service entries such as 4Router and Mimo. ToolPkg can now register AI Providers, and thinking adaptation for DeepSeek V4 has been added.
• File tools now include clearer create_file / edit_file tools, and the zip tool now supports a strict include_root_directory parameter.
• GitHub login now supports an external OAuth flow.
• Added multiple sidebar and web examples, including Account Book, Model Sites, OpenCode, SillyTavern, and Dino Runner. These have been published to the plugin marketplace.
• Chat messages can now show more complete token, duration, and message-time information, making it easier to inspect model-call behavior.

Fixed

• Fixed an issue where various_search video search could return homepages or invalid links. It now tries to return the actual target links.
• Fixed excessive memory usage when importing or exporting large chat histories by changing chat backup to streaming processing.
• Fixed instability in long chat histories, including message locating, scroll navigation, summary insertion, and variant loading.
• Fixed Compose DSL text boxes sometimes falling out of sync during focus changes, external value updates, or rerendering.
• Fixed marketplace pages not refreshing installed status, list data, and MCP plugin status after navigating back.
• Fixed oversized tool results taking up too much context. Final tool results are now truncated more conservatively.
• Fixed path references and root-directory packaging behavior in zip and debugger file tools.
• Fixed inaccurate statistics for unbound workspaces. Internal and external workspaces are now recognized separately.
• Fixed several browser-tool stability issues involving iframes, clicks, session context, and action settling.
• Fixed TTS playback and added pause/resume support for read-aloud.
• Fixed workspace refresh logic. HTML previews now refresh only when files change.

Optimized

• Removed task-completion and waiting markers. Memory summaries are now generated based on accumulated quantity and time, and the incorrect description of "memory attachment" behavior has been removed.
• Unified the marketplace browsing, detail, management, installation, and publishing flows to improve the overall marketplace experience.
• Drawing plugins now use their own plugin storage directories, reducing file interference between different plugins.
• Added OpenAI Local, and optimized local tool-calling for llama.cpp and MNN.
• Some built-in packages have been removed and published to the marketplace. They can be downloaded there when needed.
• Optimized the app sidebar and app logo.

What's Changed

• feat: add Skill selector menu to chat input 输入框的加号菜单skills选项 by @feikukuai in #524
• 为consume hooks 添加 clearInput 选项 by @Yinglianchun in #538

New Contributors

• @feikukuai made their first contribution in #524
• @Yinglianchun made their first contribution in #538

Full Changelog: v1.10.1...v1.11.0

Contributors

• @feikukuai
• @Yinglianchun

新增更完整的内置浏览器与浏览器脚本支持、主题液态玻璃效果与侧边栏自定义、FBX 虚拟形象与升级后的 MMD 预览、配置编辑器调试和自动编写 Operit 插件、工作区重命名、历史跳转与双向分页,以及更多模型、语音、绘图和工作流能力。

增加

• 支持 FBX 虚拟形象,并升级了 MMD 预览体验
• 支持通过本地 HTTP 接口从外部发起对话,提供健康检查、流式回调,并可按需拉起悬浮聊天
• 模型配置现在支持单独设置自定义请求头,接入特殊网关或第三方服务更方便
• 角色卡可以限制可用工具,角色卡和角色组也可以拥有各自独立的自定义表情
• 系统分享文件或链接时,可以直接选择发到当前会话、最近会话或新会话
• 内置浏览器能力大幅增强,支持标签页、历史、书签、权限申请、多窗口、最小化和视口控制
• 浏览器脚本支持导入、安装、启停、存储和页面菜单,网页自动化和增强能力更完整
• 支持利用配置编辑器调试和自动编写 Operit 插件
• 新增 PDF 逐页识图、App 使用时长读取,以及应用冷启动时自动触发工作流
• 发消息前可自动补充更多上下文,包括当前前台应用、当前页面 Activity、近期 App 使用时长、屏幕 OCR 文本等
• 现在会自动读取工作区根目录规则文件,让项目专属约束更容易生效
• 支持工作区重命名,并会同步更新关联聊天标题
• 聊天历史支持快速滚动、消息跳转定位和双向分页浏览
• 支持 GIF、HTML 换行和更稳的公式渲染,复杂消息显示效果更自然
• 主题新增液态玻璃效果,聊天气泡、输入栏和侧边栏都支持更多外观自定义
• 侧边栏支持独立背景色、强调色、液态玻璃按钮和展开动画设置
• 新增更多模型和服务接入,包括 Novita、OpenRouter、MiniMax 语音/绘图,以及 Kimi Coding 端点
• Waifu 模式支持自定义附加提示词,也可以自定义回复提示音和震动
• 新增更多插件示例,方便扩展上下文裁剪、聊天调试等高级玩法

修复

• 修复工作区 @ 文件补全卡顿、候选太乱的问题,搜索更快也更准
• 修复工具权限判断不准确的问题,package_proxy 权限检查和角色卡工具限制更可靠
• 修复语音测试报错信息不清楚的问题,网络错误、超时和 HTTP 失败更容易排查
• 修复 HTTP TTS 对 JSON 返回、多跳请求和 base64 音频结果的兼容问题
• 修复 SSH 长输出和 tmux 失败导致的崩溃,超长结果现在会更稳地保存到文件
• 修复聊天历史中的滚动定位、上下文跳转、工作流空参数、图片解析和图片上传卡住等问题
• 修复 GIF 和文件导入时的扩展名识别问题,以及公式遮挡、重叠和 Markdown 换行不一致的问题
• 修复 token 估算污染缓存状态、统计不准等问题,相关使用统计和缓存命中记录更完整
• 修复 MCP 和远程服务配置崩溃、脏配置残留、插件冲突来源难定位等问题
• 修复消息时间重复时可能导致的排序和跳转异常
• 修复从聊天界面改动记忆后,Memory 页面返回不刷新的问题
• 修复占位目录影响记忆检索权重,以及超长网页内容直接塞进上下文的问题
• 修复群聊、角色卡续聊、总结后自动续聊、实时渲染等多处稳定性问题

优化

• 对话历史和工具调用链路做了统一整理,复杂对话、工具调用和长上下文续聊场景更稳定
• 本地模型的工具调用链路继续打磨,Llama 和 MNN 的行为更加一致
• 消息现在可以记录更完整的模型、token 和耗时信息,单条消息详情更容易查看
• 插件扩展能力继续完善,第三方功能更容易接入,也更容易拿到当前会话和调用来源信息
• 深度搜索现在可以并发处理更多可执行任务,中间过程和总结输出也更清晰
• 内置浏览器和网页工具继续打磨,复杂页面交互、局部快照、缩放和会话状态更稳定
• 记忆系统加强了标签检索、重建进度展示和聊天界面同步
• 包管理器扫描更快,能直接查看冲突来源,也支持删除外部冲突源文件
• 大型文件修改结果和超长网页结果会更克制地返回给模型,减少无意义的上下文占用
• 聊天渲染、浮动输入栏、历史过滤、批量删除聊天记录和整体浏览体验继续优化,整体更顺滑、可定制性更高
• 原 all_about_myself 工具包正式更名为 operit_editor

It adds a more complete in-app browser with browser-script support, liquid-glass theme effects and sidebar customization, FBX virtual avatars with an upgraded MMD preview experience, support for debugging and automatically writing Operit plugins with the configuration editor, workspace renaming, history jumping and bidirectional paging, plus more model, voice, image and workflow options.

Added

• Added FBX virtual avatar support and upgraded the MMD preview experience
• Added a local HTTP chat entry point with health check, streaming callback support and optional floating-chat launch
• Model configs can now define their own extra request headers
• Character cards can now restrict available tools, and character cards/groups can keep their own custom emoji sets
• Shared files and links can now be sent to the current chat, a recent chat or a new chat
• Greatly expanded the in-app browser with tabs, history, bookmarks, permissions, multi-window support, minimization and viewport control
• Added browser-script import, installation, storage, runtime and page-menu support for richer web automation
• Added support for debugging and automatically writing Operit plugins with the configuration editor
• Added page-by-page PDF vision parsing, app-usage-time reading and cold-start workflow triggers
• Expanded automatic message context injection with current app info, current activity info, recent app usage and on-screen OCR text
• Workspace root rule files can now be picked up automatically so project-specific rules can take effect more easily
• Added workspace renaming, with linked chat titles updated together
• Added fast chat-history scrolling, message jumping and bidirectional history paging
• Added GIF support, HTML line breaks and more reliable formula rendering for richer message display
• Added liquid-glass theme effects and more appearance options for bubbles, input bars and the sidebar
• The sidebar now supports its own colors, highlight styles and animation options
• Added more model and service choices including Novita, OpenRouter, MiniMax voice/image options and the Kimi Coding endpoint
• Added editable Waifu extra prompts plus custom reply sound/vibration settings
• Added more plugin examples for context limiting, chat debugging and similar advanced scenarios

Fixed

• Fixed laggy and noisy workspace @ file suggestions so they are faster and more accurate
• Fixed inaccurate tool-permission checks, making package-proxy permission checks and character tool restrictions more reliable
• Fixed unclear TTS test errors so network failures, timeouts and HTTP errors are easier to diagnose
• Fixed HTTP TTS compatibility for JSON responses, chained requests and base64 audio payloads
• Fixed SSH long-output and tmux failures so large results no longer crash the tool flow and can be saved to files more safely
• Fixed multiple chat-history problems including scroll positioning, context jumps, empty workflow params, image parsing and stuck uploads
• Fixed GIF/file extension detection plus formula overlap/covering problems and inconsistent Markdown line-break rendering
• Fixed token estimation polluting cache state and improved token statistics and cache-hit tracking
• Fixed MCP/remote-service config crashes, invalid config leftovers and hard-to-locate plugin conflict sources
• Fixed message ordering and jump issues caused by duplicate timestamps
• Fixed the Memory page not refreshing after changes made from the chat interface
• Fixed placeholder folders affecting memory search ranking and oversized web-page content being pushed directly into context
• Fixed multiple stability problems across group chat, role-card continuation, post-summary continuation and live rendering

Optimized

• Unified the conversation and tool-call flow so complex chats, tool use and long-context continuation behave more reliably
• Continued improving local model tool calls so Llama and MNN behave more consistently
• Messages can now retain clearer model, token and timing details, making per-message diagnostics easier
• Continued improving plugin extensibility so third-party features can integrate more cleanly and understand more chat context
• Deep search can now run more ready tasks in parallel, with clearer intermediate output and summaries
• Further refined the in-app browser and web tools for complex-page interaction, partial snapshots, scaling and session handling
• Improved memory search, rebuild progress display and chat-side synchronization
• Package scanning is faster, package conflict sources are clearer, and external conflict files can be removed directly from the UI
• Long file diffs and oversized web-page results are handled more carefully to reduce unnecessary context bloat
• Continued polishing chat rendering, floating input, history filtering, batch chat deletion and overall browsing smoothness
• The former all_about_myself package has been officially renamed to operit_editor

What's Changed

• feat: add Novita AI as LLM provider by @Alex-wuhu in #425
• 功能:修复nanobanana插件的轮询机制 by @profitevillager in #441
• fix: sync memory UI updates from chat interface by @MimicHunterZ in #449
• feat: add context_limiter_c and debug_msg_dump toolpkg examples by @HateCandy in #466

New Contributors

• @Alex-wuhu made their first contribution in #425
• @profitevillager made their first contribution in #441
• @MimicHunterZ made their first contribution in #449
• @HateCandy made their first contribution in #466

Full Changelog: v1.10.0...v1.10.1

Contributors

• @Alex-yang00
• @profitevillager
• @HateCandy
• @MimicHunterZ

本次更新增加了主题的自定义程度,增加了液态玻璃、自定义气泡字体等效果,增加了角色卡群聊功能,优化了agent执行的任务稳定性,减少中断和波动。同时扩大toolpkg软件插件的开放接口,修复了若干的bug。

增加

• 增加群组功能,可以将多个角色卡拉到一个群组对话,同时支持@功能
• 增加ai自我设置的功能,能够协助用户配置软件mcp、skill、STT、TTS、模型配置等
• 增加分组折叠消息,能够折叠部分工具和全部工具
• 增加ai chat工具包的重命名和删除工具
• 增加长按app图标进入软件设置和语音模式
• 增加ollama供应商
• 增加nvidia供应商
• 增加单独的SSH插件工具包,支持使用tmux在服务器不间断运行任务,能够在工具箱中查看任务状态
• 为js引擎增加Java Bridge功能
• 增加拓展http工具的忽视ssl选项
• 增加回车发送选项,可以在显示与行为里面开启
• 增加nanobanana的模型选择
• 增加气泡式对话主题,自定义气泡背景,支持.9
• 支持气泡式下对ai和用户消息的字体和颜色的分别调整
• 增加语音悬浮窗中关闭朗读的选项
• 将深度搜索功能从内置独立到插件,同时toolpkg格式大改
• 增加apktool插件
• 支持update memory工具对tag进行修改
• 支持气泡模式下的新显示样式: 更宽的气泡
• 增加web自动化操作的文件下载能力
• 增加markdown的音视频渲染能力
• 增加命令式下的用户输入以及输入框的液态玻璃效果,可以在主题里开启
• 将助手形象正式投入使用,开启后可以在语音模式下显示,增加mp4的虚拟形象支持
• 增加xai工具包的视频生成功能
• 增加工作流运行时的取消按钮
• 增加终端自定义按键
• 增加手动解终端chroot挂载功能
• 增加了消息队列功能
• 增加openai response 通用模式供应商

修复

• 修复OpenAI兼容模式下对于glm5丢失内容的bug
• 修复autoglm的覆盖状态条遮挡点击的问题
• 修复语音识别模块的崩溃bug
• 修复记忆获取工具并发闪退的bug
• 修复上传记忆pdf节点的卡顿
• 修复悬浮窗模式下,唤起输入法后点击别的区域无法取消焦点的问题
• 修复Shizuku root启动的uid判定问题
• 包管理扫描现在不会因为一个包报错导致其他的异常了
• 修复在低版本上终端的tab显示异常的bug
• 修复文件编辑工具导致的行首空格丢失的问题
• 修复悬浮窗下点击ai生成的图片的报错
• 修复气泡模式下消息过长时后半段的内容无法进行交互的问题
• 修复隐藏状态栏下状态栏变黑的bug
• 修复web自动化时启动web导致的全屏无法交互问题
• 修复mnn的toolcall

优化

• 增强记忆召回能力
• 增强skill和mcp市场的搜索功能
• 思考显示现在的效果更好,支持思考内容里面md渲染
• 优化工作区的安卓模板出现的网络和aapt问题,升级模板的gradle和kotlin版本,同时增强工作区的按钮点击后的反馈效果
• 优化唤起输入法的卡顿问题
• 优化grep code和grep context工具,现在会有明确的timeout,同时提高了搜索代码的性能10倍以上
• 优化警告显示的ui反馈为"ai犯了一个错误"
• 现在点击启动工作区会有一个短暂的加载避免卡顿
• 优化ai对于终端是否运行现在会有更明确的感知了
• 增强ai的重试逻辑,现在的agent执行会更加稳定,不会因为工具重试导致bug
• 现在ai在输出过长导致截断时,会自动有明确的警告,不会断开任务
• 现在ai在用户离开软件进入后台后,不会马上断开连接了
• 增强mnn和llamacpp对于qwen 3.5的支持

This update enhances theme customization, adding effects like liquid glass and custom bubble fonts. It introduces character card group chat functionality, improves agent task stability with fewer interruptions and fluctuations, expands the open interfaces for toolpkg software plugins, and fixes several bugs.

Added

• Added group functionality to bring multiple character cards into a group conversation, with @mention support
• Added AI self‐configuration capability to help users set up software MCP, skills, STT, TTS, model configurations, etc.
• Added grouped/collapsible messages, allowing partial or full tool outputs to be folded
• Added rename and delete tools in the AI chat tool package
• Added long‐press on app icon to enter software settings and voice mode
• Added Ollama provider
• Added NVIDIA provider
• Added standalone SSH plugin tool package supporting tmux for uninterrupted task execution on servers, with task status view in the toolbox
• Added Java Bridge functionality for the JavaScript engine
• Added ignore‐SSL option for extended HTTP tools
• Added Enter‐to‐send option (can be enabled in Display & Behavior)
• Added model selection for nanobanana
• Added bubble‐style conversation theme with custom bubble backgrounds (supports .9‐patch)
• Support separate font and color adjustments for AI and user messages in bubble mode
• Added option to turn off speech in the voice floating window
• Separated deep search from built‐in to a plugin, with major format changes for toolpkg
• Added APK tool plugin
• Support modifying tags with the update memory tool
• Added new display style in bubble mode: wider bubbles
• Added file download capability for web automation operations
• Added audio/video rendering support in Markdown
• Added user input in command mode and liquid glass effect for input box (can be enabled in themes)
• Officially deployed assistant avatars; when enabled, they appear in voice mode; added MP4 virtual avatar support
• Added video generation functionality in xAI tool package
• Added cancel button during workflow execution
• Added custom terminal keys
• Added manual terminal chroot mount/unmount function
• Added message queue functionality
• Added OpenAI‐response universal‐mode provider

Fixed

• Fixed bug of content loss with glm5 in OpenAI‐compatible mode
• Fixed issue where autoglm overlay status bar blocked clicks
• Fixed crash bug in speech recognition module
• Fixed concurrency crash bug in memory retrieval tool
• Fixed lag when uploading PDF nodes to memory
• Fixed issue in floating window mode where tapping elsewhere after invoking the keyboard did not dismiss the focus
• Fixed UID determination issue for Shizuku/root startup
• Package management scanning no longer fails due to errors from a single package
• Fixed abnormal tab display in terminals on lower Android versions
• Fixed loss of leading spaces in file editing tool
• Fixed error when clicking AI‐generated images in floating window mode
• Fixed issue in bubble mode where interactive elements in the latter part of overly long messages were unresponsive
• Fixed black status bar bug when status bar was hidden
• Fixed full‐screen interaction issue when web automation started a web view
• Fixed mnn tool call

Optimized

• Enhanced memory recall capability
• Enhanced search functionality in skill and MCP marketplace
• Improved thought display rendering, now supporting Markdown rendering within thought content
• Optimized Android template in workspace to resolve network and aapt issues; upgraded template's Gradle and Kotlin versions; enhanced button‐click feedback in workspace
• Optimized keyboard invocation to reduce lag
• Optimized grep‐code and grep‐context tools with explicit timeouts, improving code search performance by over ×ばつ
• Changed warning UI feedback to "AI made a mistake"
• Added brief loading when starting a workspace to prevent perceived lag
• AI now has clearer awareness of whether the terminal is running
• Enhanced AI retry logic, making agent execution more stable and less prone to bugs from tool retries
• AI now shows explicit warnings when output is truncated due to length, preventing task disconnection
• AI no longer immediately disconnects when the user leaves the app and switches to background
• Enhanced mnn and llama.cpp support for Qwen 3.5

What's Changed

• fix: 修复后台点击卡顿及跨设备窗口解析兼容性问题 by @chrysoljq in #414

New Contributors

• @chrysoljq made their first contribution in #414

Full Changelog: v1.9.1...v1.10.0

Contributors

• @chrysoljq

本次更新主要修复了1.9.0下的诸多bug。

• 增强终端工具,修复终端被交互ui卡住的情况
• 修复严格工具调用下,历史工具导致请求依旧报错
• 修复远程mcp无法关闭的问题
• 移除sleep工具的时间最大限制
• 修复多种情况下的ai输出工具截断/重复错误
• 增加未绑定角色卡的聊天记录的删除
• 增加工作流的批量删除
• 增加工作流的执行日志查看
• 重做记忆库的写入逻辑,支持外接向量模型,增加连接修改工具
• 增加浏览器自动化的上传逻辑
• 修复一些输入ime问题
• 修复黑夜模式下的新输入框显示问题
• 修复新输入框的菜单说明消失的问题
• 修复界面导航bug,增加流畅性
• 修复主题的无法配置背景透明度问题
• 修复进入工具箱后整个包管理错乱问题
• 修复windows控制器下提示raw命令无法执行以及连接引导问题.

This update primarily fixes numerous bugs in version 1.9.0.

• Enhanced the terminal tool and fixed the issue where the terminal was stuck due to interactive UI.
• Fixed the issue where historical tools still caused request errors under strict tool invocation.
• Fixed the issue where remote MCP could not be closed.
• Removed the maximum time limit for the sleep tool.
• Fixed various cases of AI output being truncated or repeated incorrectly.
• Added the ability to delete chat records not bound to character cards.
• Added batch deletion for workflows.
• Added execution log viewing for workflows.
• Reworked the memory bank writing logic, added support for external vector models, and included tools for modifying connections.
• Added upload logic for browser automation.
• Fixed various input method (IME) issues.
• Fixed display issues with the new input box in dark mode.
• Fixed the issue where menu instructions for the new input box disappeared.
• Fixed interface navigation bugs and improved smoothness.
• Fixed the issue where background transparency could not be configured in themes.
• Fixed the package management disorder issue after entering the toolbox.
• Fixed the issue where raw commands could not be executed under the Windows controller and improved connection guidance.

Full Changelog: v1.9.0...v1.9.1

本次更新修复了一些bug,主要增加了移动端网页自动操作功能,以及Windows的终端操作功能,能够控制codex之类的cli。

新的输入框不习惯可以在主题里面改回老输入框。

增加

增加工具箱sql查看器
新增工作区Web项目的CORS绕过,能够访问外部网页
新增工作区的安卓模板
新增OpenAI的response兼容供应商
支持自定义su
新增网页操作
增加主界面的角色卡排序查看
增加更多的内置提示词禁用选项
增加内置工具的禁用选项
新增导出web项目打包apk时的icon剪切编辑
新增严格工具调用模式,作为toolcall在部分模型上bug的补充
增加skill的直接输入添加
增加windows命令操作功能
升级包系统,增加Toolpkg并由js动态增加工具箱界面
全新的输入框,可以通过主题里面改回老的显示形式
增加统计界面的饼图

修复

修复1.8.1下文件通过识图模型读取的bug
修复上下文总结的时候内容带图片的报错问题
修复openai兼容下的工具输出带特殊符号截断问题
修复在小米手机上主聊天的autoglm无法使用问题
修复绘图包的绘图超时问题
修复ffmpeg调用出现的部分问题
修复历史记录划出出现的闪烁问题

优化

增强模型测试连通性的测试输出
增强mcp加载提示,优化启动体验
增强内置的价格自动识别

This update fixes some bugs and primarily adds automatic mobile web operation features, along with terminal operation functionality for Windows, enabling control over CLI tools like codex.
If you're not accustomed to the new input box, you can switch back to the old one in the theme settings.

Additions

Add SQL viewer in toolbox
Add CORS bypass for Web workspace projects, enabling access to external websites
Add Android template for workspace
Add OpenAI response compatible provider
Add custom su support
Add web page operations
Add character card sorting view on main interface
Add more built-in prompt disable options
Add built-in tool disable options
Add icon cropping and editing when exporting web projects as APK
Add strict tool call mode as a supplement for toolcall bugs on some models
Add skill direct input
Add Windows command operations
Upgrade package system, add Toolpkg with dynamically added toolbox interface via JS
Add new input box, can switch back to old display style in themes
Add pie chart in statistics interface

Fixes

Fix file reading via image recognition model bug in version 1.8.1
Fix error when summarizing context with image content
Fix special character truncation issue in tool output under OpenAI compatibility
Fix autoglm unusability in main chat on Xiaomi phones
Fix drawing package timeout issue
Fix some ffmpeg call issues
Fix flickering issue when swiping out history

Improvements

Enhance test output for model connectivity
Enhance MCP loading prompts, optimize startup experience
Enhance built-in price auto-recognition

本次更新以小内容增加和bug修复为主。支持llama.cpp本地推理,优化和增加一些工具,允许角色卡相互对话/查看聊天历史,工作区允许sftp/saf绑定。实现英文覆盖95%。

增加内容

• 软件完成95%的英文覆盖(国际化支持)
• 增加工具包:qwen绘图和查看历史对话记录
• 增加气泡模式下的头像隐藏功能
• 增加token统计中的模型饼图
• 增加密钥池功能,支持批量测试密钥和导入大密钥池文本文件
• 增加skill开启/关闭选项
• 增加思考链折叠(思考+工具)显示
• 增加图片搜索工具和图片下载通道
• 增加软件的全局大备份功能(备份除MCP/skill/终端/包之外的所有设置)
• 增加所有角色卡的备份功能
• 增加llama.cpp本地推理支持,能够运行gguf本地模型
• 为visit web/download file工具增加header支持
• 增加新的唤醒方式:特定音频唤醒
• 增加html块的直接预览功能,支持全屏显示
• 增加代码块/思考块的限制最大高度
• 增加intent调用chat接口和workflow外部调用模板
• 增加角色卡导出为酒馆角色卡和JSON格式
• 增加角色卡通过多维码直接分享功能
• 增加AI之间相互对话的工具
• 增加工具包用于AI了解自身
• 增加日常工具包中的定时触发任务,AI可在特定时间提醒用户或执行任务
• 支持终端的Ctrl组合操作
• 增加LaTeX块渲染的左右滚动功能
• 气泡模式下的左右头像可设置隐藏
• 支持kimi thinking模式
• 支持读写saf文件并绑定工作区

优化内容

• 为skill下载增加缓存
• 大幅度优化发送消息前处理消息的速度,以及优化加载消息的速度
• 迁移部分文件工具/HTTP请求工具到单独的工具包(架构调整)
• 优化subagent/autoglm的效果,使AI能够意识到是否在操作主屏幕,改进取消逻辑
• 增强终端复制体验

修复异常

• 修复AI朗读回声重新被录制的问题
• 修复一个可能的AI请求头错误
• 修复悬浮窗下的token用量更新统计错误
• 修复悬浮窗对主界面的同步问题
• 修复角色卡编辑时键盘挡住内容的问题
• 修复深度搜索模式下的token爆炸问题
• 修复Mermaid移动遮挡问题
• 修复消息总结的卡死显示问题
• 修复MCP启动运行异常
• 修复工作流启动的悬浮窗无法人为退出的bug
• 修复终端的SSE调用问题
• 修复异常的弹出通知问题
• 修复表格的渲染显示截断问题
• 修复硅基流动下自动朗读和语音交流出现的打断问题

This update focuses on minor content additions and bug fixes. It adds support for llama.cpp local inference, optimizes and adds several tools, allows character cards to chat with each other/view chat history, and achieves 95% English coverage.

New Content

• Software achieves 95% English coverage (internationalization support)
• Added tool packages: Qwen drawing and viewing historical conversation records
• Added avatar hiding feature in bubble mode
• Added model pie chart in token statistics
• Added key pool functionality, supporting batch key testing and importing large key pool text files
• Added skill enable/disable options
• Added thinking chain folding display (thinking + tools)
• Added image search tool and image download channels
• Added global backup function for the software (backup all settings except MCP/skill/terminal/packages)
• Added backup functionality for all character cards
• Added llama.cpp local inference support, capable of running gguf local models
• Added header support for visit_web/download_file tools
• Added new wake-up method: specific audio wake-up
• Added direct preview functionality for HTML blocks with full-screen display support
• Added maximum height limits for code blocks/thinking blocks
• Added intent invocation for chat interface and workflow external call templates
• Added character card export to tavern character card and JSON formats
• Added direct sharing functionality for character cards via multidimensional codes
• Added tools for AI-to-AI conversation
• Added tool package for AI self-awareness
• Added scheduled trigger tasks in daily tools package, allowing AI to remind users or execute tasks at specific times
• Supported Ctrl key combinations in terminal
• Added left/right scrolling functionality for LaTeX block rendering
• Left/right avatars in bubble mode can be set to hidden
• Supported kimi thinking mode

Optimizations

• Added cache for skill downloads
• Significantly optimized message pre-processing speed before sending and message loading speed
• Migrated some file tools/HTTP request tools to separate tool packages (architecture adjustment)
• Optimized subagent/autoglm effectiveness, enabling AI to recognize whether operating on the main screen, improved cancellation logic
• Enhanced terminal copy experience

Bug Fixes

• Fixed AI voice echo re-recording issue
• Fixed a possible AI request header error
• Fixed token usage statistics update error in floating window
• Fixed synchronization issues between floating window and main interface
• Fixed keyboard blocking content issue when editing character cards
• Fixed token explosion issue in deep search mode
• Fixed Mermaid mobile blocking issue
• Fixed crash display issue with message summarization
• Fixed MCP startup and operation exceptions
• Fixed bug where workflow startup floating window couldn't be manually exited
• Fixed terminal SSE invocation issue
• Fixed abnormal pop-up notification issues
• Fixed truncated display issue in table rendering
• Fixed interruption issues with auto-reading and voice communication in silicon flow mode

Full Changelog: v1.8.0...v1.8.1

本次更新增加了工作流、语音唤醒的功能,支持了对话并行,修复了若干bug。

新功能:

• 记忆查询工具,支持按照时间查询记忆
• 工作流系统,支持语音唤醒触发,支持计算、传入传出、执行等功能
• 语音唤醒,直接进入语音对话模式,支持语音下的快速关键词附件附着
• 悬浮窗圈选功能,支持悬浮窗下的圈选屏幕修改范围
• 内置github js pack
• OpenAI格式的TTS供应商,新增两个新的STT供应商
• OpenAI绘图工具包
• 自动备份机制,每天自动备份数据库,支持数据损坏恢复
• 对话并行处理
• 角色卡展示排序选择功能
• 工具包state机制,能够动态决定工具
• 带路径的skill仓库解析

修复:

• Anthropic API兼容性问题
• Waifu模式markdown乱码占位问题
• 终端在大输出下的OOM闪退
• HTTP TTS朗读不等待导致重叠BUG
• 记忆库的/开头记忆无法正常关联到文件夹的问题
• 插入总结的BUG
• 并行JS工具的取消问题
• 虚拟屏幕下的输入不清空输入框的bug
• Ubuntu最开始的时候概率出现tar解压不完全的问题

增强:

• MCP启动逻辑和速度优化,支持点击查看mcp部署报错
• 支持token统计的单个重置
• 支持总结显示具体的总结进度
• 前台服务优化
• 终端的chroot支持,能够在root下显著提高性能
• Grep context工具增强

This update adds workflow and voice wake-up functionality, supports parallel conversations, and fixes several bugs.

New Features:

• Memory query tool supporting time-based memory retrieval
• Workflow system supporting voice wake-up triggers, with computation, input/output, and execution capabilities
• Voice wake-up directly entering voice conversation mode, supporting quick keyword attachment in voice mode
• Floating window selection function supporting screen area modification within floating window
• Built-in GitHub JS pack
• OpenAI format TTS provider, adding two new STT providers
• OpenAI drawing toolkit
• Automatic backup mechanism with daily database backup and data corruption recovery support
• Parallel conversation processing
• Character card display sorting and selection function
• Toolkit state mechanism enabling dynamic tool determination
• Path-based skill repository parsing

Fixes:

• Anthropic API compatibility issues
• Waifu mode markdown garbled placeholder issues
• Terminal OOM crashes with large outputs
• HTTP TTS playback not waiting causing overlapping BUG
• Memory library issue where memories starting with "/" couldn't properly associate with folders
• Insert summary BUG
• Parallel JS tool cancellation issues
• Virtual screen input not clearing input box bug
• Ubuntu initial tar extraction incomplete probability issue

Enhancements:

• MCP startup logic and speed optimization, supporting click-to-view MCP deployment errors
• Support for single token statistics reset
• Support for displaying specific summary progress
• Foreground service optimization
• Terminal chroot support, significantly improving performance under root
• Grep context tool enhancement

Full Changelog: v1.7.1...v1.8.0

本次更新主要支持root启动虚拟屏幕,以及AutoGLM并发虚拟屏幕操作软件。其次增加了SKILL(类似于工具包)的支持。

在关于界面中可以选择加入BETA计划,来持续追踪软件的nightly版本。

各位新年快乐!

一、自动操作相关(autoglm类)

1. 增加一键autoglm配置的自定义选项
2. 增强自动点击稳定性,增加自定义码率设置
3. 增加多autoglm实例支持,支持并发多窗口执行任务
4. 现在root下也支持虚拟屏幕的autoglm操作了

二、核心功能修复与优化

1. 修复消息总结自动拉起不受开关控制的bug,改为异步处理
2. 修复工作区网页预览的返回文字显示问题
3. 修复API调用toolcall错误问题
4. 修复用户消息大图片导致的崩溃问题
5. 修复AI对文件工具环境不敏感的问题
6. 修复代码块换行显示问题
7. 修复全屏悬浮窗下的输出朗读重复问题
8. 增强虚拟屏幕稳定性
9. 优化软件启动速度

三、新功能与用户体验

1. 增加消息总结编辑功能
2. 修改网页访问工具为悬浮窗模式(非全屏占用)
3. 增加图片对话回合有效回合数设置
4. md渲染支持font标签
5. 增加beta测试计划
6. 增加音视频读取处理和图片输出处理
7. 增加skill协议支持
8. 增加skill市场功能
9. 增强悬浮窗全屏模式,增加圈选识屏功能
10. 增加对话锁定功能

What's Changed

• Add AI chat interaction module by @weinicole14 in #301
• Add beeimg_image_uploader for image uploads by @weinicole14 in #303

Full Changelog: v1.7.0...v1.7.1

Contributors

• @weinicole14

本次更新主要针对gui自动化,为全网首个Autoglm+虚拟屏幕的App
使用出现问题可以在设置-显示与行为中关闭虚拟屏幕
同时希望各位能够聚焦于一些别的功能,不仅仅是自动化操作

增加autoglm一键配置和单独执行器
增加虚拟屏幕的是否开启逻辑以及自定义截图质量
密钥显示现在会在非聚焦的时候显示为星号
现在强制不允许autoglm设置为主模型
增加nanobanana绘图包支持
增加apply file的非覆盖支持

What's Changed

• mnn STT支持,优化 UI 样式和布局,其他顺带 by @ic-timon in #276
• 修复: 将启动应用的monkey命令替换为am start,避免修改系统设置(如屏幕自动旋转) by @ic-timon in #279
• 使用resolve-activity获取主Activity,解决unable to resolve Intent错误 by @ic-timon in #280
• Implement Nano Banana API for image generation by @weinicole14 in #288

New Contributors

• @weinicole14 made their first contribution in #288

Full Changelog: v1.6.4...v1.7.0

Contributors

• @ic-timon
• @weinicole14

本次更新支持AutoGLM的接入,优化了自动化显示的UI指示。具体使用请参照用户手册。

1. 支持autoglm的接入
2. 优化自动化过程的UI指示,优化自动化包提示词
3. 支持包管理的env环境输入,修改tavily依赖环境变量,增加xai绘图包
4. 增加长按输入支持
5. 增加工具箱的详细日志导出
6. 修复工作区深色模式显示
7. 微调grep code和apply file工具,使其表现更好
8. 修复时间包
9. 增加非自动化操作的daily life包发送消息逻辑
10. 修复测试模型时多选模型测试失败的问题

Full Changelog: v1.6.3...v1.6.4