中文版 | English

Operit AI - Your Intelligent Assistant

📱 The first fully functional, standalone AI assistant for mobile devices with powerful tool-calling capabilities 📱

--- ## 🌟 Project Introduction **Operit AI** is the first fully functional AI assistant application on mobile devices that runs **completely standalone** (except for external API calls) on your Android device. It features powerful **tool-calling capabilities**, **Deep Search**, **workflows & automation**, an **intelligent Memory Vault**, and highly customizable **User Personality** and **Character Cards**, with **MNN/llama.cpp local inference**, an **MCP/Skill ecosystem**, and a **multi-language UI**. It's more than just a chat interface—it's an **all-in-one assistant** deeply integrated with Android permissions and various tools, now featuring a built-in **Ubuntu 24 Environment** for unprecedented functionality. --- ## ⚡ Core Highlights

### 🖥️ Ubuntu 24 Environment Complete Ubuntu 24 system with support for vim, MCP, Python, and more—run complex Linux commands and automation tasks on your phone ### 🧠 Intelligent Memory System AI automatically categorizes memories, supports time-based query/import/export/auto-summary, and intelligently searches conversation history ### 🗣️ Voice Interaction Natural continuous conversations with local/cloud TTS + local STT, custom voice tones, voice/specific-audio wake-up, and auto read-aloud

### 🤖 Local AI Models Support for MNN / llama.cpp local models (GGUF)—run AI completely offline to protect your privacy ### 🎭 Personality & Character Cards Customize AI personality and speaking style, with character card backup/export (Tavern/JSON)/QR sharing and independent histories; supports AI-to-AI chats ### 🔌 Rich Tool Ecosystem 40+ built-in tools + MCP/Skill marketplace plugins + toolkits/workflows, including AutoGLM auto-click agent, drawing/image search, AI-to-AI & self-awareness tools, scheduled tasks, and a prompt marketplace

--- ## 🛠️ Feature Overview

📦 Built-in Tool System (Click to expand)

| Tool Category | Features | |--------------|----------| | 🐧 **Linux Environment** | Complete Ubuntu 24 with apt package management, Python/Node.js runtime, custom repositories | | 📁 **File System** | Read/write files, search, compress/decompress, format conversion, Git integration | | 🌐 **Network Tools** | HTTP requests, web access, file upload/download, web development and export | | ⚙️ **System Operations** | App installation, permission management, UI automation via Accessibility / ADB / Root (includes AutoGLM auto-click agent, supports adb root virtual display / multi-display) | | 🎬 **Media Processing** | Video conversion, frame extraction, OCR/vision understanding, camera capture, audio/video reading | | 🧑‍💻 **Developer & Terminal** | Web workspace/one-click packaging, code editing/syntax highlight, terminal SSH/Chroot/vim, Ctrl shortcuts | | 🎨 **AI Creation** | Drawing toolkits (OpenAI/Qwen/NanoBanana), image search/download | | 🔍 **Search Engines** | Deep search, DuckDuckGo, Tavily, Google Scholar, Bing, Sogou, Quark, Baidu Maps | | 🧩 **Toolkits & Workflows** | Toolkit ecosystem/package management, workflow automation, scheduled triggers, voice wake-up triggers |

🎨 Interface Customization (Click to expand)

- ✨ **Theme System**: Customize colors, fonts, spacing, padding - � **Multi-language UI**: Chinese/English coverage with system locale auto-switch - 🔤 **Fonts & Typography**: Global font size and chat padding controls - �🎭 **Desktop Pet**: WebP animation support, custom expressions, floating window display - 📱 **Layout Optimization**: Hide status bar, custom toolbar, tablet adaptation - 🎨 **Markdown Rendering**: LaTeX formulas (left/right scroll), code highlighting, tables, Mermaid diagrams - 🧾 **Information Display**: Thinking fold, HTML block preview, code/thinking max height - 🪟 **Floating Window UX**: Avatar hide in bubble mode, screen selection OCR, full-screen preview - 🧮 **Analytics**: Token usage stats with model pie chart

🔗 Integration Capabilities (Click to expand)

- 🤖 **Tasker Integration**: Trigger custom AI agent events for deep automation - 🌐 **MCP/Skill Marketplace**: One-click plugin installation, remote MCP, auto-description, uvx/npx support - 🔌 **Multi-Model Support**: OpenAI, Claude, Gemini, Bailing, OpenRouter, LMStudio - 🧪 **Model & Prompt Management**: Multi configs/param customization, prompt marketplace - 🔐 **Permissions**: Per-tool authorization and safety prompts - 🔑 **Key Pool & Stats**: Batch test/import, token usage stats - 🗂️ **Workspace Binding**: SAF / SFTP / SSH workspace binding and file access - 🖱️ **Auto-Click Agent**: AutoGLM + UI Tree dual pipeline for automation - 📊 **Parallel Tools**: Read-only tools execute in parallel for faster response

💬 Chat & Memory Management (Click to expand)

- 🧠 **Memory Vault**: Auto categorization/search, time queries, import/export, attachment memories - 💬 **Chat Management**: Auto-summary + editor, history folders/branches/migration, chat lock, per-character history - ⚡ **Parallel Conversations**: Parallel chat processing with toolkit state decisions - 🤖 **Character Interaction**: Character card chats and history, thinking fold display - 📦 **Chat Records**: Multi-format import/export, history backup and restore

💾 Data & Backup (Click to expand)

- 🗂️ **Global/Auto Backup**: Scheduled DB backup with corruption recovery (excluding MCP/Skill/terminal/packages) - 🎭 **Character Cards**: Backup, export (Tavern/JSON), QR sharing - 🧷 **Workspaces**: SAF/SFTP/SSH binding, code editing/syntax highlight, Git ignore - 🧰 **Skill Management**: Skill toggles, repo parsing, cached downloads

--- ## 📸 Feature Showcase

Web Development
Design web pages on your phone and export as standalone apps

Floating Window & Attachments
Access anytime, share conveniently

Plugin Marketplace
Rich MCP ecosystem

--- ## 🚀 Quick Start | Item | Description | |------|-------------| | 📋 **System Requirements** | Android 8.0+ (API 26+), 4GB+ RAM recommended, 200MB+ storage | | 📥 **Download & Install** | Download the latest APK from the [Release Page](https://github.com/AAswordman/Operit/releases) | | 📖 **User Guide** | [Complete Documentation](https://aaswordman.github.io/OperitWeb) with detailed tutorials and examples | > **Security Warning:** To ensure your data security, please download this application exclusively from the official [Release Page](https://github.com/AAswordman/Operit/releases) or the [Official Website](https://aaswordman.github.io/OperitWeb/). Installation packages from unofficial sources may be maliciously modified, potentially leading to privacy leaks or device monitoring. **Installation Steps:** Download APK → Install & Launch → Follow Setup Guide → Start Using ✨ --- ## 🔮 Roadmap / TODO - **UI automation & screenshot pipeline** - ✅ UI automation supported via Accessibility / ADB / Root permission modes - ✅ Virtual display / multi-display supported in adb root scenarios (`display` parameter) - ✅ Dual UI Tree approaches: AutoGLM + local uiautomator dump --- ## 📅 Version History

Version	Release Date	Core Updates
v1.10.1 _Latest	2026-04-17	• In-App Browser & Web Automation: Greatly expanded the built-in browser with tabs, history, bookmarks, permissions, multi-window support, minimization, and viewport control, plus browser-script import, install, storage, runtime, and page-menu support • Avatars & UI Customization: Added FBX virtual avatars, upgraded the MMD preview experience, introduced liquid-glass theme effects, and expanded sidebar, chat-bubble, and input-bar appearance customization • Plugins, Workspace & Context Enhancements: Added configuration-editor support for debugging and automatically writing Operit plugins, a local HTTP chat entry point, workspace renaming and rule-file pickup, plus stronger history jumping, bidirectional paging, and automatic context injection • Stability & Performance Improvements: Fixed tool-permission checks, HTTP TTS compatibility, SSH/tmux long-output crashes, history jumping, GIF/formula/Markdown rendering, MCP config and stats issues, while continuing to improve conversation flow, deep search, memory, browser behavior, and package management
v1.10.0	2026-03-18	• Character Group Chat & AI Self-Setup: Added group conversations with multiple character cards and @mentions, plus AI self-configuration to help set up MCP, skills, STT, TTS, and model settings • Theme & Interaction Upgrades: Added grouped message folding, bubble-theme customization for fonts/colors/backgrounds, wider bubbles, liquid-glass input styling, long-press app-icon shortcuts, and assistant/MP4 virtual avatar support • Tools & Platform Expansion: Added Ollama, NVIDIA, and OpenAI Response universal-mode providers, a standalone SSH plugin toolkit, Java Bridge, APKTool plugin, web-automation downloads, Markdown audio/video rendering, xAI video generation, workflow cancelation, custom terminal keys, and message queues • Fixes & Performance Improvements: Fixed speech-recognition, memory-concurrency, floating-window interaction, terminal display, full-screen web automation, and MNN tool-call issues; also improved memory recall, marketplace search, workspace templates, grep performance, and agent retry stability
v1.9.1	2026-02-20	• Stability & Bugfix Release: Focused fixes for multiple regressions from v1.9.0, improving overall usability and smoothness • Terminal & Tool Calling: Enhanced terminal tool; fixed terminal freezes caused by interactive UI, strict-tool-call history-triggered request errors, and Windows controller raw command execution issues • MCP & Memory Vault: Fixed remote MCP not closing; reworked memory write logic, added external vector model support, and introduced connection-editing tools • Feature Additions & UI Fixes: Added deletion of chats not bound to character cards, workflow batch deletion and execution-log viewing; fixed IME/input box/dark mode/theme transparency/toolbox package-management issues
v1.9.0	2026-02-17	• Mobile Web Automation: Added web operation capabilities, including CORS bypass for web workspace projects to access external websites • Windows Terminal Operations: Added Windows command operations for controlling CLI tools like Codex; introduced strict tool-call mode as a compatibility supplement • Tooling & System Expansion: Added toolbox SQL viewer, Android workspace template, OpenAI response-compatible provider, direct skill input, and chart enhancements in analytics • Fixes & Improvements: Fixed image-reading/context-summary/special-character truncation/ffmpeg-related issues; improved model-connectivity test output and MCP loading prompts
v1.8.1	2026-02-03	• Local Llama.cpp: GGUF local inference support and related tools • Tools & UI: Image search/download, HTML block preview, code/thinking max height, avatar hide in bubble mode, token pie chart, thinking fold display • Data & Backup: Global backup (excluding MCP/skill/terminal/packages) + character card backup/export/share, skill enable/disable, key pool import/test, workspace SAF binding support • Fixes: Voice echo capture, token stats in floating window, character editor keyboard overlap, deep search token spike, MCP startup, workflow floating window exit, table truncation, silicon flow voice interruptions
v1.8.0	2026-01-13	• Workflow System: Supports computation/input/output/execution with voice wake-up triggers • Voice Wake-up: Enter voice conversation mode directly, with keyword-based quick attachments • Parallel Conversations: Parallel processing + toolkit state mechanism for dynamic tool decisions • Additions & Improvements: Time-based memory query, auto backups, OpenAI drawing + new TTS/STT providers, MCP startup optimizations, terminal chroot, and multiple bug fixes
v1.7.1	2025-12-31	• Root Virtual Display Automation: Virtual screen supported under root, AutoGLM concurrent multi-window tasks • Skill Ecosystem: Added Skill protocol + Skill marketplace; optional BETA plan for nightly builds • UX Enhancements: Summary editing, web access in floating window mode, selection OCR, chat lock • Fixes & Optimizations: Large image crashes, ToolCall errors, code block line breaks, startup speed and virtual-screen stability
v1.7.0	2025-12-19	• GUI Automation Milestone: Autoglm + virtual display (virtual screen can be toggled in settings) • Automation Enhancements: One-click Autoglm setup + standalone executor, virtual screen toggle logic, custom screenshot quality • UX Tweaks: API key masked when unfocused; Autoglm is no longer allowed as the main model • Tooling: NanoBanana image generation pack, apply file no-overwrite support, MNN STT and more
v1.6.3	2025-12-08	• Native ToolCall Support: Added support for native model tool calls and DeepSeek's thinking tools • Workspace & Terminal Enhancements: Select project type on creation, connect to external SSH filesystems, accessibility support for terminal • Model & Message Display: Multi-select for model configs, display model name/provider on messages • Optimizations & Fixes: Improved floating window, fixed terminal lag, migrated workspace to internal data
v1.6.2	2025-11-20	• Conversation Management: Long-press branching, classified history display, batch migration • Model Configuration: Rename configs, context binding, Google native search • Bug Fixes: UI switching, bold line breaks, bubble mode issues • Added crossref academic paper search, upgraded code editor
v1.6.1	2025-11-05	• Performance Boost: Redesigned UI rendering, significantly improved fluidity • AI Vision Enhanced: Direct image recognition, indirect recognition capabilities • Terminal SSH: SSH connection support and reverse-mount phone filesystem • Auto-summary mechanism, deep search, new authorization system
v1.6.0	2025-10-21	• MNN Local Model support • Memory System Overhaul: AI auto-categorization, smart search, import/export • Terminal Enhancements: vim support, progress bars, custom repositories • Tasker integration, desktop pets, storyline tags
v1.5.2	2025-10-05	• MCP enhancements: uvx/npx support, faster startup • Workspace Git ignore • Camera capture, HTML rendering, regex filtering
v1.5.0	2025-09	• Ubuntu 24 Terminal full integration • MCP Marketplace launch • Desktop pets, deep search mode
v1.4.0	2025-08	• Parallel tool execution • Character card system, character selector • PNG character card import
v1.3.0	2025-08	• Web development features • Theme selector, custom UI • Anthropic Claude support
v1.2.x	2025-07	• Voice conversation system • Knowledge base functionality • DragonBones animation support
v1.1.x	2025-06	• MCP protocol support • OCR recognition, floating window • Full Gemini support
v1.0.0	2025-05	• First official release • Basic AI chat, tool calling • Shizuku/Root integration

> 📝 **Full Changelog**: Visit the [Releases Page](https://github.com/AAswordman/Operit/releases) for detailed updates on each version --- ## 👨‍💻 Open Source & Collaboration Join the Operit open-source ecosystem! We welcome all types of contributions: third-party scripts, MCP plugins, and core feature development. **Developer Notes:** - 📚 [Contribution Guide](docs/CONTRIBUTING.md) | [Script Development Guide (Chinese)](docs/SCRIPT_DEV_GUIDE.md) - 📦 Building requires downloading dependency archives from [Google Drive](https://drive.google.com/drive/folders/1g-Q_i7cf6Ua4KX9ZM6V282EEZvTVVfF7?usp=sharing) (`models.zip`, `subpack.zip`, `jniLibs.zip`, `libs.zip`) - 💬 Join community discussions: [QQ Group](https://qm.qq.com/q/Sa4fKEH7sO) | [Discord](https://discord.gg/YnV9MWurRF) ### 💖 Contributors Thanks to all who have contributed to Operit AI!

## 💖 Support Development If Operit AI has been helpful to you, you can voluntarily support its ongoing development and basic project maintenance: - For international support, use [Patreon](https://www.patreon.com/c/aaswordsman) - For mainland China support, use [Afdian](https://afdian.com/a/aaswordsman) - Support is entirely voluntary and does not unlock any features, quotas, updates, support replies, or other perks - Choosing not to support does not affect normal usage, updates, or access to the open-source code - You can also use the `Sponsor` button at the top of this GitHub repository to open the support page --- ## 📄 License This project is licensed under the [GNU LGPLv3](https://www.gnu.org/licenses/lgpl-3.0.html) license. In simple terms, this means: - You are free to use, modify, and distribute the code of this project. - If you modify the code and distribute it, you must also open-source your modified parts under the LGPLv3 license. - For more details, please see the [LICENSE](LICENSE) file. --- ## 📝 Feedback Encountered a problem or have a suggestion? Welcome to [submit an Issue](https://github.com/AAswordman/Operit/issues)! **Submission Guidelines:** - 📝 Clearly describe the problem/suggestion with reproduction steps - 📱 Include device model, system version, and other information - 📸 If possible, provide screenshots or screen recordings ---

⭐ If you like this project, please give us a Star ⭐

🚀 Help us promote and let more people discover Operit AI 🚀

_{Made with ❤️ by the Operit Team}