[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-74210":3},{"id":4,"name":5,"fullName":6,"owner":5,"repo":5,"description":7,"homepage":8,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":14,"contributorsCount":15,"subscribersCount":15,"size":15,"stars1d":13,"stars7d":16,"stars30d":17,"stars90d":15,"forks30d":15,"starsTrendScore":18,"compositeScore":19,"rankGlobal":9,"rankLanguage":9,"license":20,"archived":21,"fork":21,"defaultBranch":22,"hasWiki":21,"hasPages":21,"topics":23,"createdAt":9,"pushedAt":9,"updatedAt":33,"readmeContent":34,"aiSummary":35,"trendingCount":15,"starSnapshotCount":15,"syncStatus":36,"lastSyncTime":37,"discoverSource":38},74210,"openakita","openakita\u002Fopenakita","An open-source AI assistant framework with skills and agent architecture","http:\u002F\u002Fopenakita.ai",null,"Python",1809,256,13,108,0,28,72,39,96.43,"GNU Affero General Public License v3.0",false,"main",[24,25,26,27,28,29,30,31,32],"agent","ai","assistant","automation","claw","clawd","clawdbot","openclaw","python","2026-06-12 04:01:13","\u003Cp align=\"center\">\n  \u003Cimg src=\"docs\u002Fassets\u002Flogo.png\" alt=\"OpenAkita Logo\" width=\"200\" \u002F>\n\u003C\u002Fp>\n\n\u003Ch1 align=\"center\">OpenAkita\u003C\u002Fh1>\n\n\u003Cp align=\"center\">\n  \u003Cstrong>Open-Source Multi-Agent AI Assistant — Not Just Chat, an AI Team That Gets Things Done\u003C\u002Fstrong>\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fopenakita.ai\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F🌐_Website-openakita.ai-orange?style=for-the-badge\" alt=\"Official Website\" height=\"28\" \u002F>\u003C\u002Fa>\n   \n  \u003Ca href=\"https:\u002F\u002Fopenakita.ai\u002Fdownload\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F📥_Download-Desktop_App-blue?style=for-the-badge\" alt=\"Download\" height=\"28\" \u002F>\u003C\u002Fa>\n   \n  \u003Ca href=\"https:\u002F\u002Fdiscord.gg\u002FvFwxNVNH\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F💬_Discord-Join_Community-5865F2?style=for-the-badge\" alt=\"Discord\" height=\"28\" \u002F>\u003C\u002Fa>\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Flicense-AGPL--3.0--only-blue.svg?style=flat-square\" alt=\"License\" height=\"20\" \u002F>\n  \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fpython-3.11+-blue.svg?style=flat-square\" alt=\"Python Version\" height=\"20\" \u002F>\n  \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fv\u002Frelease\u002Fopenakita\u002Fopenakita?color=green&style=flat-square\" alt=\"Version\" height=\"20\" \u002F>\n  \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Fopenakita?color=green&style=flat-square\" alt=\"PyPI\" height=\"20\" \u002F>\n  \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Factions\u002Fworkflow\u002Fstatus\u002Fopenakita\u002Fopenakita\u002Fci.yml?branch=main&style=flat-square\" alt=\"Build Status\" height=\"20\" \u002F>\n  \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fopenakita\u002Fopenakita?style=flat-square\" alt=\"Stars\" height=\"20\" \u002F>\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  Multi-Agent Collaboration · Organization Orchestration · Plugin System · Sandbox Security · 30+ LLMs · 6 IM Platforms · 89+ Tools · Desktop \u002F Web \u002F Mobile\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  \u003Ca href=\"#quick-start-guide\">Quick-Start\u003C\u002Fa> •\n  \u003Ca href=\"#organization-orchestration\">Org Orchestration\u003C\u002Fa> •\n  \u003Ca href=\"#im-scan-to-bind\">Scan-to-Bind\u003C\u002Fa> •\n  \u003Ca href=\"#plugin-system\">Plugins\u003C\u002Fa> •\n  \u003Ca href=\"#sandbox-security\">Security\u003C\u002Fa> •\n  \u003Ca href=\"#documentation\">Docs\u003C\u002Fa>\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  \u003Cstrong>English\u003C\u002Fstrong> | \u003Ca href=\"README_CN.md\">中文\u003C\u002Fa>\n\u003C\u002Fp>\n\n---\n\n## What is OpenAkita?\n\n**Other AIs just chat. OpenAkita gets things done.**\n\nOpenAkita is an open-source, all-in-one AI assistant — multiple AI Agents work together, build an **AI company** that runs autonomously, search the web, operate your computer, manage files, run scheduled tasks, and respond instantly across Telegram \u002F Feishu \u002F WeCom \u002F DingTalk \u002F QQ. **Scan a QR code to bind your chat app in 30 seconds.** It remembers your preferences, teaches itself new skills, and never gives up on a task. Extend anything through the **plugin system**, protected by **6-layer sandbox security**.\n\n**Fully GUI-based setup. Ready in 5 minutes. Zero command line required.**\n\n\u003Cp align=\"center\">\n  🌐 \u003Ca href=\"https:\u002F\u002Fopenakita.ai\">\u003Cb>Website openakita.ai\u003C\u002Fb>\u003C\u002Fa>  | \n  📥 \u003Ca href=\"https:\u002F\u002Fopenakita.ai\u002Fdownload\">\u003Cb>Download Desktop App\u003C\u002Fb>\u003C\u002Fa>  | \n  📖 \u003Ca href=\"https:\u002F\u002Fopenakita.ai\u002Fdocs\">\u003Cb>Documentation\u003C\u002Fb>\u003C\u002Fa>  | \n  💬 \u003Ca href=\"https:\u002F\u002Fdiscord.gg\u002FvFwxNVNH\">\u003Cb>Discord Community\u003C\u002Fb>\u003C\u002Fa>\n\u003C\u002Fp>\n\n---\n\n## Quick-Start Guide\n\n### 🚀 For First-Time Users (3 Minutes)\n\n**No installation required** — download the desktop app and start chatting:\n\n1. **Download** the installer from [GitHub Releases](https:\u002F\u002Fopenakita.ai\u002Fdownload)\n2. **Install** and follow the onboarding wizard\n3. **Enter your API Key** (get one from [Anthropic](https:\u002F\u002Fconsole.anthropic.com\u002F) or [DeepSeek](https:\u002F\u002Fplatform.deepseek.com\u002F))\n4. **Try your first task**: Type \"Create a calculator\" and watch it work\n\n### 💻 For Developers (5 Minutes)\n\n```bash\n# Install\npip install openakita[all]\n\n# Quick setup (interactive wizard)\nopenakita init\n\n# Run your first task\nopenakita run \"Build a weather scraper\"\n```\n\n### ✨ What You Can Do Right Away\n\n| Category | Examples |\n|----------|----------|\n| **💬 Chat** | Text + images + files, voice messages, stickers |\n| **🤖 Multi-Agent** | \"Create a competitive analysis\" → research + analysis + writing agents work together |\n| **🏢 Organization** | Build an AI company — CEO, CTO, marketing, finance — agents run it autonomously |\n| **🌐 Web** | Search news, scrape websites, automate browser tasks |\n| **📁 Files** | Read\u002Fwrite\u002Fedit files, batch rename, search content |\n| **🖥️ Desktop** | Click buttons, type text, take screenshots, automate apps |\n| **⏰ Scheduled** | \"Remind me every Monday at 9am\" — cron-based reminders |\n\n### ➡️ Next Steps\n\n- **Configure LLM**: Add multiple providers for automatic failover\n- **Set up IM channels**: Scan QR code to bind WeChat\u002FFeishu\u002FWeCom in 30 seconds\n- **Explore skills**: Install from marketplace or create your own\n- **Try org mode**: Build an AI company and watch it run\n- **Join community**: [Discord](https:\u002F\u002Fdiscord.gg\u002FvFwxNVNH) | [WeChat Group](docs\u002Fassets\u002Fwechat_group.jpg)\n\n---\n\n## Core Capabilities\n\n\u003Ctable>\n\u003Ctr>\u003Ctd>\n\n### 🤝 Multi-Agent Collaboration\nMultiple AI agents with specialized skills work in parallel.\nSay one thing — a coding Agent writes code, a writing Agent drafts docs, a testing Agent verifies — all at the same time.\n\n### 🏢 Organization Orchestration\nNot just multi-Agent — build an **AI company**. CEO, CTO, CFO, marketing director… each role operates independently. Blackboard sharing, message routing, deadlock detection.\n\n### 📋 Plan Mode\nComplex tasks auto-decomposed into step-by-step plans with real-time progress tracking and automatic rollback on failure.\n\n### 🧠 ReAct Reasoning Engine\nThink → Act → Observe. Explicit three-phase reasoning with checkpoint\u002Frollback. Fails? Tries a different strategy.\n\n\u003C\u002Ftd>\u003Ctd>\n\n### 🔌 Plugin System\n8 plugin types, 3-tier permission model, 10 lifecycle hooks. Tools, channels, RAG, memory, LLM — everything is extensible.\n\n### 🛡️ 6-Layer Sandbox Security\nPath zoning · Confirmation gates · Command interception · File snapshots · Self-protection · OS-level sandbox. High-risk commands run in isolation automatically.\n\n### 📱 IM Scan-to-Bind\nWeChat, Feishu, WeCom — scan a QR code, 30 seconds to bind, use AI directly in your chat app.\n\n### 💾 Dual-Mode Memory — Smarter Over Time\nFragment memory + MDRM relational graph (causal chains · timelines · entity graph · 3D visualization), auto smart switching.\n\n\u003C\u002Ftd>\u003C\u002Ftr>\n\u003C\u002Ftable>\n\n---\n\n## Full Feature List\n\n| | Feature | Description |\n|:---:|---------|-------------|\n| 🤝 | **Multi-Agent** | Specialized agents, parallel delegation, automatic handoff, failover, real-time visual dashboard |\n| 🏢 | **Org Orchestration** | Hierarchical org structure, CEO\u002FCTO\u002FCFO roles, blackboard memory, message routing, deadlock detection, heartbeat, auto-scaling |\n| 📋 | **Plan Mode** | Auto task decomposition, per-step tracking, floating progress bar in UI |\n| 🧠 | **ReAct Reasoning** | Explicit 3-phase loop, checkpoint\u002Frollback, loop detection, strategy switching |\n| 🚀 | **Zero-Barrier Setup** | Full GUI config, onboarding wizard, 5 minutes from install to chat, zero CLI |\n| 🔧 | **89+ Built-in Tools** | 16 categories: Shell \u002F Files \u002F Browser \u002F Desktop \u002F Search \u002F Scheduler \u002F MCP … |\n| 🔌 | **Plugin System** | 8 types (tool\u002Fchannel\u002FRAG\u002Fmemory\u002FLLM\u002Fhook\u002Fskill\u002FMCP), 3-tier permissions, 10 lifecycle hooks, fault isolation |\n| 🛡️ | **6-Layer Security** | Path zoning, confirmation gates, command blocklist, file snapshots, self-protection, OS sandbox (Linux bwrap \u002F macOS seatbelt \u002F Windows MIC) |\n| 📱 | **IM Scan-to-Bind** | WeChat\u002FFeishu\u002FWeCom scan-to-bind, 30-second setup, no developer configuration needed |\n| 🛒 | **Skill Marketplace** | Search & one-click install, GitHub direct install, AI-generated skills on the fly |\n| 🌐 | **30+ LLM Providers** | Anthropic \u002F OpenAI \u002F DeepSeek \u002F Qwen \u002F Kimi \u002F MiniMax \u002F Gemini … smart failover |\n| 💬 | **6 IM Platforms** | Telegram \u002F Feishu \u002F WeCom \u002F DingTalk \u002F QQ \u002F OneBot, voice recognition, smart group chat |\n| 🔗 | **MCP Integration** | Standard MCP client, stdio \u002F HTTP \u002F SSE transports, multi-directory scan, dynamic server management |\n| 💾 | **Dual-Mode Memory** | Mode 1 fragments (3 layers + 7 types + multi-path recall) + Mode 2 MDRM relational graph (5 dimensions + multi-hop traversal + 3D visualization), auto smart switching |\n| 🎭 | **8 Personas** | Default \u002F Tech Expert \u002F Boyfriend \u002F Girlfriend \u002F Jarvis \u002F Butler \u002F Business \u002F Family |\n| 🤖 | **Proactive Engine** | Greetings, task follow-ups, idle chat, goodnight — adapts frequency to your feedback |\n| 🧬 | **Self-Evolution** | Daily self-check & repair, failure root cause analysis, auto skill generation |\n| 🔍 | **Deep Thinking** | Controllable thinking mode, real-time chain-of-thought display, IM streaming |\n| 🖥️ | **Multi-Platform** | Desktop (Win\u002FMac\u002FLinux) · Web (PC & mobile browser) · Mobile App (Android\u002FiOS), 11 panels, dark theme |\n| 📊 | **Observability** | 12 trace span types, full-chain token statistics panel |\n| 😄 | **Stickers** | 5700+ stickers, mood-aware, persona-matched |\n\n---\n\n## 5-Minute Setup\n\n### Option 1: Desktop App (Recommended)\n\n**Fully GUI-based, no command line** — this is what sets OpenAkita apart from other open-source AI assistants:\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"docs\u002Fassets\u002Fen\u002Fdesktop_quick_config.gif\" alt=\"OpenAkita Quick Setup\" width=\"800\" \u002F>\n\u003C\u002Fp>\n\n| Step | What You Do | Time |\n|:----:|-------------|:----:|\n| 1 | Download installer, double-click to install | 1 min |\n| 2 | Follow the onboarding wizard, enter API Key | 2 min |\n| 3 | Start chatting | Now |\n\n- No Python installation, no git clone, no config file editing\n- Isolated runtime — won't mess with your existing system\n- Chinese users get automatic mirror switching\n- Models, IM channels, skills, schedules — all configured in the GUI\n\n> **Download**: [GitHub Releases](https:\u002F\u002Fopenakita.ai\u002Fdownload) — Windows (.exe) \u002F macOS (.dmg) \u002F Linux (.deb)\n>\n> Learn more at **[openakita.ai](https:\u002F\u002Fopenakita.ai)**\n\n### Option 2: pip Install\n\n```bash\npip install openakita[all]    # Install with all optional features\nopenakita init                # Run setup wizard\nopenakita                     # Launch interactive CLI\n```\n\n### Option 3: Source Install\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fopenakita\u002Fopenakita.git\ncd openakita\npython -m venv venv && source venv\u002Fbin\u002Factivate\npip install -e \".[all]\"\nopenakita init\n```\n\n### Commands\n\n```bash\nopenakita                              # Interactive chat\nopenakita run \"Build a calculator\"     # Execute a single task\nopenakita serve                        # Service mode (IM channels)\nopenakita serve --dev                  # Dev mode with hot reload\nopenakita daemon start                 # Background daemon\nopenakita status                       # Check status\n```\n\n---\n\n## Multi-Platform Access\n\nOpenAkita supports **Desktop, Web, and Mobile** — use it anywhere, on any device:\n\n| Platform | Details |\n|----------|---------|\n| 🖥️ **Desktop App** | Windows \u002F macOS \u002F Linux — native app built with Tauri 2.x |\n| 🌐 **Web Access** | PC & mobile browser — enable remote access, open in any browser |\n| 📱 **Mobile App** | Android (APK) \u002F iOS (TestFlight) — native wrapper via Capacitor |\n\n### Desktop App\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"docs\u002Fassets\u002Fdesktop_terminal_en.png\" alt=\"OpenAkita Desktop App\" width=\"800\" \u002F>\n\u003C\u002Fp>\n\nCross-platform desktop app built with **Tauri 2.x + React + TypeScript**:\n\n| Panel | Function |\n|-------|----------|\n| **Chat** | AI chat, streaming output, Thinking display, drag & drop upload, image lightbox |\n| **Agent Dashboard** | Neural network visualization, real-time multi-Agent status tracking |\n| **Agent Manager** | Create, manage, and configure multiple Agents |\n| **IM Channels** | One-stop setup for all 6 platforms, scan-to-bind |\n| **Skills** | Marketplace search, install, enable\u002Fdisable |\n| **MCP** | MCP server management |\n| **Memory** | Memory management + LLM-powered review |\n| **Scheduler** | Scheduled task management |\n| **Token Stats** | Token usage statistics |\n| **Config** | LLM endpoints, system settings, advanced options |\n| **Feedback** | Bug reports + feature requests |\n\nDark\u002Flight theme · Onboarding wizard · Auto-update · Bilingual (EN\u002FCN) · Start on boot\n\n### Mobile App\n\n\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fb23.tv\u002FpWki3Vw\">\n    \u003Cimg src=\"docs\u002Fassets\u002Fen\u002Fmobile_app_demo_cover.png\" alt=\"▶ Watch Mobile App Demo\" width=\"720\" \u002F>\n  \u003C\u002Fa>\n  \u003Cbr\u002F>\n  \u003Csub>▶ Click to watch the Mobile App demo on Bilibili\u003C\u002Fsub>\n\u003C\u002Fp>\n\n- Connect your phone to the desktop backend over local network\n- Full-featured: chat, multi-Agent collaboration, memory, skills, MCP — all on mobile\n- Supports real-time streaming and Thinking chain display\n- Preview mode available without connecting to a server\n\n---\n\n## Organization Orchestration\n\n\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fb23.tv\u002FjvoWpgj\">\n    \u003Cimg src=\"docs\u002Fassets\u002Fen\u002Forg.jpg\" alt=\"▶ Watch Organization Orchestration Demo\" width=\"720\" \u002F>\n  \u003C\u002Fa>\n  \u003Cbr\u002F>\n  \u003Csub>▶ Click to watch: Built a company on OpenAkita, and it runs autonomously (Bilibili)\u003C\u002Fsub>\n\u003C\u002Fp>\n\nBeyond multi-Agent collaboration — build an entire **AI company**. OpenAkita includes a full organization orchestration engine (AgentOrg) that lets you visually design company structures in the GUI, where AI agents operate like a real company:\n\n```\n┌───────────────────────────────────────────────┐\n│                CEO \u002F Executive                 │\n│       Sets company strategy, coordinates all   │\n└───┬───────────┬───────────┬───────────┬───────┘\n    ▼           ▼           ▼           ▼\n  CTO        Product     Marketing     CFO\n Tech arch   Planning    Strategy    Budget ctrl\n    │           │           │           │\n    ▼           ▼           ▼           ▼\n Dev team    Design      Content      Analytics\n```\n\n### Key Features\n\n| Feature | Description |\n|---------|-------------|\n| **Visual Org Chart** | Drag-and-drop org builder in the GUI — nodes, edges, hierarchies |\n| **Autonomous Roles** | Each node is an independent Agent with its own identity, skills, policies, and memory |\n| **Blackboard Sharing** | 3-level blackboard memory (org \u002F department \u002F node) for secure cross-team info sharing |\n| **Message Routing** | Priority message queues with edge bandwidth control and deadlock detection |\n| **Heartbeat Checks** | Periodic health monitoring of all nodes, auto-remediation on anomalies |\n| **Auto-Scaling** | Automatically recruits new Agents under heavy load, dismisses when idle |\n| **External Tools** | Nodes can request research \u002F browser \u002F code tools on demand with approval workflow |\n| **Org Templates** | Pre-built templates (tech company, content team, etc.) — deploy with one click |\n| **Projects & Tasks** | Tree-structured task decomposition, timeline tracking, org-wide coordination |\n\n---\n\n## Multi-Agent Collaboration\n\n\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1psP5zTEE7\">\n    \u003Cimg src=\"docs\u002Fassets\u002Fen\u002Fmulti_agent_demo_cover.png\" alt=\"▶ Watch Multi-Agent Collaboration Demo\" width=\"720\" \u002F>\n  \u003C\u002Fa>\n  \u003Cbr\u002F>\n  \u003Csub>▶ Click to watch the Multi-Agent collaboration demo on Bilibili\u003C\u002Fsub>\n\u003C\u002Fp>\n\nOpenAkita has a built-in multi-Agent orchestration system — not just one AI, but an **AI team**:\n\n```\nYou: \"Create a competitive analysis report\"\n    │\n    ▼\n┌──────────────────────────────────────┐\n│      AgentOrchestrator (Director)     │\n│   Decomposes task → Assigns to Agents │\n└───┬────────────┬──────────────┬──────┘\n    ▼            ▼              ▼\n Search Agent  Analysis Agent  Writing Agent\n (web research) (data crunching) (report drafting)\n    │            │              │\n    └────────────┴──────────────┘\n                 ▼\n         Results merged, delivered to you\n```\n\n- **Specialization**: Different Agents for different domains, auto-matched to tasks\n- **Parallel Processing**: Multiple Agents work simultaneously\n- **Auto Handoff**: If one Agent gets stuck, it hands off to a better-suited one\n- **Failover**: Agent failure triggers automatic switch to backup\n- **Depth Control**: Max 5 delegation levels to prevent runaway recursion\n- **Visual Tracking**: Agent Dashboard shows real-time status of every Agent\n- **Instance Pooling**: Agent instance pool with LRU eviction for efficient resource use\n\n---\n\n## IM Scan-to-Bind\n\n\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fb23.tv\u002FdkKTjO5\">\n    \u003Cimg src=\"docs\u002Fassets\u002Fen\u002Fim_qrcode.jpg\" alt=\"▶ Watch Scan-to-Bind Tutorial\" width=\"720\" \u002F>\n  \u003C\u002Fa>\n  \u003Cbr\u002F>\n  \u003Csub>▶ Click to watch: OpenAkita scan-to-bind for WeChat, Feishu, WeCom (Bilibili)\u003C\u002Fsub>\n\u003C\u002Fp>\n\n**No developer account needed. No callback URL configuration. No technical knowledge required** — scan a QR code, 30 seconds to bind:\n\n| Platform | How to Bind | Time |\n|----------|-------------|:----:|\n| **WeChat** | Open IM Channels → Click WeChat → Scan QR code | 30 sec |\n| **Feishu** | Open IM Channels → Click Feishu → Scan to authorize | 30 sec |\n| **WeCom** | Open IM Channels → Click WeCom → Scan to bind | 30 sec |\n\nOnce bound, just @AI in your chat app — send messages, images, files, voice — AI handles it all.\n\n---\n\n## 6 IM Platforms\n\nTalk to your AI right inside the chat tools you already use:\n\n| Platform | Connection | Highlights |\n|----------|-----------|------------|\n| **WeChat** | Scan-to-bind (iLink) | Personal account, no official account needed, 30-second setup |\n| **Feishu** | WebSocket \u002F Webhook | Card messages, event subscriptions, scan-to-bind |\n| **WeCom** | Smart Robot callback \u002F WebSocket | Streaming replies, proactive push, scan-to-bind |\n| **DingTalk** | Stream WebSocket | No public IP needed |\n| **Telegram** | Webhook \u002F Long Polling | Pairing verification, Markdown, proxy support |\n| **QQ Official** | WebSocket \u002F Webhook | Groups, DMs, channels |\n| **OneBot** | WebSocket | Compatible with NapCat \u002F Lagrange \u002F go-cqhttp |\n\n- 📷 **Vision**: Send screenshots\u002Fphotos — AI understands them\n- 🎤 **Voice**: Send voice messages — auto-transcribed and processed\n- 📎 **File Delivery**: AI-generated files pushed directly to chat\n- 👥 **Group Chat**: Replies when @mentioned, stays quiet otherwise\n- 💭 **Chain-of-Thought**: Real-time reasoning process streamed to IM\n- 🔄 **Message Interrupts**: Insert new instructions between tool calls without waiting\n\n---\n\n## Plugin System\n\nOpenAkita provides a complete plugin architecture with `plugin.json` manifest declarations, a 3-tier permission model for security, and 10 lifecycle hooks for deep integration:\n\n### 8 Plugin Types\n\n| Type | Description | Example |\n|------|-------------|---------|\n| 🔧 **Tool** | Register custom tools for LLM to call | Database queries, API calls |\n| 💬 **Channel** | Add new IM channel adapters | Slack, Discord adapters |\n| 📚 **RAG** | Add external knowledge retrieval sources | Notion, Confluence retrieval |\n| 🧠 **Memory** | Extend memory storage backends | Redis, PostgreSQL storage |\n| 🤖 **LLM** | Connect new LLM providers | Private model deployments |\n| 🪝 **Hook** | Inject logic into the lifecycle | Message auditing, content filtering |\n| ⚡ **Skill** | Wrap Skills as plugins | Package skills for distribution |\n| 🔗 **MCP** | Wrap MCP Servers as plugins | Simplify MCP deployment |\n\n### 3-Tier Permission Model\n\n| Tier | Description | Example |\n|------|-------------|---------|\n| **Basic** | Auto-granted on install | Read config, register tools |\n| **Advanced** | Requires user confirmation on install | File I\u002FO, network requests |\n| **System** | Must be manually granted per-permission | Shell execution, system config |\n\n### Lifecycle Hooks\n\n`on_init` → `on_message_received` → `on_tool_result` → `on_prompt_build` → `on_retrieve` → `on_session_start` → `on_session_end` → `on_schedule` → `on_shutdown`\n\nPlugins have **automatic fault isolation**: error count exceeding threshold triggers auto-disable, preventing a single plugin from crashing the system.\n\n### Bundled AI-Media Plugins (20)\n\nBuilt on the `openakita-plugin-sdk` (`contrib\u002F` 6 modules: `verification`, `quality_gates`, `intent_verifier`, `provider_score`, `error_coach`, `slideshow_risk` \u002F `source_review`). All ship with `SKILL.md` + `README.md` + tests.\n\n| Plugin | Capability | Tests |\n|--------|-----------|-------|\n| `tongyi-image` | DashScope text-to-image \u002F image-edit \u002F outpaint \u002F background-gen | 107 |\n| `seedance-video` | Volcengine Ark Seedance text\u002Fimage-to-video, long-form chaining | 58 |\n| `tts-studio` | Multi-provider TTS dialogue rendering | 6 |\n| `avatar-speaker` | Digital-human avatar + TTS | 7 |\n| `image-edit` | Image editing dispatcher | 5 |\n| `subtitle-maker` | Generate subtitles | 4 |\n| `video-translator` | Translate video subtitles | 14 |\n| `highlight-cutter` | Highlight reel generator | 5 |\n| `poster-maker` | Static poster composition | 8 |\n| `storyboard` | Scene plan generator (uses `slideshow_risk` + `intent_verifier`) | 39 |\n| `bgm-suggester` | LLM-driven BGM brief (uses `verification`) | 45 |\n| `bgm-mixer` | madmom beat-aware BGM mixing | 68 |\n| `transcribe-archive` | Long-form transcription with `parallel_executor` + `checkpoint` | 95 |\n| `video-bg-remove` | RVM background removal (uses `dep_gate` + `checkpoint`) | 72 |\n| `video-color-grade` | ffmpeg auto color grading via `auto_color_grade_filter` | 49 |\n| `smart-poster-grid` | 4-aspect poster batches (uses `verification`) | 50 |\n| `ppt-to-video` | LibreOffice + tts-studio narrated slideshows | 79 |\n| `local-sd-flux` | Local ComfyUI (SD 1.5 \u002F SDXL \u002F FLUX) + provider ranker | 99 |\n| `shorts-batch` | Batch shorts orchestrator (uses `slideshow_risk`) | 51 |\n| `dub-it` | Video dubbing pipeline (uses `source_review`) | 52 |\n| **Total** | | **913** |\n\n> Developer docs: [Plugin System Overview](docs\u002Fplugin-system-overview.md) · [SDK contrib catalog](openakita-plugin-sdk\u002Fdocs\u002Fcontrib.md) · [Sprint 18 Cleanup Assessment](docs\u002Fsprint18-cleanup-assessment.md)\n\n---\n\n## Sandbox Security\n\nOpenAkita implements a **6-layer defense-in-depth** security model, from path management to OS-level isolation:\n\n```\nL1  Path Zoning         workspace \u002F controlled \u002F protected \u002F forbidden\nL2  Confirmation Gate   Dangerous ops (delete files, system commands) require user approval\nL3  Command Intercept   regedit, format, rm -rf — blocked outright\nL4  File Snapshots      Auto-checkpoint before writes, rollback available\nL5  Self-Protection     data\u002F, src\u002F, identity\u002F — core dirs locked from modification\nL6  OS-Level Sandbox    Linux bwrap \u002F macOS seatbelt \u002F Windows MIC\n```\n\n### Sandbox Execution\n\nWhen the policy engine classifies a shell command as **HIGH risk**, it automatically runs in an OS-level sandbox:\n\n| Platform | Sandbox Backend | Description |\n|----------|----------------|-------------|\n| **Linux** | bubblewrap (bwrap) | User-space container isolation, restricted filesystem and network |\n| **macOS** | sandbox-exec (seatbelt) | System-level sandbox policies |\n| **Windows** | Low Integrity (MIC) | Mandatory Integrity Control, low-privilege process isolation |\n\n### Additional Security Mechanisms\n\n- **Policy Engine**: `POLICIES.yaml` for tool permissions, shell command blocklist, path restrictions\n- **Resource Budgets**: Token \u002F cost \u002F duration \u002F iteration \u002F tool call limits per task\n- **Runtime Supervision**: Auto-detection of tool thrashing, reasoning loops, token anomalies\n- **Local Data**: Memory, config, and chat history stored on your machine only\n- **Open Source**: AGPL-3.0-only, fully transparent codebase\n\n---\n\n## 30+ LLM Providers\n\n**No vendor lock-in. Mix and match freely:**\n\n| Category | Providers |\n|----------|-----------|\n| **International** | Anthropic · OpenAI · Google Gemini · xAI (Grok) · Mistral · OpenRouter · NVIDIA NIM · Groq · Together AI · Fireworks · Cohere |\n| **China** | Alibaba DashScope · Kimi (Moonshot) · Xiaomi MiMo · MiniMax · DeepSeek · SiliconFlow · Volcengine · Zhipu AI · Baidu Qianfan · Tencent Hunyuan · Yunwu · Meituan LongCat · iFlow |\n| **Local** | Ollama · LM Studio (⚠️ Small models have limited tool-calling ability — not recommended yet, pending optimization) |\n\n**7 capability dimensions**: Text · Vision · Video · Tool use · Thinking · Audio · PDF\n\n**Smart failover**: One model goes down, the next picks up seamlessly.\n\n### Recommended Models\n\n**International Models (in order of recommendation):**\n\n| Model | Provider | Notes |\n|-------|----------|-------|\n| `claude-opus-4-6` | Anthropic | One of the best — top-tier coding & long-task capability, 1M context |\n| `gpt-5.4` | OpenAI | Flagship — native computer-use, 1M context, strong reasoning |\n| `claude-sonnet-4-6` | Anthropic | Best value — fully upgraded default model, 1M context |\n| `gpt-5.3-instant` | OpenAI | Best for everyday chat — significantly fewer hallucinations, natural flow |\n| `claude-opus-4-5` | Anthropic | Previous flagship, still extremely capable |\n| `claude-sonnet-4-5` | Anthropic | Stable and reliable for everyday use |\n\n**Chinese Models (recommended):**\n\n| Model | Provider | Notes |\n|-------|----------|-------|\n| `kimi-k2.5` | Moonshot | 1T MoE, Agent Swarm with up to 100 parallel sub-agents, 256K context, open-source |\n| `qwen3.5-plus` | Alibaba | 397B MoE, 1M context, 201 languages, extremely cost-effective |\n| `mimo-v2-pro` | Xiaomi | 1T MoE, 1M context, global Top 8 ranking, affordable pricing |\n| `deepseek-v3` | DeepSeek | Cost-effective benchmark, strong Chinese support |\n\n> For complex reasoning, enable Thinking mode — add `-thinking` suffix to the model name.\n>\n> ⚠️ **Local small models not recommended** (e.g. 7B\u002F14B quantized): Small models have limited tool-calling and agent collaboration capabilities, prone to hallucinations and format errors. Use API-hosted flagship models for the best experience.\n\n---\n\n## Memory System\n\nNot just a \"context window\" — true long-term memory. Supports **dual modes** with automatic switching:\n\n### Mode 1: Fragment Memory (Classic)\n\n- **Three layers**: Working memory (current task) + Core memory (user profile) + Dynamic retrieval (past experience)\n- **7 memory types**: Fact \u002F Preference \u002F Skill \u002F Error \u002F Rule \u002F Persona trait \u002F Experience\n- **Multi-path recall**: Semantic + full-text + temporal + attachment search\n- **Gets smarter over time**: Preferences you mentioned two months ago? Still remembered.\n\n### Mode 2: MDRM Relational Graph Memory (New)\n\nOn top of fragment memory, builds **causal chains, timelines, and entity relationship graphs** — letting AI truly understand connections between events:\n\n| Dimension | Description | Example |\n|-----------|-------------|---------|\n| **Temporal** | Event chronology and timelines | \"What did I do last week?\" → auto-constructs timeline |\n| **Causal** | Cause-and-effect chains | \"What caused this bug?\" → traces causal chain |\n| **Entity** | Relationships between people\u002Fprojects\u002Fconcepts | \"Which projects did Alice work on?\" → entity graph |\n| **Action** | Dependencies, prerequisites, compositions | \"What else is needed to finish X?\" → dependency analysis |\n| **Context** | Project\u002Fsession attribution | \"All discussions about this project\" → cross-session aggregation |\n\n- **4 node types**: Event \u002F Fact \u002F Decision \u002F Goal\n- **Multi-hop graph traversal**: Starting from seed nodes, expands along relationship edges to find deep connections\n- **3-layer encoding**: Fast rule-based encoding → summary backfill → session-end batch LLM encoding\n- **3D visualization**: Frontend supports 3D visualization of the memory graph\n\n### Smart Mode Switching\n\nSet `memory_mode` to `auto` (default) and the system auto-routes based on query characteristics: causal\u002Ftimeline\u002Fcross-session questions use **Mode 2 graph traversal**, preference\u002Ffact queries use **Mode 1 semantic retrieval**.\n\n- **AI-driven extraction**: Automatically distills valuable information after each conversation, dual-track writes to both modes\n- **3D memory graph**: Visualize memory nodes and relationships, intuitively understand AI's memory structure\n\n---\n\n## MCP Integration\n\nOpenAkita includes a full [MCP (Model Context Protocol)](https:\u002F\u002Fmodelcontextprotocol.io\u002F) client, enabling AI to connect with any external service:\n\n| Feature | Description |\n|---------|-------------|\n| **3 Transports** | stdio (default), Streamable HTTP, SSE (legacy compatible) |\n| **Multi-Dir Scan** | Auto-discovers MCP configs from built-in `mcps\u002F`, `.mcp`, `data\u002Fmcp\u002Fservers\u002F` directories |\n| **Dynamic Management** | Add\u002Fremove MCP servers at runtime, no restart needed |\n| **Tool Suite** | `call_mcp_tool`, `list_mcp_servers`, `add_mcp_server`, `connect_mcp_server`, and more |\n| **Progressive Disclosure** | MCP tool catalog + prompt templates, shown on demand |\n| **GUI Management** | Desktop MCP panel for one-stop configuration |\n\nConnect to GitHub, databases, Playwright browser, filesystem, or any MCP Server.\n\n---\n\n## Self-Evolution\n\nOpenAkita keeps getting stronger:\n\n```\nDaily 04:00   →  Self-check: analyze error logs → AI diagnosis → auto-fix → push report\nAfter failure →  Root cause analysis (context loss \u002F tool limitation \u002F loop \u002F budget) → suggestions\nMissing skill →  Auto-search GitHub for skills, or AI generates one on the spot\nMissing dep   →  Auto pip install, auto mirror switching for China\nEvery chat    →  Extract preferences and experience → long-term memory\n```\n\n---\n\n## Architecture\n\n```\nDesktop App (Tauri + React)\n    │\nIdentity ─── SOUL.md · AGENT.md · POLICIES.yaml · 8 Persona Presets\n    │\nCore     ─── ReasoningEngine(ReAct) · Brain(LLM) · ContextManager\n    │        PromptAssembler · RuntimeSupervisor · ResourceBudget\n    │\nAgents   ─── AgentOrchestrator(Coordination) · AgentInstancePool(Pooling)\n    │        AgentFactory · FallbackResolver(Failover)\n    │\nOrg      ─── OrgRuntime(Runtime) · OrgManager(CRUD)\n    │        OrgMessenger(Routing) · Blackboard(Shared Memory)\n    │        OrgIdentity(Inheritance) · OrgPolicies(Policies)\n    │\nPlugins  ─── PluginManager(Discovery\u002FLoading) · PluginAPI(Host Interface)\n    │        HookRegistry(10 Hooks) · PluginSandbox(Fault Isolation)\n    │\nMemory   ─── Mode1: UnifiedStore(SQLite+Vector) · RetrievalEngine(Multi-path)\n    │        Mode2: RelationalStore(MDRM Graph) · GraphEngine(Multi-hop)\n    │        MemoryModeRouter(Auto Switch) · MemoryEncoder(3-Layer)\n    │\nTools    ─── Shell · File · Browser · Desktop · Web · MCP · Skills\n    │        Plan · Scheduler · Sticker · Persona · Agent Delegation\n    │\nSecurity ─── PolicyEngine(6-Layer) · SandboxExecutor(OS Sandbox)\n    │        ConfirmationGate · CommandFilter · Checkpoint\n    │\nEvolution ── SelfCheck · FailureAnalyzer · SkillGenerator · Installer\n    │\nChannels ─── CLI · Telegram · Feishu · WeCom · WeChat · DingTalk · QQ · OneBot\n    │\nTracing  ─── AgentTracer(12 SpanTypes) · DecisionTrace · TokenStats\n```\n\n---\n\n## Documentation\n\n| Document | Content |\n|----------|---------|\n| [Configuration Guide](docs\u002Fconfiguration-guide.md) | Desktop Quick Setup & Full Setup walkthrough |\n| ⭐ [LLM Provider Setup](docs\u002Fllm-provider-setup-tutorial.md) | **API Key registration + endpoint config + Failover** |\n| ⭐ [IM Channel Setup](docs\u002Fim-channel-setup-tutorial.md) | **Telegram \u002F Feishu \u002F DingTalk \u002F WeCom \u002F QQ \u002F OneBot tutorial** |\n| [Plugin System Overview](docs\u002Fplugin-system-overview.md) | Plugin types, permissions, developer guide |\n| [Org Orchestration Design](docs\u002Fagent-org-technical-design.md) | AgentOrg technical architecture and design |\n| [Org Orchestration Guide](docs\u002Fagent-org-user-guide.md) | Organization orchestration user guide |\n| [Quick Start](docs\u002Fgetting-started.md) | Installation and basics |\n| [Architecture](docs\u002Farchitecture.md) | System design and components |\n| [Configuration](docs\u002Fconfiguration.md) | All config options |\n| [Deployment](docs\u002Fdeploy.md) | Production deployment |\n| [MCP Integration](docs\u002Fmcp-integration.md) | Connecting external services |\n| [Skill System](docs\u002Fskills.md) | Creating and using skills |\n\n---\n\n## Community\n\n\u003Ctable>\n  \u003Ctr>\n    \u003Ctd align=\"center\">\n      \u003Cimg src=\"docs\u002Fassets\u002Fwechat_official.png\" width=\"180\" alt=\"WeChat Official Account\" \u002F>\u003Cbr\u002F>\n      \u003Cb>WeChat Official\u003C\u002Fb>\u003Cbr\u002F>\n      \u003Csub>Follow for updates\u003C\u002Fsub>\n    \u003C\u002Ftd>\n    \u003Ctd align=\"center\">\n      \u003Cimg src=\"docs\u002Fassets\u002Fperson_wechat.jpg\" width=\"180\" alt=\"Personal WeChat\" \u002F>\u003Cbr\u002F>\n      \u003Cb>WeChat (Personal)\u003C\u002Fb>\u003Cbr\u002F>\n      \u003Csub>Note \"OpenAkita\" to join group\u003C\u002Fsub>\n    \u003C\u002Ftd>\n    \u003Ctd align=\"center\">\n      \u003Cimg src=\"docs\u002Fassets\u002Fwechat_group.jpg\" width=\"180\" alt=\"WeChat Group\" \u002F>\u003Cbr\u002F>\n      \u003Cb>WeChat Group\u003C\u002Fb>\u003Cbr\u002F>\n      \u003Csub>Scan to join (⚠️ refreshed weekly)\u003C\u002Fsub>\n    \u003C\u002Ftd>\n    \u003Ctd align=\"center\">\n      \u003Cimg src=\"docs\u002Fassets\u002Fqq_group.png\" width=\"180\" alt=\"QQ Group\" \u002F>\u003Cbr\u002F>\n      \u003Cb>QQ Group: 854429727\u003C\u002Fb>\u003Cbr\u002F>\n      \u003Csub>Scan or search to join\u003C\u002Fsub>\n    \u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\n\u003Cp align=\"center\">\n  🌐 \u003Ca href=\"https:\u002F\u002Fopenakita.ai\">Website\u003C\u002Fa> · \n  💬 \u003Ca href=\"https:\u002F\u002Fdiscord.gg\u002FvFwxNVNH\">Discord\u003C\u002Fa> · \n  🐦 \u003Ca href=\"https:\u002F\u002Fx.com\u002Fopenakita\">X (Twitter)\u003C\u002Fa> · \n  📧 \u003Ca href=\"mailto:zacon365@gmail.com\">Email\u003C\u002Fa>\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fopenakita\u002Fopenakita\u002Fissues\">Issues\u003C\u002Fa> · \n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fopenakita\u002Fopenakita\u002Fdiscussions\">Discussions\u003C\u002Fa> · \n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fopenakita\u002Fopenakita\">⭐ Star\u003C\u002Fa>\n\u003C\u002Fp>\n\n---\n\n## Acknowledgments\n\n- [Anthropic Claude](https:\u002F\u002Fwww.anthropic.com\u002Fclaude) — Default recommended LLM, core development partner\n- [Tauri](https:\u002F\u002Ftauri.app\u002F) — Cross-platform desktop framework\n- [ChineseBQB](https:\u002F\u002Fgithub.com\u002Fzhaoolee\u002FChineseBQB) — 5700+ stickers that give AI a soul\n- [browser-use](https:\u002F\u002Fgithub.com\u002Fbrowser-use\u002Fbrowser-use) — AI browser automation\n- [AGENTS.md](https:\u002F\u002Fagentsmd.io\u002F) \u002F [Agent Skills](https:\u002F\u002Fagentskills.io\u002F) — Open standards\n\n### Community Contributors\n\n- [@948324394](https:\u002F\u002Fgithub.com\u002F948324394) — Docker deployment support\n\n## License\n\nOpenAkita source code is licensed under the GNU Affero General Public License v3.0 only (AGPL-3.0-only) — See [LICENSE](LICENSE).\n\nThe `OpenAkita` name, logos, icons, screenshots, and other brand assets are not licensed under AGPL-3.0-only. See [TRADEMARK.md](TRADEMARK.md) for brand usage rules.\n\nThird-party licenses: [THIRD_PARTY_NOTICES.md](THIRD_PARTY_NOTICES.md)\n\n## Star History\n\n\u003Ca href=\"https:\u002F\u002Fstar-history.com\u002F#openakita\u002Fopenakita&Date\">\n \u003Cpicture>\n   \u003Csource media=\"(prefers-color-scheme: dark)\" srcset=\"https:\u002F\u002Fapi.star-history.com\u002Fsvg?repos=openakita\u002Fopenakita&type=Date&theme=dark\" \u002F>\n   \u003Csource media=\"(prefers-color-scheme: light)\" srcset=\"https:\u002F\u002Fapi.star-history.com\u002Fsvg?repos=openakita\u002Fopenakita&type=Date\" \u002F>\n   \u003Cimg alt=\"Star History Chart\" src=\"https:\u002F\u002Fapi.star-history.com\u002Fsvg?repos=openakita\u002Fopenakita&type=Date\" \u002F>\n \u003C\u002Fpicture>\n\u003C\u002Fa>\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cstrong>OpenAkita — Open-Source Multi-Agent AI Assistant That Gets Things Done\u003C\u002Fstrong>\u003Cbr\u002F>\n  \u003Ca href=\"https:\u002F\u002Fopenakita.ai\">openakita.ai\u003C\u002Fa>\n\u003C\u002Fp>\n","OpenAkita 是一个开源的多智能体AI助手框架，能够实现多AI协作完成任务。该项目支持超过30种大语言模型、6个即时通讯平台以及89种以上的工具插件，并具备组织编排和沙箱安全等核心功能，通过其插件系统可以轻松扩展能力。它特别适合需要自动化处理日常事务、文件管理或跨平台消息响应的个人及企业使用场景。项目采用Python编写，遵循AGPL-3.0许可证开放源代码。",2,"2026-06-11 03:49:32","high_star"]