diff --git a/README.md b/README.md index fddfdbba..50a25c0b 100644 --- a/README.md +++ b/README.md @@ -1,7 +1,13 @@
+

+ MateClaw Logo +

+ # MateClaw +

Build AI that thinks, acts, remembers, and ships.

+ [![GitHub Repo](https://img.shields.io/badge/GitHub-Repo-black.svg?logo=github)](https://github.com/matevip/mateclaw) [![Documentation](https://img.shields.io/badge/Docs-Website-green.svg?logo=readthedocs&label=Docs)](https://claw.mate.vip/docs) [![Live Demo](https://img.shields.io/badge/Demo-Online-orange.svg?logo=vercel&label=Demo)](https://claw-demo.mate.vip) @@ -11,49 +17,177 @@ [![Vue](https://img.shields.io/badge/Vue-3-4FC08D.svg?logo=vuedotjs)](https://vuejs.org/) [![Last Commit](https://img.shields.io/github/last-commit/matevip/mateclaw)](https://github.com/matevip/mateclaw) [![License](https://img.shields.io/badge/license-Apache--2.0-red.svg?logo=opensourceinitiative&label=License)](LICENSE) -[![GitHub Stars](https://img.shields.io/github/stars/matevip/mateclaw?style=flat&logo=github&color=yellow&label=Stars)](https://github.com/matevip/mateclaw/stargazers) -[![GitHub Forks](https://img.shields.io/github/forks/matevip/mateclaw?style=flat&logo=github&color=purple&label=Forks)](https://github.com/matevip/mateclaw/network) [[Website](https://claw.mate.vip)] [[Live Demo](https://claw-demo.mate.vip)] [[Documentation](https://claw.mate.vip/docs)] [[中文](README_zh.md)] -

- MateClaw Logo -

- -

Your AI mate, always ready to lend a claw.

-
-A personal AI assistant system built with **Java + Vue 3**, powered by [Spring AI Alibaba](https://github.com/alibaba/spring-ai-alibaba). Features multi-agent orchestration, a flexible tool/skill system with MCP protocol support, multi-layer memory, and multi-channel adapters. +MateClaw is a personal AI operating system built with **Java + Vue 3** and powered by [Spring AI Alibaba](https://github.com/alibaba/spring-ai-alibaba). -> **Core capabilities:** -> -> **Multi-Agent Orchestration** — ReAct (Thought → Action → Observation loop) and Plan-and-Execute (auto-decompose complex tasks into ordered sub-steps). Create multiple independent agents, each with their own personality and tools. -> -> **Tool & Skill System** — Built-in tools (web search, date/time) + MCP protocol for external tool integration. Pre-configured GitHub and Filesystem MCP servers — enable and go. Install skill packages from ClawHub marketplace or custom sources. -> -> **Multi-Layer Memory** — Short-term context window with auto-compression, event-driven post-conversation memory extraction, workspace files (PROFILE.md / MEMORY.md / daily notes), and scheduled memory consolidation. -> -> **Every Channel** — Web console, DingTalk, Feishu, WeChat Work, Telegram, Discord, QQ. One MateClaw, connect as needed. -> -> **Multi-Provider Models** — 20+ providers: DashScope, OpenAI, Anthropic, Google Gemini, DeepSeek, Kimi, MiniMax, Zhipu AI, Volcano Engine, OpenRouter, Ollama, LM Studio, llama.cpp, MLX, and more. Configure in the web UI. -> -> **Desktop App** — Electron-based desktop application with auto-update support. Download and double-click to run. +It is not just a chat box. It is a system for building AI workers that can reason, use tools, remember context, search the live web, digest knowledge into structured Wiki pages, generate media, and show up across the channels where work actually happens. + +The idea is simple: + +- Give each agent a clear role +- Give it the right tools and guardrails +- Let it keep memory instead of starting from zero +- Let it work across chat, channels, documents, and media +- Keep the whole system deployable by one team without turning into infrastructure theater --- -## Table of Contents +## What MateClaw Is -- [Quick Start](#quick-start) -- [Screenshots](#screenshots) -- [Architecture](#architecture) -- [Tech Stack](#tech-stack) -- [Features](#features) -- [Documentation](#documentation) -- [Roadmap](#roadmap) -- [Contributing](#contributing) -- [Contact Us](#contact-us) -- [License](#license) +MateClaw sits at the intersection of four product ideas: + +1. **An AI console** for direct interaction +2. **An agent runtime** for structured reasoning and tool use +3. **A knowledge system** that turns raw information into reusable memory and Wiki pages +4. **A deployment surface** that spans web, desktop, and external channels + +Most products do one of these. MateClaw is designed to make them work as one system. + +--- + +## Product Principles + +### 1. Agents should do work, not just talk + +MateClaw supports ReAct and Plan-and-Execute agents so the model can break work down, call tools, observe results, and continue instead of stopping at a polished paragraph. + +### 2. Knowledge should be digested, not endlessly re-read + +Raw files are useful, but structured knowledge is better. MateClaw includes an LLM Wiki knowledge base that converts source materials into linked Wiki pages with summaries, backlinks, and on-demand retrieval. + +### 3. Memory should compound + +Conversations should not disappear. MateClaw combines short-term context management, post-conversation extraction, workspace memory files, and scheduled consolidation so agents can build continuity over time. + +### 4. Tools need control, not chaos + +Powerful tools without boundaries are a liability. MateClaw includes tool guard rules, approval flows, file-path protection, and runtime filtering so capability does not become recklessness. + +### 5. AI should live where work already happens + +A useful assistant cannot be trapped in one web page. MateClaw connects to desktop, browser, and external messaging/work channels so the agent can meet users where decisions are being made. + +--- + +## What You Can Build With It + +### Personal AI Workspace + +- A persistent assistant with memory, tools, and workspace files +- A desktop app with bundled backend and auto-update +- A web console for direct chat, planning, and configuration + +### Team Knowledge Assistant + +- Ingest notes, documents, PDFs, and DOCX files +- Turn source materials into structured Wiki pages +- Let agents search, summarize, and read knowledge on demand + +### Tool-Using AI Workers + +- Agents that search the web, read files, use MCP tools, and execute workflows +- Role-specific skill packages with `SKILL.md` +- Approval and security controls for sensitive actions + +### Multimodal Content Workflows + +- Text-to-speech +- Speech-to-text +- Music generation +- Image generation +- Video generation + +### Multi-Channel AI Presence + +- Web console +- DingTalk +- Feishu +- WeChat Work +- Telegram +- Discord +- QQ + +--- + +## Core Capabilities + +### Agent Runtime + +- **ReAct agents** for thought → action → observation loops +- **Plan-and-Execute agents** for decomposing complex work into ordered steps +- **Dynamic agent configuration** loaded at runtime +- **Multi-agent setup** with separate prompts, personalities, and tool scopes +- **Runtime resilience** including context pruning, smart truncation, stale stream cleanup, and recovery for longer tasks + +### Knowledge and Memory + +- **LLM Wiki knowledge base** for structured, linked, AI-digested knowledge +- **Workspace memory files** such as `AGENTS.md`, `SOUL.md`, `PROFILE.md`, `MEMORY.md`, and daily notes +- **Post-conversation extraction** to preserve useful information automatically +- **Scheduled consolidation** so memory quality improves instead of just growing +- **Dreaming and emergence workflows** for longer-horizon memory refinement + +### Tools, Skills, and Search + +- **Built-in tools** for web search, file operations, memory access, date/time, and more +- **Advanced web search** with provider chaining, fallback strategies, and live information support +- **MCP integration** across stdio, SSE, and Streamable HTTP transports +- **Skill system** with installable `SKILL.md` packages +- **ClawHub marketplace** for discovering and installing skills +- **Tool guard and approval** for sensitive operations + +### Multimodal Creation + +- **Text-to-speech** for read-aloud and voice output +- **Speech-to-text** for audio transcription +- **Music generation** +- **Image generation** with multiple providers +- **Video generation** with async task handling + +### Model Flexibility + +Configure models in the web UI. MateClaw supports cloud and local model providers including: + +- DashScope +- OpenAI +- Anthropic +- Google Gemini +- DeepSeek +- Kimi +- MiniMax +- Zhipu AI +- Volcano Engine +- OpenRouter +- Ollama +- LM Studio +- llama.cpp +- MLX + +### Surfaces + +- **Web app** for chat, agent management, MCP, models, tools, channels, and security +- **Desktop app** with bundled JRE 21 and backend +- **External channels** for production-facing assistant workflows + +--- + +## Why The Wiki Matters + +Most AI systems treat knowledge like a warehouse of raw fragments. + +MateClaw adds another layer: a structured Wiki that AI can build and maintain. Instead of retrieving arbitrary chunks from source files every time, the system can pre-digest information into clean pages with summaries and links. + +That changes the product in three ways: + +- Agents waste less context on raw material +- Knowledge becomes easier to inspect and edit by humans +- Understanding improves over time instead of resetting on every query + +This is the difference between storing information and shaping it. --- @@ -62,208 +196,103 @@ A personal AI assistant system built with **Java + Vue 3**, powered by [Spring A ### Prerequisites - Java 17+ -- Node.js 18+ & pnpm +- Node.js 18+ and pnpm - Maven 3.9+ (or use `mvnw`) -- At least one LLM API Key (e.g., [DashScope](https://dashscope.aliyun.com/)) +- At least one LLM API key such as [DashScope](https://dashscope.aliyun.com/) ### Option 1: Local Development -**1. Start the backend** +**Backend** ```bash cd mateclaw-server export DASHSCOPE_API_KEY=your-key-here mvn spring-boot:run -# Backend runs at http://localhost:18088 -# H2 Console: http://localhost:18088/h2-console -# API Docs (SpringDoc OpenAPI): http://localhost:18088/swagger-ui.html ``` -**2. Start the frontend** +Backend: + +- App: `http://localhost:18088` +- H2 Console: `http://localhost:18088/h2-console` +- Swagger UI: `http://localhost:18088/swagger-ui.html` + +**Frontend** ```bash cd mateclaw-ui pnpm install pnpm dev -# Frontend runs at http://localhost:5173 (proxies /api to :18088) ``` -**3. Log in** +Frontend: -Open http://localhost:5173 and log in with `admin` / `admin123`. +- App: `http://localhost:5173` + +**Login** + +- Username: `admin` +- Password: `admin123` ### Option 2: Docker ```bash cp .env.example .env -# Edit .env — fill in DASHSCOPE_API_KEY and other variables - docker compose up -d -# Service runs at http://localhost:18080 (MySQL + backend) ``` -### Option 3: Desktop Application +Default service: -Download the installer from [GitHub Releases](https://github.com/matevip/mateclaw/releases): +- `http://localhost:18080` -| Platform | File | Notes | -|----------|------|-------| -| **macOS (Apple Silicon)** | `MateClaw__arm64.dmg` | Recommended for M1/M2/M3/M4/M5 Mac | -| **macOS (Apple Silicon)** | `MateClaw__arm64.zip` | zip format (Apple Silicon) | -| **macOS (Intel)** | `MateClaw__x64.dmg` | For Intel-based Mac | -| **macOS (Intel)** | `MateClaw__x64.zip` | zip format (Intel) | -| **Windows (x64)** | `MateClaw__Setup.exe` | For most Windows PCs (64-bit) | -| **Windows (x64)** | `MateClaw__x64_Setup.exe` | Explicit x64 build | -| **Windows (ARM64)** | `MateClaw__arm64_Setup.exe` | For ARM-based Windows (e.g. Surface Pro X) | +### Option 3: Desktop App -Double-click to run. The app bundles JRE 21 + the Spring Boot backend, no Java installation needed. Supports auto-update via GitHub Releases. +Download installers from [GitHub Releases](https://github.com/matevip/mateclaw/releases). -> **macOS users**: If macOS blocks the app, right-click → Open → Open again, or go to System Settings → Privacy & Security → Open Anyway. +The desktop app bundles **JRE 21 + the Spring Boot backend**, so users do not need to install Java separately. ---- - -## Screenshots - -

- MateClaw Chat with ECharts Visualization -

- -

Chat console — AI self-introduction with auto-generated ECharts donut chart

+> macOS: if the app is blocked on first launch, use right-click → Open, or allow it in Privacy & Security. --- ## Architecture -``` +```text mateclaw/ -├── mateclaw-server/ # Spring Boot backend -│ ├── src/main/java/vip/mate/ -│ │ ├── agent/ # Agent engine (ReAct, Plan-and-Execute, StateGraph) -│ │ ├── planning/ # Task planning (Plan / SubPlan models) -│ │ ├── tool/ # Tool system (built-in + MCP adapters) -│ │ ├── skill/ # Skill management (workspace + ClawHub) -│ │ ├── channel/ # Channel adapters (Web, DingTalk, Feishu, etc.) -│ │ ├── workspace/ # Conversations, messages, workspace files -│ │ ├── memory/ # Memory extraction & consolidation -│ │ ├── llm/ # Multi-provider model configs -│ │ ├── cron/ # Scheduled tasks (CronJob) -│ │ ├── auth/ # Spring Security + JWT -│ │ └── config/ # Spring bean configurations -│ └── src/main/resources/ -│ ├── application.yml # Main config (H2 for dev) -│ ├── prompts/ # Prompt templates -│ └── db/ # Schema & seed data (schema.sql, data.sql) -├── mateclaw-ui/ # Vue 3 SPA frontend -│ └── src/ -│ ├── views/ # Pages (ChatConsole, AgentWorkspace, SkillMarket, etc.) -│ ├── components/ # Reusable components -│ ├── stores/ # Pinia stores (domain-driven) -│ ├── api/ # Axios HTTP client -│ ├── router/ # Vue Router -│ ├── types/ # TypeScript types -│ └── i18n/ # Internationalization (zh-CN, en-US) -├── mateclaw-desktop/ # Electron desktop app -├── docs/ # VitePress documentation (zh + en) +├── mateclaw-server/ Spring Boot backend +├── mateclaw-ui/ Vue 3 SPA frontend +├── mateclaw-desktop/ Electron desktop app +├── docs/ VitePress documentation ├── docker-compose.yml └── .env.example ``` +Backend domains include: + +- `agent/` for runtime and orchestration +- `tool/` for built-in tools and MCP integration +- `skill/` for skill installation and execution +- `memory/` for extraction, consolidation, and dreaming +- `wiki/` for knowledge base and structured Wiki processing +- `channel/` for external platform adapters +- `workspace/` for files, messages, and conversations + --- ## Tech Stack | Layer | Technology | -|-------|-----------| -| Backend Framework | Spring Boot 3.5 + Spring AI Alibaba 1.1 | -| LLM Integration | DashScope, OpenAI, Anthropic, Gemini, DeepSeek, Kimi, MiniMax, Zhipu, Volcano Engine, OpenRouter, Ollama, LM Studio, llama.cpp, MLX | -| Agent Engine | StateGraph (ReAct + Plan-and-Execute) | +|-------|------------| +| Backend | Spring Boot 3.5 + Spring AI Alibaba 1.1 | +| Agent Runtime | StateGraph | | Database | H2 (dev) / MySQL 8.0+ (prod) | | ORM | MyBatis Plus 3.5 | -| Authentication | Spring Security + JWT | -| API Docs | SpringDoc OpenAPI 3 | +| Auth | Spring Security + JWT | | Frontend | Vue 3 + TypeScript + Vite | -| State Management | Pinia | -| UI Components | Element Plus | +| State | Pinia | +| UI | Element Plus | | Styling | TailwindCSS 4 | | Desktop | Electron + electron-updater | -| Docs Site | VitePress | - ---- - -## Features - -### Agent System - -- **ReAct Agent** — Thought → Action → Observation reasoning loop with tool calling -- **Plan-and-Execute** — Auto-decompose complex tasks into ordered sub-steps with progress tracking -- **Dynamic Agent** — Load agent configs from database at runtime -- **Multi-Agent** — Create multiple independent agents, each with their own system prompt, tools, and personality - -### Tool & Skill System - -- **Built-in Tools** — Web search (Serper/Tavily), date/time, workspace memory read/write -- **MCP Protocol** — Connect external tools via Model Context Protocol (stdio, SSE, and Streamable HTTP transports). Full lifecycle management in the web UI — add, edit, enable/disable, and test connections -- **Pre-configured MCP Servers** — GitHub (`@modelcontextprotocol/server-github`) and Filesystem ship out-of-the-box. Enable from the MCP management page and fill in your token — no code changes needed -- **Skill Packages** — Install/uninstall skill packages with `SKILL.md` manifests -- **ClawHub Marketplace** — Browse and install skills from the ClawHub registry -- **Workspace Skills** — Convention-based skill directory at `~/.mateclaw/skills/{name}/` - -### Memory System - -- **Short-Term** — Conversation context window with auto-compression when token budget exceeded -- **Post-Conversation Extraction** — Event-driven async LLM analysis, writes to PROFILE.md / MEMORY.md / daily notes -- **Memory Consolidation** — Scheduled daily emergence (CronJob at 2:00 AM) merges daily notes into long-term memory -- **Workspace Files** — Per-agent AGENTS.md, SOUL.md, PROFILE.md, MEMORY.md, memory/*.md -- **Agent Memory Tool** — Agents can read/write their own workspace files during conversations - -### Multi-Channel - -- **Web Console** — SSE streaming with rich message rendering (Markdown, code, plans) -- **DingTalk** — Webhook + event subscription -- **Feishu (Lark)** — Webhook + event subscription -- **WeChat Work** — Callback API -- **Telegram** — Bot API with webhook -- **Discord** — Bot with slash commands -- **QQ** — QQ Bot API - -### Model Providers - -Configure in the web UI (Settings → Models). Supported providers: - -| Provider | Models | -|----------|--------| -| **Cloud Providers** | | -| DashScope (Alibaba) | Qwen3.5-Max, Qwen3.5-Plus, Qwen3-Max, Qwen3-Plus, Qwen-Max, Qwen-Plus, Qwen-Turbo, Qwen-Long, DeepSeek-V3.2 | -| ModelScope | Qwen3.5-122B-A10B, GLM-5 | -| Aliyun Coding Plan | Qwen3.5-Plus, Qwen3-Coder-Next, GLM-5, GLM-4.7, MiniMax-M2.5, Kimi-K2.5 | -| OpenAI | GPT-5.2, GPT-5, GPT-5-Mini, GPT-5-Nano, GPT-4.1, GPT-4.1-Mini, GPT-4.1-Nano, o3, o4-mini, GPT-4o | -| Azure OpenAI | GPT-5, GPT-4.1, GPT-4o and more | -| Anthropic | Claude Opus 4.6, Claude Sonnet 4.6 (via model discovery) | -| Google Gemini | Gemini 3.1 Pro, Gemini 3 Flash, Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 2.0 Flash | -| DeepSeek | DeepSeek-Chat, DeepSeek-Reasoner | -| Kimi (Moonshot) | Kimi-K2.5, Kimi-K2-Thinking, Kimi for Coding (CN / International / Code) | -| MiniMax | MiniMax-M2.7, MiniMax-M2.5 (International / China) | -| Zhipu AI | GLM-5.1, GLM-5, GLM-5-Turbo, GLM-5V-Turbo (CN / International) | -| Volcano Engine | Doubao-1.5-Pro-256K, Doubao-1.5-Lite, Doubao-1.5-Thinking-Pro, Doubao-1.5-Vision-Pro | -| OpenRouter | GPT-5, Claude Opus 4.6, Gemini 2.5 Pro, Llama 4 Maverick, DeepSeek R1, and 200+ more | -| **Local Providers** | | -| Ollama | Qwen3, Gemma 4, Gemma 3, Llama 3.1, DeepSeek R1, Mistral (auto-detected on startup) | -| LM Studio | Any locally-served model | -| llama.cpp | Any locally-served model | -| MLX (Apple Silicon) | Any locally-served model | - -### Security - -- **Spring Security + JWT** — Token-based authentication -- **Tool Guard** — Approval rules for sensitive tool operations -- **File Validation** — Path traversal prevention for workspace files -- **Skill Security** — Validation during skill installation - -### Scheduled Tasks - -- **CronJob System** — Create scheduled tasks with 5-field cron expressions -- **Memory Consolidation** — Auto-triggered daily for each agent -- **Custom Tasks** — Schedule any prompt to run periodically +| Docs | VitePress | --- @@ -271,80 +300,60 @@ Configure in the web UI (Settings → Models). Supported providers: | Topic | Description | |-------|-------------| -| [Introduction](https://mateclaw.mate.vip/en/intro) | What MateClaw is and core concepts | -| [Quick Start](https://mateclaw.mate.vip/en/quickstart) | Install and run (local, Docker, desktop) | -| [Console](https://mateclaw.mate.vip/en/console) | Web UI: chat and agent configuration | -| [Agents](https://mateclaw.mate.vip/en/agents) | Agent engine: ReAct, Plan-and-Execute, StateGraph | -| [Models](https://mateclaw.mate.vip/en/models) | Configure cloud, local, and custom providers | -| [Tools](https://mateclaw.mate.vip/en/tools) | Built-in tools and custom tool development | -| [Skills](https://mateclaw.mate.vip/en/skills) | Skill packages and ClawHub marketplace | +| [Introduction](https://mateclaw.mate.vip/en/intro) | Product overview and core concepts | +| [Quick Start](https://mateclaw.mate.vip/en/quickstart) | Local, Docker, and desktop setup | +| [Console](https://mateclaw.mate.vip/en/console) | Web console and day-to-day usage | +| [Agents](https://mateclaw.mate.vip/en/agents) | ReAct, Plan-and-Execute, and runtime design | +| [Models](https://mateclaw.mate.vip/en/models) | Model provider setup | +| [Tools](https://mateclaw.mate.vip/en/tools) | Built-in tools and extension model | +| [Skills](https://mateclaw.mate.vip/en/skills) | Skill packages and marketplace | | [MCP](https://mateclaw.mate.vip/en/mcp) | Model Context Protocol integration | -| [Memory](https://mateclaw.mate.vip/en/memory) | Multi-layer memory system | -| [Channels](https://mateclaw.mate.vip/en/channels) | DingTalk, Feishu, Telegram, Discord, and more | -| [Security](https://mateclaw.mate.vip/en/security) | Authentication and tool guard | +| [Memory](https://mateclaw.mate.vip/en/memory) | Memory architecture | +| [Channels](https://mateclaw.mate.vip/en/channels) | External channel integration | +| [Security](https://mateclaw.mate.vip/en/security) | Guardrails and approval | | [Desktop](https://mateclaw.mate.vip/en/desktop) | Desktop application guide | -| [API Reference](https://mateclaw.mate.vip/en/api) | REST API documentation | -| [Configuration](https://mateclaw.mate.vip/en/config) | Configuration reference | -| [FAQ](https://mateclaw.mate.vip/en/faq) | Common questions and troubleshooting | +| [API Reference](https://mateclaw.mate.vip/en/api) | REST API | +| [FAQ](https://mateclaw.mate.vip/en/faq) | Troubleshooting | --- ## Roadmap -| Area | Item | Status | -|------|------|--------| -| **Agent** | Multi-agent collaboration and delegation | Planned | -| **Agent** | Multimodal input (image, audio, video) | Planned | -| **Models** | Small + large model routing | Planned | -| **Memory** | Vector DB long-term memory (RAG) | Planned | -| **Memory** | Multimodal memory fusion | Planned | -| **Skills** | Richer ClawHub ecosystem | In Progress | -| **Channels** | WeChat personal (iLink Bot) | Planned | -| **Channels** | Email channel | Planned | -| **Desktop** | Linux support | Planned | -| **Security** | Multi-tenant support | Planned | -| **Console** | Plugin marketplace in web UI | Planned | +Current focus areas include: -_Status:_ **In Progress** — actively being worked on; **Planned** — queued or under design. +- richer multi-agent collaboration +- deeper multimodal understanding +- smarter model routing +- stronger long-term memory +- richer ClawHub ecosystem +- more channels and desktop coverage --- ## Contributing -MateClaw is open to contributions! Whether it's bug fixes, new features, documentation improvements, or new channel/tool integrations — all contributions are welcome. +MateClaw is open to product, code, docs, and integration contributions. ```bash -# Clone the repository git clone https://github.com/matevip/mateclaw.git cd mateclaw -# Backend cd mateclaw-server mvn clean compile -# Frontend cd ../mateclaw-ui pnpm install pnpm dev ``` -Please read [CONTRIBUTING.md](https://github.com/matevip/mateclaw/blob/main/CONTRIBUTING.md) (if available) before submitting a PR. - --- -## Contact Us +## Why The Name - +**Mate** means companion. +**Claw** means capability. -| Discord | X (Twitter) | DingTalk | -|---------|-------------|----------| -| Coming soon | Coming soon | Coming soon | - ---- - -## Why MateClaw? - -**Mate** — a companion, always by your side. **Claw** — sharp, capable, ready to grab any task. MateClaw is your personal AI mate that lends a claw whenever you need it. Built as a monolith with modular design, it's easy to deploy, extend, and customize. +The product is meant to feel like both: a system that stays with you, and a system that can actually grab work and move it. --- diff --git a/README_zh.md b/README_zh.md index 96171b4e..11c575f6 100644 --- a/README_zh.md +++ b/README_zh.md @@ -1,7 +1,13 @@
+

+ MateClaw Logo +

+ # MateClaw +

让 AI 真正去思考、行动、记忆，并把结果交付出来。

+ [![GitHub 仓库](https://img.shields.io/badge/GitHub-仓库-black.svg?logo=github)](https://github.com/matevip/mateclaw) [![文档](https://img.shields.io/badge/文档-在线-green.svg?logo=readthedocs&label=Docs)](https://claw.mate.vip/docs) [![在线演示](https://img.shields.io/badge/演示-在线-orange.svg?logo=vercel&label=Demo)](https://claw-demo.mate.vip) @@ -11,49 +17,177 @@ [![Vue](https://img.shields.io/badge/Vue-3-4FC08D.svg?logo=vuedotjs)](https://vuejs.org/) [![最后提交](https://img.shields.io/github/last-commit/matevip/mateclaw)](https://github.com/matevip/mateclaw) [![许可证](https://img.shields.io/badge/license-Apache--2.0-red.svg?logo=opensourceinitiative&label=License)](LICENSE) -[![GitHub Star](https://img.shields.io/github/stars/matevip/mateclaw?style=flat&logo=github&color=yellow&label=Stars)](https://github.com/matevip/mateclaw/stargazers) -[![GitHub Fork](https://img.shields.io/github/forks/matevip/mateclaw?style=flat&logo=github&color=purple&label=Forks)](https://github.com/matevip/mateclaw/network) [[官网](https://claw.mate.vip)] [[在线演示](https://claw-demo.mate.vip)] [[文档](https://claw.mate.vip/docs)] [[English](README.md)] -

- MateClaw Logo -

- -

懂你所需，利爪随行。

-
-基于 **Java + Vue 3** 的个人 AI 助手系统，由 [Spring AI Alibaba](https://github.com/alibaba/spring-ai-alibaba) 驱动。支持多 Agent 编排、灵活的工具/技能系统与 MCP 协议、多层记忆体系、多渠道接入。 +MateClaw 是一个基于 **Java + Vue 3** 构建的个人 AI 操作系统，由 [Spring AI Alibaba](https://github.com/alibaba/spring-ai-alibaba) 驱动。 -> **核心能力：** -> -> **多 Agent 编排** — ReAct（思考→行动→观察循环）和 Plan-and-Execute（自动将复杂任务拆解为有序子步骤）。创建多个独立 Agent，各有专属人格和工具。 -> -> **工具与技能系统** — 内置工具（网络搜索、日期时间）+ MCP 协议接入外部工具。预置 GitHub 和 Filesystem MCP 服务，启用即用。从 ClawHub 市场或自定义源安装技能包。 -> -> **多层记忆** — 短期上下文窗口自动压缩、事件驱动的对话后记忆提取、工作空间文件（PROFILE.md / MEMORY.md / 每日笔记）、定时记忆整合。 -> -> **全域触达** — Web 控制台、钉钉、飞书、企业微信、Telegram、Discord、QQ。一个 MateClaw，按需连接。 -> -> **多厂商模型** — 20+ 厂商：DashScope、OpenAI、Anthropic、Google Gemini、DeepSeek、Kimi、MiniMax、智谱、火山引擎、OpenRouter、Ollama、LM Studio、llama.cpp、MLX 等。在 Web 界面中配置。 -> -> **桌面应用** — 基于 Electron 的桌面应用，支持自动更新。下载即用。 +它不是另一个聊天框，而是一整套 AI 工作系统：能推理、能调工具、能记住上下文、能联网、能把原始资料消化成 Wiki、能生成多模态内容，也能出现在真正发生工作的渠道里。 + +这个产品的核心想法很简单： + +- 给每个 Agent 清晰的角色 +- 给它真正可用的工具和边界 +- 让它拥有记忆，而不是每次从零开始 +- 让它跨聊天、文档、知识、媒体和渠道协同工作 +- 保持整套系统能被一个团队部署、理解和持续迭代 --- -## 目录 +## MateClaw 到底是什么 -- [快速开始](#快速开始) -- [截图](#截图) -- [架构](#架构) -- [技术栈](#技术栈) -- [功能特性](#功能特性) -- [文档](#文档) -- [路线图](#路线图) -- [参与贡献](#参与贡献) -- [联系我们](#联系我们) -- [许可证](#许可证) +MateClaw 不是单点产品，而是 4 个产品层合成的一个系统： + +1. **AI 控制台**：人直接和 AI 交互 +2. **Agent 运行时**：让模型分步推理、调用工具、执行任务 +3. **知识系统**：把原始信息沉淀成记忆和结构化 Wiki +4. **交付面**：覆盖 Web、桌面和外部渠道 + +大多数产品只做其中一层。MateClaw 的目标，是把这四层做成一个完整工作系统。 + +--- + +## 产品思路 + +### 1. Agent 不该只会聊天，应该会干活 + +MateClaw 支持 ReAct 和 Plan-and-Execute。模型不只是生成回答，而是能拆解任务、调用工具、观察结果，再继续推进。 + +### 2. 知识不该反复原样读取，而应该被消化 + +原始文档很重要，但结构化知识更重要。MateClaw 内置 LLM Wiki 知识库，把文本、PDF、DOCX 等材料消化成可链接、可搜索、可编辑的 Wiki 页面。 + +### 3. 记忆应该越用越值钱 + +会话结束不该等于遗忘。MateClaw 把短期上下文、对话后提取、工作空间记忆文件和定时整合放进同一套体系，让 Agent 能积累连续性。 + +### 4. 工具要强，但不能失控 + +没有边界的工具系统不是能力，是事故源。MateClaw 提供工具防护、审批、路径校验和运行时过滤，让强能力能被放心使用。 + +### 5. AI 必须出现在真实工作的地方 + +真正有用的 AI 不能只困在一个网页里。MateClaw 连接桌面、Web 和外部消息渠道，让 Agent 出现在任务发生的地方。 + +--- + +## 你可以拿它做什么 + +### 个人 AI 工作台 + +- 一个有记忆、有工具、有工作空间文件的长期助手 +- 一个开箱即用的桌面应用 +- 一个用于聊天、规划和配置的 Web 控制台 + +### 团队知识助手 + +- 导入笔记、文档、PDF、DOCX +- 把原始材料转成结构化 Wiki 页面 +- 让 Agent 按需搜索、总结、阅读知识，而不是反复扫描原文 + +### 会用工具的 AI Worker + +- 能联网搜索、读文件、接 MCP 工具、执行工作流的 Agent +- 按角色安装技能包 +- 对敏感动作进行审批和防护 + +### 多模态内容生产系统 + +- 文字转语音 +- 语音转文字 +- 音乐生成 +- 图片生成 +- 视频生成 + +### 多渠道 AI 存在 + +- Web 控制台 +- 钉钉 +- 飞书 +- 企业微信 +- Telegram +- Discord +- QQ + +--- + +## 核心能力 + +### Agent 运行时 + +- **ReAct Agent**：支持思考 → 行动 → 观察循环 +- **Plan-and-Execute Agent**：适合拆解复杂任务并按步骤执行 +- **动态 Agent 配置**：运行时加载，不需要把配置写死 +- **多 Agent 体系**：每个 Agent 有自己的提示词、人格、工具范围 +- **更稳定的长任务执行**：支持上下文裁剪、智能截断、陈旧流清理、恢复机制 + +### 知识与记忆 + +- **LLM Wiki 知识库**：把原始材料转成结构化、可链接的 Wiki +- **工作空间记忆文件**：如 `AGENTS.md`、`SOUL.md`、`PROFILE.md`、`MEMORY.md`、daily notes +- **对话后自动提取**：把有价值的信息沉淀下来 +- **定时整合**：不是一味堆积，而是持续整理 +- **Dreaming / Emergence 记忆机制**：用于更长时间尺度上的记忆优化 + +### 工具、技能与搜索 + +- **内置工具**：搜索、文件、记忆、时间等能力 +- **更强的联网搜索**：支持多 Provider、回退链和实时信息获取 +- **MCP 集成**：支持 stdio、SSE、Streamable HTTP +- **技能系统**：通过 `SKILL.md` 安装和组织技能 +- **ClawHub 市场**：发现和安装技能 +- **工具防护与审批**：保障高权限操作的可控性 + +### 多模态创作 + +- **文字转语音** +- **语音转文字** +- **音乐生成** +- **图片生成** +- **视频生成** + +### 模型灵活性 + +可在 Web 界面中配置云端与本地模型，支持： + +- DashScope +- OpenAI +- Anthropic +- Google Gemini +- DeepSeek +- Kimi +- MiniMax +- 智谱 AI +- 火山引擎 +- OpenRouter +- Ollama +- LM Studio +- llama.cpp +- MLX + +### 使用入口 + +- **Web 应用**：聊天、Agent、MCP、模型、工具、渠道、安全配置 +- **桌面应用**：内置 JRE 21 和后端 +- **外部渠道**：适合真正面向业务场景的接入 + +--- + +## 为什么 Wiki 很重要 + +多数 AI 系统把知识当成“原始碎片仓库”。 + +MateClaw 多做了一层：让 AI 把知识整理成结构化 Wiki。不是每次临时从原始文档里切几段，而是先把知识变成清晰页面，再按需读取。 + +这会带来三个变化： + +- Agent 不再把大量上下文浪费在原始材料上 +- 人可以直接检查、编辑、维护知识结构 +- 知识会随着使用而变得更清晰，而不是每次查询都重新理解 + +这就是“存信息”和“塑造知识”的区别。 --- @@ -62,208 +196,103 @@ ### 前置条件 - Java 17+ -- Node.js 18+ & pnpm +- Node.js 18+ 和 pnpm - Maven 3.9+（或使用 `mvnw`） -- 至少一个 LLM API Key（如 [DashScope](https://dashscope.aliyun.com/)） +- 至少一个 LLM API Key，例如 [DashScope](https://dashscope.aliyun.com/) ### 方式一：本地开发 -**1. 启动后端** +**启动后端** ```bash cd mateclaw-server export DASHSCOPE_API_KEY=your-key-here mvn spring-boot:run -# 后端运行在 http://localhost:18088 -# H2 控制台：http://localhost:18088/h2-console -# API 文档（SpringDoc OpenAPI）：http://localhost:18088/swagger-ui.html ``` -**2. 启动前端** +后端地址： + +- 应用：`http://localhost:18088` +- H2 Console：`http://localhost:18088/h2-console` +- Swagger UI：`http://localhost:18088/swagger-ui.html` + +**启动前端** ```bash cd mateclaw-ui pnpm install pnpm dev -# 前端运行在 http://localhost:5173（代理 /api 到 :18088） ``` -**3. 登录** +前端地址： -打开 http://localhost:5173，使用 `admin` / `admin123` 登录。 +- 应用：`http://localhost:5173` -### 方式二：Docker 部署 +**登录** + +- 用户名：`admin` +- 密码：`admin123` + +### 方式二：Docker ```bash cp .env.example .env -# 编辑 .env，填写 DASHSCOPE_API_KEY 等变量 - docker compose up -d -# 服务运行在 http://localhost:18080（MySQL + 后端） ``` +默认服务地址： + +- `http://localhost:18080` + ### 方式三：桌面应用 -从 [GitHub Releases](https://github.com/matevip/mateclaw/releases) 下载安装包： +从 [GitHub Releases](https://github.com/matevip/mateclaw/releases) 下载桌面安装包。 -| 平台 | 文件 | 说明 | -|------|------|------| -| **macOS (Apple Silicon)** | `MateClaw__arm64.dmg` | 推荐 M1/M2/M3/M4/M5 Mac 使用 | -| **macOS (Apple Silicon)** | `MateClaw__arm64.zip` | zip 格式（Apple Silicon） | -| **macOS (Intel)** | `MateClaw__x64.dmg` | 适用于 Intel Mac | -| **macOS (Intel)** | `MateClaw__x64.zip` | zip 格式（Intel） | -| **Windows (x64)** | `MateClaw__Setup.exe` | 适用于大多数 Windows 电脑（64 位） | -| **Windows (x64)** | `MateClaw__x64_Setup.exe` | 显式 x64 构建 | -| **Windows (ARM64)** | `MateClaw__arm64_Setup.exe` | 适用于 ARM 架构 Windows（如 Surface Pro X） | +桌面应用内置 **JRE 21 + Spring Boot 后端**，无需额外安装 Java。 -双击运行。应用内置 JRE 21 + Spring Boot 后端，无需安装 Java。支持从 GitHub Releases 自动更新。 - -> **macOS 用户**：如果系统阻止打开，右键 → 打开 → 再次点击打开，或前往系统设置 → 隐私与安全性 → 仍要打开。 - ---- - -## 截图 - -

- MateClaw 对话界面 — ECharts 可视化 -

- -

对话控制台 — AI 自我介绍并自动生成 ECharts 环形图

+> macOS：如果首次打开被系统拦截，使用右键 → 打开，或在隐私与安全性中手动允许。 --- ## 架构 -``` +```text mateclaw/ -├── mateclaw-server/ # Spring Boot 后端 -│ ├── src/main/java/vip/mate/ -│ │ ├── agent/ # Agent 引擎（ReAct、Plan-and-Execute、StateGraph） -│ │ ├── planning/ # 任务规划（Plan / SubPlan 模型） -│ │ ├── tool/ # 工具系统（内置 + MCP 适配器） -│ │ ├── skill/ # 技能管理（工作空间 + ClawHub） -│ │ ├── channel/ # 渠道适配器（Web、钉钉、飞书等） -│ │ ├── workspace/ # 会话、消息、工作空间文件 -│ │ ├── memory/ # 记忆提取与整合 -│ │ ├── llm/ # 多厂商模型配置 -│ │ ├── cron/ # 定时任务（CronJob） -│ │ ├── auth/ # Spring Security + JWT -│ │ └── config/ # Spring Bean 配置 -│ └── src/main/resources/ -│ ├── application.yml # 主配置（开发环境用 H2） -│ ├── prompts/ # 提示词模板 -│ └── db/ # 数据库脚本（schema.sql、data.sql） -├── mateclaw-ui/ # Vue 3 SPA 前端 -│ └── src/ -│ ├── views/ # 页面（ChatConsole、AgentWorkspace、SkillMarket 等） -│ ├── components/ # 复用组件 -│ ├── stores/ # Pinia 状态管理（领域驱动） -│ ├── api/ # Axios HTTP 客户端 -│ ├── router/ # Vue Router -│ ├── types/ # TypeScript 类型 -│ └── i18n/ # 国际化（zh-CN、en-US） -├── mateclaw-desktop/ # Electron 桌面应用 -├── docs/ # VitePress 文档站（中 + 英） +├── mateclaw-server/ Spring Boot 后端 +├── mateclaw-ui/ Vue 3 SPA 前端 +├── mateclaw-desktop/ Electron 桌面端 +├── docs/ VitePress 文档 ├── docker-compose.yml └── .env.example ``` +后端核心领域包括： + +- `agent/`：Agent 运行时与编排 +- `tool/`：内置工具与 MCP 集成 +- `skill/`：技能安装与执行 +- `memory/`：提取、整合、dreaming +- `wiki/`：知识库与结构化 Wiki 处理 +- `channel/`：外部渠道适配 +- `workspace/`：文件、消息、会话 + --- ## 技术栈 -| 层次 | 技术选型 | -|------|---------| -| 后端框架 | Spring Boot 3.5 + Spring AI Alibaba 1.1 | -| 大模型接入 | DashScope、OpenAI、Anthropic、Gemini、DeepSeek、Kimi、MiniMax、智谱、火山引擎、OpenRouter、Ollama、LM Studio、llama.cpp、MLX | -| Agent 引擎 | StateGraph（ReAct + Plan-and-Execute） | +| 层次 | 技术 | +|------|------| +| 后端 | Spring Boot 3.5 + Spring AI Alibaba 1.1 | +| Agent Runtime | StateGraph | | 数据库 | H2（开发）/ MySQL 8.0+（生产） | | ORM | MyBatis Plus 3.5 | | 认证 | Spring Security + JWT | -| API 文档 | SpringDoc OpenAPI 3 | -| 前端框架 | Vue 3 + TypeScript + Vite | +| 前端 | Vue 3 + TypeScript + Vite | | 状态管理 | Pinia | -| UI 组件 | Element Plus | +| UI | Element Plus | | 样式 | TailwindCSS 4 | | 桌面端 | Electron + electron-updater | -| 文档站 | VitePress | - ---- - -## 功能特性 - -### Agent 系统 - -- **ReAct Agent** — 思考→行动→观察推理循环，支持工具调用 -- **Plan-and-Execute** — 自动将复杂任务拆解为有序子步骤，带进度追踪 -- **动态 Agent** — 运行时从数据库加载 Agent 配置 -- **多 Agent** — 创建多个独立 Agent，各有专属系统提示词、工具和人格 - -### 工具与技能系统 - -- **内置工具** — 网络搜索（Serper/Tavily）、日期时间、工作空间记忆读写 -- **MCP 协议** — 通过 Model Context Protocol 接入外部工具（stdio、SSE、Streamable HTTP 传输）。在 Web 界面中完成全生命周期管理 — 添加、编辑、启用/禁用、测试连接 -- **预置 MCP 服务** — GitHub（`@modelcontextprotocol/server-github`）和 Filesystem 开箱即用。在 MCP 管理页面启用并填入 Token 即可，无需改代码 -- **技能包** — 安装/卸载带 `SKILL.md` 清单的技能包 -- **ClawHub 市场** — 从 ClawHub 注册中心浏览和安装技能 -- **工作空间技能** — 基于约定的技能目录 `~/.mateclaw/skills/{name}/` - -### 记忆系统 - -- **短期记忆** — 会话上下文窗口，Token 超出预算时自动压缩 -- **对话后提取** — 事件驱动的异步 LLM 分析，写入 PROFILE.md / MEMORY.md / 每日笔记 -- **记忆整合** — 定时每日涌现（CronJob 凌晨 2:00），将每日笔记合并为长期记忆 -- **工作空间文件** — 每个 Agent 独立的 AGENTS.md、SOUL.md、PROFILE.md、MEMORY.md、memory/*.md -- **Agent 记忆工具** — Agent 在对话中可主动读写自己的工作空间文件 - -### 多渠道接入 - -- **Web 控制台** — SSE 流式输出，富消息渲染（Markdown、代码、计划） -- **钉钉** — Webhook + 事件订阅 -- **飞书** — Webhook + 事件订阅 -- **企业微信** — 回调接口 -- **Telegram** — Bot API + Webhook -- **Discord** — Bot + Slash Commands -- **QQ** — QQ Bot API - -### 模型厂商 - -在 Web 界面中配置（设置 → 模型）。支持的厂商： - -| 厂商 | 模型 | -|------|------| -| **云端厂商** | | -| DashScope（阿里云） | Qwen3.5-Max、Qwen3.5-Plus、Qwen3-Max、Qwen3-Plus、Qwen-Max、Qwen-Plus、Qwen-Turbo、Qwen-Long、DeepSeek-V3.2 | -| ModelScope | Qwen3.5-122B-A10B、GLM-5 | -| 阿里云百炼 Coding Plan | Qwen3.5-Plus、Qwen3-Coder-Next、GLM-5、GLM-4.7、MiniMax-M2.5、Kimi-K2.5 | -| OpenAI | GPT-5.2、GPT-5、GPT-5-Mini、GPT-5-Nano、GPT-4.1、GPT-4.1-Mini、GPT-4.1-Nano、o3、o4-mini、GPT-4o | -| Azure OpenAI | GPT-5、GPT-4.1、GPT-4o 等 | -| Anthropic | Claude Opus 4.6、Claude Sonnet 4.6（通过模型发现） | -| Google Gemini | Gemini 3.1 Pro、Gemini 3 Flash、Gemini 2.5 Pro、Gemini 2.5 Flash、Gemini 2.0 Flash | -| DeepSeek | DeepSeek-Chat、DeepSeek-Reasoner | -| Kimi（月之暗面） | Kimi-K2.5、Kimi-K2-Thinking、Kimi for Coding（国内版 / 国际版 / Code 版） | -| MiniMax | MiniMax-M2.7、MiniMax-M2.5（国际版 / 国内版） | -| 智谱 AI | GLM-5.1、GLM-5、GLM-5-Turbo、GLM-5V-Turbo（国内版 / 国际版） | -| 火山引擎 | 豆包-1.5-Pro-256K、豆包-1.5-Lite、豆包-1.5-Thinking-Pro、豆包-1.5-Vision-Pro | -| OpenRouter | GPT-5、Claude Opus 4.6、Gemini 2.5 Pro、Llama 4 Maverick、DeepSeek R1 及 200+ 模型 | -| **本地厂商** | | -| Ollama | Qwen3、Gemma 4、Gemma 3、Llama 3.1、DeepSeek R1、Mistral（启动时自动检测） | -| LM Studio | 任意本地服务的模型 | -| llama.cpp | 任意本地服务的模型 | -| MLX（Apple Silicon） | 任意本地服务的模型 | - -### 安全 - -- **Spring Security + JWT** — 基于 Token 的认证 -- **工具防护** — 敏感工具操作的审批规则 -- **文件校验** — 工作空间文件路径穿越防护 -- **技能安全** — 技能安装时的安全校验 - -### 定时任务 - -- **CronJob 系统** — 使用 5 位 cron 表达式创建定时任务 -- **记忆整合** — 每个 Agent 每日自动触发 -- **自定义任务** — 调度任意提示词定期执行 +| 文档 | VitePress | --- @@ -271,80 +300,60 @@ mateclaw/ | 主题 | 说明 | |------|------| -| [项目介绍](https://mateclaw.mate.vip/zh/intro) | MateClaw 是什么、核心概念 | -| [快速开始](https://mateclaw.mate.vip/zh/quickstart) | 安装与运行（本地、Docker、桌面） | -| [控制台](https://mateclaw.mate.vip/zh/console) | Web 界面：聊天与 Agent 配置 | -| [Agent 引擎](https://mateclaw.mate.vip/zh/agents) | ReAct、Plan-and-Execute、StateGraph | -| [模型配置](https://mateclaw.mate.vip/zh/models) | 配置云端、本地和自定义厂商 | -| [工具系统](https://mateclaw.mate.vip/zh/tools) | 内置工具与自定义工具开发 | -| [技能系统](https://mateclaw.mate.vip/zh/skills) | 技能包与 ClawHub 市场 | +| [项目介绍](https://mateclaw.mate.vip/zh/intro) | 产品定位与核心概念 | +| [快速开始](https://mateclaw.mate.vip/zh/quickstart) | 本地、Docker、桌面启动 | +| [控制台](https://mateclaw.mate.vip/zh/console) | Web 控制台日常使用 | +| [Agents](https://mateclaw.mate.vip/zh/agents) | ReAct、Plan-and-Execute 与运行时设计 | +| [模型配置](https://mateclaw.mate.vip/zh/models) | 模型 Provider 配置 | +| [工具系统](https://mateclaw.mate.vip/zh/tools) | 内置工具与扩展能力 | +| [技能系统](https://mateclaw.mate.vip/zh/skills) | 技能包与市场 | | [MCP](https://mateclaw.mate.vip/zh/mcp) | Model Context Protocol 集成 | -| [记忆系统](https://mateclaw.mate.vip/zh/memory) | 多层记忆体系 | -| [渠道接入](https://mateclaw.mate.vip/zh/channels) | 钉钉、飞书、Telegram、Discord 等 | -| [安全机制](https://mateclaw.mate.vip/zh/security) | 认证与工具防护 | -| [桌面应用](https://mateclaw.mate.vip/zh/desktop) | 桌面应用使用指南 | -| [API 参考](https://mateclaw.mate.vip/zh/api) | REST API 文档 | -| [配置指南](https://mateclaw.mate.vip/zh/config) | 配置参考 | -| [常见问题](https://mateclaw.mate.vip/zh/faq) | 常见问题与故障排查 | +| [记忆系统](https://mateclaw.mate.vip/zh/memory) | 记忆架构 | +| [渠道接入](https://mateclaw.mate.vip/zh/channels) | 外部渠道适配 | +| [安全机制](https://mateclaw.mate.vip/zh/security) | 防护与审批 | +| [桌面应用](https://mateclaw.mate.vip/zh/desktop) | 桌面端使用指南 | +| [API 参考](https://mateclaw.mate.vip/zh/api) | REST API | +| [常见问题](https://mateclaw.mate.vip/zh/faq) | 排障与说明 | --- ## 路线图 -| 方向 | 事项 | 状态 | -|------|------|------| -| **Agent** | 多 Agent 协作与任务委派 | 计划中 | -| **Agent** | 多模态输入（图片、音频、视频） | 计划中 | -| **模型** | 大小模型智能路由 | 计划中 | -| **记忆** | 向量数据库长期记忆（RAG） | 计划中 | -| **记忆** | 多模态记忆融合 | 计划中 | -| **技能** | 丰富 ClawHub 生态 | 进行中 | -| **渠道** | 微信个人号（iLink Bot） | 计划中 | -| **渠道** | 邮件渠道 | 计划中 | -| **桌面** | Linux 支持 | 计划中 | -| **安全** | 多租户支持 | 计划中 | -| **控制台** | Web 端插件市场 | 计划中 | +当前重点方向包括： -_状态说明：_ **进行中** — 正在开发；**计划中** — 排期中或设计阶段。 +- 更强的多 Agent 协作 +- 更深的多模态理解 +- 更聪明的模型路由 +- 更强的长期记忆 +- 更丰富的 ClawHub 生态 +- 更多渠道与桌面端覆盖 --- ## 参与贡献 -MateClaw 欢迎各种形式的贡献！无论是 Bug 修复、新功能、文档改进，还是新的渠道/工具集成，我们都非常欢迎。 +MateClaw 欢迎产品、代码、文档、集成四类贡献。 ```bash -# 克隆仓库 git clone https://github.com/matevip/mateclaw.git cd mateclaw -# 后端 cd mateclaw-server mvn clean compile -# 前端 cd ../mateclaw-ui pnpm install pnpm dev ``` -提交 PR 前请阅读 [CONTRIBUTING.md](https://github.com/matevip/mateclaw/blob/main/CONTRIBUTING.md)（如有）。 - --- -## 联系我们 +## 为什么叫 MateClaw - +**Mate** 是伙伴。 +**Claw** 是能力。 -| Discord | X (Twitter) | 钉钉群 | -|---------|-------------|--------| -| 即将上线 | 即将上线 | 即将上线 | - ---- - -## 为什么叫 MateClaw？ - -**Mate** — 伙伴，始终陪伴在你身边。**Claw** — 利爪，锋利有力，随时抓取任何任务。MateClaw 是你的个人 AI 伙伴，在你需要时伸出利爪。采用单体模块化设计，部署简单、扩展灵活、定制方便。 +这个产品想给人的感受，就是两者同时成立：它不是只陪你说话，也能真正抓住任务，把事情往前推进。 --- diff --git a/mateclaw-server/src/main/resources/application.yml b/mateclaw-server/src/main/resources/application.yml index 88cfa9cf..3e30a49b 100644 --- a/mateclaw-server/src/main/resources/application.yml +++ b/mateclaw-server/src/main/resources/application.yml @@ -113,3 +113,13 @@ mate: compact-trigger-ratio: 0.75 preserve-recent-pairs: 2 summary-max-tokens: 300 + # Wiki 知识库配置 + wiki: + enabled: true # 是否启用 Wiki 知识库功能 + max-chunk-size: 30000 # LLM 单次处理最大字符数（超过则分块） + max-context-chars: 10000 # 注入 Agent prompt 的 Wiki 摘要最大字符数 + max-pages-per-raw: 15 # 单个原始材料最多生成的 Wiki 页面数 + auto-process-on-upload: true # 上传原始材料后是否自动触发 AI 消化 + upload-dir: ./data/wiki-uploads # 上传文件存储目录 + max-scan-files: 500 # 目录扫描最大文件数 + max-scan-file-size: 52428800 # 扫描时跳过大于此大小的文件（字节，默认 50MB）