opensource/mateclaw
2
0
Fork 0
mirror of https://gitee.com/mateos/mateclaw.git synced 2026-09-13 03:13:41 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity
169 Commits 2 Branches 16 Tags 34 MiB
Java 73.4%
Vue 14.1%
TypeScript 7.1%
TeX 2.1%
Python 1.9%
Other 1.4%
ed37e81e7e
Go to file
matevip ed37e81e7e feat(llm): multi-model failover chain driven by per-provider priority 
Replaces the hardcoded single-DashScope fallback with a DB-driven
ordered chain. Same-provider primary deployments (e.g., DashScope
qwen-max) finally get a real fallback; if any provider in the chain
returns an empty body or transient failure, the next is tried.

Schema — DB-driven chain
- mate_model_provider gains `fallback_priority INT DEFAULT 0`. Positive
  values define try-order; 0 = not in chain. Migration V21 (h2 + mysql)
  seeds DashScope as priority 1 to preserve existing behavior.
- ModelProviderService.listFallbackChain() returns providers ordered by
  priority ascending.
- ModelProviderEntity gains the new field.

Runtime — chain walk + empty-response trigger
- AgentGraphBuilder.buildFallbackChain(primaryConfig) returns a
  List<ChatModel>, identity-filtering the primary by (providerId,
  modelName) — fixes the bug where same-provider-primary deployments got
  null fallback. Providers whose API key is missing are silently
  skipped with WARN. Old buildFallbackModel(ChatModel) kept as
  @Deprecated wrapper.
- NodeStreamingChatHelper accepts List<ChatModel>; the post-retry
  fallback block now walks the chain in priority order, single-shot
  per entry. Old single-fallback constructors retained as @Deprecated
  one-element-list wrappers so legacy callers keep working.
- New ErrorType.EMPTY_RESPONSE: when the LLM returns no content, no
  thinking, AND no tool calls, mark the result as a soft failure and
  break the same-model retry loop, handing off directly to the
  fallback chain.
- Broadcast updated to "切换到备选模型 (N/M)..." so SSE consumers see
  chain progress.

Tests
- NodeStreamingChatHelperFallbackChainTest covers constructor variants,
  chain immutability, deprecated-overload back-compat, and the
  EMPTY_RESPONSE enum exists as a compile-time contract.
- 159 tests pass (was 153 + 6 new).
2026-04-19 16:55:33 +08:00
assets docs: add preview screenshot to README 2026-04-10 18:35:13 +08:00
mateclaw-plugin-api feat(plugin): Plugin SDK + UI layout improvements 2026-04-13 18:38:03 +08:00
mateclaw-plugin-sample feat(plugin): Plugin SDK + UI layout improvements 2026-04-13 18:38:03 +08:00
mateclaw-server feat(llm): multi-model failover chain driven by per-provider priority 2026-04-19 16:55:33 +08:00
mateclaw-ui feat(ui): plain-text user messages with 8-line auto-collapse 2026-04-19 09:06:14 +08:00
mateclaw-webchat release: v1.1.0 2026-04-17 19:41:03 +08:00
.dockerignore chore: bump version to 1.1.137-SNAPSHOT 2026-04-18 21:58:54 +08:00
.env.example release: v1.1.0 2026-04-17 19:41:03 +08:00
.gitignore feat(wiki): real-time SSE progress + parallel page generation + partial-resume; fix concurrent slug collisions; fix skill overwrite + hub retry 2026-04-15 10:34:11 +08:00
docker-compose.yml chore: bump version to 1.1.137-SNAPSHOT 2026-04-18 21:58:54 +08:00
LICENSE chore: add Apache-2.0 license 2026-04-04 23:29:51 +08:00
README_zh.md docs: fact-check Claude Code/Cursor/Windsurf — correct comparison table 2026-04-10 23:34:40 +08:00
README.md docs: fact-check Claude Code/Cursor/Windsurf — correct comparison table 2026-04-10 23:34:40 +08:00
UPGRADING.md release: v1.1.0 2026-04-17 19:41:03 +08:00

README.md

MateClaw Logo

MateClaw

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

GitHub Repo Documentation Live Demo Website Java Version Spring Boot Vue Last Commit License

[Website] [Live Demo] [Documentation] [中文]

MateClaw Preview

An AI agent. A knowledge engine. A memory system. A tool runtime. A multi-channel presence.

One product. The whole widget.

MateClaw is a personal AI operating system built with Java + Vue 3, powered by Spring AI Alibaba. It's not a chatbox, not a workflow builder, not just another coding assistant. It's the entire system — from reasoning to remembering to shipping — in one deployment.

Three things make it insanely different:

  1. Agents do work, not just talk — ReAct + Plan-and-Execute. Not one-shot answers — iterative reasoning that actually completes tasks
  2. Knowledge is shaped, not just stored — An LLM Wiki that digests raw material into structured, linked pages. The difference between a warehouse and a library
  3. End-to-end, no compromises — Web console, desktop app, 7 IM channels, tool guardrails, enterprise auth. One team, one deployment, one experience

Why MateClaw

Most AI tools do one thing well. MateClaw does the whole thing.

Capability MateClaw OpenClaw CoPaw QClaw Claude Code Cursor Windsurf
Agent Orchestration ReAct + Plan-Execute Multi-agent teams Multi-agent collab Specialist agents Agent Teams + subagents Background Agents (cloud VM) Cascade engine
Knowledge System LLM Wiki (digestion) Intelligence Mode + Wiki Personal KB Knowledge graph CLAUDE.md (no RAG) Codebase indexing No
Memory Extract + Consolidate + Dream SQLite + Dreaming + Wiki ReMe (hybrid retrieval) 3-layer memory 3-layer (CLAUDE.md + auto + files) No persistent memory Memories (~48h learning)
Tool Guard & Approval RBAC + approval flow HITL + risk levels No No Permissions + Sandbox + Hooks No Turbo Mode (auto-approve)
Multi-Channel IM 7 channels 25+ channels 7 channels 5 channels 3 channels (preview) IDE only IDE only
Web Management UI Full admin dashboard Control UI Console UI Dashboard Enterprise dashboard No No
Desktop App Electron + bundled JRE macOS menu bar Electron (Beta) Win/Mac app Claude Desktop (Mac/Win) VS Code fork VS Code fork
Multimodal Creation TTS/STT/Img/Music/Video TTS/Video/Music/Image Vision input No Vision input only No No
Skill Ecosystem ClawHub marketplace ClawHub registry Python skills Templates 340+ plugins, 1300+ skills MCP marketplace MCP one-click
Enterprise Auth RBAC + JWT Basic (password) Basic auth No SSO/SCIM/RBAC SSO + Teams Teams plan
Open Source Apache 2.0 MIT Apache 2.0 Partial No (source-available) No No
Pricing Free Free Free Free (beta) $20–200/mo $0–200/mo $0–200/mo
Tech Stack Java + Vue 3 TypeScript Python + TS OpenClaw fork TypeScript Electron (VS Code) Electron (VS Code)

What makes MateClaw different?

Every product in this table is genuinely strong. Here's where MateClaw carves its own space:

  • Plan-and-Execute orchestration — Break complex work into ordered steps, execute each, adapt mid-flight. Others have multi-agent, but structured task planning with dynamic replanning is rare
  • LLM Wiki that digests, not just retrieves — Others index and search. MateClaw's Wiki turns raw material into structured, linked pages with summaries — a search engine vs. an encyclopedia
  • Java ecosystem — Built for teams already running Spring Boot in production. One JAR, one deploy. No Python runtime, no Node.js dependency chain
  • Complete admin dashboard — Agents, models, tools, skills, channels, security, cron jobs, token usage — all in one web UI. Not a CLI-first afterthought
  • Full multimodal creation — TTS, STT, image, music, and video generation as first-class built-in features. OpenClaw matches here; most others don't
  • Free and open, no asterisks — Apache 2.0. No token billing, no seat pricing, no feature gating. Claude Code starts at $20/mo, Cursor and Windsurf up to $200/mo

Architecture

Business Architecture

Technical Architecture

Technical Architecture

Core Capabilities

Agent Runtime

  • ReAct agents — Think, act, observe, repeat. Iterative reasoning that gets things done
  • Plan-and-Execute — Decompose complex work into ordered steps, then execute each one
  • Dynamic configuration — Load agent personality, tools, and constraints from the database at runtime
  • Runtime resilience — Context pruning, smart truncation, stale stream cleanup, and recovery

Knowledge & Memory

  • LLM Wiki — AI-powered knowledge base that digests raw materials into structured, linked pages with summaries
  • Workspace memory — AGENTS.md, SOUL.md, PROFILE.md, MEMORY.md, daily notes
  • Memory lifecycle — Post-conversation extraction, scheduled consolidation, dreaming workflows
  • Compound memory — Understanding improves over time instead of resetting every query

Tools, Skills & MCP

  • Built-in tools — Web search, file ops, memory access, date/time, and more
  • MCP integration — stdio, SSE, and Streamable HTTP transports
  • Skill system — Installable SKILL.md packages with ClawHub marketplace
  • Tool guard — Approval flows, file-path protection, runtime filtering

Multimodal Creation

Text-to-speech · Speech-to-text · Image generation · Music generation · Video generation

Model Flexibility

14+ providers including DashScope, OpenAI, Anthropic, Gemini, DeepSeek, Kimi, Ollama, LM Studio, MLX, and more. Configure everything in the web UI.

Surfaces

  • Web console — Chat, agents, tools, skills, knowledge, models, security, settings
  • Desktop app — Electron with bundled JRE 21, no Java installation needed
  • Channels — DingTalk, Feishu, WeChat Work, Telegram, Discord, QQ

Quick Start

Prerequisites

  • Java 17+ · Node.js 18+ · pnpm · Maven 3.9+

Local Development

# Backend
cd mateclaw-server
mvn spring-boot:run          # http://localhost:18088

# Frontend
cd mateclaw-ui
pnpm install && pnpm dev     # http://localhost:5173

Login: admin / admin123

Docker

cp .env.example .env
docker compose up -d          # http://localhost:18080

Desktop App

Download from GitHub Releases. Bundles JRE 21 — no Java needed.

Tech Stack

Layer Technology
Backend Spring Boot 3.5 · Spring AI Alibaba 1.1
Agent StateGraph Runtime
Database H2 (dev) / MySQL 8.0+ (prod)
ORM MyBatis Plus 3.5
Auth Spring Security + JWT
Frontend Vue 3 · TypeScript · Vite
UI Element Plus · TailwindCSS 4
Desktop Electron · electron-updater

Project Structure

mateclaw/
├── mateclaw-server/     Spring Boot backend
├── mateclaw-ui/         Vue 3 SPA frontend
├── mateclaw-desktop/    Electron desktop app
├── docker-compose.yml
└── .env.example

Documentation

Full docs at claw.mate.vip/docs

Roadmap

  • Richer multi-agent collaboration
  • Smarter model routing
  • Deeper multimodal understanding
  • Stronger long-term memory
  • Richer ClawHub ecosystem

Contributing

git clone https://github.com/matevip/mateclaw.git
cd mateclaw
cd mateclaw-server && mvn clean compile
cd ../mateclaw-ui && pnpm install && pnpm dev

Why The Name

Mate is companion. Claw is capability.

A system that stays with you, and a system that grabs work and moves it.

License

Apache License 2.0