A personal AI assistant built on AgentScope, featuring local/cloud deployment, multi-channel connectivity, extensible skills, and full user data sovereignty.
QwenPaw (formerly CoPaw) is a personal AI assistant framework built on AgentScope, providing three core abstraction layers: model integration, channel connectivity, and skill extensibility. The project emphasizes user data sovereignty, supporting fully local deployment (data stays on-device) or cloud deployment on user-chosen servers, with no third-party data hosting.
For model integration, it supports cloud APIs (e.g., DashScope) and local inference backends (llama.cpp, Ollama, LM Studio) without requiring API keys. The channel layer uniformly connects to DingTalk, Feishu, WeChat, Discord, Telegram, QQ, iMessage, Matrix, and Twilio, serving multiple chat frontends from a single backend. The skill system includes built-in capabilities for scheduled tasks, PDF/Office processing, and news summarization, with auto-discovery loading for custom skills and no vendor lock-in. Security features include Tool Guard, file access control, Skill Security Scanner, and Shell Evasion Guard.
Version 1.1.3 adds ACP Server capability, proactive agent messaging, backup/restore system, and a Console plugin architecture. Licensed under Apache 2.0, the project offers multiple deployment options: pip, one-click scripts, Docker, desktop app (Beta), Alibaba Cloud ECS, and ModelScope Studio.
Multi-Agent Collaboration: Create multiple independent Agents with distinct roles and skill sets, enabling cross-agent messaging and task coordination via collaboration skills.
Installation
pip install qwenpaw
qwenpaw init --defaults
qwenpaw app
Open http://127.0.0.1:8088/ in browser to access Console for model configuration. Also supports Docker, one-click scripts, desktop app (Beta), and cloud deployment.
Optional Extension Groups: local (local models), whisper (speech-to-text), full (local+whisper), sip (VoIP calling), sip-livekit (LiveKit voice).
Unconfirmed: No Hugging Face page or academic paper links found in the repository; desktop app is in Beta with known compatibility and performance issues.