CoPaw AI 助手：部署个人AI助手，支持多平台集成与自定义技能

copaw-ai-assistant by aradotso/trending-skills

325 周安装量

10 GitHub Stars

GitHub

安装命令

npx skills add https://github.com/aradotso/trending-skills --skill copaw-ai-assistant

AI/机器学习自动化命令行工具

🇨🇳中文介绍

CoPaw AI 助手技能

由 ara.so 提供的技能 — Daily 2026 技能集。

CoPaw 是一个可以部署在您自己的机器或云端的个人 AI 助手框架。它通过单一代理连接到多个聊天平台（钉钉、飞书、QQ、Discord、iMessage、Telegram、Mattermost、Matrix、MQTT），支持自定义 Python 技能、定时 cron 任务、本地和云端 LLM，并在 http://127.0.0.1:8088/ 提供一个网页控制台。

安装

pip（如果已安装 Python 3.10–3.13，推荐使用）

pip install copaw
copaw init --defaults    # 使用合理的默认值进行非交互式设置
copaw app                # 启动网页控制台 + 后端

脚本安装（无需 Python 环境）

macOS / Linux:

curl -fsSL https://copaw.agentscope.io/install.sh | bash
# 支持 Ollama：
curl -fsSL https://copaw.agentscope.io/install.sh | bash -s -- --extras ollama
# 多个额外功能：
curl -fsSL https://copaw.agentscope.io/install.sh | bash -s -- --extras ollama,llamacpp

Windows CMD:

curl -fsSL https://copaw.agentscope.io/install.bat -o install.bat && install.bat

广告位招租

在这里展示您的产品或服务

触达数万 AI 开发者，精准高效

联系我们

配置 (`config.yaml`)

copaw init 会生成此文件。可以直接编辑它或使用控制台 UI。

LLM 提供商（OpenAI 兼容）

providers:
  - id: openai-main
    type: openai
    api_key: ${OPENAI_API_KEY}        # 使用环境变量引用
    model: gpt-4o
    base_url: https://api.openai.com/v1

  - id: local-ollama
    type: ollama
    model: llama3.2
    base_url: http://localhost:11434

agent:
  name: CoPaw
  language: en                        # en, zh, ja 等
  provider_id: openai-main
  context_limit: 8000

channels:
  - type: dingtalk
    app_key: ${DINGTALK_APP_KEY}
    app_secret: ${DINGTALK_APP_SECRET}
    agent_id: ${DINGTALK_AGENT_ID}
    mention_only: true                # 仅在群聊中被@时响应

通道：飞书 (Lark)

channels:
  - type: feishu
    app_id: ${FEISHU_APP_ID}
    app_secret: ${FEISHU_APP_SECRET}
    mention_only: false

channels:
  - type: discord
    token: ${DISCORD_BOT_TOKEN}
    mention_only: true

channels:
  - type: telegram
    token: ${TELEGRAM_BOT_TOKEN}

channels:
  - type: qq
    uin: ${QQ_UIN}
    password: ${QQ_PASSWORD}

channels:
  - type: mattermost
    url: ${MATTERMOST_URL}
    token: ${MATTERMOST_TOKEN}
    team: my-team

channels:
  - type: matrix
    homeserver: ${MATRIX_HOMESERVER}
    user_id: ${MATRIX_USER_ID}
    access_token: ${MATRIX_ACCESS_TOKEN}

技能是放置在 ~/.copaw/workspace/skills/ 目录下的 Python 文件。它们会在 CoPaw 启动时自动加载——无需注册步骤。

# ~/.copaw/workspace/skills/weather.py

SKILL_NAME = "get_weather"
SKILL_DESCRIPTION = "Get current weather for a city"

# Tool schema (OpenAI function-calling format)
SKILL_SCHEMA = {
    "type": "function",
    "function": {
        "name": SKILL_NAME,
        "description": SKILL_DESCRIPTION,
        "parameters": {
            "type": "object",
            "properties": {
                "city": {
                    "type": "string",
                    "description": "City name, e.g. 'Tokyo'"
                }
            },
            "required": ["city"]
        }
    }
}


def get_weather(city: str) -> str:
    """Fetch weather data for the given city."""
    import os
    import requests

    api_key = os.environ["OPENWEATHER_API_KEY"]
    url = f"https://api.openweathermap.org/data/2.5/weather"
    resp = requests.get(url, params={"q": city, "appid": api_key, "units": "metric"})
    resp.raise_for_status()
    data = resp.json()
    temp = data["main"]["temp"]
    desc = data["weather"][0]["description"]
    return f"{city}: {temp}°C, {desc}"

支持异步的技能

# ~/.copaw/workspace/skills/summarize_url.py

SKILL_NAME = "summarize_url"
SKILL_DESCRIPTION = "Fetch and summarize the content of a URL"

SKILL_SCHEMA = {
    "type": "function",
    "function": {
        "name": SKILL_NAME,
        "description": SKILL_DESCRIPTION,
        "parameters": {
            "type": "object",
            "properties": {
                "url": {"type": "string", "description": "The URL to summarize"}
            },
            "required": ["url"]
        }
    }
}


async def summarize_url(url: str) -> str:
    import httpx

    async with httpx.AsyncClient(timeout=15) as client:
        resp = await client.get(url)
        text = resp.text[:4000]   # truncate for context limit
    return f"Content preview from {url}:\n{text}"

返回结构化数据的技能

# ~/.copaw/workspace/skills/list_files.py

import os
import json

SKILL_NAME = "list_files"
SKILL_DESCRIPTION = "List files in a directory"

SKILL_SCHEMA = {
    "type": "function",
    "function": {
        "name": SKILL_NAME,
        "description": SKILL_DESCRIPTION,
        "parameters": {
            "type": "object",
            "properties": {
                "path": {
                    "type": "string",
                    "description": "Absolute or relative directory path"
                },
                "extension": {
                    "type": "string",
                    "description": "Filter by extension, e.g. '.py'. Optional."
                }
            },
            "required": ["path"]
        }
    }
}


def list_files(path: str, extension: str = "") -> str:
    entries = os.listdir(os.path.expanduser(path))
    if extension:
        entries = [e for e in entries if e.endswith(extension)]
    return json.dumps(sorted(entries))

在 config.yaml 中定义 cron 任务，以便按计划运行技能并将结果推送到通道：

cron:
  - id: daily-digest
    schedule: "0 8 * * *"            # 每天 08:00
    skill: get_weather
    skill_args:
      city: "Tokyo"
    channel_id: dingtalk-main         # 匹配下面的通道 id
    message_template: "Good morning! Today's weather: {result}"

  - id: hourly-news
    schedule: "0 * * * *"
    skill: fetch_tech_news
    channel_id: discord-main

# Install Ollama: https://ollama.ai
ollama pull llama3.2
ollama serve   # starts on http://localhost:11434



# config.yaml
providers:
  - id: ollama-local
    type: ollama
    model: llama3.2
    base_url: http://localhost:11434

providers:
  - id: lmstudio-local
    type: lmstudio
    model: lmstudio-community/Meta-Llama-3-8B-Instruct-GGUF
    base_url: http://localhost:1234/v1

llama.cpp（需要额外安装）

pip install "copaw[llamacpp]"



providers:
  - id: llamacpp-local
    type: llamacpp
    model_path: /path/to/model.gguf

工具防护（安全）

工具防护会拦截有风险的工具调用，并在执行前需要用户批准。在 config.yaml 中配置：

agent:
  tool_guard:
    enabled: true
    risk_patterns:
      - "rm -rf"
      - "DROP TABLE"
      - "os.system"
    auto_approve_low_risk: true

当调用被拦截时，控制台会显示批准提示。用户可以在工具运行前批准或拒绝。

令牌使用量追踪

令牌使用量会自动追踪，并可在控制台仪表板中查看。也可以通过编程方式访问：

# 在技能或调试脚本中
from copaw.telemetry import get_usage_summary

summary = get_usage_summary()
print(summary)
# {'total_tokens': 142300, 'prompt_tokens': 98200, 'completion_tokens': 44100, 'by_provider': {...}}

在运行 copaw app 之前设置这些变量，或者在 config.yaml 中通过 ${VAR_NAME} 引用它们：

# LLM 提供商
export OPENAI_API_KEY=...
export ANTHROPIC_API_KEY=...

# 通道
export DINGTALK_APP_KEY=...
export DINGTALK_APP_SECRET=...
export DINGTALK_AGENT_ID=...

export FEISHU_APP_ID=...
export FEISHU_APP_SECRET=...

export DISCORD_BOT_TOKEN=...
export TELEGRAM_BOT_TOKEN=...

export QQ_UIN=...
export QQ_PASSWORD=...

export MATTERMOST_URL=...
export MATTERMOST_TOKEN=...

export MATRIX_HOMESERVER=...
export MATRIX_USER_ID=...
export MATRIX_ACCESS_TOKEN=...

# 自定义技能密钥
export OPENWEATHER_API_KEY=...

模式：向钉钉发送晨间简报

# config.yaml 节选
channels:
  - id: dingtalk-main
    type: dingtalk
    app_key: ${DINGTALK_APP_KEY}
    app_secret: ${DINGTALK_APP_SECRET}
    agent_id: ${DINGTALK_AGENT_ID}

cron:
  - id: morning-brief
    schedule: "30 7 * * 1-5"         # 工作日 07:30
    skill: daily_briefing
    channel_id: dingtalk-main



# skills/daily_briefing.py
SKILL_NAME = "daily_briefing"
SKILL_DESCRIPTION = "Compile a morning briefing with weather and news"

SKILL_SCHEMA = {
    "type": "function",
    "function": {
        "name": SKILL_NAME,
        "description": SKILL_DESCRIPTION,
        "parameters": {"type": "object", "properties": {}, "required": []}
    }
}

def daily_briefing() -> str:
    import os, requests, datetime

    today = datetime.date.today().strftime("%A, %B %d")
    # 在此处添加您自己的数据源
    return f"Good morning! Today is {today}. Have a productive day!"

模式：多通道广播

# skills/broadcast.py
SKILL_NAME = "broadcast_message"
SKILL_DESCRIPTION = "Send a message to all configured channels"

SKILL_SCHEMA = {
    "type": "function",
    "function": {
        "name": SKILL_NAME,
        "description": SKILL_DESCRIPTION,
        "parameters": {
            "type": "object",
            "properties": {
                "message": {"type": "string", "description": "Message to broadcast"}
            },
            "required": ["message"]
        }
    }
}

def broadcast_message(message: str) -> str:
    # CoPaw 处理路由；返回消息并让代理发送它
    return f"[BROADCAST] {message}"

模式：文件摘要技能

# skills/summarize_file.py
SKILL_NAME = "summarize_file"
SKILL_DESCRIPTION = "Read and summarize a local file"

SKILL_SCHEMA = {
    "type": "function",
    "function": {
        "name": SKILL_NAME,
        "description": SKILL_DESCRIPTION,
        "parameters": {
            "type": "object",
            "properties": {
                "file_path": {"type": "string", "description": "Absolute path to the file"}
            },
            "required": ["file_path"]
        }
    }
}

def summarize_file(file_path: str) -> str:
    import os

    path = os.path.expanduser(file_path)
    if not os.path.exists(path):
        return f"File not found: {path}"

    with open(path, "r", encoding="utf-8", errors="ignore") as f:
        content = f.read(8000)

    return f"File: {path}\nSize: {os.path.getsize(path)} bytes\nContent preview:\n{content}"

无法在端口 8088 访问控制台

# 使用不同的端口
copaw app --port 8090

# 检查是否有其他进程占用了 8088 端口
lsof -i :8088    # macOS/Linux
netstat -ano | findstr :8088   # Windows

确认技能文件位于 ~/.copaw/workspace/skills/ 目录下
确认 SKILL_NAME、SKILL_DESCRIPTION、SKILL_SCHEMA 和处理函数都在模块级别定义
检查 ~/.copaw/workspace/logs/ 中的导入错误
添加新技能文件后重启 copaw app

通道未收到消息

验证凭据是否正确设置（环境变量或 config.yaml）
检查控制台 → 通道页面中的连接状态
对于设置了 mention_only: true 的钉钉/飞书/Discord，机器人必须被@提及
Discord 消息超过 2000 个字符会自动分割——确保机器人拥有 发送消息 权限

LLM 提供商连接失败

# 从 CLI 测试提供商（控制台 → 提供商 → 测试连接）
# 或检查日志：
tail -f ~/.copaw/workspace/logs/copaw.log

对于 Ollama：确认 ollama serve 正在运行且 base_url 匹配
对于 OpenAI 兼容的 API：验证 base_url 以 /v1 结尾
LLM 调用会自动使用指数退避重试——暂时性故障会自动解决

Windows 编码问题

# 为 CMD 设置 UTF-8 编码
chcp 65001

或在环境中设置：

export PYTHONIOENCODING=utf-8

# 重新初始化工作空间（保留 skills/ 目录）
copaw init

# 完全重置（破坏性操作）
rm -rf ~/.copaw/workspace
copaw init --defaults

ModelScope 云端部署

无需本地设置，一键云端部署：

访问 ModelScope CoPaw Studio
将工作室分叉到您的账户
在工作室设置中设置环境变量
启动工作室——控制台可通过工作室 URL 访问

🇺🇸English

CoPaw AI Assistant Skill

Skill by ara.so — Daily 2026 Skills collection.

CoPaw is a personal AI assistant framework you deploy on your own machine or in the cloud. It connects to multiple chat platforms (DingTalk, Feishu, QQ, Discord, iMessage, Telegram, Mattermost, Matrix, MQTT) through a single agent, supports custom Python skills, scheduled cron jobs, local and cloud LLMs, and provides a web Console at http://127.0.0.1:8088/.

Installation

pip (recommended if Python 3.10–3.13 is available)

pip install copaw
copaw init --defaults    # non-interactive setup with sensible defaults
copaw app                # starts the web Console + backend

Script install (no Python setup required)

macOS / Linux:

curl -fsSL https://copaw.agentscope.io/install.sh | bash
# With Ollama support:
curl -fsSL https://copaw.agentscope.io/install.sh | bash -s -- --extras ollama
# Multiple extras:
curl -fsSL https://copaw.agentscope.io/install.sh | bash -s -- --extras ollama,llamacpp

Windows CMD:

curl -fsSL https://copaw.agentscope.io/install.bat -o install.bat && install.bat

Windows PowerShell:

irm https://copaw.agentscope.io/install.ps1 | iex

After script install, open a new terminal:

copaw init --defaults
copaw app

Install from source

git clone https://github.com/agentscope-ai/CoPaw.git
cd CoPaw
pip install -e ".[dev]"
copaw init --defaults
copaw app

CLI Reference

copaw init                  # interactive workspace setup
copaw init --defaults       # non-interactive setup
copaw app                   # start the Console (http://127.0.0.1:8088/)
copaw app --port 8090       # use a custom port
copaw --help                # list all commands

Workspace Structure

After copaw init, a workspace is created (default: ~/.copaw/workspace/):

~/.copaw/workspace/
├── config.yaml          # agent, provider, channel configuration
├── skills/              # custom skill files (auto-loaded)
│   └── my_skill.py
├── memory/              # conversation memory storage
└── logs/                # runtime logs

Configuration (`config.yaml`)

copaw init generates this file. Edit it directly or use the Console UI.

LLM Provider (OpenAI-compatible)

providers:
  - id: openai-main
    type: openai
    api_key: ${OPENAI_API_KEY}        # use env var reference
    model: gpt-4o
    base_url: https://api.openai.com/v1

  - id: local-ollama
    type: ollama
    model: llama3.2
    base_url: http://localhost:11434

Agent Settings

agent:
  name: CoPaw
  language: en                        # en, zh, ja, etc.
  provider_id: openai-main
  context_limit: 8000

Channel: DingTalk

channels:
  - type: dingtalk
    app_key: ${DINGTALK_APP_KEY}
    app_secret: ${DINGTALK_APP_SECRET}
    agent_id: ${DINGTALK_AGENT_ID}
    mention_only: true                # only respond when @mentioned in groups

Channel: Feishu (Lark)

channels:
  - type: feishu
    app_id: ${FEISHU_APP_ID}
    app_secret: ${FEISHU_APP_SECRET}
    mention_only: false

Channel: Discord

channels:
  - type: discord
    token: ${DISCORD_BOT_TOKEN}
    mention_only: true

Channel: Telegram

channels:
  - type: telegram
    token: ${TELEGRAM_BOT_TOKEN}

Channel: QQ

channels:
  - type: qq
    uin: ${QQ_UIN}
    password: ${QQ_PASSWORD}

Channel: Mattermost

channels:
  - type: mattermost
    url: ${MATTERMOST_URL}
    token: ${MATTERMOST_TOKEN}
    team: my-team

Channel: Matrix

channels:
  - type: matrix
    homeserver: ${MATRIX_HOMESERVER}
    user_id: ${MATRIX_USER_ID}
    access_token: ${MATRIX_ACCESS_TOKEN}

Custom Skills

Skills are Python files placed in ~/.copaw/workspace/skills/. They are auto-loaded when CoPaw starts — no registration step needed.

Minimal skill structure

# ~/.copaw/workspace/skills/weather.py

SKILL_NAME = "get_weather"
SKILL_DESCRIPTION = "Get current weather for a city"

# Tool schema (OpenAI function-calling format)
SKILL_SCHEMA = {
    "type": "function",
    "function": {
        "name": SKILL_NAME,
        "description": SKILL_DESCRIPTION,
        "parameters": {
            "type": "object",
            "properties": {
                "city": {
                    "type": "string",
                    "description": "City name, e.g. 'Tokyo'"
                }
            },
            "required": ["city"]
        }
    }
}


def get_weather(city: str) -> str:
    """Fetch weather data for the given city."""
    import os
    import requests

    api_key = os.environ["OPENWEATHER_API_KEY"]
    url = f"https://api.openweathermap.org/data/2.5/weather"
    resp = requests.get(url, params={"q": city, "appid": api_key, "units": "metric"})
    resp.raise_for_status()
    data = resp.json()
    temp = data["main"]["temp"]
    desc = data["weather"][0]["description"]
    return f"{city}: {temp}°C, {desc}"

Skill with async support

# ~/.copaw/workspace/skills/summarize_url.py

SKILL_NAME = "summarize_url"
SKILL_DESCRIPTION = "Fetch and summarize the content of a URL"

SKILL_SCHEMA = {
    "type": "function",
    "function": {
        "name": SKILL_NAME,
        "description": SKILL_DESCRIPTION,
        "parameters": {
            "type": "object",
            "properties": {
                "url": {"type": "string", "description": "The URL to summarize"}
            },
            "required": ["url"]
        }
    }
}


async def summarize_url(url: str) -> str:
    import httpx

    async with httpx.AsyncClient(timeout=15) as client:
        resp = await client.get(url)
        text = resp.text[:4000]   # truncate for context limit
    return f"Content preview from {url}:\n{text}"

Skill returning structured data

# ~/.copaw/workspace/skills/list_files.py

import os
import json

SKILL_NAME = "list_files"
SKILL_DESCRIPTION = "List files in a directory"

SKILL_SCHEMA = {
    "type": "function",
    "function": {
        "name": SKILL_NAME,
        "description": SKILL_DESCRIPTION,
        "parameters": {
            "type": "object",
            "properties": {
                "path": {
                    "type": "string",
                    "description": "Absolute or relative directory path"
                },
                "extension": {
                    "type": "string",
                    "description": "Filter by extension, e.g. '.py'. Optional."
                }
            },
            "required": ["path"]
        }
    }
}


def list_files(path: str, extension: str = "") -> str:
    entries = os.listdir(os.path.expanduser(path))
    if extension:
        entries = [e for e in entries if e.endswith(extension)]
    return json.dumps(sorted(entries))

Cron / Scheduled Tasks

Define cron jobs in config.yaml to run skills on a schedule and push results to a channel:

cron:
  - id: daily-digest
    schedule: "0 8 * * *"            # every day at 08:00
    skill: get_weather
    skill_args:
      city: "Tokyo"
    channel_id: dingtalk-main         # matches a channel id below
    message_template: "Good morning! Today's weather: {result}"

  - id: hourly-news
    schedule: "0 * * * *"
    skill: fetch_tech_news
    channel_id: discord-main

Local Model Setup

Ollama

# Install Ollama: https://ollama.ai
ollama pull llama3.2
ollama serve   # starts on http://localhost:11434



# config.yaml
providers:
  - id: ollama-local
    type: ollama
    model: llama3.2
    base_url: http://localhost:11434

LM Studio

providers:
  - id: lmstudio-local
    type: lmstudio
    model: lmstudio-community/Meta-Llama-3-8B-Instruct-GGUF
    base_url: http://localhost:1234/v1

llama.cpp (extra required)

pip install "copaw[llamacpp]"



providers:
  - id: llamacpp-local
    type: llamacpp
    model_path: /path/to/model.gguf

Tool Guard (Security)

Tool Guard blocks risky tool calls and requires user approval before execution. Configure in config.yaml:

agent:
  tool_guard:
    enabled: true
    risk_patterns:
      - "rm -rf"
      - "DROP TABLE"
      - "os.system"
    auto_approve_low_risk: true

When a call is blocked, the Console shows an approval prompt. The user can approve or deny before the tool runs.

Token Usage Tracking

Token usage is tracked automatically and visible in the Console dashboard. Access programmatically:

# In a skill or debug script
from copaw.telemetry import get_usage_summary

summary = get_usage_summary()
print(summary)
# {'total_tokens': 142300, 'prompt_tokens': 98200, 'completion_tokens': 44100, 'by_provider': {...}}

Environment Variables

Set these before running copaw app, or reference them in config.yaml as ${VAR_NAME}:

# LLM providers
export OPENAI_API_KEY=...
export ANTHROPIC_API_KEY=...

# Channels
export DINGTALK_APP_KEY=...
export DINGTALK_APP_SECRET=...
export DINGTALK_AGENT_ID=...

export FEISHU_APP_ID=...
export FEISHU_APP_SECRET=...

export DISCORD_BOT_TOKEN=...
export TELEGRAM_BOT_TOKEN=...

export QQ_UIN=...
export QQ_PASSWORD=...

export MATTERMOST_URL=...
export MATTERMOST_TOKEN=...

export MATRIX_HOMESERVER=...
export MATRIX_USER_ID=...
export MATRIX_ACCESS_TOKEN=...

# Custom skill secrets
export OPENWEATHER_API_KEY=...

Common Patterns

Pattern: Morning briefing to DingTalk

# config.yaml excerpt
channels:
  - id: dingtalk-main
    type: dingtalk
    app_key: ${DINGTALK_APP_KEY}
    app_secret: ${DINGTALK_APP_SECRET}
    agent_id: ${DINGTALK_AGENT_ID}

cron:
  - id: morning-brief
    schedule: "30 7 * * 1-5"         # weekdays 07:30
    skill: daily_briefing
    channel_id: dingtalk-main



# skills/daily_briefing.py
SKILL_NAME = "daily_briefing"
SKILL_DESCRIPTION = "Compile a morning briefing with weather and news"

SKILL_SCHEMA = {
    "type": "function",
    "function": {
        "name": SKILL_NAME,
        "description": SKILL_DESCRIPTION,
        "parameters": {"type": "object", "properties": {}, "required": []}
    }
}

def daily_briefing() -> str:
    import os, requests, datetime

    today = datetime.date.today().strftime("%A, %B %d")
    # Add your own data sources here
    return f"Good morning! Today is {today}. Have a productive day!"

Pattern: Multi-channel broadcast

# skills/broadcast.py
SKILL_NAME = "broadcast_message"
SKILL_DESCRIPTION = "Send a message to all configured channels"

SKILL_SCHEMA = {
    "type": "function",
    "function": {
        "name": SKILL_NAME,
        "description": SKILL_DESCRIPTION,
        "parameters": {
            "type": "object",
            "properties": {
                "message": {"type": "string", "description": "Message to broadcast"}
            },
            "required": ["message"]
        }
    }
}

def broadcast_message(message: str) -> str:
    # CoPaw handles routing; return the message and let the agent deliver it
    return f"[BROADCAST] {message}"

Pattern: File summarization skill

# skills/summarize_file.py
SKILL_NAME = "summarize_file"
SKILL_DESCRIPTION = "Read and summarize a local file"

SKILL_SCHEMA = {
    "type": "function",
    "function": {
        "name": SKILL_NAME,
        "description": SKILL_DESCRIPTION,
        "parameters": {
            "type": "object",
            "properties": {
                "file_path": {"type": "string", "description": "Absolute path to the file"}
            },
            "required": ["file_path"]
        }
    }
}

def summarize_file(file_path: str) -> str:
    import os

    path = os.path.expanduser(file_path)
    if not os.path.exists(path):
        return f"File not found: {path}"

    with open(path, "r", encoding="utf-8", errors="ignore") as f:
        content = f.read(8000)

    return f"File: {path}\nSize: {os.path.getsize(path)} bytes\nContent preview:\n{content}"

Troubleshooting

Console not accessible at port 8088

# Use a different port
copaw app --port 8090

# Check if another process is using 8088
lsof -i :8088    # macOS/Linux
netstat -ano | findstr :8088   # Windows

Skills not loading

Confirm the skill file is in ~/.copaw/workspace/skills/
Confirm SKILL_NAME, SKILL_DESCRIPTION, SKILL_SCHEMA, and the handler function are all defined at module level
Check ~/.copaw/workspace/logs/ for import errors
Restart copaw app after adding new skill files

Channel not receiving messages

Verify credentials are set correctly (env vars or config.yaml)
Check the Console → Channels page for connection status
For DingTalk/Feishu/Discord with mention_only: true, the bot must be @mentioned
Discord messages over 2000 characters are split automatically — ensure the bot has Send Messages permission

LLM provider connection fails

# Test provider from CLI (Console → Providers → Test Connection)
# Or check logs:
tail -f ~/.copaw/workspace/logs/copaw.log

For Ollama: confirm ollama serve is running and base_url matches
For OpenAI-compatible APIs: verify base_url ends with /v1
LLM calls auto-retry with exponential backoff — transient failures resolve automatically

Windows encoding issues

# Set UTF-8 encoding for CMD
chcp 65001

Or set in environment:

export PYTHONIOENCODING=utf-8

Workspace reset

# Reinitialize workspace (preserves skills/)
copaw init

# Full reset (destructive)
rm -rf ~/.copaw/workspace
copaw init --defaults

ModelScope Cloud Deployment

For one-click cloud deployment without local setup:

Visit ModelScope CoPaw Studio
Fork the studio to your account
Set environment variables in the studio settings
Start the studio — Console is accessible via the studio URL

Key Links

Documentation : https://copaw.agentscope.io/
Channel setup guides : https://copaw.agentscope.io/docs/channels
Release notes : https://agentscope-ai.github.io/CoPaw/release-notes
GitHub : https://github.com/agentscope-ai/CoPaw
PyPI : https://pypi.org/project/copaw/
Discord community : https://discord.gg/eYMpfnkG8h

Weekly Installs

325

Repository

aradotso/trending-skills

GitHub Stars

First Seen

7 days ago

Security Audits

Gen Agent Trust HubFail SocketPass SnykWarn

Installed on

gemini-cli321

github-copilot321

amp321

cline321

codex321

kimi-cli321

AI Elements：基于shadcn/ui的AI原生应用组件库，快速构建对话界面

54,900 周安装

CoPaw AI 助手：部署个人AI助手，支持多平台集成与自定义技能

🇨🇳中文介绍

CoPaw AI 助手技能

安装

pip（如果已安装 Python 3.10–3.13，推荐使用）

脚本安装（无需 Python 环境）

相关 Skills

从源码安装

CLI 参考

工作空间结构

配置 (config.yaml)

LLM 提供商（OpenAI 兼容）

代理设置

通道：钉钉

通道：飞书 (Lark)

通道：Discord

通道：Telegram

通道：QQ

通道：Mattermost

通道：Matrix

自定义技能

最小技能结构

支持异步的技能

返回结构化数据的技能

Cron / 定时任务

本地模型设置

Ollama

LM Studio

llama.cpp（需要额外安装）

工具防护（安全）

令牌使用量追踪

环境变量

常见模式

模式：向钉钉发送晨间简报

模式：多通道广播

模式：文件摘要技能

故障排除

无法在端口 8088 访问控制台

技能未加载

通道未收到消息

LLM 提供商连接失败

Windows 编码问题

工作空间重置

ModelScope 云端部署

关键链接

🇺🇸English

CoPaw AI Assistant Skill

Installation

pip (recommended if Python 3.10–3.13 is available)

Script install (no Python setup required)

Install from source

CLI Reference

Workspace Structure

Configuration (config.yaml)

LLM Provider (OpenAI-compatible)

Agent Settings

Channel: DingTalk

Channel: Feishu (Lark)

Channel: Discord

Channel: Telegram

Channel: QQ

Channel: Mattermost

Channel: Matrix

Custom Skills

Minimal skill structure

Skill with async support

Skill returning structured data

Cron / Scheduled Tasks

Local Model Setup

Ollama

LM Studio

llama.cpp (extra required)

Tool Guard (Security)

Token Usage Tracking

Environment Variables

Common Patterns

Pattern: Morning briefing to DingTalk

Pattern: Multi-channel broadcast

Pattern: File summarization skill

Troubleshooting

配置 (`config.yaml`)

Configuration (`config.yaml`)