Bytebot

import requests

# 创建任务
response = requests.post(
    "http://localhost:9991/tasks",
    json={
        "instruction": "登录我的银行账户，下载上月的对账单",
        "files": []  # 可选：上传文件供AI使用
    }
)

task = response.json()
task_id = task["id"]
print(f"任务已创建: {task_id}")

# 检查任务状态
import time
while True:
    status = requests.get(f"http://localhost:9991/tasks/{task_id}")
    task_data = status.json()

    print(f"状态: {task_data['status']}")

    if task_data['status'] in ['completed', 'failed']:
        print(f"结果: {task_data['result']}")
        break

    time.sleep(2)  # 每2秒检查一次

import requests
import base64

# 读取文件并编码为base64
with open("contract.pdf", "rb") as f:
    file_content = base64.b64encode(f.read()).decode()

# 创建带文件的任务
response = requests.post(
    "http://localhost:9991/tasks",
    json={
        "instruction": "分析这份合同，提取所有付款条款和截止日期",
        "files": [
            {
                "name": "contract.pdf",
                "content": file_content,
                "mimeType": "application/pdf"
            }
        ]
    }
)

print(response.json())

import requests

BASE_URL = "http://localhost:9990"

# 获取屏幕截图
screenshot = requests.post(f"{BASE_URL}/computer-use", json={
    "action": "screenshot"
})
print(f"截图: {screenshot.json()['base64_image'][:50]}...")

# 移动鼠标并点击
requests.post(f"{BASE_URL}/computer-use", json={
    "action": "mouse_move",
    "coordinate": [100, 200]
})

requests.post(f"{BASE_URL}/computer-use", json={
    "action": "left_click"
})

# 输入文本
requests.post(f"{BASE_URL}/computer-use", json={
    "action": "type",
    "text": "Hello, Bytebot!"
})

# 按键操作
requests.post(f"{BASE_URL}/computer-use", json={
    "action": "key",
    "text": "Return"  # 按回车键
})

# 编辑 Claude Desktop 配置文件
# macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
# Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "bytebot": {
      "url": "http://localhost:9990/mcp"
    }
  }
}

# 重启Claude Desktop后，Claude就可以：
# - 控制Bytebot的虚拟桌面
# - 执行鼠标和键盘操作
# - 截图并分析屏幕内容
# - 运行复杂的多步骤任务

# 克隆仓库
git clone https://github.com/bytebot-ai/bytebot.git
cd bytebot

# 基础安装
helm install bytebot ./helm \
  --set agent.env.ANTHROPIC_API_KEY=sk-ant-...

# 生产环境配置
helm install bytebot ./helm \
  --set agent.replicas=3 \
  --set desktop.resources.requests.memory=4Gi \
  --set desktop.resources.requests.cpu=2 \
  --set postgresql.persistence.size=50Gi \
  --set ingress.enabled=true \
  --set ingress.host=bytebot.company.com

# values.yaml
agent:
  replicas: 3
  env:
    ANTHROPIC_API_KEY: "sk-ant-..."
    # 或使用LiteLLM代理
    LITELLM_PROXY_URL: "http://litellm-proxy:4000"

desktop:
  resources:
    requests:
      memory: "4Gi"
      cpu: "2"
    limits:
      memory: "8Gi"
      cpu: "4"

postgresql:
  persistence:
    enabled: true
    size: 50Gi

ingress:
  enabled: true
  className: nginx
  host: bytebot.company.com
  tls:
    enabled: true
    secretName: bytebot-tls

# 查看服务状态
docker-compose -f docker/docker-compose.yml ps

# 查看日志
docker-compose -f docker/docker-compose.yml logs -f
docker-compose -f docker/docker-compose.yml logs -f agent
docker-compose -f docker/docker-compose.yml logs -f desktop

# 停止服务
docker-compose -f docker/docker-compose.yml stop

# 重启服务
docker-compose -f docker/docker-compose.yml restart

# 更新到最新版本
docker-compose -f docker/docker-compose.yml pull
docker-compose -f docker/docker-compose.yml up -d

# 完全重置（删除所有数据）
docker-compose -f docker/docker-compose.yml down -v
docker-compose -f docker/docker-compose.yml up -d

# 查看Pod状态
kubectl get pods -l app=bytebot

# 查看日志
kubectl logs -f deployment/bytebot-agent
kubectl logs -f deployment/bytebot-desktop

# 进入容器调试
kubectl exec -it deployment/bytebot-desktop -- bash

# 扩展副本数
kubectl scale deployment bytebot-agent --replicas=5

# 更新配置
helm upgrade bytebot ./helm -f values.yaml

# 回滚版本
helm rollback bytebot

# 卸载
helm uninstall bytebot