feat(monitor): 抖音 Web 接口客户端 —— 绕开爬虫子进程,直接发 HTTP
移植自 mac-agent-os 的 mediacrawler_adapter:不起子进程、不开页面,用浏览器里那份
登录态直接调抖音 Web 接口。产物键名照抄 store/douyin,所以 ingest 那条链路一个字不用改。
**目前能用的(真环境实测,非推断)**:
profile/other : 200, 7075 字节 —— 博主主页指标(粉丝/获赞/作品数/昵称)
aweme/detail : 200, 45425 字节 —— 单条作品详情(含点赞/评论/收藏/分享)
**目前不能用的:作品列表 `aweme/post`。** 两个互相独立的原因:
1. 这个接口被抖音单独升级成了真校验:不带 x-tt-argus 回 403「Uifid Not Found」,
带上 dummy 值回 200 + **空 body**。也就是说「头在不在」骗得过,「真校验」过不了。
同一套头打 profile/other 和 aweme/detail 都是通的 —— 抖音是挑着接口加保护的,
挑中的恰好是「批量拉作品列表」这个最敏感的动作。
2. 改走页面截获也不行:CDP 浏览器打开博主主页会落到「验证码中间页」(当天大量探测的
代价,过几小时要重测)。
所以现在的边界是:**已知作品的指标刷新能做,自动发现新作品做不了**。
**排查中控住变量后得到的两条事实**(都写进注释了):
· `Accept` / `Accept-Language` / `Referer` 才是主页接口能返回真数据的原因 —— 只有
UA+client hints+Cookie 时是 200 但仅 121 字节的空壳,补上这三个头变 7074 字节。
(我先前猜的 sec-ch-ua 不是关键。)
· 因此 UA 与 client hints 必须**成套地取自同一个浏览器**,所以 BrowserIdentity 一次
从 CDP 取齐 cookie + UA + hints,而不是各自写死。
「200 + 空 body 必须当场报错」也是刻意写死的:放过去它会在下游变成「这个博主没作品」,
把一次失败伪装成一条正常结果 —— 爬虫那条路正是这么栽的,还被翻译成「账号被封」。
测试 +11:cookie 解析、请求头成套性(含 uifid 缺失/回退)、产物键名与 store 对齐、
以及 _get 的三条失败路径(空 body / 403 带网关原话 / 正常返回)。
This commit is contained in:
@@ -0,0 +1,85 @@
|
||||
# Trae Agent 工具
|
||||
|
||||
**路径**: `05_tools/08_trae_agent/`
|
||||
**最后更新**: 2026-05-02
|
||||
|
||||
## 概述
|
||||
|
||||
[Trae Agent](https://github.com/bytedance/trae-agent) 是字节跳动开源的命令行 AI 编程助手(MIT 许可证),
|
||||
类似于 Claude Code / Gemini CLI。它能将自然语言描述的开发任务转化为可执行的软件工程工作流。
|
||||
|
||||
## 文件结构
|
||||
|
||||
```
|
||||
08_trae_agent/
|
||||
├── README.md # 本文件
|
||||
├── trae_agent.sh # 主封装脚本(入口)
|
||||
├── install_trae_agent.sh # 自动安装脚本
|
||||
└── trae_config.yaml.example # 配置文件模板
|
||||
```
|
||||
|
||||
## 系统依赖
|
||||
|
||||
- **Python** ≥ 3.12(本机使用 WorkBuddy 托管 Python 3.13.12)
|
||||
- **uv** 包管理器(`curl -LsSf https://astral.sh/uv/install.sh | sh`)
|
||||
- **Git**(用于克隆仓库)
|
||||
- **LLM 后端**:oMLX (localhost:8000) / Ollama / 远程 API
|
||||
|
||||
## 安装
|
||||
|
||||
```bash
|
||||
# 方式一:自动安装脚本
|
||||
bash 05_tools/08_trae_agent/install_trae_agent.sh
|
||||
|
||||
# 方式二:agentos CLI
|
||||
agentos tool install trae-agent
|
||||
```
|
||||
|
||||
## 配置
|
||||
|
||||
编辑 `~/workbuddy-agent-os/agent-local/tools/trae-agent/trae_config.yaml`:
|
||||
|
||||
```yaml
|
||||
model_providers:
|
||||
omix: # oMLX 后端
|
||||
api_key: omlx
|
||||
base_url: http://localhost:8000/v1
|
||||
provider: openai
|
||||
openai: # 远程 API
|
||||
api_key: sk-your-key
|
||||
base_url: https://api.openai.com/v1
|
||||
provider: openai
|
||||
```
|
||||
|
||||
## 使用方法
|
||||
|
||||
```bash
|
||||
# 运行任务(非交互)
|
||||
trae_agent.sh run "创建一个 Hello World Python 脚本"
|
||||
|
||||
# 指定后端和模型
|
||||
trae_agent.sh run -p omix -m Qwen3-8B-MLX-4bit "分析项目结构"
|
||||
|
||||
# 交互模式
|
||||
trae_agent.sh interactive
|
||||
|
||||
# 查看配置
|
||||
trae_agent.sh show-config
|
||||
|
||||
# 查看可用工具
|
||||
trae_agent.sh tools
|
||||
```
|
||||
|
||||
## agentos 集成
|
||||
|
||||
```bash
|
||||
agentos tool trae run "任务描述"
|
||||
agentos tool trae interactive
|
||||
agentos tool trae config
|
||||
```
|
||||
|
||||
## 已知问题
|
||||
|
||||
1. **oMLX Qwen3-8B Chat API**: 已知返回 500 错误,临时可用 VLM 模型或切换至远程 API
|
||||
2. **网络限制**: GitHub 可能无法直接访问,安装脚本已配置 Gitee 镜像
|
||||
3. **首次运行**: 需要根据 LLM 后端修改配置文件中的 API key 和 base_url
|
||||
@@ -0,0 +1,75 @@
|
||||
# Colima 沙箱环境配置
|
||||
|
||||
**路径**: `05_tools/08_trae_agent/colima_setup.md`
|
||||
**最后更新**: 2026-05-02
|
||||
|
||||
## 概述
|
||||
|
||||
trae-agent 默认使用 Docker 容器进行环境隔离。本机通过 Colima(轻量容器运行时)提供 Docker 环境。
|
||||
|
||||
## 已安装组件
|
||||
|
||||
| 组件 | 版本 | 路径 |
|
||||
|------|------|------|
|
||||
| Colima | v0.10.1 | `~/.local/bin/colima` |
|
||||
| Docker CLI | v29.4.2 | `~/.local/bin/docker` |
|
||||
| Lima | v2.1.1 | `~/.local/bin/limactl` |
|
||||
| VM 镜像 | - | `~/.colima/ubuntu-24.04-minimal-cloudimg-arm64-docker.qcow2` |
|
||||
|
||||
## 启动 Colima
|
||||
|
||||
```bash
|
||||
# 首次启动
|
||||
colima start --cpu 2 --memory 4 --disk 20
|
||||
|
||||
# 查看状态
|
||||
colima status
|
||||
|
||||
# 停止
|
||||
colima stop
|
||||
|
||||
# 重启
|
||||
colima restart
|
||||
```
|
||||
|
||||
## 验证 Docker
|
||||
|
||||
```bash
|
||||
# 验证 Docker 是否可用
|
||||
docker info
|
||||
|
||||
# 运行测试
|
||||
docker run hello-world
|
||||
```
|
||||
|
||||
## 与 trae-agent 集成
|
||||
|
||||
```bash
|
||||
# 默认使用 Docker 隔离
|
||||
trae_agent.sh run "任务描述"
|
||||
|
||||
# 如需跳过 Docker
|
||||
trae_agent.sh run --no-docker "任务描述"
|
||||
```
|
||||
|
||||
## 常见问题
|
||||
|
||||
### Colima 启动失败 - guest agent 找不到
|
||||
```bash
|
||||
# 确保 guest agent 文件在正确位置
|
||||
ls ~/.local/share/lima/lima-guestagent.Linux-aarch64.gz
|
||||
# 如果缺失:
|
||||
cp ~/.lima/lima-guestagent.Linux-aarch64.gz ~/.local/share/lima/
|
||||
```
|
||||
|
||||
### Colima 启动失败 - VM 镜像下载慢
|
||||
```bash
|
||||
# 手动下载镜像放到 ~/.colima/ 目录
|
||||
# 启动时指定本地镜像
|
||||
colima start --disk-image ~/.colima/ubuntu-24.04-minimal-cloudimg-arm64-docker.qcow2
|
||||
```
|
||||
|
||||
### Docker 命令找不到
|
||||
```bash
|
||||
export PATH="$HOME/.local/bin:$PATH"
|
||||
```
|
||||
@@ -0,0 +1,119 @@
|
||||
#!/bin/bash
|
||||
# trae-agent 自动安装脚本
|
||||
# 用于换机后一键安装 trae-agent 环境
|
||||
# 用法: ./install_trae_agent.sh
|
||||
# 最后更新:2026-05-02
|
||||
|
||||
set -euo pipefail
|
||||
|
||||
echo "================================================"
|
||||
echo " Trae Agent 自动安装脚本"
|
||||
echo "================================================"
|
||||
|
||||
# 配置
|
||||
TRAE_AGENT_DIR="$HOME/workbuddy-agent-os/agent-local/tools/trae-agent"
|
||||
TOOL_DIR="$HOME/workbuddy-agent-os/agent-sync/05_tools/08_trae_agent"
|
||||
UV_PATH="$HOME/.local/bin/uv"
|
||||
|
||||
# 步骤1: 检查系统依赖
|
||||
echo ""
|
||||
echo "[1/5] 检查系统依赖..."
|
||||
|
||||
# Python 3.12+
|
||||
PYTHON_VERSION=$(python3 --version 2>/dev/null | grep -oP '\d+\.\d+' | head -1 || echo "0")
|
||||
if [ "$(echo "$PYTHON_VERSION >= 3.12" | bc 2>/dev/null)" != "1" ]; then
|
||||
echo " 警告: 系统 Python 版本 ($PYTHON_VERSION) 低于 3.12"
|
||||
echo " 使用 WorkBuddy 托管 Python"
|
||||
WORKBUDDY_PYTHON="$HOME/.workbuddy/binaries/python/versions/3.13.12/bin/python3"
|
||||
if [ -f "$WORKBUDDY_PYTHON" ]; then
|
||||
echo " ✅ 使用 WorkBuddy Python 3.13.12"
|
||||
export PATH="$HOME/.workbuddy/binaries/python/versions/3.13.12/bin:$PATH"
|
||||
else
|
||||
echo " ❌ 未找到兼容的 Python 版本,请安装 Python 3.12+"
|
||||
exit 1
|
||||
fi
|
||||
else
|
||||
echo " ✅ Python $PYTHON_VERSION"
|
||||
fi
|
||||
|
||||
# uv 包管理器
|
||||
if [ ! -f "$UV_PATH" ]; then
|
||||
echo " 安装 uv..."
|
||||
curl -LsSf https://astral.sh/uv/install.sh | sh
|
||||
echo " ✅ uv 安装完成"
|
||||
else
|
||||
echo " ✅ uv $(uv --version 2>/dev/null || echo '已安装')"
|
||||
fi
|
||||
export PATH="$HOME/.local/bin:$PATH"
|
||||
|
||||
# Git
|
||||
if ! command -v git &>/dev/null; then
|
||||
echo " ❌ Git 未安装"
|
||||
exit 1
|
||||
fi
|
||||
echo " ✅ Git $(git --version | head -1)"
|
||||
|
||||
# 步骤2: 克隆 trae-agent 仓库
|
||||
echo ""
|
||||
echo "[2/5] 克隆 trae-agent 仓库..."
|
||||
if [ -d "$TRAE_AGENT_DIR" ]; then
|
||||
echo " trae-agent 已存在,更新..."
|
||||
cd "$TRAE_AGENT_DIR" && git pull
|
||||
else
|
||||
mkdir -p "$(dirname "$TRAE_AGENT_DIR")"
|
||||
# 优先 Gitee 镜像(国内),fallback GitHub
|
||||
git clone https://gitee.com/ByteDance/trae-agent.git "$TRAE_AGENT_DIR" 2>/dev/null \
|
||||
|| git clone https://github.com/bytedance/trae-agent.git "$TRAE_AGENT_DIR"
|
||||
fi
|
||||
echo " ✅ 克隆完成"
|
||||
|
||||
# 步骤3: 安装 Python 依赖
|
||||
echo ""
|
||||
echo "[3/5] 安装 Python 依赖..."
|
||||
cd "$TRAE_AGENT_DIR"
|
||||
uv sync --all-extras --index-url https://pypi.tuna.tsinghua.edu.cn/simple
|
||||
echo " ✅ 依赖安装完成"
|
||||
|
||||
# 步骤4: 创建默认配置
|
||||
echo ""
|
||||
echo "[4/5] 创建默认配置..."
|
||||
if [ ! -f "$TRAE_AGENT_DIR/trae_config.yaml" ]; then
|
||||
cp "$TOOL_DIR/trae_config.yaml.example" "$TRAE_AGENT_DIR/trae_config.yaml" 2>/dev/null \
|
||||
|| cp "$TRAE_AGENT_DIR/trae_config.yaml.example" "$TRAE_AGENT_DIR/trae_config.yaml" 2>/dev/null
|
||||
echo " ✅ 默认配置已创建"
|
||||
else
|
||||
echo " ✅ 配置已存在,跳过"
|
||||
fi
|
||||
|
||||
# 步骤5: 验证安装
|
||||
echo ""
|
||||
echo "[5/5] 验证安装..."
|
||||
if cd "$TRAE_AGENT_DIR" && uv run python3 -m trae_agent.cli --help &>/dev/null; then
|
||||
echo " ✅ trae-agent CLI 可用"
|
||||
else
|
||||
echo " ❌ 验证失败"
|
||||
exit 1
|
||||
fi
|
||||
|
||||
# Docker 检查(非强制)
|
||||
echo ""
|
||||
echo " Docker 状态:"
|
||||
if command -v docker &>/dev/null; then
|
||||
echo " ✅ Docker $(docker --version | head -1)"
|
||||
else
|
||||
echo " ⚠️ Docker 未安装(trae-agent 默认使用 Docker 隔离)"
|
||||
echo " 安装: https://www.docker.com/products/docker-desktop/"
|
||||
echo " 或轻量方案: https://github.com/abiosoft/colima"
|
||||
fi
|
||||
|
||||
echo ""
|
||||
echo "================================================"
|
||||
echo " ✅ Trae Agent 安装完成!"
|
||||
echo "================================================"
|
||||
echo ""
|
||||
echo "使用方法:"
|
||||
echo " trae_agent.sh run \"描述你的任务\""
|
||||
echo " trae_agent.sh interactive"
|
||||
echo ""
|
||||
echo "配置文件: $TRAE_AGENT_DIR/trae_config.yaml"
|
||||
echo "请编辑配置文件设置 LLM API Key"
|
||||
@@ -0,0 +1,121 @@
|
||||
#!/bin/bash
|
||||
# trae-agent 封装脚本
|
||||
# AgentOS 工具入口 - 将 trae-agent CLI 封装为标准化工具
|
||||
# 用法: trae_agent.sh <run|interactive|show-config|tools> [参数...]
|
||||
# 最后更新:2026-05-02
|
||||
|
||||
set -euo pipefail
|
||||
|
||||
TRAE_AGENT_DIR="$HOME/workbuddy-agent-os/agent-local/tools/trae-agent"
|
||||
TRAE_CONFIG="$TRAE_AGENT_DIR/trae_config.yaml"
|
||||
|
||||
if [ ! -d "$TRAE_AGENT_DIR" ]; then
|
||||
echo "错误: trae-agent 未安装" >&2
|
||||
echo "请先运行: install_trae_agent.sh" >&2
|
||||
exit 1
|
||||
fi
|
||||
|
||||
_trae() {
|
||||
export PATH="$HOME/.local/bin:$PATH"
|
||||
# 绕过系统代理(macOS 系统代理端口 6478 会拦截本地 LLM 请求)
|
||||
export NO_PROXY="localhost,127.0.0.1,::1"
|
||||
export no_proxy="localhost,127.0.0.1,::1"
|
||||
cd "$TRAE_AGENT_DIR"
|
||||
uv run python3 -m trae_agent.cli "$@"
|
||||
}
|
||||
|
||||
# Docker 默认镜像配置
|
||||
DEFAULT_DOCKER_IMAGE="python:3.12-slim"
|
||||
|
||||
# Colima 沙箱管理
|
||||
COLIMA_HOME="$HOME/.colima"
|
||||
COLIMA_LIMA_HOME="$HOME/.lima"
|
||||
|
||||
_colima_status() {
|
||||
# 检查 Colima / Docker 运行状态
|
||||
if command -v docker &>/dev/null && docker info &>/dev/null 2>&1; then
|
||||
return 0
|
||||
fi
|
||||
return 1
|
||||
}
|
||||
|
||||
_trae_docker() {
|
||||
# 在 Docker 中运行 trae-agent 以隔离环境
|
||||
if ! _colima_status; then
|
||||
echo "⚠️ Docker 沙箱未就绪(Colima 未运行)" >&2
|
||||
echo " 启动命令: colima start --cpu 2 --memory 4 --disk 20" >&2
|
||||
echo " 查看状态: colima status" >&2
|
||||
echo " 回退到本地执行(--no-docker 模式)..." >&2
|
||||
echo "" >&2
|
||||
_trae "$@"
|
||||
return
|
||||
fi
|
||||
export PATH="$HOME/.local/bin:$PATH"
|
||||
export NO_PROXY="localhost,127.0.0.1,::1"
|
||||
export no_proxy="localhost,127.0.0.1,::1"
|
||||
cd "$TRAE_AGENT_DIR"
|
||||
uv run python3 -m trae_agent.cli run --docker-image "$DEFAULT_DOCKER_IMAGE" "$@"
|
||||
}
|
||||
|
||||
case "${1:-help}" in
|
||||
run|r)
|
||||
shift
|
||||
if [ $# -eq 0 ]; then
|
||||
echo "用法: trae_agent.sh run [--no-docker] [选项] \"任务描述\"" >&2
|
||||
echo "或从文件读取: trae_agent.sh run -f /path/to/task.txt" >&2
|
||||
exit 1
|
||||
fi
|
||||
# 检查是否跳过 Docker
|
||||
if [ "${1:-}" = "--no-docker" ]; then
|
||||
shift
|
||||
_trae run --config-file "$TRAE_CONFIG" "$@"
|
||||
else
|
||||
_trae_docker --config-file "$TRAE_CONFIG" "$@"
|
||||
fi
|
||||
;;
|
||||
interactive|i|shell)
|
||||
# 交互式模式
|
||||
shift
|
||||
_trae interactive --config-file "$TRAE_CONFIG" "$@"
|
||||
;;
|
||||
show-config|config)
|
||||
# 显示当前配置
|
||||
shift
|
||||
_trae show-config --config-file "$TRAE_CONFIG" "$@"
|
||||
;;
|
||||
tools)
|
||||
# 列出可用工具
|
||||
_trae tools
|
||||
;;
|
||||
help|--help|-h)
|
||||
echo "Trae Agent - AI 编程助手 CLI"
|
||||
echo ""
|
||||
echo "用法: trae_agent.sh <命令> [参数...]"
|
||||
echo ""
|
||||
echo "命令:"
|
||||
echo " run, r 执行任务(非交互模式)"
|
||||
echo " interactive, i 交互式模式"
|
||||
echo " show-config, config 查看配置"
|
||||
echo " tools 列出可用工具"
|
||||
echo " help 本帮助"
|
||||
echo ""
|
||||
echo "示例:"
|
||||
echo " trae_agent.sh run \"创建一个 Hello World Python 脚本\""
|
||||
echo " trae_agent.sh run --no-docker -p omix -m Qwen3-8B-MLX-4bit \"分析代码\""
|
||||
echo " trae_agent.sh interactive"
|
||||
echo ""
|
||||
echo "Docker 沙箱(Colima):"
|
||||
echo " 默认在 Docker 容器中执行任务(python:3.12-slim)"
|
||||
echo " 跳过 Docker: trae_agent.sh run --no-docker \"任务\""
|
||||
echo " Colima 状态: colima status"
|
||||
echo " Colima 启动: colima start --cpu 2 --memory 4 --disk 20"
|
||||
echo ""
|
||||
echo "配置文件: $TRAE_CONFIG"
|
||||
echo "后端: oMLX (localhost:8000)"
|
||||
;;
|
||||
*)
|
||||
echo "未知命令: $1" >&2
|
||||
echo "用法: trae_agent.sh help" >&2
|
||||
exit 1
|
||||
;;
|
||||
esac
|
||||
@@ -0,0 +1,54 @@
|
||||
# trae-agent 配置文件模板
|
||||
# 复制为 trae_config.yaml 并修改
|
||||
# AgentOS 工具: 05_tools/08_trae_agent/
|
||||
# 最后更新:2026-05-02
|
||||
|
||||
agents:
|
||||
trae_agent:
|
||||
enable_lakeview: false
|
||||
model: trae_agent_model
|
||||
max_steps: 100
|
||||
tools:
|
||||
- bash
|
||||
- str_replace_based_edit_tool
|
||||
- sequentialthinking
|
||||
- task_done
|
||||
|
||||
model_providers:
|
||||
# oMLX OpenAI-compatible API
|
||||
omix:
|
||||
api_key: omlx
|
||||
base_url: http://localhost:8000/v1
|
||||
provider: openai
|
||||
# Ollama
|
||||
ollama:
|
||||
api_key: ollama
|
||||
base_url: http://localhost:11434/v1
|
||||
provider: ollama
|
||||
# 远程 API 示例 (OpenAI)
|
||||
openai:
|
||||
api_key: your_openai_api_key
|
||||
base_url: https://api.openai.com/v1
|
||||
provider: openai
|
||||
# 远程 API 示例 (Anthropic)
|
||||
anthropic:
|
||||
api_key: your_anthropic_api_key
|
||||
base_url: https://api.anthropic.com
|
||||
provider: anthropic
|
||||
|
||||
models:
|
||||
trae_agent_model:
|
||||
model_provider: omix
|
||||
model: Qwen3-8B-MLX-4bit
|
||||
max_tokens: 4096
|
||||
temperature: 0.3
|
||||
top_p: 1
|
||||
max_retries: 3
|
||||
parallel_tool_calls: false
|
||||
|
||||
# 远程模型示例:
|
||||
# remote_model:
|
||||
# model_provider: openai
|
||||
# model: gpt-4o
|
||||
# max_tokens: 128000
|
||||
# temperature: 0.3
|
||||
Reference in New Issue
Block a user