Panniantong
a9f5fd93d7
Round 1: Problem-solution structure, showcase underlying tools
Round 2: Too verbose - 4 'can't do' examples is repetitive, tightened
Round 3: Compact doctor output, restored quick-reference table,
cut code examples to essentials
Round 4: Opening paragraph tighter - one sentence for all pain points
Key: README doesn't say 'we can read URLs' (every tool does that).
It says 'here are the best open-source tools for each platform,
we glued them together so you install once and get everything.'
178 lines
5.1 KiB
Markdown
178 lines
5.1 KiB
Markdown
# 👁️ Agent Eyes
|
||
|
||
**一行命令，让你的 AI Agent 看见整个互联网。**
|
||
|
||
你的 AI Agent 很聪明，但几乎是个瞎子。Reddit 封 IP、Twitter API 要 $100/月、B站屏蔽服务器、小红书需要登录——每个平台都有自己的坑，你要一个一个去踩。
|
||
|
||
Agent Eyes 把互联网上最好的开源工具粘在一起，一次安装全部搞定：
|
||
|
||
| 工具 | 它解决了什么 | Stars |
|
||
|------|------------|:-----:|
|
||
| [Jina Reader](https://github.com/jina-ai/reader) | 任意网页 → 干净 Markdown，处理 JS 渲染，去掉广告 | ⭐ 9.8K |
|
||
| [yt-dlp](https://github.com/yt-dlp/yt-dlp) | **1800+ 个视频网站**的字幕提取（YouTube、B站、TikTok…） | ⭐ 148K |
|
||
| [Exa](https://exa.ai) | AI 语义搜索引擎，一个 Key 搜全网 + Reddit + Twitter | — |
|
||
| [birdx](https://github.com/runesleo/birdx) | 不花 $100/月，用 Cookie 就能搜 Twitter 时间线和线程 | — |
|
||
| [feedparser](https://github.com/kurtmckee/feedparser) | 万能 RSS/Atom 解析 | ⭐ 2.3K |
|
||
|
||
> 不造轮子，只做胶水。统一安装、统一配置、统一接口。
|
||
|
||
[English](docs/README_en.md)
|
||
|
||
---
|
||
|
||
## 30 秒上手
|
||
|
||
复制给你的 AI Agent（Claude Code、OpenClaw、Cursor 等）：
|
||
|
||
```
|
||
帮我安装 Agent Eyes：https://raw.githubusercontent.com/Panniantong/agent-eyes/main/docs/install.md
|
||
```
|
||
|
||
Agent 自动安装、检测环境、告诉你哪些功能已经可以用。
|
||
|
||
<details>
|
||
<summary>手动安装</summary>
|
||
|
||
```bash
|
||
pip install https://github.com/Panniantong/agent-eyes/archive/main.zip
|
||
agent-eyes install --env=auto
|
||
```
|
||
</details>
|
||
|
||
---
|
||
|
||
## 装好就能用
|
||
|
||
不需要任何配置，装完直接用：
|
||
|
||
```bash
|
||
agent-eyes read "https://任意网页" # Jina Reader 驱动
|
||
agent-eyes read "https://github.com/openai/gpt-4" # GitHub 仓库/Issue/PR
|
||
agent-eyes read "https://www.youtube.com/watch?v=xxx" # yt-dlp 驱动
|
||
agent-eyes read "https://www.bilibili.com/video/BVxxx" # B站字幕
|
||
agent-eyes read "https://x.com/elonmusk/status/xxx" # 推文
|
||
agent-eyes read "https://hnrss.org/frontpage" # RSS 订阅
|
||
agent-eyes search-github "LLM 框架" # GitHub 搜索
|
||
```
|
||
|
||
---
|
||
|
||
## 按需解锁
|
||
|
||
不用的不用配。每一步都可以跳过。
|
||
|
||
### 🔍 搜索 — 免费，30 秒
|
||
|
||
一个 [Exa](https://exa.ai) Key（免费 1000 次/月），同时解锁三个搜索：
|
||
|
||
```bash
|
||
agent-eyes configure exa-key 你的KEY
|
||
|
||
agent-eyes search "2025 最好的开源 AI 工具"
|
||
agent-eyes search-reddit "best LLM" --sub LocalLLaMA
|
||
agent-eyes search-twitter "Claude Code"
|
||
```
|
||
|
||
### 🍪 Cookie — 免费，2 分钟
|
||
|
||
解锁 Twitter 高级搜索 + 小红书。本地电脑一键导入：
|
||
|
||
```bash
|
||
agent-eyes configure --from-browser chrome
|
||
```
|
||
|
||
> 服务器用户？装个 [Cookie-Editor](https://chromewebstore.google.com/detail/cookie-editor/hlkenndednhfkekhgcdicdfddnkalmdm) 扩展，Export → Header String，粘贴给 Agent。
|
||
|
||
### 🌐 代理 — $1/月，仅服务器需要
|
||
|
||
Reddit 和 B站封服务器 IP。一个代理解决两个：
|
||
|
||
```bash
|
||
agent-eyes configure proxy http://用户名:密码@IP:端口
|
||
```
|
||
|
||
> Reddit 搜索通过 Exa 免费可用，不买代理也能搜，只是读不了完整帖子。
|
||
> 推荐 [Webshare](https://webshare.io)，$1/月。
|
||
|
||
---
|
||
|
||
## 状态一目了然
|
||
|
||
```
|
||
$ agent-eyes doctor
|
||
|
||
👁️ Agent Eyes Status
|
||
✅ Web [Jina Reader] ✅ GitHub [API] ✅ RSS [feedparser]
|
||
✅ YouTube [yt-dlp] ✅ Bilibili [API] ✅ Twitter [birdx]
|
||
⬜ Search [need Exa key] ⬜ XiaoHongShu [cookies] ⬜ Reddit [proxy]
|
||
|
||
6/9 active
|
||
```
|
||
|
||
---
|
||
|
||
## 接入方式
|
||
|
||
### CLI
|
||
```bash
|
||
agent-eyes read "URL"
|
||
agent-eyes search "关键词"
|
||
agent-eyes doctor
|
||
```
|
||
|
||
### Python
|
||
```python
|
||
from agent_eyes import AgentEyes
|
||
import asyncio
|
||
eyes = AgentEyes()
|
||
asyncio.run(eyes.read("https://example.com"))
|
||
asyncio.run(eyes.search("AI agent"))
|
||
```
|
||
|
||
### MCP Server（Claude Code / Cursor）
|
||
|
||
<details>
|
||
<summary>配置</summary>
|
||
|
||
```json
|
||
{"mcpServers": {"agent-eyes": {"command": "python", "args": ["-m", "agent_eyes.integrations.mcp_server"]}}}
|
||
```
|
||
</details>
|
||
|
||
---
|
||
|
||
## 速查
|
||
|
||
| 命令 | 作用 |
|
||
|------|------|
|
||
| `agent-eyes doctor` | 查看状态 |
|
||
| `agent-eyes configure --from-browser chrome` | 一键导入 cookies |
|
||
| `agent-eyes configure exa-key KEY` | 解锁搜索 |
|
||
| `agent-eyes configure twitter-cookies "..."` | 解锁 Twitter |
|
||
| `agent-eyes configure xhs-cookie "..."` | 解锁小红书 |
|
||
| `agent-eyes configure proxy URL` | 解锁 Reddit + B站 |
|
||
|
||
---
|
||
|
||
## 设计
|
||
|
||
胶水，不是框架。每个频道 ~50 行代码，换后端改一个文件。
|
||
|
||
<details>
|
||
<summary>架构</summary>
|
||
|
||
```
|
||
web.py → Jina Reader github.py → GitHub API
|
||
youtube.py → yt-dlp bilibili.py → Bilibili API
|
||
twitter.py → birdx reddit.py → Reddit JSON
|
||
exa.py → Exa Search rss.py → feedparser
|
||
xhs.py → XHS Web API
|
||
```
|
||
</details>
|
||
|
||
## 致谢
|
||
|
||
[Jina Reader](https://github.com/jina-ai/reader) · [yt-dlp](https://github.com/yt-dlp/yt-dlp) · [birdx](https://github.com/runesleo/birdx) · [Exa](https://exa.ai) · [feedparser](https://github.com/kurtmckee/feedparser)
|
||
|
||
MIT License
|