Yt Summarizer

Name: Yt Summarizer
Rating: 3.5 (1 reviews)
Author: lyrion88

By lyrion88 👁 24 views ▲ 0 votes

Openclaw plugin: transcribes YouTube videos via CloakBrowser and generates German summaries

GitHub

Install

npm install
npm

Configuration Example

{
  "plugins": {
    "enabled": true,
    "load": {
      "paths": ["/path/to/yt-summarizer"]
    },
    "entries": {
      "youtube-summarizer": {
        "enabled": true,
        "config": {
          "cloakBrowserUrl": "http://localhost:9222",
          "whisperUrl": "http://localhost:8000"
        }
      }
    }
  }
}

README

# yt-summarizer

An [Openclaw](https://openclaw.ai) plugin that opens YouTube videos via **CloakBrowser** (stealth Chromium), extracts the transcript (YouTube subtitles first, Whisper as fallback), and returns the full text for a structured German summary.

## What it does

1. Connects to a running CloakBrowser instance over CDP
2. Navigates to the YouTube URL in a stealth Chromium session
3. Extracts the transcript — two strategies in order:
   - **youtube-json** — parses `ytInitialData` directly (fast, no extra clicks)
   - **youtube-dom** — clicks the transcript panel in the UI (fallback)
4. If no subtitles exist, downloads audio via `yt-dlp` and transcribes with Whisper (local server or OpenAI API)
5. Returns the title + transcript; Openclaw's LLM produces the German summary

### Example output (after Openclaw summarizes)

```
**Hauptthema**: …
**Kernaussagen**:
- …
**Fazit**: …
```

## Requirements

| Dependency | Purpose |
|---|---|
| [Openclaw](https://openclaw.ai) ≥ 2026.1.0 | Plugin host |
| [CloakBrowser](https://cloakbrowser.com) | Stealth Chromium CDP server |
| [yt-dlp](https://github.com/yt-dlp/yt-dlp) | Audio download (Whisper fallback only) |
| Local Whisper server **or** OpenAI API key | Transcription fallback only |

### Local Whisper server (recommended — no API key needed)

```bash
docker run -d -p 8000:8000 fedirz/faster-whisper-server:latest-cpu
```

## Installation

### 1. Install the plugin in Openclaw

Add to your Openclaw config (e.g. `~/.openclaw/config.json`):

```json
{
  "plugins": {
    "enabled": true,
    "load": {
      "paths": ["/path/to/yt-summarizer"]
    },
    "entries": {
      "youtube-summarizer": {
        "enabled": true,
        "config": {
          "cloakBrowserUrl": "http://localhost:9222",
          "whisperUrl": "http://localhost:8000"
        }
      }
    }
  }
}
```

To use OpenAI Whisper instead of a local server, replace `whisperUrl` with `openaiApiKey`:

```json
"config": {
  "cloakBrowserUrl": "http://localhost:9222",
  "openaiApiKey": "sk-…"
}
```

### 2. Enable the optional tool

The `youtube_summarize` tool is optional (opt-in). Enable it:

```json
{
  "tools": {
    "allow": ["youtube_summarize"]
  }
}
```

### 3. Start CloakBrowser

```bash
cloakbrowser --cdp-port 9222
```

## Usage

Once the plugin is loaded and the tool is enabled, ask Openclaw:

```
Summarize https://www.youtube.com/watch?v=dQw4w9WgXcQ
```

or call the tool directly:

```
youtube_summarize("https://www.youtube.com/watch?v=dQw4w9WgXcQ")
```

## Configuration reference

| Field | Type | Default | Description |
|---|---|---|---|
| `cloakBrowserUrl` | string | `http://localhost:9222` | CDP URL of the CloakBrowser instance |
| `whisperUrl` | string | — | URL of a local Whisper server (OpenAI-compatible) |
| `openaiApiKey` | string | — | OpenAI API key (used only when `whisperUrl` is not set) |

`whisperUrl` takes precedence over `openaiApiKey`. If neither is set and the video has no subtitles, the tool returns an error.

## Building from source

```bash
npm install
npm run build   # compiles TypeScript → dist/
```

The built plugin is in `dist/`. The `openclaw.plugin.json` manifest at the project root is read by Openclaw at load time.

### Docker

```bash
docker build -t yt-summarizer .
```

## How transcript extraction works

```
YouTube page loaded
        │
        ▼
  ytInitialData present?
  ┌─────┴──────┐
  Yes          No
  │            │
  Parse JSON   Open transcript panel via DOM
  (fast)       (UI interaction fallback)
        │
        ▼
  Transcript found?
  ┌─────┴──────┐
  Yes          No
  │            │
  Return text  Whisper fallback
               (yt-dlp + local/OpenAI)
```

## License

MIT

tools