matrix-nsfwbot/README.md

# nsfwbot for Matrix

`nsfwbot` is a Matrix bot plugin that detects NSFW (Not Safe For Work) images posted in Matrix chat rooms. It uses [nudenet](https://github.com/notAI-tech/NudeNet), which includes a deep learning model that efficiently runs without requiring a GPU, with low resource demands.

## Features

- **Image Analysis**: Detects and analyzes images posted in Matrix chats using NudeNet.
- **Redaction of Spammed Images**: Automatically redacts images when users send more than the allowed number within a short time frame.
- **Text Message Parsing**: Analyzes images embedded in text messages.
- **Custom Actions**: Configurable actions for detected content, including reporting and redacting messages.

## Requirements

- **Maubot**: Runs within the Maubot framework.
- **Python Dependencies**: `nudenet`, `beautifulsoup4`, and `maubot`.
  
  To install the dependencies, you can find them listed in the `requirements.txt`:
  ```plaintext
  beautifulsoup4
  nsfwdetection
  maubot
  ```

## Installation

1. **Use the Custom Maubot Docker Image**:
   Replace the official Maubot image with a custom Debian-based image:

   ```bash
   docker pull ghcr.io/tcpipuk/maubot:debian
   ```

2. a. **Install pre-prepared plugin from [repository releases](https://git.buechner.me/nbuechner/matrix-nsfwbot/releases)**

   b. **Clone the Repository**:

      ```bash
      git clone https://git.buechner.me/nbuechner/matrix-nsfwbot
      ```

      Zip the plugin files and upload through the Maubot admin interface. Ensure the plugin is configured and enabled.

3. **Configure the Plugin**:
   Edit `base-config.yaml` to set:

   - `max_concurrent_jobs`: Number of concurrent jobs to allow.
   - `via_servers`: List of servers for `matrix.to` URLs.
   - `actions`:
     - `ignore_sfw`: Ignore SFW images (default: `true`).
     - `redact_nsfw`: Redact NSFW messages (default: `true`).
     - `direct_reply`: Reply directly in the same room (default: `false`).
     - `report_to_room`: Room ID for reporting (optional).
     - `max_images`: The maximum number of images a user can send within a certain time window (default: `3`).
     - `time_window`: The time window (in seconds) in which the image count is tracked (default: `60` seconds).

   > **Note** `report_to_room` can be a room alias (like `#room:server`), but it is more efficient to use a room ID (`!room:server`).

## Usage

Once installed and configured, `nsfwbot` will automatically analyze images posted in the chat and:

- Classify images as NSFW or SFW.
- Redact images if a user exceeds the configured spam limit (`max_images`).
- Optionally report detected NSFW content to a configured room.

Example output:

```markdown
mxc://matrix.org/abcd1234 in https://matrix.to/#/!roomid:matrix.org/$eventid?via=matrix.org appears NSFW with score 87.93%
```

If multiple images are detected in a text message:

```markdown
- mxc://matrix.org/abcd1234 in https://matrix.to/#/!roomid:matrix.org/$eventid?via=matrix.org appears SFW with score 2.45%
- mxc://matrix.org/efgh5678 in https://matrix.to/#/!roomid:matrix.org/$eventid?via=matrix.org appears NSFW with score 94.82%
```

If a user sends too many images:

```markdown
User @example:matrix.org has exceeded the image limit. Previous images redacted.
```