Claw Onichan

Name: Claw Onichan
Rating: 3.5 (1 reviews)
Author: IYENTeam

By IYENTeam 👁 54 views ▲ 0 votes

OpenClaw plugin: detect emotion keywords in agent replies and attach mapped images via Discord API

GitHub

Configuration Example

{
  "plugins": {
    "entries": {
      "emotion-image": {
        "enabled": true
      }
    }
  }
}

README

# 🦀 claw-onichan

> Auto-attach emotion images to your OpenClaw agent's Discord messages.

Your agent sends a message → the plugin detects the mood from keywords → edits the message to attach a matching image. No core patches needed — pure OpenClaw Plugin SDK.

```
Agent sends: "Task completed! All tests passed ✅"
     → Plugin detects: happy
     → Discord message gets: happy.png attached
```

## How It Works

1. Hooks into `message_sent` — fires after every outbound Discord message
2. Scans the text for emotion keywords (regex pattern matching)
3. Calls Discord REST API (`PATCH`) to edit the message with an image attachment

The agent doesn't need to know about images at all. The plugin handles everything post-delivery.

---

## For AI Agents Setting This Up

> **If you're an AI agent installing this plugin for a user, here's what you need from them before proceeding.**

### What the user must provide

1. **Emotion images** — This plugin ships with NO images. Ask the user to:
   - Provide image files (PNG recommended, 256–512px, under 512KB each)
   - One image per emotion: `happy.png`, `neutral.png`, `sorry.png`, `confused.png`, `focused.png`, `loyalty.png`
   - Or tell you what style/character they want, and generate the images for them using an image generation tool

2. **Discord bot token** — The plugin calls the Discord REST API directly. The user must either:
   - Give you a bot token to set as `EMOTION_IMAGE_DISCORD_TOKEN` env var
   - Or confirm they already have a Discord account configured in `openclaw.json` (the plugin auto-detects the first available token)

3. **Which emotions to support** — The built-in set is `happy`, `neutral`, `sorry`, `confused`, `focused`, `loyalty`. Ask if they want to:
   - Add custom emotions (e.g., `sleepy`, `angry`, `excited`)
   - Modify the trigger keywords for existing emotions
   - Change the default/fallback emotion

### What you can do without asking

- Clone the repo to `extensions/emotion-image`
- Register the plugin in `openclaw.json` → `plugins.entries.emotion-image`
- Set up the directory structure
- Configure `emotionMap` and `emotionRules` based on provided images/preferences
- Restart the gateway after setup

### Setup checklist

```
□ Plugin cloned to extensions/emotion-image
□ Registered in openclaw.json plugins.entries
□ Image files placed in assets/ directory (USER MUST PROVIDE)
□ Discord bot token configured (USER MUST CONFIRM)
□ Gateway restarted
□ Test: send a message and verify image attachment in logs
```

---

## Quick Start

### 1. Install

```bash
cd /path/to/openclaw
git clone https://github.com/IYENTeam/claw-onichan.git extensions/emotion-image
```

### 2. Register in config

Add to your `openclaw.json`:

```jsonc
{
  "plugins": {
    "entries": {
      "emotion-image": {
        "enabled": true
      }
    }
  }
}
```

### 3. Prepare your images ⚠️

**This plugin does NOT include any images.** You must create or source them yourself.

Place image files in the `assets/` directory:

```
extensions/emotion-image/assets/
├── happy.png
├── neutral.png
├── sorry.png
├── confused.png
├── focused.png
└── loyalty.png
```

#### Default emotion → filename mapping

| Emotion | File | Triggers on |
|---------|------|-------------|
| `happy` | `happy.png` | 완료, success, done, ✅, 🎉 |
| `neutral` | `neutral.png` | *(fallback when nothing else matches)* |
| `sorry` | `sorry.png` | 죄송, sorry, 실수, apolog |
| `confused` | `confused.png` | 모르, confused, 질문 |
| `focused` | `focused.png` | 분석, 작업, 디버그, 코드, 테스트 |
| `loyalty` | `loyalty.png` | 안녕, 알겠, 충성 |

#### Creating images

**Option A — AI image generation**

Use any image generator with prompts like:

| Emotion | Example prompt |
|---------|---------------|
| happy | *"Cute 2D character smiling warmly, simple flat style, transparent background, 512x512"* |
| sorry | *"Cute 2D character looking apologetic with sweat drop, soft pastel, transparent background, 512x512"* |
| confused | *"Cute 2D character with question mark above head, purple tones, transparent background, 512x512"* |
| focused | *"Cute 2D character with determined eyes and glasses, cool blue, transparent background, 512x512"* |

**Option B — Open source icons**

- [OpenMoji](https://openmoji.org/) — open source emoji
- [Twemoji](https://github.com/twitter/twemoji) — Twitter's emoji set
- [Flaticon](https://www.flaticon.com/) — free icons (check license)

**Recommended specs:**
- Format: PNG (transparent background supported)
- Size: 256×256 to 512×512 px
- Aspect ratio: 1:1
- File size: under 512KB

### 4. Set Discord bot token

The plugin needs a Discord bot token to call the edit API.

**Option A — Environment variable (recommended)**

```bash
export EMOTION_IMAGE_DISCORD_TOKEN="your-bot-token"
```

**Option B — Auto-detect from OpenClaw config**

If your `openclaw.json` has a Discord account configured, the plugin will automatically use the first available bot token:

```jsonc
{
  "channels": {
    "discord": {
      "accounts": {
        "my-bot": {
          "token": "your-bot-token"
        }
      }
    }
  }
}
```

### 5. Restart OpenClaw

```bash
openclaw gateway restart
```

That's it. Your agent's messages will now get emotion images automatically.

---

## Configuration

All options go in `plugins.entries.emotion-image.config`:

```jsonc
{
  "plugins": {
    "entries": {
      "emotion-image": {
        "enabled": true,
        "config": {
          "imageDir": "./my-custom-images",
          "defaultEmotion": "neutral",
          "emotionMap": {
            "happy": "my-happy.webp"
          },
          "emotionRules": {
            "excited": ["awesome", "amazing", "incredible"]
          }
        }
      }
    }
  }
}
```

| Option | Type | Default | Description |
|--------|------|---------|-------------|
| `enabled` | boolean | `true` | Kill switch |
| `imageDir` | string | `assets/` | Directory containing emotion images |
| `defaultEmotion` | string | `"neutral"` | Fallback when no keywords match |
| `emotionMap` | object | built-in | Emotion name → image filename override |
| `emotionRules` | object | built-in | Emotion name → keyword patterns to add |

### Adding custom emotions

1. Add keywords in `emotionRules`:
```jsonc
{ "emotionRules": { "sleepy": ["tired", "졸려", "zzz"] } }
```

2. Map to an image file in `emotionMap`:
```jsonc
{ "emotionMap": { "sleepy": "sleepy.png" } }
```

3. Place `sleepy.png` in your `imageDir`.

---

## Built-in Emotion Rules

| Emotion | Keywords |
|---------|----------|
| sorry | 죄송, 미안, 실수, 잘못, sorry, apolog |
| happy | 완료, 성공, 됐습니다, 해결, pass, done, ✅, 🎉, 축하 |
| confused | 모르, 이상하, 원인불명, confused, unclear, 질문 |
| focused | 조사, 분석, 디버그, 작업, 수정, 코드, 빌드, 배포, 테스트 |
| loyalty | 회장님, 충성, 알겠, 안녕, ㅎㅇ |

Rules are evaluated top-to-bottom. First match wins.

---

## Troubleshooting

**Hook never fires**
- Make sure the plugin uses `api.on("message_sent", ...)` — NOT `api.registerHook(...)`. These are different systems in OpenClaw. `registerHook` is for legacy/HTTP hooks and won't be called by the delivery pipeline.

**Discord edit fails (no image appears)**
- Check your bot token is valid
- The bot can only edit its own messages
- Look for `emotion-image: Discord edit failed` in gateway logs
- Check for Discord API rate limits (429 status)

**"Discord token not found" warning**
- Set `EMOTION_IMAGE_DISCORD_TOKEN` env var, or ensure your `openclaw.json` has a Discord account with a token

**Image not found**
- Verify image files exist in your `imageDir`
- Filenames must match `emotionMap` values exactly

---

## Development

```bash
# Run tests (from OpenClaw root)
npx vitest run extensions/emotion-image/index.test.ts

# 27 tests covering:
# - detectEmotion() keyword matching
# - MEDIA line cleanup
# - channel: prefix stripping
# - buildEmotionRules() config merging
# - editMessageWithImage() Discord API calls
# - Plugin entry shape validation
```

---

## License

MIT

channels