# me.byjp.emojify.suggestEmoji

> Published by [byjp.me](https://lexicon.garden/identity/did:plc:ephkzpinhaqcabtkugtbzrwu)

## Links

- [View on Lexicon Garden](https://lexicon.garden/lexicon/did:plc:ephkzpinhaqcabtkugtbzrwu/me.byjp.emojify.suggestEmoji)
- [Documentation](https://lexicon.garden/lexicon/did:plc:ephkzpinhaqcabtkugtbzrwu/me.byjp.emojify.suggestEmoji/docs)
- [Examples](https://lexicon.garden/lexicon/did:plc:ephkzpinhaqcabtkugtbzrwu/me.byjp.emojify.suggestEmoji/examples)

## Definitions

### `me.byjp.emojify.suggestEmoji`

**Type**: `query`

Suggest emoji whose meaning best matches a short passage of text.

#### Parameters

| Property | Type | Required | Description |
|----------|------|----------|-------------|
| `text` | `string` | Yes | The passage of text to match, up to 600 units. Despite the maxGraphemes field name, the limit is actually enforced in Unicode code points (runes), not user-perceived graphemes — a single multi-rune grapheme cluster (e.g. a ZWJ family emoji) counts as multiple units toward it. maxLength is in UTF-8 bytes, not runes — emoji-laden input can run up to 4 bytes per rune. Note that the underlying model reads at most 256 tokens (~1300 runes of ordinary English, far fewer for punctuation-dense text); anything beyond that is ignored rather than rejected, and a long passage whose subject is a small fraction of the text tends to yield a generic suggestion. |
| `limit` | `integer` | No | Maximum number of suggestions to return. |

#### Output

**Encoding**: `application/json`

| Property | Type | Required | Description |
|----------|------|----------|-------------|
| `suggestions` | `array` | Yes |  |

#### Errors

- **TextTooLong**: text exceeds the 600-unit limit (counted in Unicode code points/runes, not graphemes).
- **NoMatch**: no suggestion cleared the minimum score.
- **RateLimitExceeded**: too many requests from this client; slow down and retry.
- **Overloaded**: the server is at capacity and shed this request; retry after a short delay.

### `me.byjp.emojify.suggestEmoji#suggestion`

**Type**: `object`

| Property | Type | Required | Description |
|----------|------|----------|-------------|
| `name` | `string` | Yes | The emoji's CLDR label. |
| `emoji` | `string` | Yes | The suggested emoji. |
| `score` | `integer` | Yes | A similarity score — cosine similarity between the query and this emoji's meaning, adjusted by a per-emoji penalty that suppresses generically-matching emoji — expressed in per-mille (parts per thousand, 0-1000). This is not basis points (which would be parts per ten-thousand). Divide by 1000 for the 0..1 float value used internally — this is an integer because the lexicon primitive set has no float type. |

## Raw Schema

```json
{
  "id": "me.byjp.emojify.suggestEmoji",
  "defs": {
    "main": {
      "type": "query",
      "errors": [
        {
          "name": "TextTooLong",
          "description": "text exceeds the 600-unit limit (counted in Unicode code points/runes, not graphemes)."
        },
        {
          "name": "NoMatch",
          "description": "no suggestion cleared the minimum score."
        },
        {
          "name": "RateLimitExceeded",
          "description": "too many requests from this client; slow down and retry."
        },
        {
          "name": "Overloaded",
          "description": "the server is at capacity and shed this request; retry after a short delay."
        }
      ],
      "output": {
        "schema": {
          "type": "object",
          "required": [
            "suggestions"
          ],
          "properties": {
            "suggestions": {
              "type": "array",
              "items": {
                "ref": "#suggestion",
                "type": "ref"
              }
            }
          }
        },
        "encoding": "application/json"
      },
      "parameters": {
        "type": "params",
        "required": [
          "text"
        ],
        "properties": {
          "text": {
            "type": "string",
            "maxLength": 2400,
            "description": "The passage of text to match, up to 600 units. Despite the maxGraphemes field name, the limit is actually enforced in Unicode code points (runes), not user-perceived graphemes — a single multi-rune grapheme cluster (e.g. a ZWJ family emoji) counts as multiple units toward it. maxLength is in UTF-8 bytes, not runes — emoji-laden input can run up to 4 bytes per rune. Note that the underlying model reads at most 256 tokens (~1300 runes of ordinary English, far fewer for punctuation-dense text); anything beyond that is ignored rather than rejected, and a long passage whose subject is a small fraction of the text tends to yield a generic suggestion.",
            "maxGraphemes": 600
          },
          "limit": {
            "type": "integer",
            "default": 3,
            "maximum": 5,
            "minimum": 1,
            "description": "Maximum number of suggestions to return."
          }
        }
      },
      "description": "Suggest emoji whose meaning best matches a short passage of text."
    },
    "suggestion": {
      "type": "object",
      "required": [
        "emoji",
        "name",
        "score"
      ],
      "properties": {
        "name": {
          "type": "string",
          "description": "The emoji's CLDR label."
        },
        "emoji": {
          "type": "string",
          "description": "The suggested emoji."
        },
        "score": {
          "type": "integer",
          "maximum": 1000,
          "minimum": 0,
          "description": "A similarity score — cosine similarity between the query and this emoji's meaning, adjusted by a per-emoji penalty that suppresses generically-matching emoji — expressed in per-mille (parts per thousand, 0-1000). This is not basis points (which would be parts per ten-thousand). Divide by 1000 for the 0..1 float value used internally — this is an integer because the lexicon primitive set has no float type."
        }
      }
    }
  },
  "$type": "com.atproto.lexicon.schema",
  "lexicon": 1
}
```