Embedding Skill

Provides text embedding generation via the unified embedding service.

Commands

embed_texts

Generate embeddings for multiple texts.

Input:

{
  "texts": ["text1", "text2"]
}

Output:

{
  "success": true,
  "vectors": [[0.1, 0.2, ...], [0.3, 0.4, ...]],
  "count": 2,
  "dimension": 2560
}

embed_single

Generate embedding for a single text.

Input:

{
  "text": "hello world"
}

Output:

{
  "success": true,
  "vector": [0.1, 0.2, ...],
  "dimension": 2560
}

Usage

# Via MCP client
@omni("embedding.embed_texts", {"texts": ["query text"]})

# Via Python
from omni.foundation.services.embedding import embed_batch
vectors = embed_batch(["text1", "text2"])

Notes

• Uses Qwen/Qwen3-Embedding-4B model (2560 dimensions)
• Model is preloaded when MCP server starts
• Supports both batch and single text embedding