Use embeddings endpoint

Use the /api/v1/embeddings endpoint within the LLM Service for Titan Text V2. This endpoint converts input text into numerical vector embeddings that capture semantic meaning, enabling use cases such as semantic search and content clustering.

Ensure these guidelines are met when you use the embeddings endpoint:

  • inputText field is required (all other fields in the request payload are optional)
  • dimensions field supports only these values: 256, 512, or 1024
  • embeddingTypes supports these values: binary or float. When this value is not specified, the embeddings response will be generated in a float format.

Payload for using the /api/v1/embeddings endpoint

{
    "inputText": "Embeddings turn text into dense numerical vectors that capture meaning — so the model can understand similarity and relationships between sentences",
    "config": {
        "dimensions": 512,
        "normalize": true,
        "embeddingTypes": [
            "binary"
        ]
    }
}