Skip to main content
The VoyageEmbeddingOptions type defines the configuration options you can pass when generating embeddings with text embedding models.

​
Properties

​
inputType
'query' | 'document'
The input type for the embeddings.Default: queryVoyage automatically prepends a prompt to your inputs before vectorizing them:
  • For query: “Represent the query for retrieving supporting documents: ”
  • For document: “Represent the document for retrieval: ”
​
outputDimension
number
The number of dimensions for the resulting output embeddings.Default: Model-specific default dimensionSupported values vary by model:
  • voyage-3.5, voyage-3.5-lite: 2048, 1024 (default), 512, 256
  • voyage-code-3: 2048, 1024 (default), 512, 256
  • voyage-3-large: 2048, 1024 (default), 512, 256
See Voyage AI Embeddings documentation for supported values by model.
​
outputDtype
'float' | 'int8' | 'uint8' | 'binary' | 'ubinary'
The data type for the resulting output embeddings.Default: float
  • float: Each returned embedding is a list of 32-bit (4-byte) single-precision floating-point numbers. Supported by all models.
  • int8: Each returned embedding is a list of 8-bit (1-byte) integers ranging from -128 to 127. Supported by voyage-code-3.
  • uint8: Each returned embedding is a list of 8-bit (1-byte) integers ranging from 0 to 255. Supported by voyage-code-3.
  • binary: Each returned embedding is a list of 8-bit integers (int8) that represent bit-packed, quantized single-bit embedding values. The length of the returned list is 1/8 of outputDimension. Uses the offset binary method. Supported by voyage-code-3.
  • ubinary: Each returned embedding is a list of 8-bit integers (uint8) that represent bit-packed, quantized single-bit embedding values. The length of the returned list is 1/8 of outputDimension. Supported by voyage-code-3.
See Voyage AI FAQ on quantization
​
truncation
boolean
Whether to truncate the input texts to fit within the context length.

​
Usage

​
Basic usage with input type

import { voyage } from '@voyageai/ai-provider';
import { embed } from 'ai';

const { embedding } = await embed({
  model: voyage.textEmbeddingModel('voyage-3'),
  value: 'What is the capital of France?',
  experimental_providerOptions: {
    voyage: {
      inputType: 'query',
    },
  },
});

​
Using custom output dimensions

import { voyage } from '@voyageai/ai-provider';
import { embed } from 'ai';

const { embedding } = await embed({
  model: voyage.textEmbeddingModel('voyage-code-3'),
  value: 'function calculateSum(a, b) { return a + b; }',
  experimental_providerOptions: {
    voyage: {
      inputType: 'document',
      outputDimension: 512,
    },
  },
});

​
Using quantized embeddings

import { voyage } from '@voyageai/ai-provider';
import { embed } from 'ai';

const { embedding } = await embed({
  model: voyage.textEmbeddingModel('voyage-code-3'),
  value: 'const greeting = "Hello, World!";',
  experimental_providerOptions: {
    voyage: {
      outputDtype: 'int8',
      outputDimension: 1024,
    },
  },
});

​
Batch embeddings with truncation

import { voyage } from '@voyageai/ai-provider';
import { embedMany } from 'ai';

const { embeddings } = await embedMany({
  model: voyage.textEmbeddingModel('voyage-3'),
  values: [
    'First document',
    'Second document',
    'Third document',
  ],
  experimental_providerOptions: {
    voyage: {
      inputType: 'document',
      truncation: true,
    },
  },
});

Build docs developers (and LLMs) love

Get started for freeTalk to us