Embeddings

Available Models

The full list of models can be viewed on the Models page or retrieved dynamically via API endpoints: https://openai.bothub.chat/v1/models and https://bothub.chat/api/v2/replicate/v1/models

Embeddings are numerical representations of text that allow models to understand meaning and semantic proximity between words, sentences, or entire documents.

When text is converted into an embedding vector, the model preserves not the words themselves, but their semantic representation. This allows you to compare texts with each other, find similar content, and perform intelligent searches.

Embeddings are actively used in AI applications, search systems, and RAG (Retrieval-Augmented Generation) scenarios, where the model works with an external knowledge base.

---

Where Embeddings Are Used

Embeddings are especially useful in tasks where understanding the meaning of text is important, rather than just matching words. For example, they are used for semantic search, finding similar documents, content recommendations, intelligent chatbots, text clustering, and data classification.

Embeddings are also an essential part of RAG systems, where the model first searches for relevant information in a knowledge base and then uses it when generating a response.

---

Available Models

We have the following embedding models available:

• text-embedding-3-small (for fast and low-cost tasks)
• text-embedding-3-large (for higher quality and search accuracy)
• text-embedding-ada-002

---

Generating Embeddings

To create an embedding vector, send text via the API and specify the embedding model. The API will return an array of numbers that can be used for searching, comparing, or storing in a vector database.

Below is a basic example of generating embeddings using the text-embedding-3-large model.

JS

async function main() {
  const response = await openai.embeddings.create({
    model: 'text-embedding-3-large',
    input: 'The food was delicious and the waiter...',
    encoding_format: 'float'
  });

  const embedding = response.data[0].embedding;

  console.log(embedding);
}

main();

Python

def main():
    response = client.embeddings.create(
        model="text-embedding-3-large",
        input="The food was delicious and the waiter...",
        encoding_format="float"
    )

    embedding = response.data[0].embedding

    print(embedding)

main()

Go

package main

import (
	"context"
	"fmt"

	"github.com/openai/openai-go"
	"github.com/openai/openai-go/option"
)

func main() {
	client := openai.NewClient(
		option.WithAPIKey("YOUR_BOTHUB_API_KEY"),
		option.WithBaseURL("https://openai.bothub.chat/v1"),
	)

	response, err := client.Embeddings.New(context.Background(), openai.EmbeddingNewParams{
		Model:          openai.EmbeddingModelTextEmbedding3Large,
		Input:          openai.EmbeddingNewParamsInputUnion{OfString: openai.String("The food was delicious and the waiter...")},
		EncodingFormat: openai.EmbeddingNewParamsEncodingFormatFloat,
	})
	if err != nil {
		panic(err)
	}

	fmt.Println(response.Data[0].Embedding)
}

cURL

curl https://openai.bothub.chat/v1/embeddings \
  -X POST \
  -H "Authorization: Bearer <your access token>" \
  -H "Content-Type: application/json" \
  -d '{
    "input": "The food was delicious and the waiter...",
    "model": "text-embedding-3-large",
    "encoding_format": "float"
  }'

---

Response Format

The API returns an object containing the embedding vector. The vector itself is an array of floating-point numbers (float) that describes the meaning of the text in a multi-dimensional space.

The resulting embeddings are typically stored in an external vector database or used to calculate semantic proximity between texts.

Example of accessing the embedding vector:

const embedding = response.data[0].embedding;

---

Main Parameters

Parameter	Type	Description
model	string	Embedding model
input	string | array	Text or array of texts
encoding_format	string	Format of the embedding vector (float or base64)

---

Recommendations

info

For most tasks, text-embedding-3-large is recommended if you require maximum accuracy for searching and comparing texts.

If speed and cost are priorities, text-embedding-3-small is a better choice.

You can view the official OpenAI documentation on embeddings here