Embeddings

New Embedding Models

The latest embedding models from UmamiAI API, text-embedding-3-small and text-embedding-3-large, are now available. These models offer cost savings, enhanced multilingual support, and customizable parameters to manage their size.

What are Embeddings?

Embeddings from UmamiAI API quantify the similarity between text strings. These embeddings are particularly useful for:

• Search: Rank search results by their relevance to a query.
• Clustering: Group similar text strings together.
• Recommendations: Suggest items based on related text strings.
• Anomaly Detection: Identify outliers that differ significantly from the norm.
• Diversity Measurement: Analyze the spread of similarities within a dataset.
• Classification: Categorize text strings by comparing them to labeled examples.

An embedding is a vector (list) of floating-point numbers, where the distance between vectors indicates their relatedness. Smaller distances indicate higher similarity, while larger distances suggest lower similarity.

For more information on Embeddings pricing, visit our pricing page. Costs are calculated based on the number of tokens in the input.

Example: Generating Embeddings

curl https://api.umamiai.xyz/v1/embeddings \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_UMAMIAI_API_KEY" \
  -d '{
    "input": "Your text string goes here",
    "model": "text-embedding-3-small"
  }'

The response will include the embedding vector and additional metadata.

Example Embedding Response

{
  "object": "list",
  "data": [
    {
      "object": "embedding",
      "index": 0,
      "embedding": [
        -0.006929283495992422,
        -0.005336422007530928,
        // ...(omitted for spacing)
        -4.547132266452536e-05,
        -0.024047505110502243
      ]
    }
  ],
  "model": "text-embedding-3-small",
  "usage": {
    "prompt_tokens": 5,
    "total_tokens": 5
  }
}

By default, the length of the embedding vector is 1536 for text-embedding-3-small or 3072 for text-embedding-3-large. You can reduce the dimensions of the embedding using the dimensions parameter without losing its ability to represent concepts. More details on embedding dimensions can be found in the embedding use case section.

Embedding Models

UmamiAI API offers two robust third-generation embedding models (indicated by -3 in the model ID).

Model	~ Pages per Dollar	Performance on MTEB Eval	Max Input Tokens
text-embedding-3-small	62,500	62.3%	8191
text-embedding-3-large	9,615	64.6%	8191
text-embedding-ada-002	12,500	61.0%	8191

Example in Python

Here's how to use the embeddings API in Python:

import os
import json
import openai

# Initialize the API client
client = openai.OpenAI(
    base_url="https://api.umamiai.xyz/v1",
    api_key=os.getenv("UMAMIAI_API_KEY"),
)

# Define the text for which to generate an embedding
text = "Your text string goes here"

# Request the embedding
response = client.embeddings.create(
    input=text,
    model="text-embedding-3-small"
)

# Extract the embedding from the response
embedding = response['data'][0]['embedding']

# Print the embedding
print(json.dumps(embedding, indent=2))

This Python example shows how to set up an API client, send text to the embeddings API, and handle the response to extract and print the embedding vector.