Embeddings

Hopfield allows developers to easily query and validate responses from embeddings providers. You can use different APIs with type guarantees with Zod and composability.

Usage

Create and use embeddings from OpenAI, directly with an API client:

ts
import hop from "hopfield";
import openai from "hopfield/openai";
import OpenAI from "openai";
 
const hopfield = hop.client(openai).provider(new OpenAI());
 
const embeddings = hopfield.embedding("text-embedding-ada-002", 3);
 
const response = await embeddings.get({ input: ["ready", "set", "hop"] });
 
const embeddingCount = response.data.length;
            
const embeddingCount: 3
 
const embeddingLength = response.data[0].embedding.length;
            
const embeddingLength: 1536

ts
import hop from "hopfield";
import openai from "hopfield/openai";
import OpenAI from "openai";
 
const hopfield = hop.client(openai).provider(new OpenAI());
 
const embeddings = hopfield.embedding("text-embedding-ada-002", 3);
 
const response = await embeddings.get({ input: ["ready", "set", "hop"] });
 
const embeddingCount = response.data.length;
            
const embeddingCount: 3
 
const embeddingLength = response.data[0].embedding.length;
            
const embeddingLength: 1536

Or if you prefer, you can only use Hopfield's Zod validations, and use the OpenAI SDK directly:

ts
import hop from "hopfield";
import openai from "hopfield/openai";
import OpenAI from "openai";
 
const openAIClient = new OpenAI();
 
const hopfield = hop.client(openai);
 
const embeddings = hopfield.embedding("text-embedding-ada-002", 3);
 
const response = await openAIClient.embeddings.create({
  model: embeddings.model,
  input: ["ready", "set", "hop"],
});
 
const parsed = embeddings.returnType.parse(response);
 
const embeddingCount = parsed.data.length;
            
const embeddingCount: 3
 
const embeddingLength = parsed.data[0].embedding.length;
            
const embeddingLength: 1536

ts
import hop from "hopfield";
import openai from "hopfield/openai";
import OpenAI from "openai";
 
const openAIClient = new OpenAI();
 
const hopfield = hop.client(openai);
 
const embeddings = hopfield.embedding("text-embedding-ada-002", 3);
 
const response = await openAIClient.embeddings.create({
  model: embeddings.model,
  input: ["ready", "set", "hop"],
});
 
const parsed = embeddings.returnType.parse(response);
 
const embeddingCount = parsed.data.length;
            
const embeddingCount: 3
 
const embeddingLength = parsed.data[0].embedding.length;
            
const embeddingLength: 1536

Parameters

Model Name

The model name to use for the embedding. This parameter depends on the client specified. The embedding length will change based on this parameter, since different text embeddings can have varying lengths.

ts
const embeddings = hopfield.embedding("text-embedding-ada-002");
 
const response = await embeddings.get({ input: ["hop"] });
 
const embedding = response.data[0].embedding;
         
const embedding: [number, number, number, number, number, number, number, number, number, number, number, number, number, number, number, number, number, number, number, number, number, ... 1514 more ..., number]

ts
const embeddings = hopfield.embedding("text-embedding-ada-002");
 
const response = await embeddings.get({ input: ["hop"] });
 
const embedding = response.data[0].embedding;
         
const embedding: [number, number, number, number, number, number, number, number, number, number, number, number, number, number, number, number, number, number, number, number, number, ... 1514 more ..., number]

OpenAI

The OpenAI model name defaults to the value shown below. This is currently the only supported model.

ts
import type { DefaultOpenAIEmbeddingModelName } from "hopfield/openai";
                            (alias) type DefaultOpenAIEmbeddingModelName = "text-embedding-ada-002"
import DefaultOpenAIEmbeddingModelName

ts
import type { DefaultOpenAIEmbeddingModelName } from "hopfield/openai";
                            (alias) type DefaultOpenAIEmbeddingModelName = "text-embedding-ada-002"
import DefaultOpenAIEmbeddingModelName

Embedding Count

The count of text embeddings to be returned. For all providers, this defaults to 1. This is capped at 20.

ts
const embeddings = hopfield.embedding("text-embedding-ada-002", 3);
 
const response = await embeddings.get({ input: ["ready", "set", "hop"] });
 
const embeddingLength = response.data.length;
            
const embeddingLength: 3
 
const thirdEmbeddingLength = response.data[2].embedding.length;
               
const thirdEmbeddingLength: 1536

ts
const embeddings = hopfield.embedding("text-embedding-ada-002", 3);
 
const response = await embeddings.get({ input: ["ready", "set", "hop"] });
 
const embeddingLength = response.data.length;
            
const embeddingLength: 3
 
const thirdEmbeddingLength = response.data[2].embedding.length;
               
const thirdEmbeddingLength: 1536

Embeddings ​

Usage ​

Parameters ​

Model Name ​

OpenAI ​

Embedding Count ​