TypeScript library

Learn how to use the Together TypeScript library to access all of Together's REST API from server-side TypeScript or JavaScript.

Pre-requisites

Ensure you have node installed on your machine.
Create a free account to obtain a Together API Key, found in your account settings.
Install the library with npm install together-ai.

Querying a chat model

To query a chat model, first choose one of Together's open-source chat models. We'll use mistralai/Mixtral-8x7B-Instruct-v0.1 for this example.

Next, pass your query to the messages key of together.chat.completions.create:

import Together from "together-ai";

const together = new Together({
  apiKey: process.env["TOGETHER_API_KEY"], // This is the default and can be omitted
});

const response = await together.chat.completions.create({
  messages: [{ role: "user", content: "Tell me fun things to do in New York" }],
  model: "mistralai/Mixtral-8x7B-Instruct-v0.1",
});

console.log(response.choices[0].message.content);

The create() function returns a promise, which you can await get the result.

Streaming tokens from a chat model

To stream the response back, pass stream: true into together.chat.completions.create:

import Together from "together-ai";

const together = new Together({
  apiKey: process.env["TOGETHER_API_KEY"], // This is the default and can be omitted
});

const stream = await together.chat.completions.create({
  messages: [{ role: "user", content: "Tell me fun things to do in New York" }],
  model: "mistralai/Mixtral-8x7B-Instruct-v0.1",
  stream: true,
});

for await (const chunk of stream) {
  console.log(chunk.choices[0].delta.content);
}

Querying a completion model

To query a chat model, first choose one of Together's open-source code or language models. We'll use codellama/CodeLlama-34b-Python-hf for this example.

Next, pass your query to the prompt key of together.completions.create:

import Together from "together-ai";

const together = new Together({
  apiKey: process.env["TOGETHER_API_KEY"], // This is the default and can be omitted
});

const response = await together.completions.create({
  model: "codellama/CodeLlama-34b-Python-hf",
  prompt: content,
  max_tokens: 500,
});

console.log(response.choices[0].text);

Querying an image model

To query a chat model, first choose one of Together's open-source image models. We'll use runwayml/stable-diffusion-v1-5 for this example.

Next, pass your query to the prompt key of together.images.create:

import Together from "together-ai";

const together = new Together({
  apiKey: process.env["TOGETHER_API_KEY"], // This is the default and can be omitted
});

const response = await together.images.create({
  model: "runwayml/stable-diffusion-v1-5",
  prompt: content,
  steps: 10,
  n: 4,
});

console.log(response.data[0].b64_json);

Get an embedding

To get an embedding, first choose one of Together's open-source embedding models. We'll use togethercomputer/m2-bert-80M-2k-retrieval for this example.

Next, pass your content to the input key of together.embeddings.create:

import Together from "together-ai";

const together = new Together({
  apiKey: process.env["TOGETHER_API_KEY"], // This is the default and can be omitted
});

const response = await together.embeddings.create({
  model: "togethercomputer/m2-bert-80M-2k-retrieval",
  input: content,
});

console.log(response.data[0].embedding);