OpenAI

Connect to OpenAI with an API key, choose an OpenAI model ID, and use it with the standard nimgent APIs.

Make a request

import std/os
import nimgent
import nimgent/providers/openai

let model = openAI(getEnv("OPENAI_API_KEY")).model("gpt-4.1-mini")
echo generateText(model, prompt = "Say hello in one sentence.").text

Set OPENAI_API_KEY before you run the program.

openAI(...) uses the OpenAI Responses API by default. If you pass a Chat Completions endpoint, nimgent uses that format instead:

let provider = openAI(
  getEnv("OPENAI_API_KEY"),
  endpoint = "https://api.openai.com/v1/chat/completions"
)

OpenAI-specific options

Use OpenAIOptions for parallelToolCalls, store, user, and embedding dimensions.

import std/options

let response = generateText(
  model,
  prompt = "Explain this code.",
  providerOptions = ProviderOptions(
    openai: OpenAIOptions(store: some(false)))
)

Use dimensions with embed or embedMany. Changing dimensions means you need to re-embed every vector in a store.

Next steps

See Settings for portable controls, or Files and images to attach documents and images.