Completion models

A completion model turns normalized messages into assistant content. It is the model contract used by direct completions, agents, structured output, extractors, and model-driven pipeline stages.

Create a completion model

import { OpenAIClient } from '@anvia/openai'

const openai = new OpenAIClient({
  apiKey: process.env.OPENAI_API_KEY,
})

export const model = openai.completionModel('gpt-5.5')

Provider packages map Anvia requests into their SDK and return normalized responses.

Send one request

import { createCompletion } from '@anvia/core'
import { model } from './model'

const result = await createCompletion(model, {
  instructions: 'Answer clearly and concisely.',
  input: 'Summarize this incident in one sentence.',
  maxTokens: 120,
})

console.log(result.text)

Use Completions for the full request and result workflow, or pass the same model to an agent.

Capabilities

Completion models declare whether they support streaming, tools, tool choice, image input, document input, output schemas, reasoning, and provider-executed tools.

Do not infer support from the model name alone. Smoke test the exact model ID with every capability your workflow requires.