@core-ai/openai-compat

OpenAI-compatible Chat Completions provider for @core-ai/core-ai.

Installation

npm install @core-ai/core-ai @core-ai/openai-compat zod

Usage

import { generate } from '@core-ai/core-ai';
import { createOpenAICompat } from '@core-ai/openai-compat';

const provider = createOpenAICompat({
    apiKey: process.env.API_KEY,
    baseURL: 'https://gateway.example.com/v1',
});

const result = await generate({
    model: provider.chatModel('qwen3-235b'),
    messages: [{ role: 'user', content: 'Hello!' }],
});

console.log(result.content);

Known OpenAI models map maxTokens to the parameter required by Chat Completions. Unknown model IDs default to max_tokens. Set a provider-wide compatibility default when an endpoint requires max_completion_tokens:

const provider = createOpenAICompat({
    apiKey: process.env.API_KEY,
    baseURL: 'https://gateway.example.com/v1',
    maxTokensParameter: 'max_completion_tokens',
});

await generate({
    model: provider.chatModel('custom-reasoning-model'),
    messages: [{ role: 'user', content: 'Hello!' }],
    maxTokens: 1024,
});

This provider only exposes Chat Completions models. It accepts known nonstandard response fields used by OpenAI-compatible gateways, including reasoning_content and reasoning. Set reasoning: false when those fields should not be interpreted as reasoning output.

Structured output uses a forced function tool by default for broad endpoint compatibility. If the endpoint supports strict JSON Schema response formats, opt in with structuredOutputMode: 'native':

const provider = createOpenAICompat({
    apiKey: process.env.API_KEY,
    baseURL: 'https://gateway.example.com/v1',
    structuredOutputMode: 'native',
});