@chinudotdev/provider-openai

OpenAI-compatible agent-core/v1 provider adapter — Chat Completions API, SSE streaming, tools, images, reasoning extensions.

Works with OpenAI, Azure OpenAI, and any OpenAI-compatible endpoint.

Install

npm install @chinudotdev/provider-openai

Quick start

import { createOpenAIProvider } from "@chinudotdev/provider-openai";

const provider = createOpenAIProvider({
  apiKey: process.env.OPENAI_API_KEY!,
  baseURL: "https://api.openai.com", // optional
  chatCompletionsPath: "/v1/chat/completions", // optional override
});

for await (const event of provider.stream(
  {
    spec: "agent-core/v1",
    model: "gpt-4o-mini",
    messages: [{ role: "user", content: [{ type: "text", text: "Hello" }] }],
  },
  new AbortController().signal,
)) {
  if (event.type === "text_delta") process.stdout.write(event.text);
  if (event.type === "stop") break;
}

OpenAI-compatible endpoints

baseURL is the API host or prefix. chatCompletionsPath defaults to /v1/chat/completions. Override it for providers that use a different route:

const provider = createOpenAIProvider({
  apiKey: process.env.GLM_API_KEY!,
  baseURL: "https://api.z.ai/api/coding/paas/v4",
  chatCompletionsPath: "/chat/completions",
});

Environment helper

import { createProviderFromEnv } from "@chinudotdev/provider-openai";
// Reads OPENAI_API_KEY, optional OPENAI_BASE_URL, OPENAI_CHAT_COMPLETIONS_PATH

With Agent SDK

import { Agent } from "@chinudotdev/agent";
import { createProviderFromEnv } from "@chinudotdev/provider-openai";

const agent = new Agent({
  model: "gpt-4o-mini",
  provider: createProviderFromEnv(),
});

Examples

After install, a minimal smoke script ships at node_modules/@chinudotdev/provider-openai/examples/01-chat.ts.

Docs

• Provider guide
• Examples

License

MIT