@harness-kernel/provider-openai

@harness-kernel/provider-openai provides OpenAIProvider, a runtime-owned model provider under the openai namespace.

import { OpenAIProvider } from "@harness-kernel/provider-openai";

const provider = new OpenAIProvider({
  apiKeyEnv: "OPENAI_API_KEY",
  models: [{ id: "gpt-5.1", label: "GPT-5.1" }],
});

Use it in runtime composition:

const store = await createHarnessSessionStore({
  agent: { definition: agent },
  providers: [new OpenAIProvider()],
  defaultModel: "openai/gpt-5.1",
});

Options include id, apiKey, apiKeyEnv, baseURL, headers, and models. The provider reads OPENAI_API_KEY by default when no explicit key option is provided.

The package is optional runtime infrastructure; agents should not require it unless they are documenting a host setup.