OpenAI API-key and ChatGPT subscription drivers for Namzu.

Install · Usage · Documentation

Implements the kernel's LLMProvider interface over the official OpenAI SDK. The package has two explicit transports: OpenAIProvider speaks Chat Completions with an API key, while CodexProvider speaks the account-routed Responses backend with a ChatGPT subscription session. Installed only if you use either one — the kernel has no preferred vendor and no driver is a dependency of it.

Install

pnpm add @namzu/sdk @namzu/openai

@namzu/sdk is a peer dependency. Install both.

Usage

OpenAI API key

import { ProviderRegistry } from '@namzu/sdk'
import { registerOpenAI } from '@namzu/openai'

// Once, at startup.
registerOpenAI()

const { provider } = ProviderRegistry.create({
  type: 'openai',
  apiKey: process.env.OPENAI_API_KEY!,
  model: 'gpt-4o-mini', // default; overridable per call
})

for await (const chunk of provider.chatStream({
  model: 'gpt-4o-mini',
  messages: [{ role: 'user', content: 'Hello' }],
})) {
  if (chunk.delta.content) process.stdout.write(chunk.delta.content)
}

chatStream is the only model entry point; a non-streaming call is that stream collected. In practice the kernel's run loop calls it and hands you events.

ChatGPT subscription

import { ProviderRegistry } from '@namzu/sdk'
import { registerCodex } from '@namzu/openai'

declare const accessToken: string
declare const accountId: string

registerCodex()

const { provider } = ProviderRegistry.create({
  type: 'codex',
  accessToken,
  accountId,
  model: 'gpt-5.6-luna',
})

accessToken and accountId must come from a user-authorized ChatGPT session. The driver does not discover, refresh or persist credentials; that ownership belongs to the host. Namzu CLI can reuse a usable device session owned by the Codex CLI, without copying it into Namzu's credential store, or create a separate Namzu-owned device login when no external session is available.

Documentation

• The OpenAI driver — configuration, refusals and compatible endpoints
• Namzu docs

License

FSL-1.1-MIT, converting to MIT two years after each release.