The HTTP model driver for Namzu.

Install · Usage · Documentation

Implements the kernel's LLMProvider interface over a plain HTTP endpoint. Speaks two wire dialects and refuses a mismatch rather than guessing which one a gateway meant. Installed only if you use it — the kernel has no preferred vendor and no driver is a dependency of it.

Install

pnpm add @namzu/sdk @namzu/http

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

Usage

import { ProviderRegistry } from '@namzu/sdk'
import { registerHttp } from '@namzu/http'

registerHttp() // once, at startup

const { provider } = ProviderRegistry.create({
  type: 'http',
  baseURL: 'http://localhost:8000/v1',
  apiKey: process.env.INFERENCE_TOKEN ?? '',
  dialect: 'openai',
})

for await (const chunk of provider.chatStream({
  model: 'my-served-model',
  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 turn loop calls it and hands you events.

Documentation

• Namzu docs

License

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