xsAI：按协议适配的轻量 AI SDK，支持 OpenAI、Anthropic 等协议

Quick start

pnpm add xsai

import { generateText, responses } from 'xsai'

const model = responses({
  apiKey: process.env.OPENAI_API_KEY,
  baseURL: 'https://api.openai.com/v1/',
  model: 'gpt-6-luna',
})

const { text } = await generateText(model, { input: 'Say hello.' })
console.log(text)

Streaming uses the same model. Each event is a typed TextEvent in a standard ReadableStream:

import { streamText } from 'xsai'

const stream = streamText(model, { input: 'Describe a quiet forest.' })

for await (const event of stream) {
  if (event.type === 'text.delta')
    process.stdout.write(event.delta)
  if (event.type === 'step.end' && event.status === 'failed')
    throw event.error
}

Why xsAI

Small parts. Each package does one job. Every package is an ES module with no side effects, so your bundler removes the code that you do not use.
A model belongs to a protocol, not to a provider. One adapter covers every service that speaks the same protocol. Change the service, keep your code.
Web standards. fetch, ReadableStream, AbortSignal, and FormData. The source imports no Node.js module. Replace fetch to add a proxy, a test double, or your own retries.
Typed from end to end. Options, results, events, and tool inputs have TypeScript types. Schemas follow Standard Schema, so you can keep the validation library that you already use.

One adapter per protocol

xsAI does not ship a package for each provider. It ships an adapter for each protocol, and many services and gateways speak at least one of them.

Protocol	Package	Factory
OpenAI Responses	`@xsai/text-responses`	`responses()`
Chat Completions	`@xsai/text-chat`	`chat()`
Anthropic Messages	`@xsai/text-messages`	`messages()`

Every factory returns a LanguageModel, so generateText, streamText, and tool calls work the same with all three. If your service speaks another protocol, you can write a custom model.

Packages

Install xsai for everything, or only the packages that you use.

Package	Provides
`@xsai/text`	`generateText`, `streamText`, `collect`, and `tool`
`@xsai/audio`	Speech generation and transcription
`@xsai/decide`	Typed answers to yes-or-no, choice, and score questions
`@xsai/embed`	Text embeddings
`@xsai/image`	Image generation
`@xsai/model`	Model lists from a service
`@xsai/shared`	HTTP options, `sendRequest`, and error types
`xsschema`	Schema conversion and validation, as a separate package
`xsai`	One package that re-exports everything above except `xsschema`

Runs anywhere

xsAI runs on any runtime with fetch and web streams: Node.js, Deno, Bun, Cloudflare Workers, and browsers.

Community projects

xsAI is used in community and in-house projects, including:

Contributing

See CONTRIBUTING.md. To edit or build the docs, see documentation maintenance.

License

MIT