About

Simple and powerful wrapper for Qwen AI API with OAuth/PKCE authentication. Built with Bun/TypeScript, zero external dependencies.

Features:
- OAuth/PKCE device flow authentication
- Automatic token refresh
- Streaming responses
- Simple ask() method for quick prompts

Install

``bash
bun add qwen.js
`

Quick Start

`typescript
import { createQwen } from "qwen.js"

const qwen = createQwen()

await qwen.authenticate()

const answer = await qwen.ask("What is quantum computing?")
console.log(answer)
`

Usage

$3

`typescript
import { createQwen } from "qwen.js"

const qwen = createQwen({
accessToken: "your-access-token",
refreshToken: "your-refresh-token",
})

const response = await qwen.ask("Hello!")
`

$3

`typescript
import { createQwen } from "qwen.js"

const qwen = createQwen()

const { url, userCode } = await qwen.login()
console.log(Open: ${url})
console.log(Code: ${userCode})

await qwen.waitForAuth()

const answer = await qwen.ask("Hello!")
`

$3

`typescript
import { createQwen } from "qwen.js"

const qwen = createQwen({ accessToken: "..." })

for await (const chunk of qwen.chatStream("Tell me a story")) {
process.stdout.write(chunk)
}
`

$3

`typescript
import { createQwen } from "qwen.js"

const qwen = createQwen({ accessToken: "..." })

const response = await qwen.chat([
{ role: "system", content: "You are a helpful assistant" },
{ role: "user", content: "What is 2+2?" },
])

console.log(response.choices[0].message.content)
`

$3

`typescript
import { createQwen } from "qwen.js"

const qwen = createQwen()
await qwen.authenticate()

const tokens = qwen.getTokens()
await Bun.write("tokens.json", JSON.stringify(tokens))

const saved = await Bun.file("tokens.json").json()
const qwen2 = createQwen({
accessToken: saved.accessToken,
refreshToken: saved.refreshToken,
})
`

API

$3

| Option | Type | Description |
|--------|------|-------------|
| accessToken | string | Pre-existing access token |
| refreshToken | string | Pre-existing refresh token |
| model | string | Default model (default: coder-model) |

$3

| Method | Description |
|--------|-------------|
| authenticate() | Full OAuth flow with console prompts |
| login() | Get device code and auth URL |
| waitForAuth() | Poll for token after user authenticates |
| setTokens(accessToken, refreshToken?) | Set tokens manually |
| getTokens() | Get current auth state |
| ask(prompt, options?) | Simple prompt, returns string |
| chat(messages, options?) | Full chat, returns ChatResponse |
| chatStream(messages, options?) | Streaming chat, yields chunks |
| setModel(model)` | Change default model |

License

MIT