Integration with AI SDK

Elysia provides support for response streaming with ease, allowing you to integrate with Vercel AI SDKs seamlessly.

Response streaming

Elysia supports continuous streaming of a ReadableStream and Response, allowing you to return streams directly from the AI SDKs. Choose one of the return statements shown in each example.

import { Elysia } from 'elysia'
import { streamText } from 'ai'
import { openai } from '@ai-sdk/openai'

new Elysia().get('/', () => {
    const stream = streamText({
        model: openai('gpt-5'),
        system: 'You are Yae Miko from Genshin Impact',
        prompt: 'Hi! How are you doing?'
    })

    // Just return a ReadableStream
    return stream.textStream

    // UI Message Stream is also supported
    return stream.toUIMessageStream()
})

Elysia will handle the stream automatically, allowing you to use it in various ways.

Server-sent events

Elysia also supports Server-Sent Events for streaming responses by simply wrapping a ReadableStream with the sse function.

import { Elysia, sse } from 'elysia'
import { streamText } from 'ai'
import { openai } from '@ai-sdk/openai'

new Elysia().get('/', () => {
    const stream = streamText({
        model: openai('gpt-5'),
        system: 'You are Yae Miko from Genshin Impact',
        prompt: 'Hi! How are you doing?'
    })

    // Each chunk will be sent as a Server Sent Event
    return sse(stream.textStream)

    // UI Message Stream is also supported
    return sse(stream.toUIMessageStream())
})

As a response

If you don’t need type safety from the stream for further usage with Eden, you can return the stream directly as a response.

import { Elysia } from 'elysia'
import { streamText } from 'ai'
import { openai } from '@ai-sdk/openai'

new Elysia().get('/', () => {
    const stream = streamText({
        model: openai('gpt-5'),
        system: 'You are Yae Miko from Genshin Impact',
        prompt: 'Hi! How are you doing?'
    })

    return stream.toTextStreamResponse()

    // UI Message Stream Response will use SSE
    return stream.toUIMessageStreamResponse()
})

Manual streaming

If you want to have more control over the streaming, you can use a generator function to yield the chunks manually.

import { Elysia, sse } from 'elysia'
import { streamText } from 'ai'
import { openai } from '@ai-sdk/openai'

new Elysia().get('/', async function* () {
    const stream = streamText({
        model: openai('gpt-5'),
        system: 'You are Yae Miko from Genshin Impact',
        prompt: 'Hi! How are you doing?'
    })

    for await (const data of stream.textStream)
        yield sse({
            data,
            event: 'message'
        })

    yield sse({
        event: 'done'
    })
})

Fetch

If the AI SDK doesn’t support the model you’re using, you can still use the fetch function to call the provider’s HTTP API and stream the response directly.

import { Elysia } from 'elysia'

new Elysia().get('/', () => {
    return fetch('https://api.openai.com/v1/chat/completions', {
        method: 'POST',
        headers: {
            'Content-Type': 'application/json',
            Authorization: `Bearer ${process.env.OPENAI_API_KEY}`
        },
        body: JSON.stringify({
            model: 'gpt-5',
            stream: true,
            messages: [
                {
                    role: 'system',
                    content: 'You are Yae Miko from Genshin Impact'
                },
                { role: 'user', content: 'Hi! How are you doing?' }
            ]
        })
    })
})

Elysia will proxy the response from the global fetch with streaming support automatically.

For additional information, please refer to the AI SDK documentation.