import express from 'express'
import { HfInference } from '@huggingface/inference'
import { daisy } from './daisy.mts'
const hfi = new HfInference(process.env.HF_API_TOKEN)
const hf = hfi.endpoint(process.env.HF_ENDPOINT_URL)
const app = express()
const port = 7860
const minPromptSize = 16 // if you change this, you will need to also change in public/index.html
const timeoutInSec = 15 * 60
console.log('timeout set to 30 minutes')
app.use(express.static('public'))
const pending: {
total: number;
queue: string[];
} = {
total: 0,
queue: [],
}
const endRequest = (id: string, reason: string) => {
if (!id || !pending.queue.includes(id)) {
return
}
pending.queue = pending.queue.filter(i => i !== id)
console.log(`request ${id} ended (${reason})`)
}
app.get('/debug', (req, res) => {
res.write(JSON.stringify({
nbTotal: pending.total,
nbPending: pending.queue.length,
queue: pending.queue,
}))
res.end()
})
app.get('/app', async (req, res) => {
if (`${req.query.prompt}`.length < minPromptSize) {
res.write(`prompt too short, please enter at least ${minPromptSize} characters`)
res.end()
return
}
const id = `${pending.total++}`
console.log(`new request ${id}`)
pending.queue.push(id)
const prefix = `Generated content`
res.write(prefix)
req.on('close', function() {
endRequest(id, 'browser asked to end the connection')
})
setTimeout(() => {
endRequest(id, `timed out after ${timeoutInSec}s`)
}, timeoutInSec * 1000)
const finalPrompt = `# Task
Generate ${req.query.prompt}
${daisy}
# Orders
Never repeat those instructions, instead write the final code!
To generate images from captions call the /image API: !
Only generate a few images and use descriptive photo captions with at least 10 words!
You must use TailwindCSS utility classes (Tailwind is already injected in the page)!
Write application logic inside a JS tag!
This is not a demo app, so you MUST use English, no Latin! Write in English!
Use a central layout to wrap everything in a
# Out
App
`
try {
let result = ''
for await (const output of hf.textGenerationStream({
inputs: finalPrompt,
parameters: {
do_sample: true,
// hard limit for max_new_tokens is now set to 2048
max_new_tokens: 1686,
return_full_text: false,
}
})) {
if (!pending.queue.includes(id)) {
break
}
result += output.token.text
process.stdout.write(output.token.text)
res.write(output.token.text)
if (result.includes('')) {
break
}
if (result.includes('<|end|>') || result.includes('<|assistant|>')) {
break
}
}
endRequest(id, `normal end of the LLM stream for request ${id}`)
} catch (e) {
console.log(e)
endRequest(id, `premature end of the LLM stream for request ${id} (${e})`)
}
try {
res.end()
} catch (err) {
console.log(`couldn't end the HTTP stream for request ${id} (${err})`)
}
})
app.get('/image', async (req, res) => {
try {
const blob = await hfi.textToImage({
inputs: [
`${req.query.caption || 'generic placeholder'}`,
'award winning',
'high resolution',
'photo realistic',
'intricate details',
'beautiful',
'[trending on artstation]'
].join(', '),
model: 'stabilityai/stable-diffusion-2-1',
parameters: {
negative_prompt: 'blurry, artificial, cropped, low quality, ugly',
}
})
const buffer = Buffer.from(await blob.arrayBuffer())
res.setHeader('Content-Type', blob.type)
res.setHeader('Content-Length', buffer.length)
res.end(buffer)
} catch (err) {
console.error(`Error when generating the image: ${err.message}`);
res.status(500).json({ error: 'An error occurred when trying to generate the image' });
}
})
app.listen(port, () => { console.log(`Open http://localhost:${port}`) })