Container-ready via docker/ compose (frontend nginx + backend Node). Compose adjusted for Coolify on the prod server: frontend uses expose:80 (no host binding — host 8080 is taken by the Coolify proxy; Traefik routes visigine.de), backend ALLOWED_ORIGINS=https://visigine.de. Secrets stay in server/.env (git-ignored); see server/.env.example. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
57 lines
2.0 KiB
JavaScript
57 lines
2.0 KiB
JavaScript
// OpenAI chat completions provider. Uses the cheapest GPT-4-class model.
|
|
const ENDPOINT = 'https://api.openai.com/v1/chat/completions'
|
|
const MODEL = 'gpt-4o-mini'
|
|
const TIMEOUT_MS = 30_000
|
|
const SYSTEM_PROMPT =
|
|
'You are a helpful AI search assistant. Answer the user\'s question concisely, in German if the question is in German. Recommend specific companies, services, or products when relevant.'
|
|
|
|
// Approximate per-1M token rates in USD for gpt-4o-mini.
|
|
const COST = { inputPer1M: 0.15, outputPer1M: 0.60 }
|
|
|
|
async function once(prompt, signal) {
|
|
const res = await fetch(ENDPOINT, {
|
|
method: 'POST',
|
|
signal,
|
|
headers: {
|
|
'Content-Type': 'application/json',
|
|
Authorization: `Bearer ${process.env.OPENAI_KEY}`,
|
|
},
|
|
body: JSON.stringify({
|
|
model: MODEL,
|
|
temperature: 0.3,
|
|
max_tokens: 600,
|
|
messages: [
|
|
{ role: 'system', content: SYSTEM_PROMPT },
|
|
{ role: 'user', content: prompt },
|
|
],
|
|
}),
|
|
})
|
|
return res
|
|
}
|
|
|
|
export async function query(prompt) {
|
|
const t0 = Date.now()
|
|
const controller = new AbortController()
|
|
const timer = setTimeout(() => controller.abort(), TIMEOUT_MS)
|
|
try {
|
|
let res = await once(prompt, controller.signal)
|
|
if (res.status >= 500) {
|
|
res = await once(prompt, controller.signal)
|
|
}
|
|
if (!res.ok) {
|
|
return { provider: 'openai', content: '', ms: Date.now() - t0, costUsd: 0, error: `OPENAI_${res.status}` }
|
|
}
|
|
const data = await res.json()
|
|
const content = data?.choices?.[0]?.message?.content || ''
|
|
const inTok = data?.usage?.prompt_tokens || 0
|
|
const outTok = data?.usage?.completion_tokens || 0
|
|
const costUsd = (inTok * COST.inputPer1M + outTok * COST.outputPer1M) / 1_000_000
|
|
return { provider: 'openai', content, ms: Date.now() - t0, costUsd, error: null }
|
|
} catch (e) {
|
|
const code = e?.name === 'AbortError' ? 'TIMEOUT' : 'NETWORK'
|
|
return { provider: 'openai', content: '', ms: Date.now() - t0, costUsd: 0, error: code }
|
|
} finally {
|
|
clearTimeout(timer)
|
|
}
|
|
}
|