AI Horoscope API

Given a zodiac sign and an optional date, the endpoint composes a prompt and routes it through our LLM gateway to a language model, which writes a short daily horoscope. The response carries the generated horoscope text, a disclaimer to surface alongside it, the identifier of the model that produced the text, and the language of the output — localised via the language parameter, defaulting sensibly when omitted. Because a language model sits in the path, this endpoint behaves differently from the arithmetic and ephemeris routes: it is slower, typically one to three seconds, and the text is freshly generated guidance rather than a reproducible calculation, so two calls with identical inputs may return differently worded readings. Weekly, monthly and yearly horoscopes are sibling endpoints sharing the same gateway and shape.

curl -X POST https://api.astroway.info/v1/horoscope/daily \
  -H "X-Api-Key: aw_live_..." \
  -H "Content-Type: application/json" \
  -d '{
  "sign": "taurus",
  "date": "2026-06-15",
  "language": "en"
}'

{
  "ok": true,
  "data": {
    "horoscope": "Today favours steady progress…",
    "disclaimer": "For entertainment purposes…",
    "model": "gemini-2.5-flash",
    "language": "en"
  }
}

Reach for this when you want ready-to-display horoscope copy without writing or maintaining your own content pipeline. It suits daily-content surfaces — push notifications, email digests, app home screens, widgets — where a freshly worded reading per sign per day is the product. Always render the returned disclaimer with the text, and request the language you need via the language parameter rather than translating client-side. Because output is non-deterministic and the call carries model latency, cache each sign-and-date reading once generated rather than calling on every page view, and budget for the one-to-three-second response when generating on demand. For longer horizons, switch to the weekly, monthly or yearly sibling endpoints.

This is generated guidance, not a deterministic calculation — identical inputs can yield differently worded text, and the disclaimer in the response is there to be shown to end users. The model field names whichever model the gateway selected, which can vary over time. Latency of roughly one to three seconds is expected because a language model is in the request path; the math endpoints respond far faster.