Async generation and polling - Havincy

Generations are asynchronous. A generation goes through these statuses:

Status	Meaning
`queued`	Accepted, waiting for a worker
`processing`	Being generated by the model
`completed`	Done: `url`, `download_url` and `preview_url` are available
`failed`	The model could not produce the media. Credits are refunded. See `failure_reason`.

Option 1: wait in the request

Add wait_seconds (0–55) to POST /generations, GET /generations/{id}, POST /encore/creations or GET /encore/creations/{id}. The API returns as soon as the job is finished, or after the delay with the current status. Ideal for images.

Option 2: poll

import os, time, requests

HEADERS = {"Authorization": f"Bearer {os.environ['HAVINCY_API_KEY']}"}
BASE = "https://havincy.com/api/v1"

job = requests.post(f"{BASE}/generations", headers=HEADERS, json={
    "modality": "video",
    "model_id": "fal-kling-v3-pro",
    "prompt": "A paper boat sailing on a rainy street",
    "duration": 5,
}).json()
asset_id = job["assets"][0]["id"]

while True:
    gen = requests.get(f"{BASE}/generations/{asset_id}", headers=HEADERS,
                       params={"wait_seconds": 30}).json()
    if gen["status"] in ("completed", "failed"):
        break
    time.sleep(5)

print(gen["assets"][0].get("url"))

Combining a short polling loop with wait_seconds keeps the number of requests low and your rate limit free.

Files

url: stable link to the file (images, videos, audio, 3D).
download_url: same file with a download disposition.
preview_url: poster image for videos when available.

⌘I