Python urllib: POST /v1/images, 200 or 202, after Imagen 4 Fast

Check the HTTP status, not the body: 200 means data holds image URLs, and 202 means the body is a job envelope you must poll. Google lists imagen-4.0-fast-generate-001 among the Imagen 4 ids shut down on August 17, 2026, so a Python caller moving to Sume's POST /v1/images has to handle both answers.

Why both answers exist

/v1/images waits up to 30 seconds by default. Most models finish inside that and return 200. A slow request, such as 4K, high quality or a large n, can run out of wait and degrade to 202, and mode: "async" or "webhook" always returns 202.

The call

The sample uses only urllib. Non-2xx answers raise HTTPError, which is what you want for a 402 or 400. On 202 it polls /v1/jobs/{id}/status until terminal is true, honors the next_poll_after_seconds hint with a 2-second floor, then reads /result and pulls each artifact's url.

The model id is google/nano-banana-2.1, the id in Sume's image docs. If you still want an Imagen variant, read its id from GET /v1/images/models; the sample does not assume one.

import json
import os
import time
import urllib.request

BASE = os.environ.get("SUME_BASE", "https://api.sume.com")
HEADERS = {"Authorization": "Bearer " + os.environ["SUME_API_KEY"], "Content-Type": "application/json"}

def call(method, path, body=None, extra=None):
    req = urllib.request.Request(BASE + path, method=method, headers={**HEADERS, **(extra or {})},
                                 data=json.dumps(body).encode() if body else None)
    with urllib.request.urlopen(req, timeout=40) as res:
        return res.status, json.load(res)

status, body = call("POST", "/v1/images", {"model": "google/nano-banana-2.1", "prompt": "red kettle on a white shelf"},
                    {"Idempotency-Key": "kettle-0001"})
if status == 200:
    urls = [img["url"] for img in body["data"]]
else:  # 202: the 30 s wait expired, or you asked for async; the body is the job envelope
    job_id = body["data"]["job"]["id"]
    while True:
        _, st = call("GET", f"/v1/jobs/{job_id}/status")
        if st["data"]["terminal"]:
            break
        time.sleep(max(2, st["data"].get("next_poll_after_seconds") or 2))
    _, result = call("GET", f"/v1/jobs/{job_id}/result")
    urls = [a["url"] for a in result["data"]["result"]["artifacts"]]
print(urls)

Gaps

There is no deadline on the poll, no handling of a failed job (read the job's error), and no 429 backoff. The Idempotency-Key stays the same across a retry, so a repeated submit adopts the first job. Price differs from the retired model, and usage.cost on a 200 is the amount billed.