decolua/9router · error

Runway polling timeout

Error message

Runway polling timeout

What it means

The Runway generation task did not reach SUCCEEDED (or FAILED/CANCELLED) within POLL_TIMEOUT_MS, so parseResponse aborts polling and throws. Runway image/video generation is asynchronous: the code sleeps POLL_INTERVAL_MS between status checks and gives up once the deadline passes, even though the task may still be running upstream.

Source

Thrown at open-sse/handlers/imageProviders/runwayml.js:42

    if (isVideo) {
      return { promptText: body.prompt, model, ratio, duration: 5, ...(body.image ? { promptImage: body.image } : {}) };
    }
    return { promptText: body.prompt, model, ratio, ...(body.image ? { referenceImages: [{ uri: body.image }] } : {}) };
  },
  async parseResponse(response, { headers }) {
    const { id } = await response.json();
    if (!id) throw new Error("Runway: no task id returned");
    const taskUrl = `${BASE_URL}/tasks/${id}`;
    const deadline = Date.now() + POLL_TIMEOUT_MS;
    while (Date.now() < deadline) {
      await sleep(POLL_INTERVAL_MS);
      const r = await fetch(taskUrl, { headers });
      if (!r.ok) throw new Error(`Runway status ${r.status}`);
      const s = await r.json();
      if (s.status === "SUCCEEDED") return s;
      if (s.status === "FAILED" || s.status === "CANCELLED") throw new Error(s.failure || "Runway task failed");
    }
    throw new Error("Runway polling timeout");
  },
  normalize: (responseBody) => {
    const outputs = Array.isArray(responseBody.output) ? responseBody.output : [];
    return { created: nowSec(), data: outputs.map((url) => ({ url })) };
  },
};

View on GitHub (pinned to 90b52e06ff)

Solutions

1. Increase POLL_TIMEOUT_MS in open-sse/handlers/imageProviders/_base.js to cover the slowest model you use (video generation can take several minutes)
2. Fetch `${BASE_URL}/tasks/{id}` manually with the Bearer token to check the task's current status — if still running, the timeout was simply too short
3. Reduce POLL_INTERVAL_MS slightly to catch terminal transitions sooner (avoid hammering the API; keep a few seconds minimum)
4. Resubmit the generation; if tasks routinely time out on one model, switch to a faster image model (gen4_image)

Example fix

// before (open-sse/handlers/imageProviders/_base.js)
export const POLL_TIMEOUT_MS = 60_000;
// after — allow up to 5 minutes for slow video tasks
export const POLL_TIMEOUT_MS = 300_000;

try {
  const result = await runwayImageCall(params);
} catch (err) {
  if (err.message === "Runway polling timeout") {
    // task may still be running upstream — poll /tasks/{id} manually with backoff
    // before resubmitting, to avoid duplicate generations
  } else {
    throw err;
  }
}

When it happens

Trigger: parseResponse's `while (Date.now() < deadline)` loop expires — the task stays in RUNNING/QUEUED/PENDING for longer than POLL_TIMEOUT_MS while polling every POLL_INTERVAL_MS.

Common situations: Long video generations (image_to_video, 5s clips) or queues under heavy load exceeding the polling window; task stuck forever due to a Runway-side stall; POLL_TIMEOUT_MS configured too short for the chosen model; clock skew is irrelevant here but very slow networks add per-poll latency, effectively shrinking the window.

Understand the failure class

• Timeouts: ETIMEDOUT, deadlines, and hung requests — what actually expires when a request times out.

Related errors

• NanoBanana polling timeout
• Runway task failed
• BFL: no polling_url returned
• NanoBanana status ${r.status}
• NanoBanana generation failed

AI-assisted analysis of decolua/9router@90b52e06ff (2026-08-30). Data as JSON: /api/errors/ff84ef119d728fa6. Report an issue: GitHub.