ATH-MaaS/Pixelle-Video · error · RuntimeError

可灵 API 错误: code={data.get('code')}, message={data.get('messa

Error message

可灵 API 错误: code={data.get('code')}, message={data.get('message')}

What it means

Kling's submit endpoint returned HTTP 200 but the JSON body carries a non-zero business code, meaning task submission was rejected at the API level. The client raises RuntimeError with the API's code and message fields (data.code != 0).

Source

Thrown at pixelle_video/services/api_services/video_kling.py:255

        resp = self._session.post(
            url,
            json=body,
            headers=headers,
            timeout=300,
            proxies=_proxy_dict(self.local_proxy),
        )
        if not resp.ok:
            try:
                err_body = resp.json()
            except Exception:
                err_body = resp.text
            logger.error(f"KlingVideoClient: HTTP {resp.status_code}, 响应: {err_body}")
            resp.raise_for_status()
        data = resp.json()

        if data.get("code") != 0:
            raise RuntimeError(
                f"可灵 API 错误: code={data.get('code')}, message={data.get('message')}"
            )

        task_id = data["data"]["task_id"]
        logger.info(f"KlingVideoClient: 任务已提交 task_id={task_id}")
        return task_id

    # ─── 查询任务 ───

    def _query_task(self, task_id: str, endpoint: str = "image2video") -> dict:
        """
        查询单个任务状态

        Returns:
            API 响应中的 data 字段
        """
        url = f"{self.base_url}/v1/videos/{endpoint}/{task_id}"
        headers = self._auth_headers()

View on GitHub (pinned to 848b054e4f)

Solutions

  1. Read the message in the exception; it states the exact API-side rejection reason (auth, quota, or invalid parameter).
  2. Check/regenerate Kling API credentials and ensure the JWT (from _auth_headers) is fresh and not expired.
  3. Align parameters with the model: v3 supports duration 3-15, v2.x only 5/10; kling-v2-5-turbo ignores sound; v2-6 sound=on requires pro mode (the client auto-switches this).
  4. Trim prompt/negative_prompt to <=2500 characters and verify model_name is an exact Kling model id.
  5. If the code indicates throttling/quota, wait or top up the account, then retry.

Example fix

# before
body = {"model_name": "kling-v2-5-turbo", "duration": "15", ...}  # API 错误: invalid duration
# after
body = {"model_name": "kling-v3", "duration": "10", ...}  # v3 支持 3~15s
Defensive patterns

Strategy: validation

Validate before calling

assert api_key and api_secret, "缺少可灵 API 凭证"
assert model_name in ("kling-v3", "kling-v2-6", "kling-v2-5-turbo"), f"未知模型 {model_name}"
assert len(prompt) <= 2500, "prompt 超过 2500 字符限制"
assert duration in ("5", "10") or model_name.startswith("kling-v3"), "该模型仅支持 5/10 秒"

Try / catch

try:
    task_id = client.generate_video(...)
except RuntimeError as e:
    if str(e).startswith("可灵 API 错误"):
        code = str(e).split("code=")[1].split(",")[0]
        if code in ("auth_failed", "401"):
            refresh_credentials(); retry()
        else:
            logger.error(f"提交被拒绝: {e}")  # 修正参数后重试

Prevention

When it happens

Trigger: POST /v1/videos/{image2video|text2video} returns {"code": <non-zero>, "message": ...} in video_kling.py:254-257 — typically invalid api key/jwt, exceeded quota, invalid model_name/mode/duration combination, prompt too long (>2500 chars), or account suspension.

Common situations: Expired Kling access key/secret producing auth codes, using kling-v2-5-turbo with duration="15" (only v3 supports 3-15s), sending sound=on with a model that doesn't support it, or free-tier rate/quota limits hit.

Related errors

AI-assisted analysis of ATH-MaaS/Pixelle-Video@848b054e4f (2026-08-30). Data as JSON: /api/errors/64615cfce9428d78. Report an issue: GitHub.