iflytek/astron-agent · error · CustomException

RPA_REQUEST_ERROR

RPA_REQUEST_ERROR

Error message

{frame.message}

What it means

The RPA node consumes a streaming HTTP/SSE response from the RPA service. Each `data:` frame is validated into `_StreamResponse`; if the frame's business `code` is non-zero, the node raises RPA_REQUEST_ERROR carrying the frame's `message` verbatim — this is the RPA service reporting a business-level failure for the request.

Solutions

1. Read frame.message (the error text) to get the RPA service's own failure reason and fix the task parameters accordingly.
2. Verify RPA service availability and credentials (endpoint URL, API key/robot configuration) in the node settings.
3. Check RPA platform quotas/licensing if the message indicates limit exhaustion.
4. Add retry with backoff for transient upstream errors (5xx-like codes) in the node or at the RPA service.

Example fix

// before: fire-and-forget stream read
for msg in resp:
    frame = _StreamResponse.model_validate_json(msg.removeprefix("data:"))

// after: bounded retries for transient codes
for attempt in range(3):
    try:
        return await run_rpa_stream(payload)
    except CustomException as e:
        if is_transient(e) and attempt < 2:
            await asyncio.sleep(2 ** attempt)
            continue
        raise

def rpa_node_config_ok(cfg: dict) -> bool:
    return bool(cfg.get("endpoint")) and bool(cfg.get("api_key")) and bool(cfg.get("robot_id"))

def is_error_frame(frame: _StreamResponse) -> bool:
    return frame.code != 0

try:
    result = await run_rpa_node(...)
except CustomException as e:
    if e.err_code == CodeEnum.RPA_REQUEST_ERROR:
        logger.error(f"RPA service rejected task: {e.err_msg}")
        if is_transient_rpa_failure(str(e.err_msg)):
            result = await retry_with_backoff(run_rpa_node, attempts=3)
        else:
            raise
    else:
        raise

When it happens

Trigger: async_execute -> execute reads an SSE frame where `frame.code != 0` after `_StreamResponse.model_validate_json(msg.removeprefix("data:"))` — i.e. the remote RPA backend returned an error frame for the submitted task.

Common situations: RPA service downtime or internal error; invalid RPA task parameters (wrong robot ID, missing workflow key); RPA license/quota exhausted; auth token for the RPA platform expired; upstream robot execution failure at runtime.

Related errors

• Failed to establish SSE connection
• 40024
• AUDIT_OUTPUT_ERROR
• Chat ID cannot be empty
• CodeConvert.sparkLinkCode(code)

AI-assisted analysis of iflytek/astron-agent@5e758547a8 (2026-09-12). Data as JSON: /api/errors/6702e1e91226dddd. Report an issue: GitHub.

Appendix: source

Thrown at core/workflow/engine/nodes/rpa/rpa_node.py:99

                    }
                )

            async with aiohttp.ClientSession(
                timeout=ClientTimeout(total=24 * 60 * 60, sock_connect=30)
            ) as session:
                async with session.post(
                    url=url, headers=headers, json=req_body
                ) as response:
                    async for line in response.content:
                        msg = line.decode("utf-8")
                        if not msg.startswith("data:"):
                            continue
                        await span.add_info_event_async(f"recv: {msg}")
                        frame = _StreamResponse.model_validate_json(
                            msg.removeprefix("data:")
                        )
                        if frame.code != 0:
                            raise CustomException(
                                err_code=CodeEnum.RPA_REQUEST_ERROR,
                                err_msg=frame.message,
                            )
                        data = frame.data if frame.data is not None else {}
            outputs.update(
                {
                    output: data.get(output)
                    for output in self.output_identifier
                    if output in data
                }
            )

            return NodeRunResult(
                status=status,
                inputs=inputs,
                outputs=outputs,
                node_id=self.node_id,
                node_type=self.node_type,

View on GitHub (pinned to 5e758547a8)