Troubleshooting

Check these first

Confirm that your tool uses the AsterWise URL shown in its setup guide.
Check that your API key is active and was copied in full.
Check your balance in the app.
Use aster-code or aster-work as the model name.

Common problems

Problem	What to do
Authentication failed	Check your key. If it was revoked, create a new one and update your tool.
Insufficient balance	Add balance in the app, then try again.
Spending control reached	Review your account’s spending settings.
Too many simultaneous requests	Let existing requests finish, then send another. This is separate from a per-minute rate limit.
Request too large	Shorten the prompt or conversation history. See context and output limits.
Model or request not supported	Compare the model name, URL, and request fields with the guide for your tool.
`provider_empty_response`	No answer was returned. With reasoning enabled, try a larger output-token limit or lower reasoning effort; see the explanation below.
Connection interrupted	Check Usage to see whether the request completed before sending it again.

Common error codes

Error code	HTTP status before streaming	Meaning and next step
`invalid_api_key`	401	The API key is invalid or unavailable. Check it or create a replacement.
`insufficient_funds`	402	Add account balance before trying again.
`daily_limit_reached`	429	Review your account’s daily spending control.
`concurrency_limit`	429	Wait for existing requests to finish.
`invalid_output_limit`	400	Supply a valid positive integer for the output-token limit.
`unsupported_output_limit`	400	Use an output-token limit within the model’s supported range.
`provider_empty_response`	502	Aster finished without an answer or usable token usage. See the guidance below.

Read error.type to identify the error. The examples here cover AsterWise gateway errors; proxy errors and malformed requests can use a different format.

`provider_empty_response`

An aster-work request ended without an answer and without the usage information needed to account for it. The failed request is not charged. AsterWise returns this error instead of leaving the request waiting for usage information. With reasoning enabled, a small output-token allowance can be used up before an answer is produced. Reasoning and the visible answer share the output allowance. This is one possible cause of the error, not proof that every affected request has too small a limit. Try one of these changes before submitting the request again:

Increase the output-token limit within Aster Work’s supported limit. For Chat Completions, use max_completion_tokens (or max_tokens, but not both); for Responses, use max_output_tokens; for Messages, use max_tokens.
Choose a lower supported reasoning effort. For example, change Chat Completions reasoning_effort from high to low. See reasoning controls for each endpoint.

A larger limit allows more billable output on a successful request. There is no single token limit that guarantees an answer for every prompt, and AsterWise does not automatically increase your limit or lower your reasoning effort. Retrying unchanged may produce the same error. If it persists, contact support with the request ID. A nonstreaming request returns HTTP 502 with this JSON shape:

{
  "type": "error",
  "error": {
    "type": "provider_empty_response",
    "code": "provider_empty_response",
    "message": "Aster finished without returning an answer, and token usage was unavailable. If reasoning is enabled, the output-token limit may have been exhausted. Try increasing the output-token limit or lowering reasoning effort. This failed request is not charged."
  },
  "request_id": "123e4567-e89b-42d3-a456-426614174000"
}

Save the x-request-id response header or the matching request_id in the error body. For streaming requests, HTTP headers may already show 200. Read the stream for an error payload and closure rather than treating HTTP 200 as success. Chat Completions sends the error JSON in an SSE data: frame; Messages and Responses also name the event error. This failure ends the stream without a success terminal. AsterWise does not automatically retry this error; your SDK or tool may have its own retry settings. A valid token-limit completion with reliable usage is different: Chat Completions can return finish_reason: "length" and usage, even if the visible answer is empty. That completed request is billed for its recorded usage and does not use this error code.

When to retry

Correct key, balance, and request errors before retrying. For a temporary service error, wait briefly and retry a limited number of times. Avoid repeatedly submitting the same request automatically. Failed and cancelled requests are not charged. If you are unsure whether a request succeeded, check its status in Usage.

Still need help?

Contact support with the error message and request identifier. Never share your API key.

Get started

Connect your tools

API guides

Account and support

Check these first

Common problems

Common error codes

`provider_empty_response`

When to retry

Still need help?

​Check these first

​Common problems

​Common error codes

​provider_empty_response

​When to retry

​Still need help?

Check these first

Common problems

Common error codes

`provider_empty_response`

When to retry

Still need help?