Error Handling

​

1. Try to Get HTTP Error Codes to Initially Locate the Problem

a. In the code, try to print out the error codes and error messages (message). These can help you locate most of the issues.

HTTP/1.1 400 Bad Request
Date: Thu, 19 Dec 2024 08:39:19 GMT
Content-Type: application/json; charset=utf-8
Content-Length: 87
Connection: keep-alive

{"code":20012,"message":"Model does not exist. Please check it carefully.","data":null}

• Common error codes and reasons:
  • 400: Incorrect parameters. Please refer to the error message (message) to correct the invalid request parameters.
  • 401: Incorrect API Key setting.
  • 403: Insufficient permissions. The most common reason is that the model requires real-name authentication. Refer to the error message (message) for other cases.
  • 429: Triggered rate limit. Refer to the error message (message) to determine which specific type of rate limit (RPM /RPD / TPM / TPD / IPM / IPD), and refer to Rate Limits for specific rate limit policies.
  • 504 / 503: Generally due to high service system load, try again later. For chat and text-to-speech requests, try using stream mode (“stream” : true). Refer to Stream Mode.
  • 500: An unknown service error. Contact relevant personnel for investigation.

b. If the client does not output the corresponding information, you can consider running the curl command in the command line (for LLM models)

  curl --request POST \
      --url https://api.siliconflow.cn/v1/chat/completions \
      --header 'accept: application/json' \
      --header 'authorization: Bearer 改成你的apikey' \
      --header 'content-type: application/json' \
      --data '
  {
      "model": "记得改模型",
      "messages": [
        {
          "role": "user",
          "content": "你好"
        }
      ],
      "max_tokens": 128
    }' -i

​

2. Try Using a Different Model to See if the Issue Persists

​

3. If a Proxy is Enabled, Consider Disabling the Proxy and Trying Again

​

4. Try Switching the Endpoint from api.siliconflow.cn to api.siliconflow.com, which has global load balancing capabilities and routes requests to the appropriate nodes based on the source