Common Issues

Common Error Codes

4xx (Client Error Status Codes): Generally, these indicate issues like incorrect request syntax, authentication failure, or authorization failure, preventing the request from being completed.
5xx (Server Error Status Codes): Generally, these indicate server-side errors, such as server downtime or request processing timeouts.

Error Code	Possible Scenario	Solution
400	Request body format error, etc.	Check the error content returned in the conversation or view the error message in the console and follow the prompts. [Common Case 1]: If it's a Gemini model, you might need to bind a payment card; [Common Case 2]: Data volume exceeded, common in visual models where image size exceeds the upstream single request traffic limit; [Common Case 3]: Unsupported parameters added or parameters filled incorrectly. Try creating a clean assistant to test if it works properly; [Common Case 4]: Context exceeds limit. Clear context, start a new conversation, or reduce the number of context entries.
401	Authentication failed: model not supported or server account banned, etc.	Contact or check the status of the corresponding service provider's account.
403	Request operation unauthorized	Perform corresponding operations based on the error message returned in the conversation or the error message prompt in the console.
404	Requested resource not found	Check the request path, etc.
422	Request format is correct, but semantically incorrect	The server can parse such errors but cannot process them. Common in JSON semantic errors (e.g., null values; expecting a string but receiving a number or boolean, etc.).
429	Request rate limit reached	Request rate (TPM or RPM) limit reached, wait a moment before trying again.
500	Internal server error, unable to complete request	If it persists, contact the upstream service provider.
501	Server does not support the requested functionality, unable to complete request
502	As a gateway or proxy, the server received an invalid response from an upstream server while attempting to fulfill the request.
503	The server is currently unable to handle the client's request due to overload or system maintenance. The length of the delay may be included in the server's Retry-After header.
504	The server, acting as a gateway or proxy, did not receive a timely response from the upstream server.

How to View Console Errors

After clicking the BLACK HOLE client window, press the shortcut key Ctrl + Shift + I (Mac: Command + Option + I).

In the popped-up console window, click Network → click to view the last completions marked with a red × at ② (when encountering errors in conversation, translation, model connectivity checks, etc.) or generations (when encountering errors in drawing) → click Response to view the complete returned content (area ④ in the figure).

If you cannot determine the cause of the error, please send a screenshot of this interface to the support team for assistance.

This inspection method can be used not only to obtain error messages during conversations but also during model testing, knowledge base addition, drawing, etc. In any case, you need to open the debug window first and then perform the request operation to get the request information.

Formulas not rendered / Formula rendering errors

If formulas are not rendered and instead display as formula code, check if the formulas have delimiters.

Delimiter usage

Inline formulas

Use a single dollar sign: $formula$

Or use $ and $, e.g.: $formula$

Block formulas

Use double dollar signs: $$formula$$

Or use \[formula\]

Example: $$\sum_{i=1}^n x_i$$
$\sum_{i=1}^n x_i$

Formula rendering errors/garbled text often occur when formulas contain Chinese content; try switching the formula engine to KateX.

Failed to create knowledge base / Failed to get embedding dimensions

Model status unavailable

Confirm whether the service provider supports the model or whether the model service status is normal for the service provider.

Used a non-embedding model

Model cannot recognize images / Unable to upload or select images

First, you need to confirm whether the model supports image recognition. BLACK HOLE categorizes popular models; those with an eye icon after the model name support image recognition.

Image recognition models support image file uploads. If the model's functionality is not correctly matched, you can find the model in the corresponding service provider's model list, click the settings button after its name, and check the image option.

Specific model information can be found and reviewed with the corresponding service provider. Similar to embedding models, models that do not support vision do not need to have image functionality forcibly enabled, and checking the image option will have no effect.