Troubleshoot AI autodocumentation issues
This page provides solutions to common issues encountered while using the AI Autodocumentation feature. The same messages appear when you click Save and test while configuring an AI engine and when generating descriptions. If your issue is not listed here, please contact our Support team.
Connection test failed
When you click Save and test, Dataedo Portal sends a test prompt to the provider. If the provider doesn't respond correctly, the message Connection test failed. The engine was not saved. appears, followed by one of the reasons described below. The engine isn't saved until the test passes.
Incorrect API key
Incorrect API key. Please contact your administrator or check the API key in the AI engine settings.
Cause
The provider rejected the API key. The key might be mistyped, revoked, or expired, or it might belong to a different provider than the one selected for the engine.
Solutions
- Generate a new API key in the provider's console and paste it into the API key field. For a saved engine, click Change next to the field first.
- For Amazon Bedrock, check whether the API key has expired. Short-term Bedrock API keys are valid only for a limited time.
Model or deployment not found
The model or deployment was not found. Please check the engine name and endpoint in the AI engine settings.
Cause
The provider doesn't recognize the model ID or deployment name, or the endpoint points to the wrong resource.
Solutions
- Check that the model ID in the Model field is spelled exactly as in the provider's documentation.
- For Azure OpenAI, enter the deployment name, not the model ID, and check that the Endpoint points to the Azure OpenAI resource that contains the deployment.
- For Amazon Bedrock, check that the model is available in the region included in the endpoint URL and that your AWS account has access to it.
- For OpenAI-compatible providers, check that the model is loaded or available on the server and that the endpoint ends with the API's base path, for example
/v1/.
Request rejected by the AI provider
The request was rejected by the AI provider. Please check the AI engine settings.
Cause
The provider accepted the connection but rejected the request, for example because the model doesn't support chat completions or the Max tokens value exceeds the model's output limit.
Solutions
- Lower the Max tokens value in the engine's Advanced settings.
- Select a chat model. Embedding, image, and audio models can't generate descriptions.
Rate limit reached
The AI provider rate limit was exceeded. Please try again later.
Cause
The request exceeded the rate limit of your provider account, for example the number of tokens or requests per minute. It commonly happens when generating descriptions for many objects at once or for tables and views with a large number of columns. Some providers, including OpenAI, return the same error when your account has run out of credits or exceeded its quota.
Solutions
- Wait a moment and click the Rerun generation buttons under each object to generate missing descriptions.
- Check the usage limits, quota, and billing details of your account in the provider's console, and upgrade your plan if needed.
- Reduce the Max tokens setting in the engine's Advanced settings to generate shorter descriptions.
- Add another engine with a model that has higher limits, and use it for bulk generation.
Cannot reach the AI provider endpoint
Cannot reach the AI provider endpoint. Please check the endpoint and network connection.
Cause
The Dataedo Portal server couldn't connect to the provider. Requests are sent by the Dataedo Portal server, so this happens when the server has no internet access, a firewall or proxy blocks the connection, the SSL connection can't be established, or the endpoint URL is wrong.
Solutions
- Ensure internet access is enabled for the Dataedo Portal server.
- If internet access cannot be enabled, add the provider's endpoint to your firewall exception list or any other network security mechanism blocking access. See Endpoints to allow in your network.
- For Azure OpenAI, Amazon Bedrock, and OpenAI-compatible providers, check the URL in the Endpoint field. For a self-hosted server,
localhostrefers to the machine that runs Dataedo Portal.
The AI request timed out
The AI request timed out. Please try again.
Cause
The provider didn't respond in time. The connection test waits up to 30 seconds. Self-hosted models can be slow to respond to the first request while the model is loading.
Solutions
- Click Save and test again.
- For self-hosted models, load the model on the server before testing the engine.
Incomplete engine configuration
The AI engine configuration is incomplete. Please check the API key and endpoint in the AI engine settings.
Cause
The API key is missing, or the endpoint isn't a valid URL.
Solutions
- Enter the API key.
- Enter the full endpoint URL, including
https://(orhttp://for a self-hosted server without SSL).
Need help?
If you run into any problems or have questions, reach out to Dataedo support.
