Table of contents
Official Content
  • This documentation is valid for:

It is possible that some errors may occur during the RAG Assistant embedding setup. Below you can find some of these errors with possible solutions.

400 This model does not support specifying dimensions

Check that the Embeddings Provider is correctly set for the associated RAG Assistant. To do so, check that the Provider Name parameter is not empty and matches the Embeddings Provider.

Query vector dimension 3072 does not match the dimension of the index 1536

Check the embeddings Model Name used and make sure to set 1536 as Dimensions value.

By default Model Name takes the value text-embedding-ada-002.

You can set text-embedding-3-small or text-embedding-3-large as Model Name but apply it to set the value 1536 in Dimensions.

If you are using a different embedding model and cannot use the default 1536, you must contact the help desk for guidance in correcting this configuration.

Unauthorized. Access token is missing using AzureOpenAI

The following error appears when trying to upload documents configured to use AzureOpenAI

statusCode: 401
message: Unauthorized. Access token is missing, invalid, audience is incorrect (https://cognitiveservices.azure.com), or have expired.

The problem is that the API Key is missing or wrong for the embeddings model, check the global configuration and the deployed assigned models; you can assign a correct value on the embeddings apiKey section.

This model's maximum context length is 8192 tokens using openAI Embeddings

The following error appears when executing a RAG assistant

"error": {
"message": "This model's maximum context length is 8192 tokens, however you requested 12000 tokens (12000 in your prompt; 0 for the completion). Please reduce your prompt; or completion length.",
"type": "invalid_request_error",
"param": null,
"code": null
}

The problem is that the embeddings model supports up to 8192 tokens, reduce the question length to fit the maximum supported by the embeddings model.

You are not allowed to sample from this model

The following error appears when executing a query against a RAG assistant

You are not allowed to sample from this model

The embeddings model is not correctly configured; make sure you are using a correct Provider and model name; check the Requests section for more detail on the error. Just for reference if using openAI provider, a valid value could be: text-embedding-ada-002 or text-embedding-3-small.

Embeddings aborted or Embeddings Request timed out

The following error appears when indexing a long document

Embeddings aborted
Embeddings Request timed out

Check the request section, alternative options to apply to the associated assistant, add the following to the Profile Metadata section

{
    "chat": {
        "embeddings": {
            "batchSize": 60
        }
    }
}

For the case the batchSize is downsized from 100 to 60 items, upload again the document.

Image exceeds max pixels allowed

The following error appears when indexing an image

BedrockException - {"message":"400 Bad Request: Image exceeds max pixels allowed. "}

Check the image resolution and try uploading it with lower resolution.

total number of images must be at most 1

The following error appears when using a multimodal embeddings model

invalid request: total number of images must be at most 1 - received 2

Lower the multimodalBatchSize parameter to 1 as detailed here.

Embeddings Request failed with status code 401

The following error appears when calling a RAG assistant from the workspace

Embeddings Request failed with status code 401

Check the Project API token section, make sure the Default token is present and active.

Last update: March 2025 | © GeneXus. All rights reserved. GeneXus Powered by Globant