VertexAI RAG Engine does not work in Europe

Google Cloud Build with AI Custom ML & MLOps
1

Hi,

I am experimenting with Google’s RAGEngine.

According to the docs it should be possible to use the RagEngine in europe-west3: https://cloud.google.com/vertex-ai/generative-ai/docs/rag-engine/rag-overview#supported-regions

I am indeed able to create a corpus which also shows in the GCP UI in the Rag Engine View: https://console.cloud.google.com/vertex-ai/rag/corpus

Whenever I am trying to upload a document to the Rag Corpus though I am experiencing the following issue: (“RagCorpus ‘%s’ is not found: %s”, ‘projects/[our_project]/locations/europe-west3/ragCorpora/[rag_id]’, ‘https://europe-west3-aiplatform.googleapis.com/upload/v1beta1/projects/[our_project]/locations/europe-west3/ragCorpora/[rag_id]/ragFiles:upload’)

When I use us-central1 instead everything works like a charme.

In addition I am not able to specify the embedding model that I wanna use for my RagCorpus. I am following the example from the official documentation: https://cloud.google.com/vertex-ai/generative-ai/docs/rag-engine/rag-quickstart

In fact the embedding model always defaults to text-embedding-005, so setting:

embedding_model_config = rag.RagEmbeddingModelConfig(
    vertex_prediction_endpoint=rag.VertexPredictionEndpoint(
        publisher_model="publishers/google/models/[some-other-embedding-model]"
    )
)

is not changing the RAGs embedding model as of now whats quite contraintuitive and contradicts the python quickstart guide.

I am not the only one experiencing these issues ( https://www.googlecloudcommunity.com/gc/AI-ML/API-Vertex-AI-RAG-Engine-Upload-file-in-corpus-failed/m-p/885165 ) but chose to create a new thread as it might get more attention and since in the previous thread the issue with setting specific embeddings has not been described yet.

Thanks for helping me out

Lars

2

Hi LJF00,

Welcome to the Google Cloud Community!

The error “RagCorpus ‘%s’ is not found” in europe-west3 when uploading documents, suggests a regional limitation or an incomplete deployment of the RAG Engine’s document upload service in europe-west3. This means that while the corpus metadata might be created, the underlying service responsible for handling uploads in that region appears unable to correctly locate or process requests for it. Here are some suggestions you can try to resolve the issue:

  • Check Google Cloud’s Service Health Page you can monitor ongoing incidents and outages to see if europe-west3 is currently affected.
  • Since us-central1 works fine for you, you might want to continue using it until europe-west3 is confirmed to be fully operational.
  • If the issue persists despite no reported outages, consider reaching out to Google Cloud Support to clarify whether europe-west3 is fully functional for RAG Engine.

Regarding the embedding model issue, it seems like the model selection isn’t working as expected. If the embedding model always defaults to text-embedding-005, despite specifying another model, this could be a bug or a limitation in the current implementation. With this, I suggest filing a bug report so our engineering team can check whether this is a potential bug or project-specific issue only. Before filing, please check this documentation on what to expect after you’ve opened an issue.

Additionally, Keep an eye on the Vertex AI release notes for updates on RAG engine and its regional feature availability.

Was this helpful? If so, please accept this answer as “Solution”. If you need additional assistance, reply here within 2 business days and I’ll be happy to help.

3

Hi nikacalupas,

thanks for taking your time to address my issues and sorry for not replying within 2 business days.

Unfortunately the issues are not resolved yet.

Regarding european servers not being supported yet despite documentation:
By now europe-west4 is another supposedly supported region for the VertexAI RAGEngine (see first link in my original post). However for both european regions creating the corpus works, but adding files does not.

According to the Google Cloud’s Service Health Page there are no ongoing incidents/outages

Continuing with us-central1 which works fine for me, is unfortunately not an option. I can only use it for dummy data due to GDPR restrictions which only allow me to use european servers to keep out data in the EU. Therefore, having functional european servers is really important to us.

We have contacted the Google Cloud Support (we are also paying for a certain support level) but unfortunately we have not received any updates on this issue in 2 months by now, so I just try to raise the issue via all available channels to increase the chances of getting updates regarding the issue.

Thanks in advance for helping me out

Lars