Support Azure OpenAi deployments

[edubrigham]

Due to GDPR constraints, it would be great to use Azure Openai deployments.
In Europe we're not allowed to use OpenAI gpt models for public consumption
because they are hosted on American servers.

[n4ze3m]

Hey, Dialoqbase support. For the Azure API, you need to add the following variables to .env:

• AZURE_OPENAI_API_KEY
• AZURE_OPENAI_API_VERSION
• AZURE_OPENAI_API_INSTANCE_NAME
• AZURE_OPENAI_API_DEPLOYMENT_NAME

Let me know if you have any issues

[n4ze3m]

Hey, you need to add the embedding model on the same page as the chat model. Choose the embedding tag and add the model.

[edubrigham]

I'm very sorry to bother you again. unfortunately I still didn't manage to get the chatbot to connect to my azure model
You said that I need to add the following variables to .env:

AZURE_OPENAI_API_KEY
AZURE_OPENAI_API_VERSION
AZURE_OPENAI_API_INSTANCE_NAME
AZURE_OPENAI_API_DEPLOYMENT_NAME

as per screenshots, I added the following entries to the .env file in the Docker folder:
AZURE_OPENAI_API_DEPLOYMENT_NAME="gpt-4"
AZURE_OPENAI_API_INSTANCE_NAME="azureopenai-francecentral-instance"
AZURE_OPENAI_API_KEY="my_azure_openai_key"
AZURE_OPENAI_API_VERSION="2024-05-01-preview"

and deployed a gpt-4 chat model and ada-002 embeddings model, but I still get the error message:
code: 'OperationNotSupported',
2024-09-04 10:19:22 dialoqbase | message: 'The embeddings operation does not work with the specified model, gpt-4. Please choose different model and try again. You can learn more about which models can be used with each operation here: https://go.microsoft.com/fwlink/?linkid=2197993.'

What am I doing wrong?
Thanks a million for your help!

[n4ze3m]

Hey, it's okay! 😊 I can assist you. I think you need to add the AZURE_OPENAI_API_EMBEDDINGS_DEPLOYMENT_NAME variable for the embedding model, just like in chat. Let me know if it works!

[edubrigham]

Thank you for all the help, I used the following .env variables and it worked:

AZURE_OPENAI_API_DEPLOYMENT_NAME="gpt-4"
AZURE_OPENAI_API_INSTANCE_NAME="azureopenai-francecentral-instance"
AZURE_OPENAI_API_KEY="my_azure_key"
AZURE_OPENAI_API_VERSION="2024-05-01-preview"
AZURE_OPENAI_API_EMBEDDINGS_DEPLOYMENT_NAME="text-embedding-ada-002"
AZURE_OPENAI_API_EMBEDDINGS_INSTANCE_NAME="azureopenai-francecentral-instance"
AZURE_OPENAI_API_EMBEDDINGS_KEY="my_azure_key"

but this setup leads me to a query: if I need to hardcode the deployment name for a specific model in the .env file, then does this mean I can only use this one specific model for all new RAG bots I create? I can't have multiple azure deployments/models in the .env file can I? if I want to have one bot using gpt-4o and another using gpt-4o-mini, it won't work, right? Also, is it possible to declare a different the AZURE_OPENAI_API_KEY inside the Chat Model API Key input field?

[n4ze3m]

Glad it works! I will make it easier to add models from the UI in an upcoming update. :)

[edwhiskey]

Due to GDPR constraints, it would be great to use Azure Openai deployments. In Europe we're not allowed to use OpenAI gpt models for public consumption because they are hosted on American servers.

Hello edubrigham,

May I know how it work for you? I couldn't use the model after I added the Azure openAi key.
I tried to add the model in Admin portal but still not working.