ChatCompletion error when trying to use CALM with Azure OpenAi endpoint - Rasa CALM - Rasa Community Forum

ChatCompletion error when trying to use CALM with Azure OpenAi endpoint

You have selected 0 posts.

215 views 1 link

post by apo on Mar 21, 2024

Hello everyone, i tried initializing my first CALM bot today. I followed the guide. I set the correct environment variables for using the Azure endpoint but when i run rasa shell and try to interact with the bot i get the following error:

[error ] llm_command_generator.llm.error error=InvalidRequestError(message=“Must provide an ‘engine’ or ‘deployment_id’ parameter to create a <class ‘openai.api_resources.chat_completion.ChatCompletion’>”, param=‘engine’, code=None, http_status=None, request_id=None)

and this info right after it:

[info ] llm_command_generator.predict_commands.finished commands=[ErrorCommand(error_type=‘rasa_internal_error_default’, info={})]

thanks in advance

post by Amina-Derouiche on Mar 21, 2024

Hello @apo
Here is how the config should look like for the LLMCommandGenerator part while using an azure OpenAI endpoint.


- name: LLMCommandGenerator
    llm:
      engine: gpt-4-**
      deployment_id:  gpt-4-**
      request_timeout: 7
      max_tokens: 256
    embeddings:
      type: openai
      model: text-embedding-***

If this doesn’t work please share with us your config so we can have a closer look.

post by apo on Mar 22, 2024

Hi @Amina thanks for the fast answer. Using that config changed the error i get. Now i get the error that the deployment doesnt exist


- name: LLMCommandGenerator
  llm:
    engine: gpt-4-0613
    deployment_id: rasa
    request_timeout: 7
    max_tokens: 256
    embeddings:
      type: openai
      model: text-embedding-ada-002

(this is my current config)

thats my current config, my deployment is called “rasa” for engine i put gpt-4 and the modell i chose in Azure, for the model at embeddings i used one i found in forums, because there wasnt anything hinting to what that could be in my Azure Studio.

This is the error i get now

[error ] llm_command_generator.llm.error error=InvalidRequestError(message=‘The API deployment for this resource does not exist. If you created the deployment within the last 5 minutes, please wait a moment and try again.’, param=None, code=‘DeploymentNotFound’, http_status=404, request_id=None)

post by Amina-Derouiche on Mar 22, 2024

Hey @apo

Seems it is a deployment either the deployment is not correctly configured or not available.
In your .env you need the

Maybe this thread can help you as well. Let me know how it goes.