LaVague QA , Azure Open AI configuration file sample

[suryakumaran2611]

I would like to use LaVague QA with azure open ai service , can you give me a sample configuration on how to do custom context for azure ?

[lyie28]

@suryakumaran2611, yeah, if you have deployed all three models (the one you will use as multi-modal, LLM and embedding model), your context file should look like this:

from lavague.core.token_counter import TokenCounter
from lavague.contexts.openai import AzureContext

# declare the token counter before any LLMs are initialized
token_counter = TokenCounter()

# init context
context = AzureContext(
    api_key="<YOUR_API_KEY>",
    deployment="<YOUR_DEPLOYMENT_NAME>",
    llm="MODEL_NAME",
    mm_llm="M0DEL_NAME",
    endpoint="<YOUR_ENDPOINT>",
    embedding="<EMBEDDING_MODEL_NAME>",
    embedding_deployment="<YOUR_EMBEDDING_DEPLOYMENT_NAME>"
)

Then you save your file and pass the path of your context file to the lavague-qa CLI command, i.e. lavague-test -c ./my_context.py.

If you only want to use Azure for some but not all three models let me know and I can show you an example of this too.

[lyie28]

This guide gives more info on Azure Contexts - you just need to define a Context object called context in your lavague-qa config file: https://docs.lavague.ai/en/latest/docs/integrations/azure/