In Azure, deployments are often named other things than the model names, for instance "gpt-4o" could be named "wrapper-api-gpt-4o". This causes the mapping to not recognise the model since it is not in the mapping json.
I would suggest an additional parameter that differentiates the deployment name used for querying the api vs the model selector.
In Azure, deployments are often named other things than the model names, for instance "gpt-4o" could be named "wrapper-api-gpt-4o". This causes the mapping to not recognise the model since it is not in the mapping json.
I would suggest an additional parameter that differentiates the deployment name used for querying the api vs the model selector.