Command Model Provider Guides
A model deployment refers to an active instance of one of the Cohere command models. The Toolkit currently supports the following model deployment options:
Command Model Provider Options
- Cohere Platform (
model_deployments/cohere_platform.py)
- This deployment option utilizes the Cohere Python SDK to interface with the Cohere Platform. You will need a Cohere API key, which is automatically generated when you create an account. You can find your API key here.
- Azure (
model_deployments/azure.py)
- This option connects to your Azure deployment. To set up an Azure deployment, follow the instructions here. After deploying a model, retrieve the endpoint URL and API key from Azure AI Studio: navigate to https://ai.azure.com/build/ -> Project -> Deployments -> Click your deployment to access the URL and API Key. Remember to append
/v1 to the endpoint URL when using the Cohere SDK.
- SageMaker (
model_deployments/sagemaker.py)
- This deployment option connects to your SageMaker endpoint. For setup instructions, refer to the SageMaker setup guide or use the command notebook available here. Note your AWS region and endpoint name when executing the notebook, as these will be required as environment variables.
- Local Models with LlamaCPP (
community/model_deployments/local_model.py)
- This option enables the use of a local model. To implement this deployment, download a model of your choice. You can use Cohere command models or explore other available models here. Enable community features by setting
USE_COMMUNITY_FEATURES=True in your .env file.
Adding Your Own Deployment
- Create a deployment file and place it in the /community/model_deployments folder. Implement the function calls from
BaseDeployment, mirroring the existing deployment structures.
- Register your deployment in src/community/config/deployments.py.
- Add any required environment variables to the
.env template.
Adding a Cohere Private Deployment
To add a Cohere private deployment, replicate the steps above for the Cohere Platform implementation, adjusting the base_url for your private deployment and incorporating any custom authentication steps.