Documentation

# FAQ & Troubleshooting

1. What types of models can I manage in the Model Hub?

Purple Fabric's Model Hub allows you to manage three main types of models:

  • LLMs (Large Language Models) for generation tasks

  • Embedding Models for Vector-Based Retrieval

  • Rerankers for ranking search results or candidates

2. What’s the difference between PF-managed and Custom Key models?

  • PF-Managed Models are hosted and maintained by Purple Fabric with pre-configured endpoints and validated compatibility.

  • Custom Key Models are models you connect using your API keys (e.g., OpenAI, Azure OpenAI, Cohere). You’re responsible for setup, billing, and configuration.

3. How do I enable a model for use in a workspace?

Navigate to the Model Hub, find the model you want, and click "Enable”. Only Tenant Admins have permission to perform this action.

4. Can I use different models for different use cases within the same tenant?

Yes. You can. Enable the different models that you want to use for the model to be available inside the Expert Agent Studio.

5. What happens if I disable a model that is already in use?

Disabling a model will:

  • The Agent Responses will fail for the agents using the Model

  • Disable it from being selected in Expert Agent Studio

6. How can I onboard a new model using a custom key?

  1. Go to Model Hub > Add Custom Key Model

  2. Enter the required endpoint details, API key, and supported tasks

  3. Save

8. How do I decide which model is right for my use case?

You can benchmark different models using the Benchmarking Module.

9. Can I onboard models from providers outside the model providers list?

No. Currently, only models from six providers are supported -

  • AWS Bedrock for LLMs and Embeddings

  • Amazon Sagemaker for Embeddings and Reranker

  • Azure AI Studio for LLMs and Reranker

  • Azure OpenAI for LLMs and Embeddings

  • Cohere for ReRanker

  • Gemini for LLMs