Not able to manage resources

Hi Shikhar Agarwal

The "Data + Indexes" option in Azure AI Foundry has recently been deprecated or restructured, which has impacted how users implement Retrieval-Augmented Generation (RAG) workflows with their own data. This change has not been clearly documented in a changelog, leading to confusion among users. Azure AI Foundry now supports RAG through two main approaches: auto-provisioned indexes and SDK-based custom indexing. When users upload data directly into the Foundry playground, the platform automatically creates a free Azure AI Search index, enabling immediate experimentation with RAG.

Azure AI Foundry continues to support Retrieval Augmented Generation (RAG) functionality through multiple implementation approaches. While the interface has evolved, the core functionality remains intact, allowing organizations to implement RAG solutions using either built-in integrations or custom implementations.

The platform offers three primary methods for implementing RAG: Azure AI Foundry with vector index and retrieval augmentation, Azure OpenAI with search index capabilities, and Azure Machine Learning utilizing search index as a vector store. Each approach leverages Azure AI Search as the foundation for information retrieval, ensuring consistent performance and scalability across implementations.

For optimal implementation, users can configure several key parameters: the number of retrieved documents (ranging from 3 to 20), document filtering strictness (scale of 1-5), and response limitation flags. These parameters directly impact the model's ability to generate accurate responses while maintaining data privacy and relevance.

The RAG architecture operates through a coordinated workflow involving four primary components: user interface, application server/orchestrator, Azure AI Search, and Azure OpenAI. This architecture enables secure and efficient information retrieval while maintaining precise control over grounding data used by the Large Language Model (LLM).

Organizations can maximize relevance and recall through hybrid queries combining keyword and vector search capabilities. This approach provides enhanced precision while maintaining flexibility in query formulation and result processing. For detailed implementation guidance, Microsoft provides comprehensive documentation and code samples through various repositories.

The platform supports both textual and vector-based indexing strategies, enabling organizations to optimize their implementation based on specific requirements. This flexibility ensures that the solution can accommodate various data formats and querying needs while maintaining optimal performance characteristics.

Reference links: Retrieval Augmented Generation (RAG) in Azure AI Search Overview , Azure AI Foundry Runtime Parameters , Custom RAG Pattern Implementation Details

Hope it helps!

Thank you