Technical Pattern: Build Your Own AI Assistant

Source Knowledge Repository: Here, information is curated and sourced. It acts as the initial gathering point for all the data that will empower the AI assistant, ensuring a rich and diverse knowledge base.

 

Data Layer: The documents for topics are stored in scalable and secure cloud storage, while OneLake provides an integrated data lake that supports the storage and analysis of data. Data may be stored in a variety of data services based on the data type, scale, and expected access patterns. These data services include Azure Database for PostgreSQL, Azure CosmosDB, Azure SQL Database, and Storage Accounts.  A guide to selecting between these services are covered in this blog.

 

Microsoft Fabric Synapse Data Engineering Pipeline: A robust pipeline to process, manage, and orchestrate data flow. It is the foundation that supports the transformation of raw data into actionable insights.

Notebook: Data scientists and engineers use these to process, enrich, vectorize, and index data, making it machine-readable and thus understandable by AI models.

 

Azure AI Search: This component indexes knowledge articles, creating a searchable database that the AI can query to find relevant information quickly and efficiently.

 

AML Prompt Flow and Azure AI Services: This is where orchestration and evaluation happen. It involves the coordination of various AI models and services to ensure that the AI assistant can understand and respond appropriately to user queries.

 

Container Registry and App Service: These components are responsible for the deployment and management of the AI services, ensuring they are always available and responsive to user needs.

 

Web Front-end: A user-friendly interface that allows end-users to explore documents from different sources and interact with the AI to generate content.