This section outlines the components and interactions that powers the unified data analysis platform. The architecture ingests customer information, product details and transaction history and surfaces insights via an interactive web experience.
SQL scripts for the customer, product and transaction details are the primary input into the system. These tables are uploaded and stored for downstream insight generation.
Stores uploaded customer information, product details and transaction history tables. Serves as the primary knowledge source to surface insights in the Fabric Data Agent.
Provides large language model (LLM) capabilities to support naltural language querying.
Fabric Data Agent is connected to an agent in Microsoft Copilot Studio and surfaced as a channel in Microsoft Teams.
Users can explore call insights, visualize trends, ask questions in natural language, directly inside Microsoft Teams.
SQL scripts for the customer, product and transaction details are the primary input into the system. These tables are uploaded and stored for downstream insight generation.
Stores uploaded customer information, product details and transaction history tables. Serves as the primary knowledge source to surface insights in the web application. And persists chat history and session context for the web interface. Enables retrieval of past interactions.
Provides large language model (LLM) capabilities to support natural language querying.
Handles orchestration and intelligent function/tool calling for contextualized responses and multi-step reasoning over retrieved data.
Hosts the web application and API layer that interfaces with the AI services and storage layers. Manages user sessions and handles REST calls.
Stores containerized deployments for use in the hosting environment.
An interactive UI where users can explore call insights, visualize trends, ask questions in natural language, and generate charts. Connects directly to SQL Database in Fabric and App Services for real-time interaction.