How microsoft/aitour-build-your-first-agent-with-azure-ai-agent-service works
Main Components
• Workshop/Presentation Materials: Represented by the separate directories (“lab”, “session-delivery-resources”, and “img”) that provide lab instructions, presenter resources, and visual assets.
• Core Agent Code (src): Although the code is likely small and sample-based (given the workshop context), it embodies the integration representing the “brain” of the agent.
• Azure AI Agent Service Component: This is the primary service facilitating natural language processing, function calling, dynamic query generation, and agent orchestration.
• Data Storage: Although not explicit in the file tree, the README mentions a read-only SQLite database used for performing dynamic SQL queries.
• Visualization & File Generation Module: The agent uses a Code Interpreter to generate visualizations and downloadable files (Excel, CSV, JSON, image formats).
• External Services Integration: Integrations like Bing are used for performing grounded searches external to the internal data sources.
• Security and Context Management: Emphasize read-only database access and the built-in security best practices documented in the README.
Relationships and Interactions
• User Input: Begin with a user interface (could be abstracted as a client or command line interface in the workshop) that initiates requests.
• Agent Service Orchestration: The user input is passed to the Azure AI Agent Service which manages conversation context.
• SQL Query & Database: The agent dynamically generates SQL queries from user input, executes these against the read-only SQLite database, and retrieves data.
• Data Processing and Visualization: Retrieved data may be processed to generate visualizations (e.g., charts, tables) through the Code Interpreter module.
• File Generation: The system then provides options for outputting data into various file formats.
• External Data Enrichment: When needed (e.g., for competitive analysis), the agent calls the Bing search API to complement its internal data.
• Feedback Loop: The interactions are managed via a context management system ensuring that conversation remains coherent over time.
Architectural Patterns and Design Principles
• Modularity: Separate components for processing (query generation, visualization, file generation) indicate a modular system where each service or function is isolated.
• Layered Architecture: There is a clear separation between input management, business logic (agent orchestration, dynamic query generator), data access (SQLite), and external API integrations (Bing).
• Security Best Practices: Emphasize read-only access to data and safe interaction with external and internal data sources.
• Extensibility: The workshop focuses on adaptability so the agent can be repurposed for customer support, data analysis, or competitive analysis by plugging in additional tools.
Relevant Technologies and Frameworks
• Azure AI Agent Service – Core to managing the AI agent’s functionality.
• SQLite – As the read-only database used for SQL query execution.
• Code Interpreter Module – For generating visualizations and different file formats.
• External APIs like Bing – For conducting external, grounded searches.
• Python (implied from the linked Python library) – Likely used in the agent’s implementation.