AI-driven Customer Sales Analysis

This proof-of-concept demonstrates how AI models can analyze customer sales data and provide insights for business decision-making. The system uses function calling to access real customer and sales order data stored in a PostgreSQL database.

Project Structure 🏛️

├── src/
│   ├── Athena.Api/              # ASP.NET Core API with SemanticKernel
│   │   ├── Database/            # Entity Framework Core entities and DbContext
│   │   ├── Extensions/          # Database seeding and configuration extensions  
│   │   ├── Plugins/             # SemanticKernel function plugins
│   │   └── Migrations/          # EF Core database migrations
│   └── Athena.AppHost/          # .NET Aspire orchestration host
│       └── ollama-startup.sh    # Ollama container initialization script
├── .vscode/                     # VS Code launch configurations
└── README.md                    # You are here 😉

Building Blocks 🏢

New Architecture Features:

• Universal Container Support: Works seamlessly with both Docker and Podman
• API Facade Pattern: Athena API acts as a gateway/proxy for Ollama, providing OpenAI-compatible endpoints
• Integrated Documentation: Built-in Scalar API documentation for easy API exploration
• Robust Initialization: Enhanced container startup scripts with health checking and model caching

graph TD
    A[Open-WebUI<br/>Chat Interface] --> |OpenAI Compatible API<br/>HTTP/REST| B[Athena API<br/>ASP.NET Core + SemanticKernel]
    
    B --> |Entity Framework Core<br/>Npgsql<br/>Customer & Sales Entities<br/>EF Core Schema| C[(PostgreSQL Database<br/>Customers & Sales Orders)]
    
    B --> |OllamaApiFacade<br/>Proxy/Gateway<br/>OpenAI Compatible API| D[Ollama Container]
    
    D --> E[qwen3:1.7b Model<br/>• Small footprint<br/>• Function calling support<br/>• Good performance/size ratio]
    
    E --> |Function Call Responses<br/>JSON Schema| B
    
    B --> |SemanticKernel Plugin with Function Definitions<br/>GetCustomer, GetAllCustomers<br/>GetSalesForCustomer, GetAllSales| E
    
    A --> |Also supports direct<br/>API endpoints| B
    
    style A fill:#e1f5fe
    style B fill:#f3e5f5
    style C fill:#e8f5e8
    style D fill:#fff3e0
    style E fill:#ffebee

Loading

Large Language Model 🧠

The LLM in use is qwen3:1.7b as it fully supports function calling and offers a good tradeoff between context size and performance.

Container Compatibility 🐳

Universal Container Runtime Support: The application now works seamlessly with both Docker and Podman through improved Aspire endpoint management:

• Docker: Uses standard Docker networking
• Podman: Compatible with Podman's container networking
• Aspire Integration: Automatic service discovery and endpoint resolution

API Architecture 🏗️

Facade Pattern Implementation: The Athena API serves dual purposes:

1. Business Logic API: Custom endpoints for customer and sales data with AI-powered analysis
2. Ollama API Facade: Acts as a proxy/gateway providing OpenAI-compatible endpoints
   • Open-WebUI connects through this facade instead of directly to Ollama
   • Enables seamless integration with any OpenAI-compatible client
   • Maintains consistent API patterns across the application

Getting started 🚀

Prerequisites

• .NET 9.0 SDK
• Docker or Podman
• VS Code (optional, but recommended)

Quick Start

1. Launch with VS Code: Open the project and use F5 to start debugging with the "Athena AppHost (HTTP)" configuration

   • The Aspire dashboard will automatically open in your browser
   • All services (PostgreSQL, Ollama, Open-WebUI) will be orchestrated automatically

2. Or launch manually:

   cd src/Athena.AppHost
   dotnet run --launch-profile http

3. Wait for initialization:

   • PostgreSQL database will be created and seeded with sample data
   • Ollama will automatically pull the qwen3:1.7b model (first run only)
   • Open-WebUI will be available at http://localhost:8181

4. Start chatting: Open http://localhost:8181 and ask questions like:

   • "Show me all customers"
   • "What sales orders does customer CUST-001 have?"
   • "Analyze the sales patterns for our top customers"

Available Endpoints

• Aspire Dashboard: http://localhost:15230 (auto-opens with VS Code)
• Open-WebUI: http://localhost:8181 (connects through API facade)
• API Documentation: http://localhost:{api-port}/scalar (Scalar API docs)
• API Endpoints:
  • GET /api/customers - List all customers
  • GET /api/customers/{customerId} - Get specific customer
  • GET /api/sales-orders - List all sales orders
  • GET /api/sales-orders/customer/{customerId} - Get customer's orders
  • POST /api/chat - AI chat endpoint (SemanticKernel-powered)
• Ollama API Facade: Full OpenAI-compatible API proxy to Ollama
  • Accessible through the Athena API for seamless integration
  • Supports all standard OpenAI chat completion endpoints

Entity Framework Migrations 🗄️

For Entity Framework design-time operations (migrations, database updates), you need to set the connection string as an environment variable:

# Set the connection string for EF migrations
export ConnectionStrings__postgres-default="Host=localhost;Database=athena;Username=postgres;Password=postgres"

# Then run EF commands
dotnet ef migrations add YourMigrationName
dotnet ef database update

The application uses Entity Framework Core with PostgreSQL to manage customer and sales order data.

Sample Interactions 💬

Customer Analysis

👤 User: "Show me all customers"
🤖 AI: Lists all customers with their details and provides insights

👤 User: "What can you tell me about customer CUST-001?"
🤖 AI: Provides customer details and their sales history

👤 User: "Which customer has the highest total order value?"
🤖 AI: Analyzes sales data and identifies top-performing customers

Sales Pattern Analysis

👤 User: "Analyze the sales trends for the last month"
🤖 AI: Reviews recent sales orders and identifies patterns

👤 User: "What's the average order value across all customers?"
🤖 AI: Calculates and explains sales metrics with business insights

The AI assistant acts as a sales analyst providing data-driven insights and recommendations based on real customer and sales order data.

Technology Stack 🛠️

• .NET 9.0 - Application runtime
• ASP.NET Core - Web API framework with facade pattern
• Entity Framework Core - Data access layer
• PostgreSQL - Database with automatic seeding
• SemanticKernel - AI orchestration and function calling
• Ollama - Local LLM hosting (qwen3:1.7b model) with enhanced startup
• OllamaApiFacade - Excellent NuGet package by Gregor Biswanger for OpenAI-compatible Ollama integration
• .NET Aspire - Cloud-ready orchestration with universal container support
• Open-WebUI - Chat interface (connects via API facade)
• Scalar - Interactive API documentation
• Docker/Podman - Container runtime (both supported)

Acknowledgments 🙏

Special thanks to Gregor Biswanger for the fantastic OllamaApiFacade NuGet package, which provides seamless OpenAI-compatible API in ASP.NET Core services.