Azure Chatbot Integration

An ASP.NET Core MVC chatbot integration built around Azure OpenAI. The intended use case is a customer assistant inside a pizza application: system instructions define the assistant's role and scope, while the web interface sends customer questions to a configured model deployment.

This repository contains the standalone chatbot interface and integration layer, not a complete pizza ordering platform. It does not include a restaurant menu database, payment flow, or order-management API.

Features

• Azure OpenAI chat-completion integration.
• Configurable endpoint, deployment, system instructions, and output-token limit.
• A Razor chat interface with a session-scoped conversation transcript.
• MVC separation between controllers, view models, views, and session helpers.

Requirements

• .NET 8 SDK. global.json selects the .NET 8 SDK family.
• An Azure OpenAI resource with an available chat model deployment.
• The resource endpoint, API key, and deployment name.

The project references Azure.AI.OpenAI 2.1.0 and Newtonsoft.Json 13.0.3. dotnet restore installs those dependencies; do not add arbitrary newer package versions just to run the application.

Setup

git clone https://github.com/civanaban2/chatbot-project.git
cd chatbot-project
dotnet restore ChatBotProject.sln
dotnet build ChatBotProject.sln --no-restore

Set configuration using environment variables in the same terminal before starting the application. Replace placeholders with your resource values:

export OpenAIAPIKey='YOUR_AZURE_OPENAI_API_KEY'
export OpenAIEndPoint='https://YOUR_RESOURCE.openai.azure.com/'
export OpenAIDeploymentName='YOUR_DEPLOYMENT_NAME'
export MaxOutputTokenCount='100'

Use the deployment name configured in Azure, which may differ from the underlying model name. These environment variables override the placeholders in ChatBotProject/appsettings.json. Keep real credentials outside tracked files. ASP.NET Core reads environment variables directly; this project does not automatically load .env files.

The checked-in SystemMessage provides a pizza-assistant example. Override it with an environment variable if needed:

export SystemMessage='You are a customer assistant for a pizza application. Answer within the supplied application context and do not invent menu or order information.'

The example does not contain actual restaurant data. Prompts define intended behavior; they are not a substitute for an authoritative menu or order API.

Run

dotnet run --project ChatBotProject/ChatBotProject.csproj --launch-profile http

Open http://localhost:5288/Mth404Bot/Index. The root URL opens the template home page; the chat screen is a separate route. The http launch profile sets the Development environment. For an HTTPS launch, configure a development certificate and use the https profile instead.

Request flow

1. The browser submits UserInput to Mth404BotController.Index.
2. The controller loads the endpoint, credentials, deployment, system message, and token limit from ASP.NET Core configuration.
3. Each model call contains the system message and the current user message.
4. The reply is appended to the displayed transcript, serialized to JSON in HTTP session state, and rendered by the Razor view.

The stored transcript is not sent back to the model on subsequent turns. The interface displays previous messages, but the model does not have multi-turn conversational memory. Session data is held in process memory and is lost when the application restarts; it is not a durable conversation store.

Configuration keys

Key	Purpose
OpenAIAPIKey	Azure OpenAI resource API key
OpenAIEndPoint	Azure OpenAI resource endpoint
OpenAIDeploymentName	Azure model deployment name
SystemMessage	Assistant instructions and application scope
MaxOutputTokenCount	Positive integer output-token limit; default example is 100

Configuration keys are case-insensitive. The controller reads systemMessage, which resolves to the checked-in SystemMessage setting.

Structure

ChatBotProject.sln
global.json
ChatBotProject/
  Program.cs
  appsettings.json
  Controllers/MTH404BotController.cs
  Extensions/SessionExtensions.cs
  Models/ChatViewModel.cs
  Views/MTH404Bot/Index.cshtml
  Properties/launchSettings.json

The historical MTH404 names remain in the source and UI. They are project identifiers, not a separate AI model or Azure service.

Scope and limitations

This is an educational integration prototype. It does not implement RAG, menu retrieval, tools/agents, order placement, or persistent chat storage. API calls are synchronous, and API-failure handling, rate limiting, authentication, input limits, and production deployment controls require further work. Customer messages are sent to the configured Azure service; avoid using real customer data while evaluating the demo.

There is no automated test suite in the repository. Building checks local compilation; an end-to-end response also requires valid Azure configuration and a live model deployment.