pi-chat

A project showcasing a containerized chat application, which communicates and causes actions on a raspberry pi

Project Structure

• webapp/ - Python Flask web application with chat interface powered by Azure AI Foundry
• functions/ - Azure Functions for MCP endpoints (GetTelemetry and SendAction)
• infra/ - Infrastructure as Code (Bicep) for Azure resources
• raspberry-pi/ - Python applications for receiving telemetry and action messages on Raspberry Pi

Architecture

User Browser
    ↓ (HTTPS)
Web Application (Azure App Service - Flask)
    ↓ (Azure AI SDK)
Azure AI Foundry (GPT-4o-mini)
    ↓ (Function Calling / MCP Tools)
Azure Functions (MCP Endpoints)
    ↓ (Service Bus)
Azure Service Bus Topics (Telemetry & Action)
    ↓ (Subscriptions)
Raspberry Pi Services
    ├─ Telemetry Receiver → Sensor Reading
    └─ Action Receiver → Action Execution

Components

Web Application

A Python Flask-based web application that provides:

• Interactive chat interface for users
• Integration with Azure AI Foundry for intelligent conversations
• MCP tool integration for function calling to control Raspberry Pi
• Application Insights telemetry and monitoring

See webapp/README.md for details.

Azure Functions

HTTP endpoints that receive requests and forward them to Service Bus topics:

• GetTelemetry - Sends telemetry requests to the Telemetry topic
• SendAction - Sends action requests to the Action topic

See functions/README.md for details.

Raspberry Pi Services

Two Python services that run on the Raspberry Pi:

1. Telemetry Receiver - Receives messages from the Service Bus Telemetry topic and processes them based on the SensorKey field (Temperature, Light, CPU).

2. Action Receiver - Receives messages from the Service Bus Action topic and processes them based on the ActionType field (Camera, etc.).

See raspberry-pi/README.md for installation and usage.

Infrastructure

Azure resources deployed using Bicep templates, including:

• Container Registry
• App Service (for web application)
• Function App (for MCP endpoints)
• Service Bus (with topics and subscriptions)
• OpenAI Service (GPT-4o-mini)
• AI Foundry Hub
• Application Insights (monitoring and telemetry)

See infra/README.md for deployment instructions.

Quick Start

1. Deploy Infrastructure

# Create resource group
az group create --name rg-pichat-dev --location eastus

# Deploy infrastructure
cd infra
az deployment group create \
  --resource-group rg-pichat-dev \
  --template-file main.bicep \
  --parameters main.bicepparam

2. Deploy Azure Functions

cd functions
func azure functionapp publish <function-app-name>

3. Build and Deploy Web Application

# Build and push container image
cd webapp
az acr build --registry <acr-name> --image chat-app:latest .

# Configure App Service with environment variables (see webapp/README.md)

4. Configure Raspberry Pi Services

See raspberry-pi/README.md for setup instructions.

Usage

Once deployed, access the web application at your App Service URL (e.g., https://pichat-dev-app.azurewebsites.net). The chat interface allows you to:

• Ask questions about sensor data: "What's the current temperature?"
• Control devices: "Take a picture with the camera"
• Get system information: "What's the CPU usage?"

The AI will automatically call the appropriate Azure Functions to interact with your Raspberry Pi.