1. Overview

A cloud-based system that leverages GPT completions API to recommend tour itineraries for a given city

2. Architecture

# itinerary-recommender

3. To Debug locally (VS Code)

Requirements

• Azure account + subscription
• Azure Tools for VSCode
• .Net 6

Steps

1. Clone this repository to the local environment
2. Setup Azure resources on Azure cloud using the Azure tools see link. You must be signed in see link
3. Deploy the Azure Function App using the Azure tools see link
4. Download the function app configuration settings to your local project. see link
5. The downloaded local.settings.json file will contain some settings. Update the file with the following configurations - GPT_API_KEY, GPT_API_BASE_URL, GPT_API_COMPLETION_ENDPOINT, EMAIL_ACCOUNT, EMAIL_PASSWORD, SMTP_HOST, SMTP_PORT.
6. The local.settings.json file should be similar to the following template after the update.

{
  "IsEncrypted": false,
  "Values": {
    "AzureWebJobsStorage": "DefaultEndpointsProtocol=https;AccountName=<your-account-name>;AccountKey=<your-storage-account-key>;EndpointSuffix=core.windows.net",
    "FUNCTIONS_WORKER_RUNTIME": "dotnet",
    "FUNCTIONS_EXTENSION_VERSION": "~4",
    "WEBSITE_CONTENTAZUREFILECONNECTIONSTRING": "DefaultEndpointsProtocol=https;AccountName=<your-account-name>;AccountKey=<your-storage-account-key>;EndpointSuffix=core.windows.net",
    "WEBSITE_CONTENTSHARE": "<your-account-value>",
    "WEBSITE_RUN_FROM_PACKAGE": "1",
    "APPINSIGHTS_INSTRUMENTATIONKEY": "<your-instrumentation-key>",
    "GPT_API_KEY": "<your-gpt-api-key>",
    "GPT_API_BASE_URL": "https://api.openai.com",
    "GPT_API_COMPLETION_ENDPOINT": "/v1/completions",
    "EMAIL_ACCOUNT": "<your-email-to-send-itinerary-from>",
    "EMAIL_PASSWORD": "<your-email->",
    "SMTP_HOST": "<your-smtp-host>",
    "SMTP_PORT": "<your-smtp-port>"
  }
}

6. Start Debugging

NB:

• GPT_API_KEY can be obtained from OpenAI by subscribing to their API service
• If perhaps the GPT_API_BASE_URL, GPT_API_COMPLETION_ENDPOINT has changed as at the time of running this code, update the values in the template as necessary.
• The EMAIL_PASSWORD in the template refers to App Password. I had to use the App password because I could not authenticate with my google account password. It may not be the best option but it is open for improvement.

4. To run on Azure Cloud

This repository includes a GitAction workflow to automatically deploy the app on Azure automatically on push to master with some minor updates to the workflow and repository secrets.

Requirements

• GitHub Account
• Azure account + subscription

Steps

1. Create a service principal and assign a `contributor`` role to it. See Link.
2. Create a new service principal client secret. See Link. Ensure to copy the secret immediately after creation

NB: GitAction uses this service principal to authenticate to Azure cloud to setup infrastructure and deploy the function app.

3. Fork this repository and create the following repository secrets:

• AZURE_CREDENTIALS = {"clientSecret": "******", "subscriptionId": "******", "tenantId": "******", "clientId": "******"}
• EMAIL_ACCOUNT = *****
• EMAIL_PASSWORD = ****
• GPT_API_KEY = ****

4. Customize the branch to trigger to workflow and env section of the workflow file - .github/workflows/deploy_azure_function_app.yml as needed.

5. push changes to the branch for which the workflow is configured.

NB:

• The subscriptionId, tenantId values are specific to the Azure Account. The clientId refers to the ApplicationId of the service principal(App registered) in step 1 and the clientSecret refers to the secret created in step 2.
• GPT_API_KEY can be obtained from OpenAI by subscribing to their API service
• If perhaps the GPT_API_BASE_URL, GPT_API_COMPLETION_ENDPOINT has changed as at the time of running this code, update the values in the env section of the workflow file.
• The EMAIL_PASSWORD in the template refers to App Password. I had to use the App password because I could not authenticate with my google account password. It may not be the best option but it is open for improvement.
• Ensure to make the Azure names as unique as possible and stick to Azure naming conventions see link

REMEMBER TO CLEAN UP THE CLOUD RESOURCES AFTER TESTING See Link