AI News Audio Briefing

This is a demo application that converts AI news articles into audio briefings. The project demonstrates how to build an automated pipeline using Google Cloud services, including Gemini 2.5 Pro for content summarization and Cloud Text-to-Speech for audio generation.

The application uses Cloud Run Job for content processing and Cloud Run Service for the web interface. It fetches RSS feeds, creates summaries, and generates audio files for demonstration purposes.

Architecture Diagram

Features

Content Processing

• Summarizes RSS feeds from AI news sources using Gemini 2.5 Pro
• Processes multiple RSS feeds and selects recent articles
• Converts technical articles to conversational summaries

Audio Generation

• Converts summaries to MP3 audio files using Google Cloud Text-to-Speech
• Combines intro, content segments, and outro into complete episodes
• Generates standard MP3 files for web playback

Infrastructure

• Cloud Run Job handles batch processing (2 CPU, 2GB memory, 1-hour timeout)
• Cloud Run Service provides web interface for listening to episodes
• Cloud Scheduler triggers daily processing at 6 AM Europe/Rome time
• Terraform configuration for deployment

Access Control

• Optional IAP authentication via email allowlist
• Private GCS storage with signed URLs for audio file access
• Service accounts with minimal required permissions

Project Structure

/
├── job/
│   ├── main.py         # The backend Cloud Run Job logic
│   ├── Dockerfile      # Container definition for the job
│   └── requirements.txt  # Python dependencies for the job
│
├── webapp/
│   ├── main.py         # The frontend Cloud Run Service logic
│   ├── Dockerfile      # Container definition for the service
│   └── requirements.txt  # Python dependencies for the service
│
├── terraform/
│   ├── main.tf         # Terraform configuration for the entire stack
│   └── ...             # Other Terraform files
│
└── README.md           # This file

---

Cloud Deployment Guide

This demo application can be deployed using the provided Terraform configuration.

Prerequisites

1. Terraform CLI installed.
2. Google Cloud CLI installed and authenticated (gcloud auth application-default login).
3. A Google Cloud project with billing enabled.
4. Permissions to enable APIs and create all the resources defined in the Terraform configuration (e.g., Owner or Editor roles).

Deployment Steps

1. Navigate to the Terraform Directory:

   cd terraform

2. Create a terraform.tfvars file: This is the most secure way to provide your project-specific variables. Create the file and add the following content, replacing the placeholder values:

   project_id      = "your-gcp-project-id"
   gcs_bucket_name = "your-globally-unique-bucket-name"
   gemini_api_key  = "your-gemini-api-key"
   
   # Optional: Enable IAP authentication (leave empty for public access)
   iap_allowed_emails = []
   # iap_allowed_emails = ["admin@yourcompany.com"] # Uncomment to enable auth

3. Initialize Terraform: This command downloads the necessary provider plugins.

   terraform init

4. Plan the Deployment: This command shows you what resources will be created. It's a dry run and is safe to run.

   terraform plan

5. Apply the Configuration: This command builds the container images and deploys the cloud resources.

   terraform apply

   Terraform will ask for confirmation. Type yes to proceed. The deployment will take several minutes.

6. Access the Application: Once complete, Terraform will output the URL of the web application. If IAP is enabled, users will authenticate with Google before accessing the app.

Deploying Application Updates

If you only need to update the application code for the backend job or the frontend web app, you do not need to re-run terraform apply. You can deploy updates directly using the gcloud CLI.

Updating the Backend Job (rss-audio-generator-job)

After making changes to the code in the job/ directory:

gcloud run jobs deploy rss-audio-generator-job \
    --source=./job \
    --project=<YOUR_PROJECT_ID> \
    --region=europe-west1 

Updating the Frontend Web App (rss-summaries-webapp)

After making changes to the code in the webapp/ directory:

gcloud run deploy rss-summaries-webapp \
    --source=./webapp \
    --project=<YOUR_PROJECT_ID> \
    --region=europe-west1

---

Local Testing

Testing the Web App

The web application can be tested locally, but it requires you to have a GCS bucket with audio files already in it.

1. Navigate to the webapp directory:

   cd webapp

2. Activate a virtual environment:

   python3 -m venv venv && source venv/bin/activate
   pip install -r requirements.txt

3. Authenticate to Google Cloud:

   gcloud auth application-default login

4. Set Environment Variables:

   export GCS_BUCKET_NAME="<YOUR_GCS_BUCKET_NAME>"

5. Set Additional Environment Variables (for signed URLs):

   export SERVICE_ACCOUNT_EMAIL="<YOUR_WEBAPP_SERVICE_ACCOUNT_EMAIL>"

6. Run the Flask App:

   flask --app main run

   You can now access the web app at http://127.0.0.1:5000.

Testing the Backend Job

The backend job is designed to run in the cloud and relies on the Cloud Run environment's service account and metadata server. Therefore, local testing of the job is not recommended. The most reliable way to test it is to trigger it in the cloud after deployment.