• Hexagon microservice architecture - implemented as domain, adapter and service sub-projects for each microservice
• KotlinJS - NodeJS transpiling
• PubSub - event-driven messaging
• Firestore - NoSQL database
• Docker - containerised deployment
• Cloud Run - serverless deployment of microservices
• Cloud Scheduler - cron jobs
• GitHub Actions - CI automation
• Pulumi - infrastructure as code, using micro-stacks
• nginx - web reverse proxy
• Google Analytics - analytics and engagement measurement

1. Create a new Google Cloud project.
2. Create a Service Account for the infrastructure as code setup with the following roles:
   • Artifact Registry Administrator
   • Firebase Admin
   • Service Account User
   • Service Usage Admin
   • Pub/Sub Admin
   • Cloud Scheduler Admin
   • Cloud Run Admin
   • (Optional) If you're specifying a custom domain mapping, as we are, verify domain ownership and add your service account as owner.
3. Export a JSON API key for your Service Account and call it credentials-gcp-infra.json.
4. Signup and Install Pulumi locally.
5. Create a Pulumi access token and login locally using pulumi login.
6. The project uses Pulumi micro-stacks to deploy the microservices individually. Each microservice has a corresponding infra folder containing its Pulumi.yaml infrastructure program, eg search/infra. To get the project going, you will need to manually initialise each microservice on GCP using the Pulumi scripts.

The order to do this matters, so go with common/infra first, then all other microservices, then proxy-web/infra. The reason is that the resouces are created incrementally at each stage and we currently have no way synchronize them.

7. The steps to deploy a microservice's infrastructure is the same for all:
   1. Navigate to its infra folder.
   2. Paste the credentials-gcp-infra.json file.
   3. Create a new empty Pulumi project with no resources using the pulumi new command and follow the instructions:
      • you can use the prompt Empty project with no resources for Pulumi AI;
      • you can use prod as your stack name;
   4. Replace the name in the microservice Pulumi.yaml with the value you entered in the prompt.
   5. Open Pulumi.prod.yml and replace the gcp:project value with your project id.
   6. Run pulumi up to automatically create the required microservice infrastructure.
   7. Repeat for the remaining microservices.
8. Find your auto-generated firebase-adminsdk Service Account and give it the following additional roles:
   • Pub/Sub Publisher, for publishing messages to PubSub topics;
9. Export a JSON API key for your firebase-adminsdk Service Account and call it credentials-gcp-app.json - the app will need it later.

The project powers an existing Slack app, so you'll need one in order to run it.

1. Create a new Slack app.
2. You will need an SSH tunnel to your localhost for Slack's APIs. You can use serveo.net for free and configure it with this command ssh -R YOUR_DOMAIN.serveo.net:80:localhost:3000 serveo.net.
3. Point the following Slack features to the relevant project API endpoints that know how to respond to them using the url you used for serveo.net:
   • Slash commands -> YOUR_DOMAIN.serveo.net/api/slack/slash
   • OAuth -> YOUR_DOMAIN.serveo.net/api/slack/auth
   • Events -> YOUR_DOMAIN.serveo.net/api/slack/event
   • Interactivity -> YOUR_DOMAIN.serveo.net/api/slack/interactivity
4. Make a note of your Slack Client ID, Secret and Signing Secret.

1. Install Docker Desktop and start it up. No additional configuration is required as the project uses Docker Compose to run locally. Checkout the docker folder for the setup.
2. Install IntelliJ. This project has been tested with IntelliJ IDEA 2023.2.5.
3. Open the root project with IntelliJ and wait for it to initialise.
4. Create a Slack channel to receive server error messages and monitoring updates. The project is configured to post all unhandled Throwables to that channel. We use the Incoming Webhooks app to obtain a channel URL webhook.
5. Create a Google Analytics 4 property and an API secret.
6. Create the following secrets.properties files:

# Under /slack/domain/secrets.properties
SLACK_SIGNING_SECRET=YOUR_SLACK_SIGNING_SECRET
SLACK_CLIENT_ID=YOUR_SLACK_CLIENT_ID
SLACK_CLIENT_SECRET=YOUR_SLACK_CLIENT_SECRET

# Under /slack-web/domain/secrets.properties
SLACK_CLIENT_ID=YOUR_SLACK_CLIENT_ID

# Under /common/monitoring/secrets.properties
MONITORING_SLACK_URL=YOUR_MONITORING_SLACK_URL

# Under /common/analytics/secrets.properties
GOOGLE_ANALYTICS_MEASUREMENT_ID=YOUR_ANALYTICS_MEASUREMENT_ID
GOOGLE_ANALYTICS_API_SECRET=YOUR_ANALYTICS_API_SECRET

7. Copy the credentials-gcp-app.json Service Account JSON API key to the root project folder.

The project is configured to build with GitHub Actions and have a separate workflow for each microservice. Checkout the .github folder for details. Follow these steps to configure the CI environment:

1. Add your Pulumi access token as a GitHub encrypted secret with the name PULUMI_ACCESS_TOKEN.
2. Each of the variables defined in the secrets.properties files you created above should be exposed as GitHub encrypted secrets, using the same names as keys.
3. Add an additional GCP_SA_KEY_INFRA GitHub encrypted secret, containing the raw JSON API key for the above infrastructure as code Service Account.
4. Add an additional GCP_SA_KEY_APP GitHub encrypted secret, containing the raw JSON API key for the firebase-adminsdk Service Account.
5. (Optional) Install the Pulumi GitHub app to get automated summaries of your infrastructure as code changes directly on your PR.

Once this is done:

• opening pull requests against the repo will trigger build/test checks as well as infrastructure changes preview for the microservice that has been changed;
• merging pull requests to the main branch deploys the changes to the corresponding microservice to Google Cloud;