Skip to content


Repository files navigation

This repository has source code that runs, which is Google Cloud DevRel's demo site to celebrate Pi Day.

This is not an official google product.


The server is written in Go and runs in Cloud Functions.

The entry point is the Get function in functions.go.


Server architecture diagram. There's a Cloud Load Balancer in the front that redirects requests to Cloud Function instances in us-central1, europe-west1, asia-northeast1 regions. The functions connect to Cloud Storage in the US multi-region. Logging and Monitoring are used for monitoring. Cloud DNS is used for DNS resolutions.

Cloud configurations are managed with Terraform. The configuration file is, and uses the Google Cloud Platform Provider.

Most of the configurations are done by terraform with a few exceptions.

terraform init
terraform apply

Functions (deploy with gcloud)

Set environment variables (we use Bash extensions here).

export PROJECT=piaas-gcp
export REGIONS=(us-central1 europe-west1 asia-northeast1)
export STAGE_BUCKET=piaas-gcp-gcf-staging


gcloud beta functions deploy api-pi-staging --gen2 --runtime go116 --trigger-http --entry-point Get --source . \
  --stage-bucket=$STAGE_BUCKET --ingress-settings=internal-and-gclb --region=$REGIONS[1] \
  --allow-unauthenticated --service-account=$GCF_API_SA --project=$PROJECT

The staging API is accessible via e.g.

curl -v


The whole setup is just experimental but we call it "production".

for R in $REGIONS; \
  gcloud beta functions deploy api-pi --gen2 --runtime go116 --trigger-http --entry-point Get --source . \
    --stage-bucket=$STAGE_BUCKET --ingress-settings=internal-and-gclb --region=$R \
    --allow-unauthenticated --service-account=$GCF_API_SA \
    --min-instances=1  --project=$PROJECT

The production API is accessible via

curl -v

Utility commands

There are several utility commands under cmd/.


The dtob command is a small tool to generate binary data used in some of the test cases. It takes a pi digit text as input and generates hex values that you can paste in a Go source file. For example, pkg/unpack/reader_test.go have test cases that check decoding of compressed digits that use outputs of this program.

tail -c +3 pi.txt | head -c 100 | go run cmd/dtob/main.go -b 30

The output should look like this:

0x60, 0xe2, 0x3e, 0xb8, 0xae, 0x61, 0xa6, 0x13,
0x00, 0x0f, 0x58, 0xf3, 0x84, 0x66, 0xef, 0x56,
// Block Boundary
0x17, 0x3f, 0x65, 0x1a, 0x21, 0x09, 0xca, 0x45,
0x00, 0x60, 0x5b, 0x4a, 0x96, 0x06, 0x14, 0x08,
// Block Boundary
0x09, 0xfb, 0xd6, 0x59, 0x35, 0x00, 0x33, 0x52,
0x00, 0xe9, 0xe5, 0x0f, 0x83, 0xb7, 0xdf, 0x88,
// Block Boundary
0x00, 0xe6, 0xc6, 0x3d, 0x9b, 0x70, 0x7a, 0x2f,
// Block Boundary


This is a command line version of the API that uses the same code to fetch and parse ycd files. The major difference is that you can fetch as many digits as you'd like with this program.

go run ./cmd/extract -s 42 -n 2000


The indexer program is used to generate the index files that the API needs to determine which object to fetch. The generated file needs to be in gen/index/.

go run ./cmd/indexer --bucket pi50t >  gen/index/index.go


This is a command line emulator of the Functions API. Check out functions-framework-go to learn more about the framework.


The frontend is developed with Jekyll and React.


You need Yarn and Ruby to compile the frontend. We use Yarn 3 and Ruby 3.1.

To install dependencies, run



There is a Yarn script to build and monitor changes for both Jekyll and Webpack and run a development server.

yarn serve


To build minimized files for deployment, run:

yarn prod


We use Firebase Hosting to server the frontend files.

Firebase hosting is separately managed with the Firebase command. To install the tool, run the following command:

npm install -g firebase-tools
firebase login

Make sure to build the latest production build before testing or deploying with Firebase. The files will be generated in dist directory.

yarn prod

Test the configuration (such as firebase.json) locally by running:

export PROJECT=piaas-gcp
firebase emulators:start --project=${PROJECT}

You can also create preview channels to test the website on Firebase servers. CHANNEL_ID should be set to an identifier you want to use (e.g. preview)

export CHANNEL_ID=preview
firebase hosting:channel:deploy ${CHANNEL_ID} --project=${PROJECT}

After confirming the preview, run the following command to serve it live:

firebase hosting:clone ${PROJECT}:${CHANNEL_ID} ${PROJECT}:live

See the official documentation for more details on the testing and deployment process with Firebase.


No description, website, or topics provided.



Code of conduct





No releases published


No packages published