Frankfurter

Frankfurter is a free and open-source currency data API that tracks reference exchange rates published by the European Central Bank.

api.frankfurter.app hosts a public instance of the API.

Getting Started

Get the latest exchange rates.

https://api.frankfurter.app/latest

Get rates for a past date.

https://api.frankfurter.app/2000-01-03

Get rates for a period.

https://api.frankfurter.app/2010-01-01..2010-01-31

For more examples, read the docs.

Configuring your own instance

You need to enter your own API keys and username/passwords for the following files:

cp config/newrelic.yml.example config/newrelic.yml
cp .env.example .env
cp telegraf.conf.example telegraf.conf

Deployment

You can self-host Frankfurter with Docker.

Using Docker

docker build -t frankfurter .

docker run -d -p 3000:3000 \
  -e "SQLITE_DB=frankfurtdb.db" \
  -e PORT=3000 \
  --name frankfurter \
  -it frankfurter

Check out http://localhost:3000/latest

Push to Docker Hub

In order to make deployment super-simple, we can push a version to Docker Hub, I simply push to palbrattberg/frankfurter.

To push your own, first create a new repository on Docker Hub, then build and push your image:

docker login
docker buildx create --use desktop-linux
docker buildx build --push --platform linux/arm/v7,linux/arm64,linux/amd64 -t palbrattberg/frankfurter:latest .

Load testing

We use k6 for load testing the API.

For local runs, just install k6 and go for it:

brew install k6
k6 run spec/load_test/stress.js

If you want nice visualization, you might want to push the output to Grafana. Be sure to update your local copy of telegraf.conf with good values for username, password and url, see k6 with Grafana Cloud.

brew install k6 telegraf 
ulimit -n 10240 # increase the number of open files allowed
telegraf -config /usr/local/etc/telegraf.conf
k6 run --out influxdb=http://localhost:8186 spec/load_test/stress.js

Now you should be able to see some stats in Grafana Cloud.

Running tests

# must have .env file present
cp .env.example .env # only run this one time

bundle exec rake

Miscellaneous

In this fork there is no rounding of the rates in order to support better precision for downstream applications.

This fork uses sqlite instead of PostgreSQL.